research article attenuation correction effects in...
TRANSCRIPT
Research ArticleAttenuation Correction Effects in Rainfall Estimation atX-Band Dual-Polarization Radar Evaluation with a DenseRain Gauge Network
Young-A Oh1 DaeHyung Lee1 Sung-Hwa Jung2 Kyung-Yeub Nam2 and GyuWon Lee13
1Department of Astronomy and Atmospheric Sciences Research and Training Team for Future Creative Astrophysicists andCosmologists Kyungpook National University 80 Daehakro Bukgu Daegu 41566 Republic of Korea2Radar Analysis Division Weather Radar Center Korea Meteorological Administration 61 16-Gil Yeouidaebangro DongjakguSeoul 07062 Republic of Korea3Center for Atmospheric REmote Sensing (CARE) Kyungpook National University 80 Daehakro BukguDaegu 41566 Republic of Korea
Correspondence should be addressed to GyuWon Lee gyuwonknuackr
Received 25 December 2015 Accepted 24 April 2016
Academic Editor Hiroyuki Hashiguchi
Copyright copy 2016 Young-A Oh et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
The effects of attenuation correction in rainfall estimation with X-band dual-polarization radar were investigated with a denserain gauge network The calibration bias in reflectivity (119885
119867) was corrected using a self-consistency principle The attenuation
correction of 119885119867and the differential reflectivity (119885DR) were performed by a path integration method After attenuation correction
119885119867and 119885DR were significantly improved and their scatter plots matched well with the theoretical relationship between 119885
119867and
119885DR The comparisons between the radar rainfall estimation and the rain gauge rainfall were investigated using the bulk statisticswith different temporal accumulations and spatial averagesThe bias significantly improves from 70 to 0 with 119877(119885
119867) However
the improvement with 119877(119885119867 119885DR) was relatively small from 3 to 1 This indicated that rainfall estimation using a polarimetric
variable was more robust at attenuation than was a single polarimetric variable method The bias did not show improvement incomparisons between the temporal accumulations or the spatial averages in either rainfall estimationmethodHowever the randomerror improved from 68 to 25 with different temporal accumulations or spatial averages This result indicates that temporalaccumulation or spatial average (aggregation) is important to reduce random error
1 Introduction
Weather radar has provided useful information abouthydrometeors for various meteorological and hydrologicalapplications since its introduction inmeteorological observa-tionsWeather radar can observe various parameters ofmete-orological phenomena with high spatiotemporal resolutionover wide observational areas However it suffers frommanyerror sources including radar calibration beam shieldingattenuation bright band contamination beam broadeningand anomalous propagation Many researchers have studiedthe characteristics of these error sources and ways to improvethe quality of radar data [1ndash3]
Recently X-band dual-polarization radar has receivedsignificant attention due to its finer resolution ease of
mobility and lower cost compared to longer wavelengthradars However the attenuation is a very important issue inX-band since it is inversely proportional to radarwavelengthThe values of the radar parameters associated with the back-scattered power such as reflectivity (119885
119867) and differential
reflectivity (119885DR) can be significantly reduced by attenuationThe attenuation is caused by absorption and scattering in
the medium along the propagation path of the radar beamThe attenuation is severe where the medium is dense and itscomposition size is largeThe attenuation by atmospheric gasmolecules is small being about 15 dB per 100 km in the C-band [3] However media such as rain drops hail andmeltedsnow particles can cause significant attenuation especiallyin shorter wavelength radar The amount of attenuation in
Hindawi Publishing CorporationAdvances in MeteorologyVolume 2016 Article ID 9716535 20 pageshttpdxdoiorg10115520169716535
2 Advances in Meteorology
the C-band can reach about 12 dB due to strong convectivecells as compared to the S-band [3] Furthermore the attenu-ation accumulates over distances as a radar beam propagatesThe power is frequently lost completely over far rangesTherefore it is not possible to monitor and analyze severeweather such as heavy rains typhoons and heavy snowswithout the proper correction of the attenuation in the C- orX-band radars
The attenuation correction has been investigated bymanyresearchers Tuttle and Rinehart [4] suggested an attenuationcorrection method using dual-wavelength (S- and X-band)radar measurements with a relationship between specificattenuation (119860
119867) and 119885
119867 Recently the specific differential
propagation phase shift (119870DP) has been used widely forattenuation correction because 119870DP is not affected by radarpower calibration attenuation and partial beam blockingIn addition 119870DP is less sensitive to the natural variability ofdrop size distributions (DSDs) in rainfall estimation Bringiet al [1] showed that both 119860
119867and the specific differential
attenuation (119860DP) are almost linearly related to 119870DP throughscattering simulations and these relationships have beenaccepted by many researchers However the coefficients ofthese relationships derived from the scattering simulationsvary significantly with temperature DSD variability and thedrop deformationmodel Park et al [5] showed that the coef-ficients of 119860
119867-119870DP and 119860DP-119860119867 relationships vary greatly
from 0139 to 0335 dB(∘)minus1 and from 0114 to 0174 dB(∘)minus1 inthe X-band respectively due to changes in temperatures andthe different drop deformation models
While 119870DP has many advantages in the attenuationcorrection as described above estimating 119870DP from themeasured total differential phase shift (ΨDP) is challengingbecause of the backscatter differential phase shift (120575) and themeasurement errors for ΨDP [1] Scarchilli et al [6] used aniterationmethod with a 120575-119885DR relationship to remove 120575 fromΨDP and corrected the attenuation using estimated 119870DP inthe C-band Anagnostou et al [7] Kalogiros et al [8] andChang et al [9] used an iteration method in the X-bandAs a different method for removing 120575 Hubbert and Bringi[10] separated 120575 and the differential propagation phase (ΦDP)from ΨDP through an iterative filtering technique (FIR) andcalculated the hail signal from 120575 Park et al [5] also usedthe same filtering technique to extractΦDP for correcting theattenuation and the differential attenuation in the X-bandRecently Mishra et al [11] investigated dual-polarimetricproducts and three types of rainfall estimation results withhigh spatiotemporal resolution data from X-band radar S-band radar and disdrometer data with an intercomparisonand T-matrix simulation method [12] Mishra et al [11] useda very similar attenuation correction process in their workwhere 119860
119867was calculated by an FIR filtered ΦDP and 119860DP
was calculated by an 119860DP-119860119867 relationship based on a self-consistency method
Two rainfall estimationmethods119877(119885119867) and119877(119885
119867 119885DR)
were used for evaluation of the attenuation correction effectThe self-consistency based correction of the attenuation andthe differential attenuation from the FIR filtered ΦDP wereused to improve the accuracy in the rainfall estimation
methods by X-band polarimetric radar The accuracy eval-uation of the rainfall estimation methods was performedby comparisons with rain gauge measurements with highspatial resolution However there were instrumental uncer-tainties and errors in the representativeness of the raingauge measurements The instrumental error has severalsources including the gauge calibration wind effects andwettingevaporation loss [13 14]The representativeness erroris linked with the spatial sampling area and the tempo-ral accumulation of the rain gauge [15] Habib et al [16]investigated sampling error of the tipping bucket gauge asfunction of accumulation times in a simulation study Ciach[17] verified local random errors in the tipping bucket gaugethrough experimental data Habib et al [18] investigatedthe correlation with separation distances in a rain gaugecluster In this study we considered the calibration effectsin the instrumental error of the rain gauge Instrumentalerror from the calibration effect was removed by laboratoryexperiments to providemore accurate and reliable referencesThe representativeness error was investigated by the temporalaccumulation and spatial average of rainfall amount
In Section 2 the radar and rain gauge data used inthis study are presented The methods for the attenuationcorrection byΦDP the rainfall estimation and its verificationand the analysis of the natural variability of rain fields usingrain gauge data are described in Section 3 The analysisresults are shown in Section 4 The results are summarizedin Section 5
2 Data
21 Radar Data The radar data from the X-band dual-polarimetric radar (NIMS-XPOL) of the National Institute ofMeteorological Sciences (NIMS) was used in this study Thegeneral characteristics of the NIMS-XPOL data are shownin Table 1 The NIMS-XPOL can be installed onto a five-tontruck bed (Figure 1(a)) and mounted on a steel tower bed(Figure 1(b))
The NIMS-XPOL was operated during June 2010 atMuan-gun Jeollanam-do which is on the southwest coastof Korea (latitude = N 350940∘ longitude = E 126285∘) forthe observation of heavy rains during the summer monsoonseason The rain estimation in Section 32 was performedwith the NIMS-X radar data collected during 55 hours from1600UTC to 2130UTC on July 10 2010 During this rainevent the NIMS-X was mounted on a steel tower bed inthe Muan-gun observation center to minimize the effect ofbeam shielding and ground clutter and a full volume scanwith 14 Plan Position Indicators (PPIs) was performed every25minThe gate spacing was 150m and the Pulse RepetitionFrequency (PRF) was 999Hz And the total rainfall amountwas over 50mm The evolution of the rain field is shown inFigure 2 The echoes moved toward the northeast and thesmall and weak rain cells near the radar site at 1601UTCwerereplaced by the strong convective cells approaching from thesouthwest
Advances in Meteorology 3
Table 1 General characteristics of the NIMS-XPOL radar [21]
X-band polarimetric mobile Doppler weather radar (ARC-X250MP) specification
Transmitter
Type MagnetronPeak power (kW) 250Frequency (MHz) 9360
PRF (MHz) 0sim2500
Antenna
Diameter (m) 244Polarization Orthogonal linear simultaneous H amp VBeamwidth (∘) 095Gain (dB) gt43
Data processingNumber of gates 30001000 PRFGate spacing 30m to 1000mRaw data 119885
119867 119881119905 119878119882 119885DR ΦDP 120588HV
(a) (b)
Figure 1 Photographs of the NIMS-XPOL radar installed onto (a) a truck bed and (b) a steel tower bed (courtesy of NIMS)
22 Rain Gauge Data The rain gauge data was collectedby ten tipping bucket (TB) rain gauges operated by theKyungpook National University (KNU) Korea The tip sizeof all of the gauges was 02mm and the diameter of thereceiving orifice was 1539 cm The time resolution was 05 sand the data was recorded based on the event (tipping) timeTo compare with the estimated rain rate (119877) from the NIMS-XPOL radar the rain accumulation from ten gauges wascollected during the month of July 2010 near Unnam-myeonwhich is located at a distance of about 16 km at 160∘ azimuthangle from the NIMS-XPOL radarThe area was only slightlyaffected by beam shielding of the NIMS-XPOL Consideringthe geographical situation and the limited locations for theinstallation all gauges were installed on the rooftops ofbuildings Figure 3 shows the location of the NIMS-XPOLand the deployment of the ten gauges (R is the radar siteand a number represents a gauge location) The total rainfallamount for the month at each gauge site is shown in Table 2At all gauge sites the total rainfall accumulation was over300mm for themonthThe rainfall amountmeasured duringthe same period with the NIMS-XPOL radar data used inthe rainfall estimation was about 50mm per 55 hours (seeTable 2)
The instrumental biases of all gauges were calculatedthrough an ideal experiment using laboratory and fieldobservations for 9 months from October 2009 to June 2010(except April 2010) at KNU The rain gauge dataset was
corrected for the instrumental bias prior to the detailedanalysis
3 Methodology
31 Attenuation Correction AlgorithmUsing Differential Prop-agation Phase Shift The attenuation correction algorithmwas performed by the procedure as shown in Figure 4 whichincluded (1) quality control of the radar data (2) estimationofΦDP (3) calculation of the attenuation amount and (4) theattenuation correction of 119885
119867and 119885DR
As a part of the quality control of the radar data 119885119867and
119885DR were corrected for instrumental biases 119885119867calibration
bias of about 37 dB was obtained by comparison with thereflectivity from the particle size velocity disdrometer (PAR-SIVEL) and from theNIMS-XPOL radarThe119885DR calibrationbias of minus15 dB was calculated from the vertical pointing data[19]The isolated point echo shows that the119885
119867values in their
surrounding pixels existed by less than 50 of the total pixelsto be removed The 120588HV threshold of 09 was applied to theentire radar measurement field
After applying the quality control measures ΦDP wasestimated from measured ΨDP by eliminating 120575 and theobservational noise through an iterative filtering techniqueIn this study the 20th FIR filter of Hubbert and Bringi [10]was used This filter preserved a monotonic increasing trendofΦDP due to the propagation medium while it removed the
4 Advances in Meteorology
1801 1816 1831
19011846
1931 1946 2001
1916
605550454030 35 65 702015100 25minus10
(dBZ)
Figure 2 PPI images of attenuation corrected 119885119867from 1801UTC July 10 2010 with an interval of 15min
Table 2 Total rainfall amount for the month and per analysis time
Gauge ID RG1 RG2 RG3 RG4 RG5 RG6 RG7 RG8 RG9 RG10Total rainfall amount for the month (mm) 3214 3717 3308 3139 3218 3426 3321 3239 3743 3245Total rainfall amount for the analysis time (mm) 476 560 510 529 552 571 544 544 513 531
smaller scale fluctuations compared to the filtering window(25 km when the gate size is 0125m) due to 120575 and theobservation noise Hubbert and Bringi [10] also found thatrepeating the filtering process 10 times produced good results
The additional problem in the estimation of ΦDP wasthe variable offset of ΦDP ΦDP observation can be noisyin near range Radar in general has ΦDP offset to avoidobserving negative value of ΦDP in near range Ideally theoffset should not change with the azimuthal angles Howeverthe offsets of measured ΨDP of the NIMS-X varied withthe azimuthal angles and ΦDP values decreased near theradar when rain cells existed above the radar Therefore
the offset was determined as the minimum value of ΦDPwithin 3 km of the radar and estimated filtered ΨDP wasadjusted accordingly
The calculation of 119860119867
and 119860DP was based on thefollowing two relationships [1 5 20]
119860119867 (119903) = 120590119870DP (119903)
119860DP (119903) = 120576119860119867 (119903) (1)
where 119860119867(119903) and 119860DP(119903) are the specific attenuation and
the specific differential attenuation at range 119903 from the radar
Advances in Meteorology 5
3510
3506
3502
3498
3494 Unnam
12628 12636 12644
1 234 96
8 75
10
R
Figure 3 Layout of NIMS-XPOL and the 10 rain gauges nearUnnam-myeon Muan-gun Jeollanam-do Korea R shows thelocation of the NIMS-XPOL the numbered points indicate theobservational sites of the 10 rain gauges
respectively119870DP(119903) is the specific differential one-way phaseshift calculated using the following equation
119870DP (119903) =ΦDP (1199032) minus ΦDP (1199031)
2 (1199032minus 1199031)
(2)
where 119903 is between 1199031and 1199032 The coefficients 120590 (= 03445)
and 120576 (= 01705) were derived from 119860119867 119860DP and 119870DP
obtained by scattering simulations with the DSDs data col-lected by the PrecipitationOccurrence Sensor System (POSS)from March to September of 2001 in Pusan Korea [21] Thetotal path integrated attenuation and differential attenuationwere calculated withΦDP by integrating the following
2int
119903
0
119860119867(1199031015840) 1198891199031015840= 120590ΦDP (119903) (3)
2int
119903
0
119860DP (1199031015840) 1198891199031015840= 120576 times 2int
119903
0
119860119867(1199031015840) 1198891199031015840 (4)
Number 2 in the equations indicates that the values are two-way values
Finally the measured reflectivity 119885119867-meas(119903) [dBZ] and
the differential reflectivity 119885DR-meas(119903) [dB] at 119903 were cor-rected with the calculated two-way path integrated attenua-tion and differential attenuation
119885119867-corr (119903) = 119885119867-meas (119903) + 2int
119903
0
119860119867(1199031015840) 1198891199031015840
119885DR-corr (119903) = 119885DR-meas (119903) + 2int119903
0
119860DP (1199031015840) 1198891199031015840
(5)
32 Radar Rain Estimation and Comparison with Rain GaugeData A comparison between estimated 119877 from the radar
Measured dual-polarization parameters(ZH ZDR ΦDP 120588HV)
Quality control
(ii) Speckling echo removal
(i) Calibration for instrumental bias of ZH ZDR
(iii) 120588HV threshold application
(ii) Adjustment of its offset(i) 20th FIR filter application to ΨDP
Estimation of ΦDP
Calculation of AH and ADP
Correction of ZH and ZDR
from DSDs120590 and 120576
Figure 4 The schematic diagram of the attenuation correctionalgorithm
parameters and the rain rate (119866) from the rain gauge data wasperformed to evaluate the attenuation correction algorithmIn addition to the rain attenuation the difference between119877 and 119866 was caused by the errors due to the instrumentaluncertainty and representativeness errors of the rain gaugedata the radar observational noise and the errors in the radarrainfall estimation The bias due to the instrumental uncer-tainty in gauge data was removed by properly calibrating thegauges through laboratory tests and a gauge intercomparisonThe representativeness errors were due to the mismatch ofthe sampling area (volume) between the gauge and the radarThe random error in the gauge data could be minimized bytime and areal integrationThe errors due to themeasurementnoise in the radar data could be reduced in a similar mannerThe time offset between the rain gauge and the radar samplingarea was ignored due to the close distance (about 16 km) andthe low elevation angle (3∘)
Therefore 119877 and 119866 were derived with two different radarsampling areas and with increasing accumulation time
(1)119877(038 km2) versus119866(0071198982)119877(038 km2)was estimated
from the average radar measurement with 9 pixels (038 km2at 16 km range) The gauge rainfall rate 119866(007m2) wasderived from the rainfall accumulation from an individualrain gauge that had a sampling area of 007m2
(2) 119877(6 1198961198982) versus 119866(6 1198961198982
) 119877(6 km2) was estimated fromthe areal averaged radar parameters at a 16 km times 31 km radar
6 Advances in Meteorology
Table 3 Two rain rate retrieval algorithms and boundary conditions [22]
119877 (119885119867) = 120572119885
120573
119867119877 (119885DR) = 120572119885
120573
1198671001120574119885DR
Z gt 35 dBZ Z lt 35 dBZ Z gt 35 dBZ 119885DR gt 03 dB Z lt 35 dBZ 119885DR gt 03 dB120572 353 times 10
minus2383 times 10
minus2103 times 10
minus2145 times 10
minus2
120573 621 times 10minus2
608 times 10minus2
985 times 10minus1
890 times 10minus1
120574 mdash mdash minus6479 minus5295
measurement resolution 119866(6 km2) was the average value of119866(007m2) from the ten gauges within the same area
Estimated 119877 from radar parameters was calculated withtwo rain estimators as shown in Table 3These rain estimatorswere derived by a least squares fit with the theoretical 119885
119867
119885DR and 119877 obtained by scattering simulations with theDSD data collected by POSS from March to Septemberof 2001 in Pusan Korea [21] To improve the estimationaccuracy each estimator was derived for two categoriesThe reflectivity threshold of 35 dBZ was used as a stan-dard stratiformconvective rain classification [22] The 119885DRthreshold of 03 dB was used to minimize the effect of the119885DR measurement noiseThePPI data with an elevation angleof 3∘ was used to suppress the beam shielding and groundclutterThe119885
119867and119885DR fields from the radar were smoothed
by the linear average of values at 3 times 3 pixels to reduce theobservational noise
The bias in 119866 from the ten rain gauges was corrected bythe laboratory test and the intercomparison with field dataThe detail is presented in Appendix To evaluate the atten-uation correction algorithm and verify the rainfall retrievalaccuracy the comparisons of119877 and119866were represented by thegeneral statistics such as normalized bias (NB) normalizedstandard deviation (NSD) root-mean square error (rmse)relative rms error (r rmse) and Spearman correlation (corr)as follows
NB119877-119866 (Δ119905) =
(1119899)sum119899minus1
119896=0[119877119896 (Δ119905) minus 119866119896 (Δ119905)]
119866 (Δ119905)
NSD119877-119866 (Δ119905) =
radic(1119899)sum119899minus1
119896=0[119877119896 (Δ119905) minus 119866119896 (Δ119905)]
2
119866 (Δ119905)
rmse119877-119866 (Δ119905) =
radicsum119899minus1
119896=0[119877119896 (Δ119905) minus 119866119896 (Δ119905)]
2
119899
rmse119877-119866 (Δ119905) = radic
sum119899minus1
119896=0[119877119896 (Δ119905) minus 119866119896 (Δ119905)]
2
sum119899minus1
119896=0[119866119896 (Δ119905)]
2
corr119877-119866 (Δ119905)
=
sum119899minus1
119896=0[119877119894119896 (Δ119905) minus 119877119894 (Δ119905)] [119866119895119896 (Δ119905) minus 119866119895 (Δ119905)]
radicsum119899minus1
119896=0[119877119894119896 (Δ119905) minus 119877119894 (Δ119905)]
2radicsum119899minus1
119896=0[119866119895119896 (Δ119905) minus 119866119895 (Δ119905)]
2
(6)
where 119899 is the number of 119877 and 119866 pairs at a given temporalintegrationΔ119905 and the overbar indicates the temporal average
33 Spatial Correlation and Variability of the Rain FieldsRain gauges are widely used to measure rainfall amountsdue tomany advantages including easy installation mobilitylow price and ease of data processing The rain gauge datais used as a reference for the ground rainfall amount forthe evaluation of rain retrieval algorithms adjustment ofother instruments data assimilation and other hydrologicalapplications However it is affected by instrumental errorsand representativeness errors due to the variability of rainfields
In a TB rain gauge the instrumental errors are due toimproper calibration the time delay of the tips leakage dueto its measuring principle and so forthThis error is exploredin the Appendix
The representativeness errors are subdivided into spatialand temporal representativeness They are mainly caused byhaving a small sampling area and a limited sampling time dueto the gauge measuring principle The actual rain fields varywith different spatiotemporal scales that are not equivalentto the spatial and temporal resolution of the rain gaugeTherefore the rain gauge data represent the actual rain fieldsin a very limited manner
These errors should be considered and minimized priorto using rain gauge data as ground truth Furthermore thedegree of these errors has to be understood quantitativelysince there can be other error sources in various applicationsIn this section the spatial and temporal variability of the rainfields are investigated using the rain data collected by the tenrain gauges for the month of July 2010 at Muan Korea
First the Spearman correlation coefficient 119903119894119895(Δ119905)
between a pair of 119866 values from the 119894th and 119895th individualrain gauge with integration time Δ119905 was calculated fromthe following equation to obtain the Spearman correlationcoefficient of the rain fields
119903119894119895 (Δ119905)
=
sum119899minus1
119896=0[119866119894119896 (Δ119905) minus 119866119894 (Δ119905)] [119866119895119896 (Δ119905) minus 119866119895 (Δ119905)]
radicsum119899minus1
119896=0[119866119894119896 (Δ119905) minus 119866119894 (Δ119905)]
2radicsum119899minus1
119896=0[119866119895119896 (Δ119905) minus 119866119895 (Δ119905)]
2
(7)
where 119899 is the number of gauge data pairs at given Δ119905 and theoverbar indicates the temporal average of the rainfall rate Itwas derived withΔ119905 to research the spatial and temporal scaleof the variation for the rain field
In addition the effect of the natural variability of the rainfields within two different areas (038 km2 and 6 km2) wasexplored by analyzing the bias and random error between thespatially averaged rainfall rate ⟨119866⟩ and 119866 from the individual
Advances in Meteorology 7
32km
21 km
11 km
20km40km60km
minus180 18020 30 40 50 60 70 80 90 100120140160100(a)
32 km
21 km
11 km
20km40km60km
minus180 18020 30 40 50 60 70 80 90 100120140160100(b)
Figure 5 PPIs of measured (a) and filtered (b) ΨDP of the NIMS-XPOL at 2101UTC July 10 2010
gauge The NB and NSD of 119866119894with ⟨119866⟩ were calculated as
follows
NB (119894) =(sum119899minus1
119896=0[119866119894119896 (Δ119905) minus ⟨119866119896 (Δ119905)⟩]) 119899
⟨119866 (Δ119905)⟩
NSD (119894) =(1119899)radic(sum
119899minus1
119896=0[119866119894119896 (Δ119905) minus ⟨119866119896 (Δ119905)⟩]
2)
⟨119866 (Δ119905)⟩
(8)
where 119894 indicates the 119894th gauge and the overbar indicates thetemporal average
This analysis provided a reasonable spatial average andtemporal accumulation resolution for the calculation ofmoreaccurate and realistic rainfall amounts from the gauge dataand a quantitative guideline regarding the errors in rainestimation due to the variability of the rain fields
4 Results
41 Attenuation Correction of Reflectivity and DifferentialReflectivity The attenuation correction algorithm in Sec-tion 31 was applied to the NIMS-XPOL radar data ΦDPwas estimated by eliminating 120575 and the observational noisefrom measured ΨDP using an iterative FIR filter The offsetof ΦDP was then adjusted Figure 5 shows the PPIs ofmeasured ΨDP and filtered ΦDP After applying the FIR filterthe small scale variation due to 120575 and observational noisein measured ΨDP was reduced and filtered ΨDP (estimatedΦDP) monotonically increased with an increase in rangeThe maximum value of estimated ΦDP reached to 140∘ Inaddition estimated ΦDP at the azimuth angles of 150∘ and200∘ sharply increased with the loss of signal beyond thesesharp gradient areas This reflected that strong convective
cells existed in these directions and that these cells caused thesevere signal attenuation EstimatedΦDP values between 250
∘
and 360∘ azimuth angle decreased after filtering since ΦDPoffsets were positive values in these directions Figure 6 showsthe range profile of measured and filtered ΨDP for estimationof ΦDP at 71
∘ 160∘ and 334∘ azimuth angles The unexpectedlarge values near the radarwere removed by the120588HV threshold(red line)The small fluctuation inmeasuredΨDP due to 120575 andthe observational noise was effectively removed after filteringThe value of estimated ΦDP was adjusted to 0 by the offset(blue line)
Estimated ΦDP was used to calculate the path integrated119860119867 and the measured119885
119867values were corrected as shown in
Figure 7 Figure 7(a) represents measured-attenuated119885119867and
relatively small values of 119885119867at far ranges This may indicate
that the 119885119867values at far ranges were affected by the strong
attenuation due to strong convective cells near the radarThisspatial pattern of 119885
119867was the typical feature of the attenuated
radar signal due to the rain After correcting the attenuationthe value of the 119885
119867field became larger over the entire area
and the radially attenuated 119885119867
was corrected by showinga typical rain cell pattern in 119885
119867(Figure 7(b)) The largest
difference between the uncorrected and the corrected 119885119867at
2101UTCwas over 50 dB However the white area behind thestrong echoes was not detected as there was a complete loss ofthe signal powerThese regions could not be corrected by thisalgorithmThe circle pattern of119885
119867fields within about a 4 km
range from theNIMS-XPOL could not present the actual rainechoes due to the instrumental issues (the same was true for119885DR) These results are also presented in the range profiles inFigure 8
Similarly measured 119885DR were corrected with the pathintegrated 119860DP derived from the path integrated 119860
119867using
(4) (see Figure 9) Before applying attenuation correction
8 Advances in Meteorology
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
0
50
100
150
200
ΦDP (∘) azimuth angle = 71∘
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
0
50
100
150
200
ΦDP (∘) azimuth angle = 160∘
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
0
50
100
150
200
ΦDP (∘) azimuth angle = 334∘
(c)
Figure 6 Range profiles of measured (black) 120588HV threshold isolated point removing (red) and FIR filter and offset adjusted (blue) ΨDP forestimation of ΦDP at 71
∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
32km
21 km
11 km
20km40km60km
minus10 10 15 20 25 30 35 40 45 50 55 60 65 700(a)
32 km
21 km
11 km
20km40km60km
minus10 10 15 20 25 30 35 40 45 50 55 60 65 700(b)
Figure 7 PPIs of measured (a) and attenuation corrected (b) 119885119867of the NIMS-XPOL for the case in Figure 5
Advances in Meteorology 9
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 71∘
minus20
0
20
40
60
80
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 160∘
minus20
0
20
40
60
80
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 334∘
minus20
0
20
40
60
80
(c)
Figure 8 Range profiles of measured (line) calibration bias 120588HV threshold and isolated point removing (dot) and attenuation corrected(dash) 119885
119867at 71∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
the attenuation signature in119885DR was determined Ideally the119885DR value for rain should be positive since rain drops areoblate due to aerodynamic force However themeasured119885DRfield in Figure 9 showed a negative value over wide areasAfter correcting the attenuation the119885DR field was larger than0 dB and showed a realistic rain patternThis is also presentedin the range profile of Figure 10The largest difference of119885DRbetween before and after attenuation correction at 2101UTCwas about 8 dB The 119885DR values around the 220∘ azimuthangle beyond 40 km were the largest They did not matchwith the area of the largest 119885DR values (around the 210∘azimuth angle with a range of 20 km and 40 km) This wasan unexpected result because 119885DR is proportional to 119885
119867in
theory This mismatch between 119885119867
and 119885DR at far rangemay be due to contamination due to beam broadening andaccumulated errors due to uncertainty in the attenuationcorrection with different ranges
Figures 11 and 12 show the time series of 119885119867
and119885DR at the ten gauge sites before and after applying theattenuation correction The differences between uncorrectedand corrected values existed after 1800UTC and were at amaximumaround 2100UTCThis attenuation correctionwas
evaluated through comparison with the theoretical values of119885119867and 119885DR In Figure 13 the uncorrected 119885
119867-119885DR scatter
plots are apart from the theoretical119885119867-119885DR values [23] After
attenuation correction the119885119867-119885DR scatter plots were close to
the calculated valuesThis indicated that119885119867and119885DR over the
gauge sites were corrected with reliable values
42 Variability of the Rain Fields The Spearman correlationcoefficient of the rainfall rate from the 10 spatially dis-tributed rain gauges was calculated with accumulation timeand separation distance (Figure 14) The overall Spearmancorrelation coefficient decreased with increasing separationdistance at all accumulation timesThis tendency was similarto the result from Habib and Krajewski [24] However thecorrelation value was higher in this study Note that the gaugetip resolution in Habib and Krajewski [24] was 0254mmwhile it was 02mm in this study In addition the rainfall ratesfrom the gauge data used in this study mainly consisted ofmoderate rainfall rates while Habib and Krajewski [24] useddata containing heavy rainfall rates
For accumulation times of less than 10min the corre-lation at the same distance was dramatically increased with
10 Advances in Meteorology
32km
21 km
11 km
20km40km60km
18
14
00
02
04
06
26
40
minus10
22
minus20
30
10
50
minus30
(a)
32 km
21 km
11 km
20km40km60km
18
14
00
02
04
06
26
40
minus10
22
minus20
30
10
50
minus30
(b)
Figure 9 PPIs of measured (a) and attenuation corrected (b) 119885DR of NIMS-XPOL for the case in Figure 5
increasing accumulation time and the slope of the correlationfunction was steep This was due to the reduction of theinstrumental errors of the gauge data by time accumulationThis indicated that rainfall accumulated for less than 10minfrom single gauge data cannot represent the rainfall over itssurrounding area due to instrumental errors For accumula-tion times longer than 10min the slope became less steepThis reflected the remaining variability of the rain field Inaddition the accumulated rainfall rate of about 10min wassuitable for minimizing the instrumental uncertainty whilepreserving the variation in the rain field The correlationreached about 098 at 05 km in 10min of accumulationand this approached 098 at 10 km at 60min accumulationTherefore a single rain gauge mostly represented the rainfallrate within 05 km and 10 km with 10min and 60minaccumulation respectively
The NB between an areal averaged rainfall rate ⟨119866⟩ anda single gauge rainfall rate 119866 is shown with accumulationtime in Figures 15(a) and 16(a) Where the average area was038 km2 (corresponding to 9 pixels of NIMS-XPOL radarat 16 km range) the NB reached a 6ndash8 maximum Wherethe averaged area is 6 km2 (corresponding to the coverage forall of the gauges) the NB reached about a 10 maximumFurthermore the NB hardly changed with accumulationtime Note that this statistic was derived after removinginstrumental bias This indicated that the rainfall rate froma single gauge was biased due to the spatial variability of therain fields and this bias was maintained with accumulationtime
The NSD of a single gauge for a mean area of 038 km2was over 40 and it reached about 70ndash90 for a mean areaof 6 km2 at an accumulation time of 1min (see Figures 15(b)and 16(b)) This statistic included instrumental errors (about
40ndash60 at 16mmhminus1 and 60ndash80 at 10mmhminus1 as shownin the Appendix) as well as the spatial variability of the rainfields This indicated that the variance between the point andarea values from the gauge data at short accumulation timeswas similar to the errors due to the instrumental uncertaintyof the rain gaugeTheNSD in 60min accumulation decreasedto over 10 and about 15ndash45 with a mean area of 038 km2and 6 km2 respectively including the NSD of 6ndash10 due toinstrumental uncertainty This indicated that the rainfall ratefrom single gauges may differ significantly from the actualrainfall rate due to the variability of the rain fields and theinstrumental uncertainty in short accumulation times
43 Verification of Rainfall Retrieval The rainfall rate fromthe NIMS-XPOL radar measurements was retrieved withtwo rainfall rate estimators 119877(119885
119867) and 119877(119885
119867 119885DR) The
estimated rainfall rate was verified through comparison withthe rainfall rate (119866) from the rain gauge data These twoestimators were applied before and after the attenuation cor-rection First 119877(038 km2) was compared with 119877(007 km2)at 10min accumulation (Figure 17) Using 119877(119885
119867) the NB
NSD rmse and r rmse without the attenuation correctionwere significantly large at 70 127 127 (mmhr) and78 respectively However they were reduced to minus152 521 (mmhr) and 32 respectively after applying theattenuation correction Using 119877(119885
119867 119885DR) the NB was just
3 NSD was 49 rmse was 496 (mmhr) and r rmsewas 31 before applying attenuation correction This wassimilar to the result from 119877(119885
119867) after attenuation correction
However there was no significant change in119877(119885119867 119885DR)with
attenuation correction This means that 119877(119885119867 119885DR) was less
sensitive to rain attenuation than was 119877(119885119867) because 119885
119867is
proportional to 119877 and 119885DR is inversely proportional to 119877
Advances in Meteorology 11
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 71∘
minus10
minus5
0
5
10
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 160∘
minus10
minus5
0
5
10
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 334∘
minus10
minus5
0
5
10
(c)
Figure 10 Range profiles of measured (line) calibration bias 120588HV threshold and isolated point removing (dot) and attenuation corrected(dash) 119885DR at 71∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
in this rain estimator (see Table 3) Furthermore comparingthe analysis of variance results between areal (038 km2) andpoint (007m2) values from the gauge data (Section 42)half of the errors after attenuation correction were producedby the small scale variability of the rain fields and theinstrumental uncertainty of the rain gauge
A similar comparison is shown in Figure 18 for a 60minaccumulation Using 119877(119885
119867) the NB NSD rmse and r rmse
were decreased from minus65 86 71 (mmhr) and 70to minus5 34 28 (mmhr) and 28 by the attenuationcorrection respectively Using 119877(119885
119867 119885DR) the NB NSD
rmse and r rmse were only slightly decreased after theattenuation correction However the NSD rmse and r rmsewere reduced when compared at 10min accumulation In thesame accumulation time the errors in the gauge data reachedto over 15
This comparison was affected by different sampling areasof the gauges and the radar Therefore the areal averagevalues were compared (Figure 19) Using119877(119885
119867)with the areal
averaged radar measurements the NB was reduced fromminus69 to 7 NSD was reduced from 122 to 42 rmsewas reduced from 39 (mmhr) to 13 (mmhr) and r rmse
was reduced from 77 to 26 at an accumulation time of10min Using 119877(119885
119867 119885DR) the NB NSD rmse and r rmse
were decreased fromminus24 67 21 (mmhr) and 42 to 539 12 (mmhr) and 24 respectively
TheNB and r rmsewithmatched sampling areas betweenthe gauge and the radar are shown in Figures 20 and 21 Thereduction of NB due to temporal accumulation was smallin both of the rain estimators The NB was reduced to lessthan 7 by attenuation correction for all accumulation timeswith both119877(119885
119867) and119877(119885
119867 119885DR)The attenuation correction
effect was significantly greater in119877(119885119867) Similarly the r rmse
was reduced significantly in 119877(119885119867) by attenuation correction
(from 68 to 25 at a 60min accumulation) The reductionof the r rmse due to temporal accumulation was larger in119877(119885119867 119885DR) because the significant observational noise of
119885DR was smoothed The r rmse was decreased from 18to 15 in 119877(119885
119867 119885DR) at a 60min accumulation time after
attenuation correction
5 Conclusions
In this study an attenuation correction algorithm based onΦDP was developed to improve the accuracy of the rainfall
12 Advances in Meteorology
1st gauge site 2nd gauge site
3rd gauge site 4th gauge site
5th gauge site 6th gauge site
7th gauge site 8th gauge site
9th gauge site 10th gauge site
17 18 19 20 2116Time (h)
17 18 19 20 2116Time (h)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
UncorrectedCorrected
UncorrectedCorrected
Figure 11 Time series of measured (red circle) and attenuation corrected (blue circle) 119885119867at the 10 rain gauge sites
Advances in Meteorology 13
UncorrectedCorrected
UncorrectedCorrected
17 18 19 20 2116Time (h)
17 18 19 20 2116Time (h)
1st gauge site 2nd gauge site
3rd gauge site 4th gauge site
5th gauge site 6th gauge site
7th gauge site 8th gauge site
9th gauge site 10th gauge site
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
Figure 12 Time series of measured (red circle) and attenuation corrected (blue circle) 119885DR at the 10 rain gauge sites
14 Advances in Meteorology
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Uncorrected(a)
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Corrected Simulated from DSD(b)
Figure 13 Scatter plots of 119885119867and 119885DR before (a) and after (b) attenuation correction at the 10 rain gauge sites DSD observation results (dot
[21])
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
04
05
06
07
08
09
10
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
Accum time = 1min
Accum time = 10min
Accum time = 2min
Accum time = 20min
Accum time = 5min
Accum time = 60min
Figure 14 The Spearman correlation coefficient 119903 between data from two rain gauges at different accumulation times
estimation using X-band radar This algorithm was evalu-ated with radar measurements from NIMS-XPOL Howevermeasured ΨDP contained 120575 and observation noise as wellas ΦDP Furthermore the ΨDP offset varied with azimuthangles and decreased near the radar when rain cells were
above the radar In this study ΦDP was estimated by aniterative filtering technique and by adjusting with an offsetdetermined as the minimum value ofΦDP within 3 km of theradar As a result estimated ΦDP monotonically increasedThe path integrated value of119860
119867and119860DP was calculated from
Advances in Meteorology 15
2nd 6th 9th gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(038km2) versus G
(a) Normalized bias
⟨G⟩(038km2) versus G
2nd 6th 9th gaugesMean NSD
Mean R
0
20
40
60
80
100
NSD
()
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 15 (a) The NB and (b) NSD between ⟨119866⟩(038 km2) and 119866 from three rain gauges with accumulation time ⟨119866⟩ is the temporallyaveraged rainfall rate of ⟨119866⟩
All gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(6km2) versus G
(a) Normalized bias
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
0
20
40
60
80
100
NSD
()
All gaugesMean NSD
Mean R
⟨G⟩(6km2) versus G
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 16 (a) The NB and (b) NSD between ⟨119866⟩(6 km2) and 119866 from 10 rain gauges with accumulation time ⟨119866⟩ is the temporally averaged⟨119866⟩ with accumulation time
estimated ΦDP using 119860119867-ΦDP and 119860DP-ΦDP relationships
derived from the DSDs data Measured 119885119867
and 119885DR werecorrected with the calculated path integrated 119860
119867and 119860DP
by each gate The attenuated signal patterns of 119885119867and 119885DR
were restored to a typical rain echo pattern However theattenuation corrected 119885
119867and 119885DR fields were not matched
at distant ranges This was likely because of contaminationdue to beam broadening and the accumulation of errorswith increasing range Comparing the 119885
119867-119885DR scatter plots
over the 10 rain gauge sites with the results obtained froma scattering simulation with the DSD data the attenuation
corrected 119885119867-119885DR scatter plots are in good agreement with
the 119885119867-119885DR scatter plots from the DSDs
The validation for the improvement of accuracy in therainfall estimation through attenuation correction was per-formed by comparing with dense rain gauge network data119877(119885119867) from the attenuated (observed)119885
119867produced a severe
underestimation with an NB NSD rmse and r rmse of70 127 127 (mmhr) and 78 at 10min accumulationrespectively while 119877(119885
119867) from the attenuation corrected 119885
119867
showed better agreement with the gauge measurements withan NB NSD rmse and r rmse of minus1 52 521 (mmhr)
16 Advances in Meteorology
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus070
NSD = 127
rmse = 1279
corr = 0858
Corrected
NB = minus001
NSD = 052
rmse = 521
corr = 0927
r_rmse = 0786 r_rmse = 0320
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = 003
NSD = 049
rmse = 496
r_rmse = 0305
corr = 0936
Corrected
NB = 001
NSD = 050
rmse = 502
r_rmse = 0308
corr = 0936
(b) 119877(119885119867119885DR)
Figure 17 The scatter plots of 119877(038 km2 10min) versus 119866(007m2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
CorrectedUncorrected
R(m
mh
)
0
10
20
30
40
10 20 30 400G (mmh)
Uncorrected
NB = minus065
NSD = 086
rmse = 706
r_rmse = 0702
corr = 0878
Corrected
NB = minus005
NSD = 034
rmse = 281
r_rmse = 0279
corr = 0866
(a) 119877(119885119867)
CorrectedUncorrected
R(m
mh
)
10 20 30 400G (mmh)
0
10
20
30
40
Uncorrected
NB = 008
NSD = 023
rmse = 189
r_rmse = 0188
corr = 0942
Corrected
NB = 007
NSD = 025
rmse = 205
r_rmse = 0203
corr = 0940
(b) 119877(119885119867119885DR)
Figure 18 The scatter plots of 119877(038 km2 60min) versus 119866(007m2 60min) (a) using 119885119867and (b) using 119885
119867and 119885DR
and 32 at 10min accumulation respectively At 60minaccumulation theNBNSD rmse and r rmsewere decreasedfrom minus65 86 71 (mmhr) and 70 to minus5 3428 (mmhr) and 28 respectively Using 119877(119885
119867 119885DR) the
NB NSD rmse and r rmse were only slightly reduced even if
attenuation was corrected but were smaller than when using119877(119885119867) This indicated that 119877(119885
119867 119885DR) was less sensitive to
the attenuation than 119877(119885119867) In addition half of the errors
that remained after attenuation correction were caused by thesmall scale variability of the rain fields and the instrumental
Advances in Meteorology 17
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus069
NSD = 122
rmse = 391
r_rmse = 0770
corr = 0893
Corrected
NB = 007
NSD = 042
rmse = 133
r_rmse = 0262
corr = 0960
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = minus024
NSD = 067
rmse = 215
r_rmse = 0424
corr = 0718
Corrected
NB = 005
NSD = 039
rmse = 124
r_rmse = 0245
corr = 0891
(b) 119877(119885119867119885DR)
Figure 19 The scatter plots of 119877(6 km2 10min) versus 119866(6 km2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
R(ZH)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(a) 119877(119885119867)
R(ZH ZDR)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(b) 119877(119885119867119885DR)
Figure 20 The normalized bias of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and 119885DR The red
point-line and the blue point-line indicate without and with the attenuation correction respectively
uncertainty of the rain gauges Using an areal averagedvalue the NB was reduced to less than 7 with attenua-tion correction using either rainfall estimator regardless ofaccumulation timeThe r rmse were significantly reduced byattenuation correction in 119877(119885
119867) and the reduction of r rmse
with 119877(119885119867 119885DR) was even more dramatic
As summarized above the rain attenuation for 119885119867
and 119885DR in the X-band was corrected by ΦDP and theresults showed a good agreement with the value obtainedby scattering simulations with the DSD datasets The resultsof comparison with rain gauge data also showed a signifi-cant improvement in the accuracy of rainfall estimation by
18 Advances in Meteorology
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10r_
rmse
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH)
(a) 119877(119885119867)
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10
r_rm
se
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH ZDR)
(b) 119877(119885119867119885DR)
Figure 21 The relative root-mean square error of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and
119885DR The red point-line and the blue point-line indicate without and with the attenuation correction respectively
attenuation correction which supports the quantitative useof the X-band radar in rain data retrieval
Appendix
A Instrumental Uncertainty of the Rain Gauge
The instrumental uncertainty of the TB rain gauge is asso-ciated with improper calibration leakage and precipitationmissing the gauge due to the measurement principle andwetting and evaporation losses In addition external sourcessuch as wind turbulence and installed position also con-tribute to the instrumental uncertainty Ciach [17] reportedthat the TB gauge data suffers a significant uncertainty asso-ciated with random differences between closely collocatedTB gauges through the analysis of the error as a functionof the rainfall intensity for three timescales In addition theimproper calibration induces a bias in the rainfall measure-ment and this bias cannot be removed by increasing accu-mulation timeTherefore the proper correction or processingof TB gauge data is necessary to obtain accurate and reliablereading of the rainfall amount
In this appendix instrumental biases were calculatedto reduce the instrumental uncertainty through an idealexperiment in the laboratory and a field intercomparisonwith 10 TB rain gauges In addition the random errors werequantitatively calculated as a function of rainfall rate andaccumulation time
A1 Correction of Instrumental Bias The instrumental bias119861119868(119894) of the 119894th gauge was defined as a combination
of the absolute bias 119861119860(ref) of the reference gauge and
the relative bias 119861119877(119894) of the 119894th gauge according to the
following equation
119861119868 (119894) = 119861119860 (ref) times 119861119877 (119894) (A1)
119861119860(ref) signified the ratio of the calculated value from
the instrumental resolution of the reference gauge to the truevalue measured by other instruments
119861119860 (ref) =
rainfalltruthrainfallgauge (ref)
(A2)
In this study 119861119860(ref) was derived by an ideal experiment
using an electronic scale First the reference gauge was puton the scale and the scale was initialized The water dropsflowed out into the receiving orifice of the reference gaugeuntil the tipping number became 60Themeasured weight of60 tips was 2144 g while the theoretical (or expected) valuewas 2244 g for the reference gauge (receiving orifice size of1545 cm 60 tips and 02mm tip resolution) Therefore theabsolute bias of the reference gauge from (A2) was about0955 This indicates that the rainfall amount measured bythis reference gauge was overestimated by about 45 andthis bias should be corrected prior to any interpretation
119861119877(119894) represented the ratio of the observed (recorded)
value from the 119894th gauge to the observed (recorded) valuefrom the reference gauge
119861119877 (119894) =
rainfallgauge (ref)rainfallgauge (119894)
(A3)
119861119877(119894) was calculated from ten collocated rain gauges
including the reference gauge located at an observation field
Advances in Meteorology 19
Table 4 The instrumental and relative biases of the 10 tipping bucket rain gauges
Gauge ID RG1 RG2 RG3 RG4 RG5 RG6 RG7 RG8 RG9 RG10Instrumental bias 094 108 099 100 096 104 098 102 108 098Relative bias 098 113 103 105 100 109 102 107 113 102
013610152230406080120150200250300400
()
Rain
rate
(mm
h)
100101
Accumulation time (min)
10
100NSD ()
Figure 22 Two-dimensional normalized standard deviationbetween the standard rainfall and each rain gauge
at KNUKoreaThe observational period for intercomparisonwas from October 2009 to June 2010 (14 rainy days) andthe total accumulated rainfall for this period was about300mm The results are presented in Table 4 All relativebiases were mainly larger than 10 This indicated that themost of gauges underestimated the rainfall amount comparedto the reference rain gauge The relative bias of the referencegauge (119894 = 5) was 10
Finally 119861119860(119894) was calculated from (A1) (see Table 4) All
instrumental biases were between 090 and 11 and the meanvalue of the biases was nearly 10 This indicated that therainfall rate from a single gauge data had a maximum 10bias These biases were corrected for each gauge to removeexisting instrumental errors
A2 Quantification of Random Errors To quantify the ran-dom error due to instrumental uncertainty the NSD of theten gauges was calculated with accumulation time and rainrate using data from an intercomparison The NSD of the 119894thgauge compared to the reference gauge was derived for 20rain rate classes (Δ119877) and 17 accumulation time classes (Δ119905)as follows
NSD119894 (Δ119905 Δ119877)
=
(1119899)radicsum119899minus1
119896=0[1198661015840
119870(Δ119905 Δ119877) minus 119866119894119896 (Δ119905 Δ119877)]
2
1198661015840
(A4)
where 1198661015840 and 119866 indicate the reference and individual gaugesrespectively The overbar means the temporal average ofthe rainfall rate The NSDs of nine different gauges (exceptthe reference gauge) were averaged to obtain the two-dimensional average NSD (Figure 22) The average NSD
tended to decrease with increasing accumulation time andwith increasing rain rate The NSD was about 10ndash15when accumulation time was 10min and the rain rate was10mmhminus1 The instrumental errors were quite significant forlonger accumulation times and higher rainfall rates Theseerrors should be used as a guideline when a single gauge isused as a reference for any validation study
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper has been revised and extended fromMs Young-AOhrsquos MS thesis [25] of Kyungpook National University Thisresearch is supported by Development and Application ofCross Governmental Dual-Pol Radar Harmonization (WRC-2013-A-1) project of theWeather Radar Center KoreaMeteo-rological AdministrationThis work was funded by the KoreaMeteorological Administration Research and DevelopmentProgram under Grant KMIPA2015-1010
References
[1] V N Bringi V Chandrasekar N Balakrishnan and D SZrnic ldquoAn examination of propagation effects in rainfall onradar measurements at microwave frequenciesrdquo Journal ofAtmospheric and Oceanic Technology vol 7 no 6 pp 829ndash8401990
[2] A Huggel W Schmid and AWaldvogel ldquoRaindrop size distri-butions and the radar bright bandrdquo Journal of AppliedMeteorol-ogy vol 35 no 10 pp 1688ndash1701 1996
[3] G W Lee Errors in rain measurement by radar effect of vari-ability of drop size distributions [PhD thesis]McGill University2003
[4] J D Tuttle and R E Rinehart ldquoAttenuation correction in dual-wavelength analysesrdquo Journal of ClimateampAppliedMeteorologyvol 22 no 11 pp 1914ndash1921 1983
[5] S-G Park V N Bringi V Chandrasekar M Maki andK Iwanami ldquoCorrection of radar reflectivity and differentialreflectivity for rain attenuation at X band Part I theoretical andempirical basisrdquo Journal of Atmospheric and Oceanic Technol-ogy vol 22 no 11 pp 1621ndash1632 2005
[6] G Scarchilli E Gorgucci V Chandrasekar and T A SeligaldquoRainfall estimation using polarimetric techniques at C-bandfrequenciesrdquo Journal of Applied Meteorology vol 32 no 6 pp1150ndash1160 1993
[7] M N Anagnostou E N Anagnostou and J VivekanandanldquoCorrection for rain path specific and differential attenuation ofX-band dual-polarization observationsrdquo IEEE Transactions onGeoscience and Remote Sensing vol 44 no 9 pp 2470ndash24802006
20 Advances in Meteorology
[8] J Kalogiros M N Anagnostou E N Anagnostou M Mon-topoli E Picciotti and F S Marzano ldquoEvaluation of a newpolarimetric algorithm for rain-path attenuation correction ofX-band radar observations against disdrometerrdquo IEEE Transac-tions on Geoscience and Remote Sensing vol 52 no 2 pp 1369ndash1380 2014
[9] W-Y Chang J Vivekanandan and T-C C Wang ldquoEstimationof X-band polarimetric radar attenuation and measurementuncertainty using a variational methodrdquo Journal of AppliedMeteorology and Climatology vol 53 no 4 pp 1099ndash1119 2014
[10] J Hubbert and V N Bringi ldquoAn iterative filtering technique forthe analysis of copolar differential phase and dual-frequencyradar measurementsrdquo Journal of Atmospheric and OceanicTechnology vol 12 no 3 pp 643ndash648 1995
[11] K V Mishra W F Krajewski R Goska et al ldquoDeploy-ment and performance analyses of high-resolution iowa XPOLradar system during the NASA IFloodS campaignrdquo Journal ofHydrometeorology vol 17 no 2 pp 455ndash479 2016
[12] M Thurai G J Huang V N Bringi W L Randeu and MSchonhuber ldquoDrop shapes model comparisons and calcu-lations of polarimetric radar parameters in rainrdquo Journal ofAtmospheric and Oceanic Technology vol 24 no 6 pp 1019ndash1032 2007
[13] M D Humphrey J D Istok J Y Lee J A Hevesi and AL Flint ldquoA new method for automated dynamic calibration oftipping-bucket rain gaugesrdquo Journal of Atmospheric andOceanicTechnology vol 14 no 6 pp 1513ndash1519 1997
[14] E HabibW F Krajewski V Nespor andA Kruger ldquoNumericalsimulation studies of rain gage data correction due to windeffectrdquo Journal of Geophysical ResearchAtmospheres vol 104 no16 Article ID 1999JD900228 pp 19723ndash19733 1999
[15] G J Ciach and W F Krajewski ldquoOn the estimation of radarrainfall error variancerdquoAdvances inWater Resources vol 22 no6 pp 585ndash595 1999
[16] E Habib W F Krajewski and A Kruger ldquoSampling errors oftipping-bucket rain gaugemeasurementsrdquo Journal of HydrologicEngineering vol 6 no 2 pp 159ndash166 2001
[17] G J Ciach ldquoLocal random errors in tipping-bucket rain gaugemeasurementsrdquo Journal of Atmospheric andOceanic Technologyvol 20 no 5 pp 752ndash759 2003
[18] EHabibW F Krajewski andG J Ciach ldquoEstimation of rainfallinterstation correlationrdquo Journal ofHydrometeorology vol 2 no6 pp 621ndash629 2001
[19] NIMS ldquoStudy on estimation of rainfall amount and classifi-cation of hydrometeor species using dual-polarimetric radar(III)rdquo pp 1ndash20 2010
[20] E Gorgucci V Chandrasekar and L Baldini ldquoCorrection ofX-band radar observation for propagation effects based on theself-consistency principlerdquo Journal of Atmospheric and OceanicTechnology vol 23 pp 1668ndash1681 2006
[21] NIMS ldquoStudy on estimating raindrop size distribution forquantitative estimation of radar rainfall amountrdquo pp 1ndash55 2009
[22] NIMS ldquoQuantitative precipitation estimation and hydrometeoridentification using dual-polarization radar-Phase IIrdquo pp 1ndash742009
[23] S-G Park M Maki K Iwanami V N Bringi and V Chan-drasekar ldquoCorrection of radar reflectivity and differential reflec-tivity for rain attenuation at X band Part II Evaluation andapplicationrdquo Journal of Atmospheric and Oceanic Technologyvol 22 no 11 pp 1633ndash1655 2005
[24] E Habib and W F Krajewski ldquoUncertainty analysis of theTRMM ground-validation radar-rainfall products applicationto the TEFLUN-B field campaignrdquo Journal of Applied Meteorol-ogy vol 41 no 5 pp 558ndash572 2002
[25] Y-A Oh Rainfall estimation from X-band dual-polarizationradar measurements effects of attenuation correction and evalu-ation with a dense rain gauge network [MS thesis] KyungpookNational University 2011
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom
Applied ampEnvironmentalSoil Science
Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal of
Geophysics
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
2 Advances in Meteorology
the C-band can reach about 12 dB due to strong convectivecells as compared to the S-band [3] Furthermore the attenu-ation accumulates over distances as a radar beam propagatesThe power is frequently lost completely over far rangesTherefore it is not possible to monitor and analyze severeweather such as heavy rains typhoons and heavy snowswithout the proper correction of the attenuation in the C- orX-band radars
The attenuation correction has been investigated bymanyresearchers Tuttle and Rinehart [4] suggested an attenuationcorrection method using dual-wavelength (S- and X-band)radar measurements with a relationship between specificattenuation (119860
119867) and 119885
119867 Recently the specific differential
propagation phase shift (119870DP) has been used widely forattenuation correction because 119870DP is not affected by radarpower calibration attenuation and partial beam blockingIn addition 119870DP is less sensitive to the natural variability ofdrop size distributions (DSDs) in rainfall estimation Bringiet al [1] showed that both 119860
119867and the specific differential
attenuation (119860DP) are almost linearly related to 119870DP throughscattering simulations and these relationships have beenaccepted by many researchers However the coefficients ofthese relationships derived from the scattering simulationsvary significantly with temperature DSD variability and thedrop deformationmodel Park et al [5] showed that the coef-ficients of 119860
119867-119870DP and 119860DP-119860119867 relationships vary greatly
from 0139 to 0335 dB(∘)minus1 and from 0114 to 0174 dB(∘)minus1 inthe X-band respectively due to changes in temperatures andthe different drop deformation models
While 119870DP has many advantages in the attenuationcorrection as described above estimating 119870DP from themeasured total differential phase shift (ΨDP) is challengingbecause of the backscatter differential phase shift (120575) and themeasurement errors for ΨDP [1] Scarchilli et al [6] used aniterationmethod with a 120575-119885DR relationship to remove 120575 fromΨDP and corrected the attenuation using estimated 119870DP inthe C-band Anagnostou et al [7] Kalogiros et al [8] andChang et al [9] used an iteration method in the X-bandAs a different method for removing 120575 Hubbert and Bringi[10] separated 120575 and the differential propagation phase (ΦDP)from ΨDP through an iterative filtering technique (FIR) andcalculated the hail signal from 120575 Park et al [5] also usedthe same filtering technique to extractΦDP for correcting theattenuation and the differential attenuation in the X-bandRecently Mishra et al [11] investigated dual-polarimetricproducts and three types of rainfall estimation results withhigh spatiotemporal resolution data from X-band radar S-band radar and disdrometer data with an intercomparisonand T-matrix simulation method [12] Mishra et al [11] useda very similar attenuation correction process in their workwhere 119860
119867was calculated by an FIR filtered ΦDP and 119860DP
was calculated by an 119860DP-119860119867 relationship based on a self-consistency method
Two rainfall estimationmethods119877(119885119867) and119877(119885
119867 119885DR)
were used for evaluation of the attenuation correction effectThe self-consistency based correction of the attenuation andthe differential attenuation from the FIR filtered ΦDP wereused to improve the accuracy in the rainfall estimation
methods by X-band polarimetric radar The accuracy eval-uation of the rainfall estimation methods was performedby comparisons with rain gauge measurements with highspatial resolution However there were instrumental uncer-tainties and errors in the representativeness of the raingauge measurements The instrumental error has severalsources including the gauge calibration wind effects andwettingevaporation loss [13 14]The representativeness erroris linked with the spatial sampling area and the tempo-ral accumulation of the rain gauge [15] Habib et al [16]investigated sampling error of the tipping bucket gauge asfunction of accumulation times in a simulation study Ciach[17] verified local random errors in the tipping bucket gaugethrough experimental data Habib et al [18] investigatedthe correlation with separation distances in a rain gaugecluster In this study we considered the calibration effectsin the instrumental error of the rain gauge Instrumentalerror from the calibration effect was removed by laboratoryexperiments to providemore accurate and reliable referencesThe representativeness error was investigated by the temporalaccumulation and spatial average of rainfall amount
In Section 2 the radar and rain gauge data used inthis study are presented The methods for the attenuationcorrection byΦDP the rainfall estimation and its verificationand the analysis of the natural variability of rain fields usingrain gauge data are described in Section 3 The analysisresults are shown in Section 4 The results are summarizedin Section 5
2 Data
21 Radar Data The radar data from the X-band dual-polarimetric radar (NIMS-XPOL) of the National Institute ofMeteorological Sciences (NIMS) was used in this study Thegeneral characteristics of the NIMS-XPOL data are shownin Table 1 The NIMS-XPOL can be installed onto a five-tontruck bed (Figure 1(a)) and mounted on a steel tower bed(Figure 1(b))
The NIMS-XPOL was operated during June 2010 atMuan-gun Jeollanam-do which is on the southwest coastof Korea (latitude = N 350940∘ longitude = E 126285∘) forthe observation of heavy rains during the summer monsoonseason The rain estimation in Section 32 was performedwith the NIMS-X radar data collected during 55 hours from1600UTC to 2130UTC on July 10 2010 During this rainevent the NIMS-X was mounted on a steel tower bed inthe Muan-gun observation center to minimize the effect ofbeam shielding and ground clutter and a full volume scanwith 14 Plan Position Indicators (PPIs) was performed every25minThe gate spacing was 150m and the Pulse RepetitionFrequency (PRF) was 999Hz And the total rainfall amountwas over 50mm The evolution of the rain field is shown inFigure 2 The echoes moved toward the northeast and thesmall and weak rain cells near the radar site at 1601UTCwerereplaced by the strong convective cells approaching from thesouthwest
Advances in Meteorology 3
Table 1 General characteristics of the NIMS-XPOL radar [21]
X-band polarimetric mobile Doppler weather radar (ARC-X250MP) specification
Transmitter
Type MagnetronPeak power (kW) 250Frequency (MHz) 9360
PRF (MHz) 0sim2500
Antenna
Diameter (m) 244Polarization Orthogonal linear simultaneous H amp VBeamwidth (∘) 095Gain (dB) gt43
Data processingNumber of gates 30001000 PRFGate spacing 30m to 1000mRaw data 119885
119867 119881119905 119878119882 119885DR ΦDP 120588HV
(a) (b)
Figure 1 Photographs of the NIMS-XPOL radar installed onto (a) a truck bed and (b) a steel tower bed (courtesy of NIMS)
22 Rain Gauge Data The rain gauge data was collectedby ten tipping bucket (TB) rain gauges operated by theKyungpook National University (KNU) Korea The tip sizeof all of the gauges was 02mm and the diameter of thereceiving orifice was 1539 cm The time resolution was 05 sand the data was recorded based on the event (tipping) timeTo compare with the estimated rain rate (119877) from the NIMS-XPOL radar the rain accumulation from ten gauges wascollected during the month of July 2010 near Unnam-myeonwhich is located at a distance of about 16 km at 160∘ azimuthangle from the NIMS-XPOL radarThe area was only slightlyaffected by beam shielding of the NIMS-XPOL Consideringthe geographical situation and the limited locations for theinstallation all gauges were installed on the rooftops ofbuildings Figure 3 shows the location of the NIMS-XPOLand the deployment of the ten gauges (R is the radar siteand a number represents a gauge location) The total rainfallamount for the month at each gauge site is shown in Table 2At all gauge sites the total rainfall accumulation was over300mm for themonthThe rainfall amountmeasured duringthe same period with the NIMS-XPOL radar data used inthe rainfall estimation was about 50mm per 55 hours (seeTable 2)
The instrumental biases of all gauges were calculatedthrough an ideal experiment using laboratory and fieldobservations for 9 months from October 2009 to June 2010(except April 2010) at KNU The rain gauge dataset was
corrected for the instrumental bias prior to the detailedanalysis
3 Methodology
31 Attenuation Correction AlgorithmUsing Differential Prop-agation Phase Shift The attenuation correction algorithmwas performed by the procedure as shown in Figure 4 whichincluded (1) quality control of the radar data (2) estimationofΦDP (3) calculation of the attenuation amount and (4) theattenuation correction of 119885
119867and 119885DR
As a part of the quality control of the radar data 119885119867and
119885DR were corrected for instrumental biases 119885119867calibration
bias of about 37 dB was obtained by comparison with thereflectivity from the particle size velocity disdrometer (PAR-SIVEL) and from theNIMS-XPOL radarThe119885DR calibrationbias of minus15 dB was calculated from the vertical pointing data[19]The isolated point echo shows that the119885
119867values in their
surrounding pixels existed by less than 50 of the total pixelsto be removed The 120588HV threshold of 09 was applied to theentire radar measurement field
After applying the quality control measures ΦDP wasestimated from measured ΨDP by eliminating 120575 and theobservational noise through an iterative filtering techniqueIn this study the 20th FIR filter of Hubbert and Bringi [10]was used This filter preserved a monotonic increasing trendofΦDP due to the propagation medium while it removed the
4 Advances in Meteorology
1801 1816 1831
19011846
1931 1946 2001
1916
605550454030 35 65 702015100 25minus10
(dBZ)
Figure 2 PPI images of attenuation corrected 119885119867from 1801UTC July 10 2010 with an interval of 15min
Table 2 Total rainfall amount for the month and per analysis time
Gauge ID RG1 RG2 RG3 RG4 RG5 RG6 RG7 RG8 RG9 RG10Total rainfall amount for the month (mm) 3214 3717 3308 3139 3218 3426 3321 3239 3743 3245Total rainfall amount for the analysis time (mm) 476 560 510 529 552 571 544 544 513 531
smaller scale fluctuations compared to the filtering window(25 km when the gate size is 0125m) due to 120575 and theobservation noise Hubbert and Bringi [10] also found thatrepeating the filtering process 10 times produced good results
The additional problem in the estimation of ΦDP wasthe variable offset of ΦDP ΦDP observation can be noisyin near range Radar in general has ΦDP offset to avoidobserving negative value of ΦDP in near range Ideally theoffset should not change with the azimuthal angles Howeverthe offsets of measured ΨDP of the NIMS-X varied withthe azimuthal angles and ΦDP values decreased near theradar when rain cells existed above the radar Therefore
the offset was determined as the minimum value of ΦDPwithin 3 km of the radar and estimated filtered ΨDP wasadjusted accordingly
The calculation of 119860119867
and 119860DP was based on thefollowing two relationships [1 5 20]
119860119867 (119903) = 120590119870DP (119903)
119860DP (119903) = 120576119860119867 (119903) (1)
where 119860119867(119903) and 119860DP(119903) are the specific attenuation and
the specific differential attenuation at range 119903 from the radar
Advances in Meteorology 5
3510
3506
3502
3498
3494 Unnam
12628 12636 12644
1 234 96
8 75
10
R
Figure 3 Layout of NIMS-XPOL and the 10 rain gauges nearUnnam-myeon Muan-gun Jeollanam-do Korea R shows thelocation of the NIMS-XPOL the numbered points indicate theobservational sites of the 10 rain gauges
respectively119870DP(119903) is the specific differential one-way phaseshift calculated using the following equation
119870DP (119903) =ΦDP (1199032) minus ΦDP (1199031)
2 (1199032minus 1199031)
(2)
where 119903 is between 1199031and 1199032 The coefficients 120590 (= 03445)
and 120576 (= 01705) were derived from 119860119867 119860DP and 119870DP
obtained by scattering simulations with the DSDs data col-lected by the PrecipitationOccurrence Sensor System (POSS)from March to September of 2001 in Pusan Korea [21] Thetotal path integrated attenuation and differential attenuationwere calculated withΦDP by integrating the following
2int
119903
0
119860119867(1199031015840) 1198891199031015840= 120590ΦDP (119903) (3)
2int
119903
0
119860DP (1199031015840) 1198891199031015840= 120576 times 2int
119903
0
119860119867(1199031015840) 1198891199031015840 (4)
Number 2 in the equations indicates that the values are two-way values
Finally the measured reflectivity 119885119867-meas(119903) [dBZ] and
the differential reflectivity 119885DR-meas(119903) [dB] at 119903 were cor-rected with the calculated two-way path integrated attenua-tion and differential attenuation
119885119867-corr (119903) = 119885119867-meas (119903) + 2int
119903
0
119860119867(1199031015840) 1198891199031015840
119885DR-corr (119903) = 119885DR-meas (119903) + 2int119903
0
119860DP (1199031015840) 1198891199031015840
(5)
32 Radar Rain Estimation and Comparison with Rain GaugeData A comparison between estimated 119877 from the radar
Measured dual-polarization parameters(ZH ZDR ΦDP 120588HV)
Quality control
(ii) Speckling echo removal
(i) Calibration for instrumental bias of ZH ZDR
(iii) 120588HV threshold application
(ii) Adjustment of its offset(i) 20th FIR filter application to ΨDP
Estimation of ΦDP
Calculation of AH and ADP
Correction of ZH and ZDR
from DSDs120590 and 120576
Figure 4 The schematic diagram of the attenuation correctionalgorithm
parameters and the rain rate (119866) from the rain gauge data wasperformed to evaluate the attenuation correction algorithmIn addition to the rain attenuation the difference between119877 and 119866 was caused by the errors due to the instrumentaluncertainty and representativeness errors of the rain gaugedata the radar observational noise and the errors in the radarrainfall estimation The bias due to the instrumental uncer-tainty in gauge data was removed by properly calibrating thegauges through laboratory tests and a gauge intercomparisonThe representativeness errors were due to the mismatch ofthe sampling area (volume) between the gauge and the radarThe random error in the gauge data could be minimized bytime and areal integrationThe errors due to themeasurementnoise in the radar data could be reduced in a similar mannerThe time offset between the rain gauge and the radar samplingarea was ignored due to the close distance (about 16 km) andthe low elevation angle (3∘)
Therefore 119877 and 119866 were derived with two different radarsampling areas and with increasing accumulation time
(1)119877(038 km2) versus119866(0071198982)119877(038 km2)was estimated
from the average radar measurement with 9 pixels (038 km2at 16 km range) The gauge rainfall rate 119866(007m2) wasderived from the rainfall accumulation from an individualrain gauge that had a sampling area of 007m2
(2) 119877(6 1198961198982) versus 119866(6 1198961198982
) 119877(6 km2) was estimated fromthe areal averaged radar parameters at a 16 km times 31 km radar
6 Advances in Meteorology
Table 3 Two rain rate retrieval algorithms and boundary conditions [22]
119877 (119885119867) = 120572119885
120573
119867119877 (119885DR) = 120572119885
120573
1198671001120574119885DR
Z gt 35 dBZ Z lt 35 dBZ Z gt 35 dBZ 119885DR gt 03 dB Z lt 35 dBZ 119885DR gt 03 dB120572 353 times 10
minus2383 times 10
minus2103 times 10
minus2145 times 10
minus2
120573 621 times 10minus2
608 times 10minus2
985 times 10minus1
890 times 10minus1
120574 mdash mdash minus6479 minus5295
measurement resolution 119866(6 km2) was the average value of119866(007m2) from the ten gauges within the same area
Estimated 119877 from radar parameters was calculated withtwo rain estimators as shown in Table 3These rain estimatorswere derived by a least squares fit with the theoretical 119885
119867
119885DR and 119877 obtained by scattering simulations with theDSD data collected by POSS from March to Septemberof 2001 in Pusan Korea [21] To improve the estimationaccuracy each estimator was derived for two categoriesThe reflectivity threshold of 35 dBZ was used as a stan-dard stratiformconvective rain classification [22] The 119885DRthreshold of 03 dB was used to minimize the effect of the119885DR measurement noiseThePPI data with an elevation angleof 3∘ was used to suppress the beam shielding and groundclutterThe119885
119867and119885DR fields from the radar were smoothed
by the linear average of values at 3 times 3 pixels to reduce theobservational noise
The bias in 119866 from the ten rain gauges was corrected bythe laboratory test and the intercomparison with field dataThe detail is presented in Appendix To evaluate the atten-uation correction algorithm and verify the rainfall retrievalaccuracy the comparisons of119877 and119866were represented by thegeneral statistics such as normalized bias (NB) normalizedstandard deviation (NSD) root-mean square error (rmse)relative rms error (r rmse) and Spearman correlation (corr)as follows
NB119877-119866 (Δ119905) =
(1119899)sum119899minus1
119896=0[119877119896 (Δ119905) minus 119866119896 (Δ119905)]
119866 (Δ119905)
NSD119877-119866 (Δ119905) =
radic(1119899)sum119899minus1
119896=0[119877119896 (Δ119905) minus 119866119896 (Δ119905)]
2
119866 (Δ119905)
rmse119877-119866 (Δ119905) =
radicsum119899minus1
119896=0[119877119896 (Δ119905) minus 119866119896 (Δ119905)]
2
119899
rmse119877-119866 (Δ119905) = radic
sum119899minus1
119896=0[119877119896 (Δ119905) minus 119866119896 (Δ119905)]
2
sum119899minus1
119896=0[119866119896 (Δ119905)]
2
corr119877-119866 (Δ119905)
=
sum119899minus1
119896=0[119877119894119896 (Δ119905) minus 119877119894 (Δ119905)] [119866119895119896 (Δ119905) minus 119866119895 (Δ119905)]
radicsum119899minus1
119896=0[119877119894119896 (Δ119905) minus 119877119894 (Δ119905)]
2radicsum119899minus1
119896=0[119866119895119896 (Δ119905) minus 119866119895 (Δ119905)]
2
(6)
where 119899 is the number of 119877 and 119866 pairs at a given temporalintegrationΔ119905 and the overbar indicates the temporal average
33 Spatial Correlation and Variability of the Rain FieldsRain gauges are widely used to measure rainfall amountsdue tomany advantages including easy installation mobilitylow price and ease of data processing The rain gauge datais used as a reference for the ground rainfall amount forthe evaluation of rain retrieval algorithms adjustment ofother instruments data assimilation and other hydrologicalapplications However it is affected by instrumental errorsand representativeness errors due to the variability of rainfields
In a TB rain gauge the instrumental errors are due toimproper calibration the time delay of the tips leakage dueto its measuring principle and so forthThis error is exploredin the Appendix
The representativeness errors are subdivided into spatialand temporal representativeness They are mainly caused byhaving a small sampling area and a limited sampling time dueto the gauge measuring principle The actual rain fields varywith different spatiotemporal scales that are not equivalentto the spatial and temporal resolution of the rain gaugeTherefore the rain gauge data represent the actual rain fieldsin a very limited manner
These errors should be considered and minimized priorto using rain gauge data as ground truth Furthermore thedegree of these errors has to be understood quantitativelysince there can be other error sources in various applicationsIn this section the spatial and temporal variability of the rainfields are investigated using the rain data collected by the tenrain gauges for the month of July 2010 at Muan Korea
First the Spearman correlation coefficient 119903119894119895(Δ119905)
between a pair of 119866 values from the 119894th and 119895th individualrain gauge with integration time Δ119905 was calculated fromthe following equation to obtain the Spearman correlationcoefficient of the rain fields
119903119894119895 (Δ119905)
=
sum119899minus1
119896=0[119866119894119896 (Δ119905) minus 119866119894 (Δ119905)] [119866119895119896 (Δ119905) minus 119866119895 (Δ119905)]
radicsum119899minus1
119896=0[119866119894119896 (Δ119905) minus 119866119894 (Δ119905)]
2radicsum119899minus1
119896=0[119866119895119896 (Δ119905) minus 119866119895 (Δ119905)]
2
(7)
where 119899 is the number of gauge data pairs at given Δ119905 and theoverbar indicates the temporal average of the rainfall rate Itwas derived withΔ119905 to research the spatial and temporal scaleof the variation for the rain field
In addition the effect of the natural variability of the rainfields within two different areas (038 km2 and 6 km2) wasexplored by analyzing the bias and random error between thespatially averaged rainfall rate ⟨119866⟩ and 119866 from the individual
Advances in Meteorology 7
32km
21 km
11 km
20km40km60km
minus180 18020 30 40 50 60 70 80 90 100120140160100(a)
32 km
21 km
11 km
20km40km60km
minus180 18020 30 40 50 60 70 80 90 100120140160100(b)
Figure 5 PPIs of measured (a) and filtered (b) ΨDP of the NIMS-XPOL at 2101UTC July 10 2010
gauge The NB and NSD of 119866119894with ⟨119866⟩ were calculated as
follows
NB (119894) =(sum119899minus1
119896=0[119866119894119896 (Δ119905) minus ⟨119866119896 (Δ119905)⟩]) 119899
⟨119866 (Δ119905)⟩
NSD (119894) =(1119899)radic(sum
119899minus1
119896=0[119866119894119896 (Δ119905) minus ⟨119866119896 (Δ119905)⟩]
2)
⟨119866 (Δ119905)⟩
(8)
where 119894 indicates the 119894th gauge and the overbar indicates thetemporal average
This analysis provided a reasonable spatial average andtemporal accumulation resolution for the calculation ofmoreaccurate and realistic rainfall amounts from the gauge dataand a quantitative guideline regarding the errors in rainestimation due to the variability of the rain fields
4 Results
41 Attenuation Correction of Reflectivity and DifferentialReflectivity The attenuation correction algorithm in Sec-tion 31 was applied to the NIMS-XPOL radar data ΦDPwas estimated by eliminating 120575 and the observational noisefrom measured ΨDP using an iterative FIR filter The offsetof ΦDP was then adjusted Figure 5 shows the PPIs ofmeasured ΨDP and filtered ΦDP After applying the FIR filterthe small scale variation due to 120575 and observational noisein measured ΨDP was reduced and filtered ΨDP (estimatedΦDP) monotonically increased with an increase in rangeThe maximum value of estimated ΦDP reached to 140∘ Inaddition estimated ΦDP at the azimuth angles of 150∘ and200∘ sharply increased with the loss of signal beyond thesesharp gradient areas This reflected that strong convective
cells existed in these directions and that these cells caused thesevere signal attenuation EstimatedΦDP values between 250
∘
and 360∘ azimuth angle decreased after filtering since ΦDPoffsets were positive values in these directions Figure 6 showsthe range profile of measured and filtered ΨDP for estimationof ΦDP at 71
∘ 160∘ and 334∘ azimuth angles The unexpectedlarge values near the radarwere removed by the120588HV threshold(red line)The small fluctuation inmeasuredΨDP due to 120575 andthe observational noise was effectively removed after filteringThe value of estimated ΦDP was adjusted to 0 by the offset(blue line)
Estimated ΦDP was used to calculate the path integrated119860119867 and the measured119885
119867values were corrected as shown in
Figure 7 Figure 7(a) represents measured-attenuated119885119867and
relatively small values of 119885119867at far ranges This may indicate
that the 119885119867values at far ranges were affected by the strong
attenuation due to strong convective cells near the radarThisspatial pattern of 119885
119867was the typical feature of the attenuated
radar signal due to the rain After correcting the attenuationthe value of the 119885
119867field became larger over the entire area
and the radially attenuated 119885119867
was corrected by showinga typical rain cell pattern in 119885
119867(Figure 7(b)) The largest
difference between the uncorrected and the corrected 119885119867at
2101UTCwas over 50 dB However the white area behind thestrong echoes was not detected as there was a complete loss ofthe signal powerThese regions could not be corrected by thisalgorithmThe circle pattern of119885
119867fields within about a 4 km
range from theNIMS-XPOL could not present the actual rainechoes due to the instrumental issues (the same was true for119885DR) These results are also presented in the range profiles inFigure 8
Similarly measured 119885DR were corrected with the pathintegrated 119860DP derived from the path integrated 119860
119867using
(4) (see Figure 9) Before applying attenuation correction
8 Advances in Meteorology
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
0
50
100
150
200
ΦDP (∘) azimuth angle = 71∘
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
0
50
100
150
200
ΦDP (∘) azimuth angle = 160∘
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
0
50
100
150
200
ΦDP (∘) azimuth angle = 334∘
(c)
Figure 6 Range profiles of measured (black) 120588HV threshold isolated point removing (red) and FIR filter and offset adjusted (blue) ΨDP forestimation of ΦDP at 71
∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
32km
21 km
11 km
20km40km60km
minus10 10 15 20 25 30 35 40 45 50 55 60 65 700(a)
32 km
21 km
11 km
20km40km60km
minus10 10 15 20 25 30 35 40 45 50 55 60 65 700(b)
Figure 7 PPIs of measured (a) and attenuation corrected (b) 119885119867of the NIMS-XPOL for the case in Figure 5
Advances in Meteorology 9
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 71∘
minus20
0
20
40
60
80
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 160∘
minus20
0
20
40
60
80
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 334∘
minus20
0
20
40
60
80
(c)
Figure 8 Range profiles of measured (line) calibration bias 120588HV threshold and isolated point removing (dot) and attenuation corrected(dash) 119885
119867at 71∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
the attenuation signature in119885DR was determined Ideally the119885DR value for rain should be positive since rain drops areoblate due to aerodynamic force However themeasured119885DRfield in Figure 9 showed a negative value over wide areasAfter correcting the attenuation the119885DR field was larger than0 dB and showed a realistic rain patternThis is also presentedin the range profile of Figure 10The largest difference of119885DRbetween before and after attenuation correction at 2101UTCwas about 8 dB The 119885DR values around the 220∘ azimuthangle beyond 40 km were the largest They did not matchwith the area of the largest 119885DR values (around the 210∘azimuth angle with a range of 20 km and 40 km) This wasan unexpected result because 119885DR is proportional to 119885
119867in
theory This mismatch between 119885119867
and 119885DR at far rangemay be due to contamination due to beam broadening andaccumulated errors due to uncertainty in the attenuationcorrection with different ranges
Figures 11 and 12 show the time series of 119885119867
and119885DR at the ten gauge sites before and after applying theattenuation correction The differences between uncorrectedand corrected values existed after 1800UTC and were at amaximumaround 2100UTCThis attenuation correctionwas
evaluated through comparison with the theoretical values of119885119867and 119885DR In Figure 13 the uncorrected 119885
119867-119885DR scatter
plots are apart from the theoretical119885119867-119885DR values [23] After
attenuation correction the119885119867-119885DR scatter plots were close to
the calculated valuesThis indicated that119885119867and119885DR over the
gauge sites were corrected with reliable values
42 Variability of the Rain Fields The Spearman correlationcoefficient of the rainfall rate from the 10 spatially dis-tributed rain gauges was calculated with accumulation timeand separation distance (Figure 14) The overall Spearmancorrelation coefficient decreased with increasing separationdistance at all accumulation timesThis tendency was similarto the result from Habib and Krajewski [24] However thecorrelation value was higher in this study Note that the gaugetip resolution in Habib and Krajewski [24] was 0254mmwhile it was 02mm in this study In addition the rainfall ratesfrom the gauge data used in this study mainly consisted ofmoderate rainfall rates while Habib and Krajewski [24] useddata containing heavy rainfall rates
For accumulation times of less than 10min the corre-lation at the same distance was dramatically increased with
10 Advances in Meteorology
32km
21 km
11 km
20km40km60km
18
14
00
02
04
06
26
40
minus10
22
minus20
30
10
50
minus30
(a)
32 km
21 km
11 km
20km40km60km
18
14
00
02
04
06
26
40
minus10
22
minus20
30
10
50
minus30
(b)
Figure 9 PPIs of measured (a) and attenuation corrected (b) 119885DR of NIMS-XPOL for the case in Figure 5
increasing accumulation time and the slope of the correlationfunction was steep This was due to the reduction of theinstrumental errors of the gauge data by time accumulationThis indicated that rainfall accumulated for less than 10minfrom single gauge data cannot represent the rainfall over itssurrounding area due to instrumental errors For accumula-tion times longer than 10min the slope became less steepThis reflected the remaining variability of the rain field Inaddition the accumulated rainfall rate of about 10min wassuitable for minimizing the instrumental uncertainty whilepreserving the variation in the rain field The correlationreached about 098 at 05 km in 10min of accumulationand this approached 098 at 10 km at 60min accumulationTherefore a single rain gauge mostly represented the rainfallrate within 05 km and 10 km with 10min and 60minaccumulation respectively
The NB between an areal averaged rainfall rate ⟨119866⟩ anda single gauge rainfall rate 119866 is shown with accumulationtime in Figures 15(a) and 16(a) Where the average area was038 km2 (corresponding to 9 pixels of NIMS-XPOL radarat 16 km range) the NB reached a 6ndash8 maximum Wherethe averaged area is 6 km2 (corresponding to the coverage forall of the gauges) the NB reached about a 10 maximumFurthermore the NB hardly changed with accumulationtime Note that this statistic was derived after removinginstrumental bias This indicated that the rainfall rate froma single gauge was biased due to the spatial variability of therain fields and this bias was maintained with accumulationtime
The NSD of a single gauge for a mean area of 038 km2was over 40 and it reached about 70ndash90 for a mean areaof 6 km2 at an accumulation time of 1min (see Figures 15(b)and 16(b)) This statistic included instrumental errors (about
40ndash60 at 16mmhminus1 and 60ndash80 at 10mmhminus1 as shownin the Appendix) as well as the spatial variability of the rainfields This indicated that the variance between the point andarea values from the gauge data at short accumulation timeswas similar to the errors due to the instrumental uncertaintyof the rain gaugeTheNSD in 60min accumulation decreasedto over 10 and about 15ndash45 with a mean area of 038 km2and 6 km2 respectively including the NSD of 6ndash10 due toinstrumental uncertainty This indicated that the rainfall ratefrom single gauges may differ significantly from the actualrainfall rate due to the variability of the rain fields and theinstrumental uncertainty in short accumulation times
43 Verification of Rainfall Retrieval The rainfall rate fromthe NIMS-XPOL radar measurements was retrieved withtwo rainfall rate estimators 119877(119885
119867) and 119877(119885
119867 119885DR) The
estimated rainfall rate was verified through comparison withthe rainfall rate (119866) from the rain gauge data These twoestimators were applied before and after the attenuation cor-rection First 119877(038 km2) was compared with 119877(007 km2)at 10min accumulation (Figure 17) Using 119877(119885
119867) the NB
NSD rmse and r rmse without the attenuation correctionwere significantly large at 70 127 127 (mmhr) and78 respectively However they were reduced to minus152 521 (mmhr) and 32 respectively after applying theattenuation correction Using 119877(119885
119867 119885DR) the NB was just
3 NSD was 49 rmse was 496 (mmhr) and r rmsewas 31 before applying attenuation correction This wassimilar to the result from 119877(119885
119867) after attenuation correction
However there was no significant change in119877(119885119867 119885DR)with
attenuation correction This means that 119877(119885119867 119885DR) was less
sensitive to rain attenuation than was 119877(119885119867) because 119885
119867is
proportional to 119877 and 119885DR is inversely proportional to 119877
Advances in Meteorology 11
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 71∘
minus10
minus5
0
5
10
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 160∘
minus10
minus5
0
5
10
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 334∘
minus10
minus5
0
5
10
(c)
Figure 10 Range profiles of measured (line) calibration bias 120588HV threshold and isolated point removing (dot) and attenuation corrected(dash) 119885DR at 71∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
in this rain estimator (see Table 3) Furthermore comparingthe analysis of variance results between areal (038 km2) andpoint (007m2) values from the gauge data (Section 42)half of the errors after attenuation correction were producedby the small scale variability of the rain fields and theinstrumental uncertainty of the rain gauge
A similar comparison is shown in Figure 18 for a 60minaccumulation Using 119877(119885
119867) the NB NSD rmse and r rmse
were decreased from minus65 86 71 (mmhr) and 70to minus5 34 28 (mmhr) and 28 by the attenuationcorrection respectively Using 119877(119885
119867 119885DR) the NB NSD
rmse and r rmse were only slightly decreased after theattenuation correction However the NSD rmse and r rmsewere reduced when compared at 10min accumulation In thesame accumulation time the errors in the gauge data reachedto over 15
This comparison was affected by different sampling areasof the gauges and the radar Therefore the areal averagevalues were compared (Figure 19) Using119877(119885
119867)with the areal
averaged radar measurements the NB was reduced fromminus69 to 7 NSD was reduced from 122 to 42 rmsewas reduced from 39 (mmhr) to 13 (mmhr) and r rmse
was reduced from 77 to 26 at an accumulation time of10min Using 119877(119885
119867 119885DR) the NB NSD rmse and r rmse
were decreased fromminus24 67 21 (mmhr) and 42 to 539 12 (mmhr) and 24 respectively
TheNB and r rmsewithmatched sampling areas betweenthe gauge and the radar are shown in Figures 20 and 21 Thereduction of NB due to temporal accumulation was smallin both of the rain estimators The NB was reduced to lessthan 7 by attenuation correction for all accumulation timeswith both119877(119885
119867) and119877(119885
119867 119885DR)The attenuation correction
effect was significantly greater in119877(119885119867) Similarly the r rmse
was reduced significantly in 119877(119885119867) by attenuation correction
(from 68 to 25 at a 60min accumulation) The reductionof the r rmse due to temporal accumulation was larger in119877(119885119867 119885DR) because the significant observational noise of
119885DR was smoothed The r rmse was decreased from 18to 15 in 119877(119885
119867 119885DR) at a 60min accumulation time after
attenuation correction
5 Conclusions
In this study an attenuation correction algorithm based onΦDP was developed to improve the accuracy of the rainfall
12 Advances in Meteorology
1st gauge site 2nd gauge site
3rd gauge site 4th gauge site
5th gauge site 6th gauge site
7th gauge site 8th gauge site
9th gauge site 10th gauge site
17 18 19 20 2116Time (h)
17 18 19 20 2116Time (h)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
UncorrectedCorrected
UncorrectedCorrected
Figure 11 Time series of measured (red circle) and attenuation corrected (blue circle) 119885119867at the 10 rain gauge sites
Advances in Meteorology 13
UncorrectedCorrected
UncorrectedCorrected
17 18 19 20 2116Time (h)
17 18 19 20 2116Time (h)
1st gauge site 2nd gauge site
3rd gauge site 4th gauge site
5th gauge site 6th gauge site
7th gauge site 8th gauge site
9th gauge site 10th gauge site
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
Figure 12 Time series of measured (red circle) and attenuation corrected (blue circle) 119885DR at the 10 rain gauge sites
14 Advances in Meteorology
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Uncorrected(a)
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Corrected Simulated from DSD(b)
Figure 13 Scatter plots of 119885119867and 119885DR before (a) and after (b) attenuation correction at the 10 rain gauge sites DSD observation results (dot
[21])
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
04
05
06
07
08
09
10
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
Accum time = 1min
Accum time = 10min
Accum time = 2min
Accum time = 20min
Accum time = 5min
Accum time = 60min
Figure 14 The Spearman correlation coefficient 119903 between data from two rain gauges at different accumulation times
estimation using X-band radar This algorithm was evalu-ated with radar measurements from NIMS-XPOL Howevermeasured ΨDP contained 120575 and observation noise as wellas ΦDP Furthermore the ΨDP offset varied with azimuthangles and decreased near the radar when rain cells were
above the radar In this study ΦDP was estimated by aniterative filtering technique and by adjusting with an offsetdetermined as the minimum value ofΦDP within 3 km of theradar As a result estimated ΦDP monotonically increasedThe path integrated value of119860
119867and119860DP was calculated from
Advances in Meteorology 15
2nd 6th 9th gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(038km2) versus G
(a) Normalized bias
⟨G⟩(038km2) versus G
2nd 6th 9th gaugesMean NSD
Mean R
0
20
40
60
80
100
NSD
()
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 15 (a) The NB and (b) NSD between ⟨119866⟩(038 km2) and 119866 from three rain gauges with accumulation time ⟨119866⟩ is the temporallyaveraged rainfall rate of ⟨119866⟩
All gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(6km2) versus G
(a) Normalized bias
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
0
20
40
60
80
100
NSD
()
All gaugesMean NSD
Mean R
⟨G⟩(6km2) versus G
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 16 (a) The NB and (b) NSD between ⟨119866⟩(6 km2) and 119866 from 10 rain gauges with accumulation time ⟨119866⟩ is the temporally averaged⟨119866⟩ with accumulation time
estimated ΦDP using 119860119867-ΦDP and 119860DP-ΦDP relationships
derived from the DSDs data Measured 119885119867
and 119885DR werecorrected with the calculated path integrated 119860
119867and 119860DP
by each gate The attenuated signal patterns of 119885119867and 119885DR
were restored to a typical rain echo pattern However theattenuation corrected 119885
119867and 119885DR fields were not matched
at distant ranges This was likely because of contaminationdue to beam broadening and the accumulation of errorswith increasing range Comparing the 119885
119867-119885DR scatter plots
over the 10 rain gauge sites with the results obtained froma scattering simulation with the DSD data the attenuation
corrected 119885119867-119885DR scatter plots are in good agreement with
the 119885119867-119885DR scatter plots from the DSDs
The validation for the improvement of accuracy in therainfall estimation through attenuation correction was per-formed by comparing with dense rain gauge network data119877(119885119867) from the attenuated (observed)119885
119867produced a severe
underestimation with an NB NSD rmse and r rmse of70 127 127 (mmhr) and 78 at 10min accumulationrespectively while 119877(119885
119867) from the attenuation corrected 119885
119867
showed better agreement with the gauge measurements withan NB NSD rmse and r rmse of minus1 52 521 (mmhr)
16 Advances in Meteorology
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus070
NSD = 127
rmse = 1279
corr = 0858
Corrected
NB = minus001
NSD = 052
rmse = 521
corr = 0927
r_rmse = 0786 r_rmse = 0320
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = 003
NSD = 049
rmse = 496
r_rmse = 0305
corr = 0936
Corrected
NB = 001
NSD = 050
rmse = 502
r_rmse = 0308
corr = 0936
(b) 119877(119885119867119885DR)
Figure 17 The scatter plots of 119877(038 km2 10min) versus 119866(007m2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
CorrectedUncorrected
R(m
mh
)
0
10
20
30
40
10 20 30 400G (mmh)
Uncorrected
NB = minus065
NSD = 086
rmse = 706
r_rmse = 0702
corr = 0878
Corrected
NB = minus005
NSD = 034
rmse = 281
r_rmse = 0279
corr = 0866
(a) 119877(119885119867)
CorrectedUncorrected
R(m
mh
)
10 20 30 400G (mmh)
0
10
20
30
40
Uncorrected
NB = 008
NSD = 023
rmse = 189
r_rmse = 0188
corr = 0942
Corrected
NB = 007
NSD = 025
rmse = 205
r_rmse = 0203
corr = 0940
(b) 119877(119885119867119885DR)
Figure 18 The scatter plots of 119877(038 km2 60min) versus 119866(007m2 60min) (a) using 119885119867and (b) using 119885
119867and 119885DR
and 32 at 10min accumulation respectively At 60minaccumulation theNBNSD rmse and r rmsewere decreasedfrom minus65 86 71 (mmhr) and 70 to minus5 3428 (mmhr) and 28 respectively Using 119877(119885
119867 119885DR) the
NB NSD rmse and r rmse were only slightly reduced even if
attenuation was corrected but were smaller than when using119877(119885119867) This indicated that 119877(119885
119867 119885DR) was less sensitive to
the attenuation than 119877(119885119867) In addition half of the errors
that remained after attenuation correction were caused by thesmall scale variability of the rain fields and the instrumental
Advances in Meteorology 17
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus069
NSD = 122
rmse = 391
r_rmse = 0770
corr = 0893
Corrected
NB = 007
NSD = 042
rmse = 133
r_rmse = 0262
corr = 0960
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = minus024
NSD = 067
rmse = 215
r_rmse = 0424
corr = 0718
Corrected
NB = 005
NSD = 039
rmse = 124
r_rmse = 0245
corr = 0891
(b) 119877(119885119867119885DR)
Figure 19 The scatter plots of 119877(6 km2 10min) versus 119866(6 km2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
R(ZH)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(a) 119877(119885119867)
R(ZH ZDR)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(b) 119877(119885119867119885DR)
Figure 20 The normalized bias of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and 119885DR The red
point-line and the blue point-line indicate without and with the attenuation correction respectively
uncertainty of the rain gauges Using an areal averagedvalue the NB was reduced to less than 7 with attenua-tion correction using either rainfall estimator regardless ofaccumulation timeThe r rmse were significantly reduced byattenuation correction in 119877(119885
119867) and the reduction of r rmse
with 119877(119885119867 119885DR) was even more dramatic
As summarized above the rain attenuation for 119885119867
and 119885DR in the X-band was corrected by ΦDP and theresults showed a good agreement with the value obtainedby scattering simulations with the DSD datasets The resultsof comparison with rain gauge data also showed a signifi-cant improvement in the accuracy of rainfall estimation by
18 Advances in Meteorology
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10r_
rmse
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH)
(a) 119877(119885119867)
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10
r_rm
se
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH ZDR)
(b) 119877(119885119867119885DR)
Figure 21 The relative root-mean square error of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and
119885DR The red point-line and the blue point-line indicate without and with the attenuation correction respectively
attenuation correction which supports the quantitative useof the X-band radar in rain data retrieval
Appendix
A Instrumental Uncertainty of the Rain Gauge
The instrumental uncertainty of the TB rain gauge is asso-ciated with improper calibration leakage and precipitationmissing the gauge due to the measurement principle andwetting and evaporation losses In addition external sourcessuch as wind turbulence and installed position also con-tribute to the instrumental uncertainty Ciach [17] reportedthat the TB gauge data suffers a significant uncertainty asso-ciated with random differences between closely collocatedTB gauges through the analysis of the error as a functionof the rainfall intensity for three timescales In addition theimproper calibration induces a bias in the rainfall measure-ment and this bias cannot be removed by increasing accu-mulation timeTherefore the proper correction or processingof TB gauge data is necessary to obtain accurate and reliablereading of the rainfall amount
In this appendix instrumental biases were calculatedto reduce the instrumental uncertainty through an idealexperiment in the laboratory and a field intercomparisonwith 10 TB rain gauges In addition the random errors werequantitatively calculated as a function of rainfall rate andaccumulation time
A1 Correction of Instrumental Bias The instrumental bias119861119868(119894) of the 119894th gauge was defined as a combination
of the absolute bias 119861119860(ref) of the reference gauge and
the relative bias 119861119877(119894) of the 119894th gauge according to the
following equation
119861119868 (119894) = 119861119860 (ref) times 119861119877 (119894) (A1)
119861119860(ref) signified the ratio of the calculated value from
the instrumental resolution of the reference gauge to the truevalue measured by other instruments
119861119860 (ref) =
rainfalltruthrainfallgauge (ref)
(A2)
In this study 119861119860(ref) was derived by an ideal experiment
using an electronic scale First the reference gauge was puton the scale and the scale was initialized The water dropsflowed out into the receiving orifice of the reference gaugeuntil the tipping number became 60Themeasured weight of60 tips was 2144 g while the theoretical (or expected) valuewas 2244 g for the reference gauge (receiving orifice size of1545 cm 60 tips and 02mm tip resolution) Therefore theabsolute bias of the reference gauge from (A2) was about0955 This indicates that the rainfall amount measured bythis reference gauge was overestimated by about 45 andthis bias should be corrected prior to any interpretation
119861119877(119894) represented the ratio of the observed (recorded)
value from the 119894th gauge to the observed (recorded) valuefrom the reference gauge
119861119877 (119894) =
rainfallgauge (ref)rainfallgauge (119894)
(A3)
119861119877(119894) was calculated from ten collocated rain gauges
including the reference gauge located at an observation field
Advances in Meteorology 19
Table 4 The instrumental and relative biases of the 10 tipping bucket rain gauges
Gauge ID RG1 RG2 RG3 RG4 RG5 RG6 RG7 RG8 RG9 RG10Instrumental bias 094 108 099 100 096 104 098 102 108 098Relative bias 098 113 103 105 100 109 102 107 113 102
013610152230406080120150200250300400
()
Rain
rate
(mm
h)
100101
Accumulation time (min)
10
100NSD ()
Figure 22 Two-dimensional normalized standard deviationbetween the standard rainfall and each rain gauge
at KNUKoreaThe observational period for intercomparisonwas from October 2009 to June 2010 (14 rainy days) andthe total accumulated rainfall for this period was about300mm The results are presented in Table 4 All relativebiases were mainly larger than 10 This indicated that themost of gauges underestimated the rainfall amount comparedto the reference rain gauge The relative bias of the referencegauge (119894 = 5) was 10
Finally 119861119860(119894) was calculated from (A1) (see Table 4) All
instrumental biases were between 090 and 11 and the meanvalue of the biases was nearly 10 This indicated that therainfall rate from a single gauge data had a maximum 10bias These biases were corrected for each gauge to removeexisting instrumental errors
A2 Quantification of Random Errors To quantify the ran-dom error due to instrumental uncertainty the NSD of theten gauges was calculated with accumulation time and rainrate using data from an intercomparison The NSD of the 119894thgauge compared to the reference gauge was derived for 20rain rate classes (Δ119877) and 17 accumulation time classes (Δ119905)as follows
NSD119894 (Δ119905 Δ119877)
=
(1119899)radicsum119899minus1
119896=0[1198661015840
119870(Δ119905 Δ119877) minus 119866119894119896 (Δ119905 Δ119877)]
2
1198661015840
(A4)
where 1198661015840 and 119866 indicate the reference and individual gaugesrespectively The overbar means the temporal average ofthe rainfall rate The NSDs of nine different gauges (exceptthe reference gauge) were averaged to obtain the two-dimensional average NSD (Figure 22) The average NSD
tended to decrease with increasing accumulation time andwith increasing rain rate The NSD was about 10ndash15when accumulation time was 10min and the rain rate was10mmhminus1 The instrumental errors were quite significant forlonger accumulation times and higher rainfall rates Theseerrors should be used as a guideline when a single gauge isused as a reference for any validation study
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper has been revised and extended fromMs Young-AOhrsquos MS thesis [25] of Kyungpook National University Thisresearch is supported by Development and Application ofCross Governmental Dual-Pol Radar Harmonization (WRC-2013-A-1) project of theWeather Radar Center KoreaMeteo-rological AdministrationThis work was funded by the KoreaMeteorological Administration Research and DevelopmentProgram under Grant KMIPA2015-1010
References
[1] V N Bringi V Chandrasekar N Balakrishnan and D SZrnic ldquoAn examination of propagation effects in rainfall onradar measurements at microwave frequenciesrdquo Journal ofAtmospheric and Oceanic Technology vol 7 no 6 pp 829ndash8401990
[2] A Huggel W Schmid and AWaldvogel ldquoRaindrop size distri-butions and the radar bright bandrdquo Journal of AppliedMeteorol-ogy vol 35 no 10 pp 1688ndash1701 1996
[3] G W Lee Errors in rain measurement by radar effect of vari-ability of drop size distributions [PhD thesis]McGill University2003
[4] J D Tuttle and R E Rinehart ldquoAttenuation correction in dual-wavelength analysesrdquo Journal of ClimateampAppliedMeteorologyvol 22 no 11 pp 1914ndash1921 1983
[5] S-G Park V N Bringi V Chandrasekar M Maki andK Iwanami ldquoCorrection of radar reflectivity and differentialreflectivity for rain attenuation at X band Part I theoretical andempirical basisrdquo Journal of Atmospheric and Oceanic Technol-ogy vol 22 no 11 pp 1621ndash1632 2005
[6] G Scarchilli E Gorgucci V Chandrasekar and T A SeligaldquoRainfall estimation using polarimetric techniques at C-bandfrequenciesrdquo Journal of Applied Meteorology vol 32 no 6 pp1150ndash1160 1993
[7] M N Anagnostou E N Anagnostou and J VivekanandanldquoCorrection for rain path specific and differential attenuation ofX-band dual-polarization observationsrdquo IEEE Transactions onGeoscience and Remote Sensing vol 44 no 9 pp 2470ndash24802006
20 Advances in Meteorology
[8] J Kalogiros M N Anagnostou E N Anagnostou M Mon-topoli E Picciotti and F S Marzano ldquoEvaluation of a newpolarimetric algorithm for rain-path attenuation correction ofX-band radar observations against disdrometerrdquo IEEE Transac-tions on Geoscience and Remote Sensing vol 52 no 2 pp 1369ndash1380 2014
[9] W-Y Chang J Vivekanandan and T-C C Wang ldquoEstimationof X-band polarimetric radar attenuation and measurementuncertainty using a variational methodrdquo Journal of AppliedMeteorology and Climatology vol 53 no 4 pp 1099ndash1119 2014
[10] J Hubbert and V N Bringi ldquoAn iterative filtering technique forthe analysis of copolar differential phase and dual-frequencyradar measurementsrdquo Journal of Atmospheric and OceanicTechnology vol 12 no 3 pp 643ndash648 1995
[11] K V Mishra W F Krajewski R Goska et al ldquoDeploy-ment and performance analyses of high-resolution iowa XPOLradar system during the NASA IFloodS campaignrdquo Journal ofHydrometeorology vol 17 no 2 pp 455ndash479 2016
[12] M Thurai G J Huang V N Bringi W L Randeu and MSchonhuber ldquoDrop shapes model comparisons and calcu-lations of polarimetric radar parameters in rainrdquo Journal ofAtmospheric and Oceanic Technology vol 24 no 6 pp 1019ndash1032 2007
[13] M D Humphrey J D Istok J Y Lee J A Hevesi and AL Flint ldquoA new method for automated dynamic calibration oftipping-bucket rain gaugesrdquo Journal of Atmospheric andOceanicTechnology vol 14 no 6 pp 1513ndash1519 1997
[14] E HabibW F Krajewski V Nespor andA Kruger ldquoNumericalsimulation studies of rain gage data correction due to windeffectrdquo Journal of Geophysical ResearchAtmospheres vol 104 no16 Article ID 1999JD900228 pp 19723ndash19733 1999
[15] G J Ciach and W F Krajewski ldquoOn the estimation of radarrainfall error variancerdquoAdvances inWater Resources vol 22 no6 pp 585ndash595 1999
[16] E Habib W F Krajewski and A Kruger ldquoSampling errors oftipping-bucket rain gaugemeasurementsrdquo Journal of HydrologicEngineering vol 6 no 2 pp 159ndash166 2001
[17] G J Ciach ldquoLocal random errors in tipping-bucket rain gaugemeasurementsrdquo Journal of Atmospheric andOceanic Technologyvol 20 no 5 pp 752ndash759 2003
[18] EHabibW F Krajewski andG J Ciach ldquoEstimation of rainfallinterstation correlationrdquo Journal ofHydrometeorology vol 2 no6 pp 621ndash629 2001
[19] NIMS ldquoStudy on estimation of rainfall amount and classifi-cation of hydrometeor species using dual-polarimetric radar(III)rdquo pp 1ndash20 2010
[20] E Gorgucci V Chandrasekar and L Baldini ldquoCorrection ofX-band radar observation for propagation effects based on theself-consistency principlerdquo Journal of Atmospheric and OceanicTechnology vol 23 pp 1668ndash1681 2006
[21] NIMS ldquoStudy on estimating raindrop size distribution forquantitative estimation of radar rainfall amountrdquo pp 1ndash55 2009
[22] NIMS ldquoQuantitative precipitation estimation and hydrometeoridentification using dual-polarization radar-Phase IIrdquo pp 1ndash742009
[23] S-G Park M Maki K Iwanami V N Bringi and V Chan-drasekar ldquoCorrection of radar reflectivity and differential reflec-tivity for rain attenuation at X band Part II Evaluation andapplicationrdquo Journal of Atmospheric and Oceanic Technologyvol 22 no 11 pp 1633ndash1655 2005
[24] E Habib and W F Krajewski ldquoUncertainty analysis of theTRMM ground-validation radar-rainfall products applicationto the TEFLUN-B field campaignrdquo Journal of Applied Meteorol-ogy vol 41 no 5 pp 558ndash572 2002
[25] Y-A Oh Rainfall estimation from X-band dual-polarizationradar measurements effects of attenuation correction and evalu-ation with a dense rain gauge network [MS thesis] KyungpookNational University 2011
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom
Applied ampEnvironmentalSoil Science
Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal of
Geophysics
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
Advances in Meteorology 3
Table 1 General characteristics of the NIMS-XPOL radar [21]
X-band polarimetric mobile Doppler weather radar (ARC-X250MP) specification
Transmitter
Type MagnetronPeak power (kW) 250Frequency (MHz) 9360
PRF (MHz) 0sim2500
Antenna
Diameter (m) 244Polarization Orthogonal linear simultaneous H amp VBeamwidth (∘) 095Gain (dB) gt43
Data processingNumber of gates 30001000 PRFGate spacing 30m to 1000mRaw data 119885
119867 119881119905 119878119882 119885DR ΦDP 120588HV
(a) (b)
Figure 1 Photographs of the NIMS-XPOL radar installed onto (a) a truck bed and (b) a steel tower bed (courtesy of NIMS)
22 Rain Gauge Data The rain gauge data was collectedby ten tipping bucket (TB) rain gauges operated by theKyungpook National University (KNU) Korea The tip sizeof all of the gauges was 02mm and the diameter of thereceiving orifice was 1539 cm The time resolution was 05 sand the data was recorded based on the event (tipping) timeTo compare with the estimated rain rate (119877) from the NIMS-XPOL radar the rain accumulation from ten gauges wascollected during the month of July 2010 near Unnam-myeonwhich is located at a distance of about 16 km at 160∘ azimuthangle from the NIMS-XPOL radarThe area was only slightlyaffected by beam shielding of the NIMS-XPOL Consideringthe geographical situation and the limited locations for theinstallation all gauges were installed on the rooftops ofbuildings Figure 3 shows the location of the NIMS-XPOLand the deployment of the ten gauges (R is the radar siteand a number represents a gauge location) The total rainfallamount for the month at each gauge site is shown in Table 2At all gauge sites the total rainfall accumulation was over300mm for themonthThe rainfall amountmeasured duringthe same period with the NIMS-XPOL radar data used inthe rainfall estimation was about 50mm per 55 hours (seeTable 2)
The instrumental biases of all gauges were calculatedthrough an ideal experiment using laboratory and fieldobservations for 9 months from October 2009 to June 2010(except April 2010) at KNU The rain gauge dataset was
corrected for the instrumental bias prior to the detailedanalysis
3 Methodology
31 Attenuation Correction AlgorithmUsing Differential Prop-agation Phase Shift The attenuation correction algorithmwas performed by the procedure as shown in Figure 4 whichincluded (1) quality control of the radar data (2) estimationofΦDP (3) calculation of the attenuation amount and (4) theattenuation correction of 119885
119867and 119885DR
As a part of the quality control of the radar data 119885119867and
119885DR were corrected for instrumental biases 119885119867calibration
bias of about 37 dB was obtained by comparison with thereflectivity from the particle size velocity disdrometer (PAR-SIVEL) and from theNIMS-XPOL radarThe119885DR calibrationbias of minus15 dB was calculated from the vertical pointing data[19]The isolated point echo shows that the119885
119867values in their
surrounding pixels existed by less than 50 of the total pixelsto be removed The 120588HV threshold of 09 was applied to theentire radar measurement field
After applying the quality control measures ΦDP wasestimated from measured ΨDP by eliminating 120575 and theobservational noise through an iterative filtering techniqueIn this study the 20th FIR filter of Hubbert and Bringi [10]was used This filter preserved a monotonic increasing trendofΦDP due to the propagation medium while it removed the
4 Advances in Meteorology
1801 1816 1831
19011846
1931 1946 2001
1916
605550454030 35 65 702015100 25minus10
(dBZ)
Figure 2 PPI images of attenuation corrected 119885119867from 1801UTC July 10 2010 with an interval of 15min
Table 2 Total rainfall amount for the month and per analysis time
Gauge ID RG1 RG2 RG3 RG4 RG5 RG6 RG7 RG8 RG9 RG10Total rainfall amount for the month (mm) 3214 3717 3308 3139 3218 3426 3321 3239 3743 3245Total rainfall amount for the analysis time (mm) 476 560 510 529 552 571 544 544 513 531
smaller scale fluctuations compared to the filtering window(25 km when the gate size is 0125m) due to 120575 and theobservation noise Hubbert and Bringi [10] also found thatrepeating the filtering process 10 times produced good results
The additional problem in the estimation of ΦDP wasthe variable offset of ΦDP ΦDP observation can be noisyin near range Radar in general has ΦDP offset to avoidobserving negative value of ΦDP in near range Ideally theoffset should not change with the azimuthal angles Howeverthe offsets of measured ΨDP of the NIMS-X varied withthe azimuthal angles and ΦDP values decreased near theradar when rain cells existed above the radar Therefore
the offset was determined as the minimum value of ΦDPwithin 3 km of the radar and estimated filtered ΨDP wasadjusted accordingly
The calculation of 119860119867
and 119860DP was based on thefollowing two relationships [1 5 20]
119860119867 (119903) = 120590119870DP (119903)
119860DP (119903) = 120576119860119867 (119903) (1)
where 119860119867(119903) and 119860DP(119903) are the specific attenuation and
the specific differential attenuation at range 119903 from the radar
Advances in Meteorology 5
3510
3506
3502
3498
3494 Unnam
12628 12636 12644
1 234 96
8 75
10
R
Figure 3 Layout of NIMS-XPOL and the 10 rain gauges nearUnnam-myeon Muan-gun Jeollanam-do Korea R shows thelocation of the NIMS-XPOL the numbered points indicate theobservational sites of the 10 rain gauges
respectively119870DP(119903) is the specific differential one-way phaseshift calculated using the following equation
119870DP (119903) =ΦDP (1199032) minus ΦDP (1199031)
2 (1199032minus 1199031)
(2)
where 119903 is between 1199031and 1199032 The coefficients 120590 (= 03445)
and 120576 (= 01705) were derived from 119860119867 119860DP and 119870DP
obtained by scattering simulations with the DSDs data col-lected by the PrecipitationOccurrence Sensor System (POSS)from March to September of 2001 in Pusan Korea [21] Thetotal path integrated attenuation and differential attenuationwere calculated withΦDP by integrating the following
2int
119903
0
119860119867(1199031015840) 1198891199031015840= 120590ΦDP (119903) (3)
2int
119903
0
119860DP (1199031015840) 1198891199031015840= 120576 times 2int
119903
0
119860119867(1199031015840) 1198891199031015840 (4)
Number 2 in the equations indicates that the values are two-way values
Finally the measured reflectivity 119885119867-meas(119903) [dBZ] and
the differential reflectivity 119885DR-meas(119903) [dB] at 119903 were cor-rected with the calculated two-way path integrated attenua-tion and differential attenuation
119885119867-corr (119903) = 119885119867-meas (119903) + 2int
119903
0
119860119867(1199031015840) 1198891199031015840
119885DR-corr (119903) = 119885DR-meas (119903) + 2int119903
0
119860DP (1199031015840) 1198891199031015840
(5)
32 Radar Rain Estimation and Comparison with Rain GaugeData A comparison between estimated 119877 from the radar
Measured dual-polarization parameters(ZH ZDR ΦDP 120588HV)
Quality control
(ii) Speckling echo removal
(i) Calibration for instrumental bias of ZH ZDR
(iii) 120588HV threshold application
(ii) Adjustment of its offset(i) 20th FIR filter application to ΨDP
Estimation of ΦDP
Calculation of AH and ADP
Correction of ZH and ZDR
from DSDs120590 and 120576
Figure 4 The schematic diagram of the attenuation correctionalgorithm
parameters and the rain rate (119866) from the rain gauge data wasperformed to evaluate the attenuation correction algorithmIn addition to the rain attenuation the difference between119877 and 119866 was caused by the errors due to the instrumentaluncertainty and representativeness errors of the rain gaugedata the radar observational noise and the errors in the radarrainfall estimation The bias due to the instrumental uncer-tainty in gauge data was removed by properly calibrating thegauges through laboratory tests and a gauge intercomparisonThe representativeness errors were due to the mismatch ofthe sampling area (volume) between the gauge and the radarThe random error in the gauge data could be minimized bytime and areal integrationThe errors due to themeasurementnoise in the radar data could be reduced in a similar mannerThe time offset between the rain gauge and the radar samplingarea was ignored due to the close distance (about 16 km) andthe low elevation angle (3∘)
Therefore 119877 and 119866 were derived with two different radarsampling areas and with increasing accumulation time
(1)119877(038 km2) versus119866(0071198982)119877(038 km2)was estimated
from the average radar measurement with 9 pixels (038 km2at 16 km range) The gauge rainfall rate 119866(007m2) wasderived from the rainfall accumulation from an individualrain gauge that had a sampling area of 007m2
(2) 119877(6 1198961198982) versus 119866(6 1198961198982
) 119877(6 km2) was estimated fromthe areal averaged radar parameters at a 16 km times 31 km radar
6 Advances in Meteorology
Table 3 Two rain rate retrieval algorithms and boundary conditions [22]
119877 (119885119867) = 120572119885
120573
119867119877 (119885DR) = 120572119885
120573
1198671001120574119885DR
Z gt 35 dBZ Z lt 35 dBZ Z gt 35 dBZ 119885DR gt 03 dB Z lt 35 dBZ 119885DR gt 03 dB120572 353 times 10
minus2383 times 10
minus2103 times 10
minus2145 times 10
minus2
120573 621 times 10minus2
608 times 10minus2
985 times 10minus1
890 times 10minus1
120574 mdash mdash minus6479 minus5295
measurement resolution 119866(6 km2) was the average value of119866(007m2) from the ten gauges within the same area
Estimated 119877 from radar parameters was calculated withtwo rain estimators as shown in Table 3These rain estimatorswere derived by a least squares fit with the theoretical 119885
119867
119885DR and 119877 obtained by scattering simulations with theDSD data collected by POSS from March to Septemberof 2001 in Pusan Korea [21] To improve the estimationaccuracy each estimator was derived for two categoriesThe reflectivity threshold of 35 dBZ was used as a stan-dard stratiformconvective rain classification [22] The 119885DRthreshold of 03 dB was used to minimize the effect of the119885DR measurement noiseThePPI data with an elevation angleof 3∘ was used to suppress the beam shielding and groundclutterThe119885
119867and119885DR fields from the radar were smoothed
by the linear average of values at 3 times 3 pixels to reduce theobservational noise
The bias in 119866 from the ten rain gauges was corrected bythe laboratory test and the intercomparison with field dataThe detail is presented in Appendix To evaluate the atten-uation correction algorithm and verify the rainfall retrievalaccuracy the comparisons of119877 and119866were represented by thegeneral statistics such as normalized bias (NB) normalizedstandard deviation (NSD) root-mean square error (rmse)relative rms error (r rmse) and Spearman correlation (corr)as follows
NB119877-119866 (Δ119905) =
(1119899)sum119899minus1
119896=0[119877119896 (Δ119905) minus 119866119896 (Δ119905)]
119866 (Δ119905)
NSD119877-119866 (Δ119905) =
radic(1119899)sum119899minus1
119896=0[119877119896 (Δ119905) minus 119866119896 (Δ119905)]
2
119866 (Δ119905)
rmse119877-119866 (Δ119905) =
radicsum119899minus1
119896=0[119877119896 (Δ119905) minus 119866119896 (Δ119905)]
2
119899
rmse119877-119866 (Δ119905) = radic
sum119899minus1
119896=0[119877119896 (Δ119905) minus 119866119896 (Δ119905)]
2
sum119899minus1
119896=0[119866119896 (Δ119905)]
2
corr119877-119866 (Δ119905)
=
sum119899minus1
119896=0[119877119894119896 (Δ119905) minus 119877119894 (Δ119905)] [119866119895119896 (Δ119905) minus 119866119895 (Δ119905)]
radicsum119899minus1
119896=0[119877119894119896 (Δ119905) minus 119877119894 (Δ119905)]
2radicsum119899minus1
119896=0[119866119895119896 (Δ119905) minus 119866119895 (Δ119905)]
2
(6)
where 119899 is the number of 119877 and 119866 pairs at a given temporalintegrationΔ119905 and the overbar indicates the temporal average
33 Spatial Correlation and Variability of the Rain FieldsRain gauges are widely used to measure rainfall amountsdue tomany advantages including easy installation mobilitylow price and ease of data processing The rain gauge datais used as a reference for the ground rainfall amount forthe evaluation of rain retrieval algorithms adjustment ofother instruments data assimilation and other hydrologicalapplications However it is affected by instrumental errorsand representativeness errors due to the variability of rainfields
In a TB rain gauge the instrumental errors are due toimproper calibration the time delay of the tips leakage dueto its measuring principle and so forthThis error is exploredin the Appendix
The representativeness errors are subdivided into spatialand temporal representativeness They are mainly caused byhaving a small sampling area and a limited sampling time dueto the gauge measuring principle The actual rain fields varywith different spatiotemporal scales that are not equivalentto the spatial and temporal resolution of the rain gaugeTherefore the rain gauge data represent the actual rain fieldsin a very limited manner
These errors should be considered and minimized priorto using rain gauge data as ground truth Furthermore thedegree of these errors has to be understood quantitativelysince there can be other error sources in various applicationsIn this section the spatial and temporal variability of the rainfields are investigated using the rain data collected by the tenrain gauges for the month of July 2010 at Muan Korea
First the Spearman correlation coefficient 119903119894119895(Δ119905)
between a pair of 119866 values from the 119894th and 119895th individualrain gauge with integration time Δ119905 was calculated fromthe following equation to obtain the Spearman correlationcoefficient of the rain fields
119903119894119895 (Δ119905)
=
sum119899minus1
119896=0[119866119894119896 (Δ119905) minus 119866119894 (Δ119905)] [119866119895119896 (Δ119905) minus 119866119895 (Δ119905)]
radicsum119899minus1
119896=0[119866119894119896 (Δ119905) minus 119866119894 (Δ119905)]
2radicsum119899minus1
119896=0[119866119895119896 (Δ119905) minus 119866119895 (Δ119905)]
2
(7)
where 119899 is the number of gauge data pairs at given Δ119905 and theoverbar indicates the temporal average of the rainfall rate Itwas derived withΔ119905 to research the spatial and temporal scaleof the variation for the rain field
In addition the effect of the natural variability of the rainfields within two different areas (038 km2 and 6 km2) wasexplored by analyzing the bias and random error between thespatially averaged rainfall rate ⟨119866⟩ and 119866 from the individual
Advances in Meteorology 7
32km
21 km
11 km
20km40km60km
minus180 18020 30 40 50 60 70 80 90 100120140160100(a)
32 km
21 km
11 km
20km40km60km
minus180 18020 30 40 50 60 70 80 90 100120140160100(b)
Figure 5 PPIs of measured (a) and filtered (b) ΨDP of the NIMS-XPOL at 2101UTC July 10 2010
gauge The NB and NSD of 119866119894with ⟨119866⟩ were calculated as
follows
NB (119894) =(sum119899minus1
119896=0[119866119894119896 (Δ119905) minus ⟨119866119896 (Δ119905)⟩]) 119899
⟨119866 (Δ119905)⟩
NSD (119894) =(1119899)radic(sum
119899minus1
119896=0[119866119894119896 (Δ119905) minus ⟨119866119896 (Δ119905)⟩]
2)
⟨119866 (Δ119905)⟩
(8)
where 119894 indicates the 119894th gauge and the overbar indicates thetemporal average
This analysis provided a reasonable spatial average andtemporal accumulation resolution for the calculation ofmoreaccurate and realistic rainfall amounts from the gauge dataand a quantitative guideline regarding the errors in rainestimation due to the variability of the rain fields
4 Results
41 Attenuation Correction of Reflectivity and DifferentialReflectivity The attenuation correction algorithm in Sec-tion 31 was applied to the NIMS-XPOL radar data ΦDPwas estimated by eliminating 120575 and the observational noisefrom measured ΨDP using an iterative FIR filter The offsetof ΦDP was then adjusted Figure 5 shows the PPIs ofmeasured ΨDP and filtered ΦDP After applying the FIR filterthe small scale variation due to 120575 and observational noisein measured ΨDP was reduced and filtered ΨDP (estimatedΦDP) monotonically increased with an increase in rangeThe maximum value of estimated ΦDP reached to 140∘ Inaddition estimated ΦDP at the azimuth angles of 150∘ and200∘ sharply increased with the loss of signal beyond thesesharp gradient areas This reflected that strong convective
cells existed in these directions and that these cells caused thesevere signal attenuation EstimatedΦDP values between 250
∘
and 360∘ azimuth angle decreased after filtering since ΦDPoffsets were positive values in these directions Figure 6 showsthe range profile of measured and filtered ΨDP for estimationof ΦDP at 71
∘ 160∘ and 334∘ azimuth angles The unexpectedlarge values near the radarwere removed by the120588HV threshold(red line)The small fluctuation inmeasuredΨDP due to 120575 andthe observational noise was effectively removed after filteringThe value of estimated ΦDP was adjusted to 0 by the offset(blue line)
Estimated ΦDP was used to calculate the path integrated119860119867 and the measured119885
119867values were corrected as shown in
Figure 7 Figure 7(a) represents measured-attenuated119885119867and
relatively small values of 119885119867at far ranges This may indicate
that the 119885119867values at far ranges were affected by the strong
attenuation due to strong convective cells near the radarThisspatial pattern of 119885
119867was the typical feature of the attenuated
radar signal due to the rain After correcting the attenuationthe value of the 119885
119867field became larger over the entire area
and the radially attenuated 119885119867
was corrected by showinga typical rain cell pattern in 119885
119867(Figure 7(b)) The largest
difference between the uncorrected and the corrected 119885119867at
2101UTCwas over 50 dB However the white area behind thestrong echoes was not detected as there was a complete loss ofthe signal powerThese regions could not be corrected by thisalgorithmThe circle pattern of119885
119867fields within about a 4 km
range from theNIMS-XPOL could not present the actual rainechoes due to the instrumental issues (the same was true for119885DR) These results are also presented in the range profiles inFigure 8
Similarly measured 119885DR were corrected with the pathintegrated 119860DP derived from the path integrated 119860
119867using
(4) (see Figure 9) Before applying attenuation correction
8 Advances in Meteorology
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
0
50
100
150
200
ΦDP (∘) azimuth angle = 71∘
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
0
50
100
150
200
ΦDP (∘) azimuth angle = 160∘
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
0
50
100
150
200
ΦDP (∘) azimuth angle = 334∘
(c)
Figure 6 Range profiles of measured (black) 120588HV threshold isolated point removing (red) and FIR filter and offset adjusted (blue) ΨDP forestimation of ΦDP at 71
∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
32km
21 km
11 km
20km40km60km
minus10 10 15 20 25 30 35 40 45 50 55 60 65 700(a)
32 km
21 km
11 km
20km40km60km
minus10 10 15 20 25 30 35 40 45 50 55 60 65 700(b)
Figure 7 PPIs of measured (a) and attenuation corrected (b) 119885119867of the NIMS-XPOL for the case in Figure 5
Advances in Meteorology 9
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 71∘
minus20
0
20
40
60
80
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 160∘
minus20
0
20
40
60
80
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 334∘
minus20
0
20
40
60
80
(c)
Figure 8 Range profiles of measured (line) calibration bias 120588HV threshold and isolated point removing (dot) and attenuation corrected(dash) 119885
119867at 71∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
the attenuation signature in119885DR was determined Ideally the119885DR value for rain should be positive since rain drops areoblate due to aerodynamic force However themeasured119885DRfield in Figure 9 showed a negative value over wide areasAfter correcting the attenuation the119885DR field was larger than0 dB and showed a realistic rain patternThis is also presentedin the range profile of Figure 10The largest difference of119885DRbetween before and after attenuation correction at 2101UTCwas about 8 dB The 119885DR values around the 220∘ azimuthangle beyond 40 km were the largest They did not matchwith the area of the largest 119885DR values (around the 210∘azimuth angle with a range of 20 km and 40 km) This wasan unexpected result because 119885DR is proportional to 119885
119867in
theory This mismatch between 119885119867
and 119885DR at far rangemay be due to contamination due to beam broadening andaccumulated errors due to uncertainty in the attenuationcorrection with different ranges
Figures 11 and 12 show the time series of 119885119867
and119885DR at the ten gauge sites before and after applying theattenuation correction The differences between uncorrectedand corrected values existed after 1800UTC and were at amaximumaround 2100UTCThis attenuation correctionwas
evaluated through comparison with the theoretical values of119885119867and 119885DR In Figure 13 the uncorrected 119885
119867-119885DR scatter
plots are apart from the theoretical119885119867-119885DR values [23] After
attenuation correction the119885119867-119885DR scatter plots were close to
the calculated valuesThis indicated that119885119867and119885DR over the
gauge sites were corrected with reliable values
42 Variability of the Rain Fields The Spearman correlationcoefficient of the rainfall rate from the 10 spatially dis-tributed rain gauges was calculated with accumulation timeand separation distance (Figure 14) The overall Spearmancorrelation coefficient decreased with increasing separationdistance at all accumulation timesThis tendency was similarto the result from Habib and Krajewski [24] However thecorrelation value was higher in this study Note that the gaugetip resolution in Habib and Krajewski [24] was 0254mmwhile it was 02mm in this study In addition the rainfall ratesfrom the gauge data used in this study mainly consisted ofmoderate rainfall rates while Habib and Krajewski [24] useddata containing heavy rainfall rates
For accumulation times of less than 10min the corre-lation at the same distance was dramatically increased with
10 Advances in Meteorology
32km
21 km
11 km
20km40km60km
18
14
00
02
04
06
26
40
minus10
22
minus20
30
10
50
minus30
(a)
32 km
21 km
11 km
20km40km60km
18
14
00
02
04
06
26
40
minus10
22
minus20
30
10
50
minus30
(b)
Figure 9 PPIs of measured (a) and attenuation corrected (b) 119885DR of NIMS-XPOL for the case in Figure 5
increasing accumulation time and the slope of the correlationfunction was steep This was due to the reduction of theinstrumental errors of the gauge data by time accumulationThis indicated that rainfall accumulated for less than 10minfrom single gauge data cannot represent the rainfall over itssurrounding area due to instrumental errors For accumula-tion times longer than 10min the slope became less steepThis reflected the remaining variability of the rain field Inaddition the accumulated rainfall rate of about 10min wassuitable for minimizing the instrumental uncertainty whilepreserving the variation in the rain field The correlationreached about 098 at 05 km in 10min of accumulationand this approached 098 at 10 km at 60min accumulationTherefore a single rain gauge mostly represented the rainfallrate within 05 km and 10 km with 10min and 60minaccumulation respectively
The NB between an areal averaged rainfall rate ⟨119866⟩ anda single gauge rainfall rate 119866 is shown with accumulationtime in Figures 15(a) and 16(a) Where the average area was038 km2 (corresponding to 9 pixels of NIMS-XPOL radarat 16 km range) the NB reached a 6ndash8 maximum Wherethe averaged area is 6 km2 (corresponding to the coverage forall of the gauges) the NB reached about a 10 maximumFurthermore the NB hardly changed with accumulationtime Note that this statistic was derived after removinginstrumental bias This indicated that the rainfall rate froma single gauge was biased due to the spatial variability of therain fields and this bias was maintained with accumulationtime
The NSD of a single gauge for a mean area of 038 km2was over 40 and it reached about 70ndash90 for a mean areaof 6 km2 at an accumulation time of 1min (see Figures 15(b)and 16(b)) This statistic included instrumental errors (about
40ndash60 at 16mmhminus1 and 60ndash80 at 10mmhminus1 as shownin the Appendix) as well as the spatial variability of the rainfields This indicated that the variance between the point andarea values from the gauge data at short accumulation timeswas similar to the errors due to the instrumental uncertaintyof the rain gaugeTheNSD in 60min accumulation decreasedto over 10 and about 15ndash45 with a mean area of 038 km2and 6 km2 respectively including the NSD of 6ndash10 due toinstrumental uncertainty This indicated that the rainfall ratefrom single gauges may differ significantly from the actualrainfall rate due to the variability of the rain fields and theinstrumental uncertainty in short accumulation times
43 Verification of Rainfall Retrieval The rainfall rate fromthe NIMS-XPOL radar measurements was retrieved withtwo rainfall rate estimators 119877(119885
119867) and 119877(119885
119867 119885DR) The
estimated rainfall rate was verified through comparison withthe rainfall rate (119866) from the rain gauge data These twoestimators were applied before and after the attenuation cor-rection First 119877(038 km2) was compared with 119877(007 km2)at 10min accumulation (Figure 17) Using 119877(119885
119867) the NB
NSD rmse and r rmse without the attenuation correctionwere significantly large at 70 127 127 (mmhr) and78 respectively However they were reduced to minus152 521 (mmhr) and 32 respectively after applying theattenuation correction Using 119877(119885
119867 119885DR) the NB was just
3 NSD was 49 rmse was 496 (mmhr) and r rmsewas 31 before applying attenuation correction This wassimilar to the result from 119877(119885
119867) after attenuation correction
However there was no significant change in119877(119885119867 119885DR)with
attenuation correction This means that 119877(119885119867 119885DR) was less
sensitive to rain attenuation than was 119877(119885119867) because 119885
119867is
proportional to 119877 and 119885DR is inversely proportional to 119877
Advances in Meteorology 11
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 71∘
minus10
minus5
0
5
10
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 160∘
minus10
minus5
0
5
10
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 334∘
minus10
minus5
0
5
10
(c)
Figure 10 Range profiles of measured (line) calibration bias 120588HV threshold and isolated point removing (dot) and attenuation corrected(dash) 119885DR at 71∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
in this rain estimator (see Table 3) Furthermore comparingthe analysis of variance results between areal (038 km2) andpoint (007m2) values from the gauge data (Section 42)half of the errors after attenuation correction were producedby the small scale variability of the rain fields and theinstrumental uncertainty of the rain gauge
A similar comparison is shown in Figure 18 for a 60minaccumulation Using 119877(119885
119867) the NB NSD rmse and r rmse
were decreased from minus65 86 71 (mmhr) and 70to minus5 34 28 (mmhr) and 28 by the attenuationcorrection respectively Using 119877(119885
119867 119885DR) the NB NSD
rmse and r rmse were only slightly decreased after theattenuation correction However the NSD rmse and r rmsewere reduced when compared at 10min accumulation In thesame accumulation time the errors in the gauge data reachedto over 15
This comparison was affected by different sampling areasof the gauges and the radar Therefore the areal averagevalues were compared (Figure 19) Using119877(119885
119867)with the areal
averaged radar measurements the NB was reduced fromminus69 to 7 NSD was reduced from 122 to 42 rmsewas reduced from 39 (mmhr) to 13 (mmhr) and r rmse
was reduced from 77 to 26 at an accumulation time of10min Using 119877(119885
119867 119885DR) the NB NSD rmse and r rmse
were decreased fromminus24 67 21 (mmhr) and 42 to 539 12 (mmhr) and 24 respectively
TheNB and r rmsewithmatched sampling areas betweenthe gauge and the radar are shown in Figures 20 and 21 Thereduction of NB due to temporal accumulation was smallin both of the rain estimators The NB was reduced to lessthan 7 by attenuation correction for all accumulation timeswith both119877(119885
119867) and119877(119885
119867 119885DR)The attenuation correction
effect was significantly greater in119877(119885119867) Similarly the r rmse
was reduced significantly in 119877(119885119867) by attenuation correction
(from 68 to 25 at a 60min accumulation) The reductionof the r rmse due to temporal accumulation was larger in119877(119885119867 119885DR) because the significant observational noise of
119885DR was smoothed The r rmse was decreased from 18to 15 in 119877(119885
119867 119885DR) at a 60min accumulation time after
attenuation correction
5 Conclusions
In this study an attenuation correction algorithm based onΦDP was developed to improve the accuracy of the rainfall
12 Advances in Meteorology
1st gauge site 2nd gauge site
3rd gauge site 4th gauge site
5th gauge site 6th gauge site
7th gauge site 8th gauge site
9th gauge site 10th gauge site
17 18 19 20 2116Time (h)
17 18 19 20 2116Time (h)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
UncorrectedCorrected
UncorrectedCorrected
Figure 11 Time series of measured (red circle) and attenuation corrected (blue circle) 119885119867at the 10 rain gauge sites
Advances in Meteorology 13
UncorrectedCorrected
UncorrectedCorrected
17 18 19 20 2116Time (h)
17 18 19 20 2116Time (h)
1st gauge site 2nd gauge site
3rd gauge site 4th gauge site
5th gauge site 6th gauge site
7th gauge site 8th gauge site
9th gauge site 10th gauge site
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
Figure 12 Time series of measured (red circle) and attenuation corrected (blue circle) 119885DR at the 10 rain gauge sites
14 Advances in Meteorology
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Uncorrected(a)
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Corrected Simulated from DSD(b)
Figure 13 Scatter plots of 119885119867and 119885DR before (a) and after (b) attenuation correction at the 10 rain gauge sites DSD observation results (dot
[21])
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
04
05
06
07
08
09
10
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
Accum time = 1min
Accum time = 10min
Accum time = 2min
Accum time = 20min
Accum time = 5min
Accum time = 60min
Figure 14 The Spearman correlation coefficient 119903 between data from two rain gauges at different accumulation times
estimation using X-band radar This algorithm was evalu-ated with radar measurements from NIMS-XPOL Howevermeasured ΨDP contained 120575 and observation noise as wellas ΦDP Furthermore the ΨDP offset varied with azimuthangles and decreased near the radar when rain cells were
above the radar In this study ΦDP was estimated by aniterative filtering technique and by adjusting with an offsetdetermined as the minimum value ofΦDP within 3 km of theradar As a result estimated ΦDP monotonically increasedThe path integrated value of119860
119867and119860DP was calculated from
Advances in Meteorology 15
2nd 6th 9th gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(038km2) versus G
(a) Normalized bias
⟨G⟩(038km2) versus G
2nd 6th 9th gaugesMean NSD
Mean R
0
20
40
60
80
100
NSD
()
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 15 (a) The NB and (b) NSD between ⟨119866⟩(038 km2) and 119866 from three rain gauges with accumulation time ⟨119866⟩ is the temporallyaveraged rainfall rate of ⟨119866⟩
All gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(6km2) versus G
(a) Normalized bias
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
0
20
40
60
80
100
NSD
()
All gaugesMean NSD
Mean R
⟨G⟩(6km2) versus G
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 16 (a) The NB and (b) NSD between ⟨119866⟩(6 km2) and 119866 from 10 rain gauges with accumulation time ⟨119866⟩ is the temporally averaged⟨119866⟩ with accumulation time
estimated ΦDP using 119860119867-ΦDP and 119860DP-ΦDP relationships
derived from the DSDs data Measured 119885119867
and 119885DR werecorrected with the calculated path integrated 119860
119867and 119860DP
by each gate The attenuated signal patterns of 119885119867and 119885DR
were restored to a typical rain echo pattern However theattenuation corrected 119885
119867and 119885DR fields were not matched
at distant ranges This was likely because of contaminationdue to beam broadening and the accumulation of errorswith increasing range Comparing the 119885
119867-119885DR scatter plots
over the 10 rain gauge sites with the results obtained froma scattering simulation with the DSD data the attenuation
corrected 119885119867-119885DR scatter plots are in good agreement with
the 119885119867-119885DR scatter plots from the DSDs
The validation for the improvement of accuracy in therainfall estimation through attenuation correction was per-formed by comparing with dense rain gauge network data119877(119885119867) from the attenuated (observed)119885
119867produced a severe
underestimation with an NB NSD rmse and r rmse of70 127 127 (mmhr) and 78 at 10min accumulationrespectively while 119877(119885
119867) from the attenuation corrected 119885
119867
showed better agreement with the gauge measurements withan NB NSD rmse and r rmse of minus1 52 521 (mmhr)
16 Advances in Meteorology
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus070
NSD = 127
rmse = 1279
corr = 0858
Corrected
NB = minus001
NSD = 052
rmse = 521
corr = 0927
r_rmse = 0786 r_rmse = 0320
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = 003
NSD = 049
rmse = 496
r_rmse = 0305
corr = 0936
Corrected
NB = 001
NSD = 050
rmse = 502
r_rmse = 0308
corr = 0936
(b) 119877(119885119867119885DR)
Figure 17 The scatter plots of 119877(038 km2 10min) versus 119866(007m2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
CorrectedUncorrected
R(m
mh
)
0
10
20
30
40
10 20 30 400G (mmh)
Uncorrected
NB = minus065
NSD = 086
rmse = 706
r_rmse = 0702
corr = 0878
Corrected
NB = minus005
NSD = 034
rmse = 281
r_rmse = 0279
corr = 0866
(a) 119877(119885119867)
CorrectedUncorrected
R(m
mh
)
10 20 30 400G (mmh)
0
10
20
30
40
Uncorrected
NB = 008
NSD = 023
rmse = 189
r_rmse = 0188
corr = 0942
Corrected
NB = 007
NSD = 025
rmse = 205
r_rmse = 0203
corr = 0940
(b) 119877(119885119867119885DR)
Figure 18 The scatter plots of 119877(038 km2 60min) versus 119866(007m2 60min) (a) using 119885119867and (b) using 119885
119867and 119885DR
and 32 at 10min accumulation respectively At 60minaccumulation theNBNSD rmse and r rmsewere decreasedfrom minus65 86 71 (mmhr) and 70 to minus5 3428 (mmhr) and 28 respectively Using 119877(119885
119867 119885DR) the
NB NSD rmse and r rmse were only slightly reduced even if
attenuation was corrected but were smaller than when using119877(119885119867) This indicated that 119877(119885
119867 119885DR) was less sensitive to
the attenuation than 119877(119885119867) In addition half of the errors
that remained after attenuation correction were caused by thesmall scale variability of the rain fields and the instrumental
Advances in Meteorology 17
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus069
NSD = 122
rmse = 391
r_rmse = 0770
corr = 0893
Corrected
NB = 007
NSD = 042
rmse = 133
r_rmse = 0262
corr = 0960
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = minus024
NSD = 067
rmse = 215
r_rmse = 0424
corr = 0718
Corrected
NB = 005
NSD = 039
rmse = 124
r_rmse = 0245
corr = 0891
(b) 119877(119885119867119885DR)
Figure 19 The scatter plots of 119877(6 km2 10min) versus 119866(6 km2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
R(ZH)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(a) 119877(119885119867)
R(ZH ZDR)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(b) 119877(119885119867119885DR)
Figure 20 The normalized bias of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and 119885DR The red
point-line and the blue point-line indicate without and with the attenuation correction respectively
uncertainty of the rain gauges Using an areal averagedvalue the NB was reduced to less than 7 with attenua-tion correction using either rainfall estimator regardless ofaccumulation timeThe r rmse were significantly reduced byattenuation correction in 119877(119885
119867) and the reduction of r rmse
with 119877(119885119867 119885DR) was even more dramatic
As summarized above the rain attenuation for 119885119867
and 119885DR in the X-band was corrected by ΦDP and theresults showed a good agreement with the value obtainedby scattering simulations with the DSD datasets The resultsof comparison with rain gauge data also showed a signifi-cant improvement in the accuracy of rainfall estimation by
18 Advances in Meteorology
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10r_
rmse
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH)
(a) 119877(119885119867)
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10
r_rm
se
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH ZDR)
(b) 119877(119885119867119885DR)
Figure 21 The relative root-mean square error of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and
119885DR The red point-line and the blue point-line indicate without and with the attenuation correction respectively
attenuation correction which supports the quantitative useof the X-band radar in rain data retrieval
Appendix
A Instrumental Uncertainty of the Rain Gauge
The instrumental uncertainty of the TB rain gauge is asso-ciated with improper calibration leakage and precipitationmissing the gauge due to the measurement principle andwetting and evaporation losses In addition external sourcessuch as wind turbulence and installed position also con-tribute to the instrumental uncertainty Ciach [17] reportedthat the TB gauge data suffers a significant uncertainty asso-ciated with random differences between closely collocatedTB gauges through the analysis of the error as a functionof the rainfall intensity for three timescales In addition theimproper calibration induces a bias in the rainfall measure-ment and this bias cannot be removed by increasing accu-mulation timeTherefore the proper correction or processingof TB gauge data is necessary to obtain accurate and reliablereading of the rainfall amount
In this appendix instrumental biases were calculatedto reduce the instrumental uncertainty through an idealexperiment in the laboratory and a field intercomparisonwith 10 TB rain gauges In addition the random errors werequantitatively calculated as a function of rainfall rate andaccumulation time
A1 Correction of Instrumental Bias The instrumental bias119861119868(119894) of the 119894th gauge was defined as a combination
of the absolute bias 119861119860(ref) of the reference gauge and
the relative bias 119861119877(119894) of the 119894th gauge according to the
following equation
119861119868 (119894) = 119861119860 (ref) times 119861119877 (119894) (A1)
119861119860(ref) signified the ratio of the calculated value from
the instrumental resolution of the reference gauge to the truevalue measured by other instruments
119861119860 (ref) =
rainfalltruthrainfallgauge (ref)
(A2)
In this study 119861119860(ref) was derived by an ideal experiment
using an electronic scale First the reference gauge was puton the scale and the scale was initialized The water dropsflowed out into the receiving orifice of the reference gaugeuntil the tipping number became 60Themeasured weight of60 tips was 2144 g while the theoretical (or expected) valuewas 2244 g for the reference gauge (receiving orifice size of1545 cm 60 tips and 02mm tip resolution) Therefore theabsolute bias of the reference gauge from (A2) was about0955 This indicates that the rainfall amount measured bythis reference gauge was overestimated by about 45 andthis bias should be corrected prior to any interpretation
119861119877(119894) represented the ratio of the observed (recorded)
value from the 119894th gauge to the observed (recorded) valuefrom the reference gauge
119861119877 (119894) =
rainfallgauge (ref)rainfallgauge (119894)
(A3)
119861119877(119894) was calculated from ten collocated rain gauges
including the reference gauge located at an observation field
Advances in Meteorology 19
Table 4 The instrumental and relative biases of the 10 tipping bucket rain gauges
Gauge ID RG1 RG2 RG3 RG4 RG5 RG6 RG7 RG8 RG9 RG10Instrumental bias 094 108 099 100 096 104 098 102 108 098Relative bias 098 113 103 105 100 109 102 107 113 102
013610152230406080120150200250300400
()
Rain
rate
(mm
h)
100101
Accumulation time (min)
10
100NSD ()
Figure 22 Two-dimensional normalized standard deviationbetween the standard rainfall and each rain gauge
at KNUKoreaThe observational period for intercomparisonwas from October 2009 to June 2010 (14 rainy days) andthe total accumulated rainfall for this period was about300mm The results are presented in Table 4 All relativebiases were mainly larger than 10 This indicated that themost of gauges underestimated the rainfall amount comparedto the reference rain gauge The relative bias of the referencegauge (119894 = 5) was 10
Finally 119861119860(119894) was calculated from (A1) (see Table 4) All
instrumental biases were between 090 and 11 and the meanvalue of the biases was nearly 10 This indicated that therainfall rate from a single gauge data had a maximum 10bias These biases were corrected for each gauge to removeexisting instrumental errors
A2 Quantification of Random Errors To quantify the ran-dom error due to instrumental uncertainty the NSD of theten gauges was calculated with accumulation time and rainrate using data from an intercomparison The NSD of the 119894thgauge compared to the reference gauge was derived for 20rain rate classes (Δ119877) and 17 accumulation time classes (Δ119905)as follows
NSD119894 (Δ119905 Δ119877)
=
(1119899)radicsum119899minus1
119896=0[1198661015840
119870(Δ119905 Δ119877) minus 119866119894119896 (Δ119905 Δ119877)]
2
1198661015840
(A4)
where 1198661015840 and 119866 indicate the reference and individual gaugesrespectively The overbar means the temporal average ofthe rainfall rate The NSDs of nine different gauges (exceptthe reference gauge) were averaged to obtain the two-dimensional average NSD (Figure 22) The average NSD
tended to decrease with increasing accumulation time andwith increasing rain rate The NSD was about 10ndash15when accumulation time was 10min and the rain rate was10mmhminus1 The instrumental errors were quite significant forlonger accumulation times and higher rainfall rates Theseerrors should be used as a guideline when a single gauge isused as a reference for any validation study
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper has been revised and extended fromMs Young-AOhrsquos MS thesis [25] of Kyungpook National University Thisresearch is supported by Development and Application ofCross Governmental Dual-Pol Radar Harmonization (WRC-2013-A-1) project of theWeather Radar Center KoreaMeteo-rological AdministrationThis work was funded by the KoreaMeteorological Administration Research and DevelopmentProgram under Grant KMIPA2015-1010
References
[1] V N Bringi V Chandrasekar N Balakrishnan and D SZrnic ldquoAn examination of propagation effects in rainfall onradar measurements at microwave frequenciesrdquo Journal ofAtmospheric and Oceanic Technology vol 7 no 6 pp 829ndash8401990
[2] A Huggel W Schmid and AWaldvogel ldquoRaindrop size distri-butions and the radar bright bandrdquo Journal of AppliedMeteorol-ogy vol 35 no 10 pp 1688ndash1701 1996
[3] G W Lee Errors in rain measurement by radar effect of vari-ability of drop size distributions [PhD thesis]McGill University2003
[4] J D Tuttle and R E Rinehart ldquoAttenuation correction in dual-wavelength analysesrdquo Journal of ClimateampAppliedMeteorologyvol 22 no 11 pp 1914ndash1921 1983
[5] S-G Park V N Bringi V Chandrasekar M Maki andK Iwanami ldquoCorrection of radar reflectivity and differentialreflectivity for rain attenuation at X band Part I theoretical andempirical basisrdquo Journal of Atmospheric and Oceanic Technol-ogy vol 22 no 11 pp 1621ndash1632 2005
[6] G Scarchilli E Gorgucci V Chandrasekar and T A SeligaldquoRainfall estimation using polarimetric techniques at C-bandfrequenciesrdquo Journal of Applied Meteorology vol 32 no 6 pp1150ndash1160 1993
[7] M N Anagnostou E N Anagnostou and J VivekanandanldquoCorrection for rain path specific and differential attenuation ofX-band dual-polarization observationsrdquo IEEE Transactions onGeoscience and Remote Sensing vol 44 no 9 pp 2470ndash24802006
20 Advances in Meteorology
[8] J Kalogiros M N Anagnostou E N Anagnostou M Mon-topoli E Picciotti and F S Marzano ldquoEvaluation of a newpolarimetric algorithm for rain-path attenuation correction ofX-band radar observations against disdrometerrdquo IEEE Transac-tions on Geoscience and Remote Sensing vol 52 no 2 pp 1369ndash1380 2014
[9] W-Y Chang J Vivekanandan and T-C C Wang ldquoEstimationof X-band polarimetric radar attenuation and measurementuncertainty using a variational methodrdquo Journal of AppliedMeteorology and Climatology vol 53 no 4 pp 1099ndash1119 2014
[10] J Hubbert and V N Bringi ldquoAn iterative filtering technique forthe analysis of copolar differential phase and dual-frequencyradar measurementsrdquo Journal of Atmospheric and OceanicTechnology vol 12 no 3 pp 643ndash648 1995
[11] K V Mishra W F Krajewski R Goska et al ldquoDeploy-ment and performance analyses of high-resolution iowa XPOLradar system during the NASA IFloodS campaignrdquo Journal ofHydrometeorology vol 17 no 2 pp 455ndash479 2016
[12] M Thurai G J Huang V N Bringi W L Randeu and MSchonhuber ldquoDrop shapes model comparisons and calcu-lations of polarimetric radar parameters in rainrdquo Journal ofAtmospheric and Oceanic Technology vol 24 no 6 pp 1019ndash1032 2007
[13] M D Humphrey J D Istok J Y Lee J A Hevesi and AL Flint ldquoA new method for automated dynamic calibration oftipping-bucket rain gaugesrdquo Journal of Atmospheric andOceanicTechnology vol 14 no 6 pp 1513ndash1519 1997
[14] E HabibW F Krajewski V Nespor andA Kruger ldquoNumericalsimulation studies of rain gage data correction due to windeffectrdquo Journal of Geophysical ResearchAtmospheres vol 104 no16 Article ID 1999JD900228 pp 19723ndash19733 1999
[15] G J Ciach and W F Krajewski ldquoOn the estimation of radarrainfall error variancerdquoAdvances inWater Resources vol 22 no6 pp 585ndash595 1999
[16] E Habib W F Krajewski and A Kruger ldquoSampling errors oftipping-bucket rain gaugemeasurementsrdquo Journal of HydrologicEngineering vol 6 no 2 pp 159ndash166 2001
[17] G J Ciach ldquoLocal random errors in tipping-bucket rain gaugemeasurementsrdquo Journal of Atmospheric andOceanic Technologyvol 20 no 5 pp 752ndash759 2003
[18] EHabibW F Krajewski andG J Ciach ldquoEstimation of rainfallinterstation correlationrdquo Journal ofHydrometeorology vol 2 no6 pp 621ndash629 2001
[19] NIMS ldquoStudy on estimation of rainfall amount and classifi-cation of hydrometeor species using dual-polarimetric radar(III)rdquo pp 1ndash20 2010
[20] E Gorgucci V Chandrasekar and L Baldini ldquoCorrection ofX-band radar observation for propagation effects based on theself-consistency principlerdquo Journal of Atmospheric and OceanicTechnology vol 23 pp 1668ndash1681 2006
[21] NIMS ldquoStudy on estimating raindrop size distribution forquantitative estimation of radar rainfall amountrdquo pp 1ndash55 2009
[22] NIMS ldquoQuantitative precipitation estimation and hydrometeoridentification using dual-polarization radar-Phase IIrdquo pp 1ndash742009
[23] S-G Park M Maki K Iwanami V N Bringi and V Chan-drasekar ldquoCorrection of radar reflectivity and differential reflec-tivity for rain attenuation at X band Part II Evaluation andapplicationrdquo Journal of Atmospheric and Oceanic Technologyvol 22 no 11 pp 1633ndash1655 2005
[24] E Habib and W F Krajewski ldquoUncertainty analysis of theTRMM ground-validation radar-rainfall products applicationto the TEFLUN-B field campaignrdquo Journal of Applied Meteorol-ogy vol 41 no 5 pp 558ndash572 2002
[25] Y-A Oh Rainfall estimation from X-band dual-polarizationradar measurements effects of attenuation correction and evalu-ation with a dense rain gauge network [MS thesis] KyungpookNational University 2011
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom
Applied ampEnvironmentalSoil Science
Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal of
Geophysics
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
4 Advances in Meteorology
1801 1816 1831
19011846
1931 1946 2001
1916
605550454030 35 65 702015100 25minus10
(dBZ)
Figure 2 PPI images of attenuation corrected 119885119867from 1801UTC July 10 2010 with an interval of 15min
Table 2 Total rainfall amount for the month and per analysis time
Gauge ID RG1 RG2 RG3 RG4 RG5 RG6 RG7 RG8 RG9 RG10Total rainfall amount for the month (mm) 3214 3717 3308 3139 3218 3426 3321 3239 3743 3245Total rainfall amount for the analysis time (mm) 476 560 510 529 552 571 544 544 513 531
smaller scale fluctuations compared to the filtering window(25 km when the gate size is 0125m) due to 120575 and theobservation noise Hubbert and Bringi [10] also found thatrepeating the filtering process 10 times produced good results
The additional problem in the estimation of ΦDP wasthe variable offset of ΦDP ΦDP observation can be noisyin near range Radar in general has ΦDP offset to avoidobserving negative value of ΦDP in near range Ideally theoffset should not change with the azimuthal angles Howeverthe offsets of measured ΨDP of the NIMS-X varied withthe azimuthal angles and ΦDP values decreased near theradar when rain cells existed above the radar Therefore
the offset was determined as the minimum value of ΦDPwithin 3 km of the radar and estimated filtered ΨDP wasadjusted accordingly
The calculation of 119860119867
and 119860DP was based on thefollowing two relationships [1 5 20]
119860119867 (119903) = 120590119870DP (119903)
119860DP (119903) = 120576119860119867 (119903) (1)
where 119860119867(119903) and 119860DP(119903) are the specific attenuation and
the specific differential attenuation at range 119903 from the radar
Advances in Meteorology 5
3510
3506
3502
3498
3494 Unnam
12628 12636 12644
1 234 96
8 75
10
R
Figure 3 Layout of NIMS-XPOL and the 10 rain gauges nearUnnam-myeon Muan-gun Jeollanam-do Korea R shows thelocation of the NIMS-XPOL the numbered points indicate theobservational sites of the 10 rain gauges
respectively119870DP(119903) is the specific differential one-way phaseshift calculated using the following equation
119870DP (119903) =ΦDP (1199032) minus ΦDP (1199031)
2 (1199032minus 1199031)
(2)
where 119903 is between 1199031and 1199032 The coefficients 120590 (= 03445)
and 120576 (= 01705) were derived from 119860119867 119860DP and 119870DP
obtained by scattering simulations with the DSDs data col-lected by the PrecipitationOccurrence Sensor System (POSS)from March to September of 2001 in Pusan Korea [21] Thetotal path integrated attenuation and differential attenuationwere calculated withΦDP by integrating the following
2int
119903
0
119860119867(1199031015840) 1198891199031015840= 120590ΦDP (119903) (3)
2int
119903
0
119860DP (1199031015840) 1198891199031015840= 120576 times 2int
119903
0
119860119867(1199031015840) 1198891199031015840 (4)
Number 2 in the equations indicates that the values are two-way values
Finally the measured reflectivity 119885119867-meas(119903) [dBZ] and
the differential reflectivity 119885DR-meas(119903) [dB] at 119903 were cor-rected with the calculated two-way path integrated attenua-tion and differential attenuation
119885119867-corr (119903) = 119885119867-meas (119903) + 2int
119903
0
119860119867(1199031015840) 1198891199031015840
119885DR-corr (119903) = 119885DR-meas (119903) + 2int119903
0
119860DP (1199031015840) 1198891199031015840
(5)
32 Radar Rain Estimation and Comparison with Rain GaugeData A comparison between estimated 119877 from the radar
Measured dual-polarization parameters(ZH ZDR ΦDP 120588HV)
Quality control
(ii) Speckling echo removal
(i) Calibration for instrumental bias of ZH ZDR
(iii) 120588HV threshold application
(ii) Adjustment of its offset(i) 20th FIR filter application to ΨDP
Estimation of ΦDP
Calculation of AH and ADP
Correction of ZH and ZDR
from DSDs120590 and 120576
Figure 4 The schematic diagram of the attenuation correctionalgorithm
parameters and the rain rate (119866) from the rain gauge data wasperformed to evaluate the attenuation correction algorithmIn addition to the rain attenuation the difference between119877 and 119866 was caused by the errors due to the instrumentaluncertainty and representativeness errors of the rain gaugedata the radar observational noise and the errors in the radarrainfall estimation The bias due to the instrumental uncer-tainty in gauge data was removed by properly calibrating thegauges through laboratory tests and a gauge intercomparisonThe representativeness errors were due to the mismatch ofthe sampling area (volume) between the gauge and the radarThe random error in the gauge data could be minimized bytime and areal integrationThe errors due to themeasurementnoise in the radar data could be reduced in a similar mannerThe time offset between the rain gauge and the radar samplingarea was ignored due to the close distance (about 16 km) andthe low elevation angle (3∘)
Therefore 119877 and 119866 were derived with two different radarsampling areas and with increasing accumulation time
(1)119877(038 km2) versus119866(0071198982)119877(038 km2)was estimated
from the average radar measurement with 9 pixels (038 km2at 16 km range) The gauge rainfall rate 119866(007m2) wasderived from the rainfall accumulation from an individualrain gauge that had a sampling area of 007m2
(2) 119877(6 1198961198982) versus 119866(6 1198961198982
) 119877(6 km2) was estimated fromthe areal averaged radar parameters at a 16 km times 31 km radar
6 Advances in Meteorology
Table 3 Two rain rate retrieval algorithms and boundary conditions [22]
119877 (119885119867) = 120572119885
120573
119867119877 (119885DR) = 120572119885
120573
1198671001120574119885DR
Z gt 35 dBZ Z lt 35 dBZ Z gt 35 dBZ 119885DR gt 03 dB Z lt 35 dBZ 119885DR gt 03 dB120572 353 times 10
minus2383 times 10
minus2103 times 10
minus2145 times 10
minus2
120573 621 times 10minus2
608 times 10minus2
985 times 10minus1
890 times 10minus1
120574 mdash mdash minus6479 minus5295
measurement resolution 119866(6 km2) was the average value of119866(007m2) from the ten gauges within the same area
Estimated 119877 from radar parameters was calculated withtwo rain estimators as shown in Table 3These rain estimatorswere derived by a least squares fit with the theoretical 119885
119867
119885DR and 119877 obtained by scattering simulations with theDSD data collected by POSS from March to Septemberof 2001 in Pusan Korea [21] To improve the estimationaccuracy each estimator was derived for two categoriesThe reflectivity threshold of 35 dBZ was used as a stan-dard stratiformconvective rain classification [22] The 119885DRthreshold of 03 dB was used to minimize the effect of the119885DR measurement noiseThePPI data with an elevation angleof 3∘ was used to suppress the beam shielding and groundclutterThe119885
119867and119885DR fields from the radar were smoothed
by the linear average of values at 3 times 3 pixels to reduce theobservational noise
The bias in 119866 from the ten rain gauges was corrected bythe laboratory test and the intercomparison with field dataThe detail is presented in Appendix To evaluate the atten-uation correction algorithm and verify the rainfall retrievalaccuracy the comparisons of119877 and119866were represented by thegeneral statistics such as normalized bias (NB) normalizedstandard deviation (NSD) root-mean square error (rmse)relative rms error (r rmse) and Spearman correlation (corr)as follows
NB119877-119866 (Δ119905) =
(1119899)sum119899minus1
119896=0[119877119896 (Δ119905) minus 119866119896 (Δ119905)]
119866 (Δ119905)
NSD119877-119866 (Δ119905) =
radic(1119899)sum119899minus1
119896=0[119877119896 (Δ119905) minus 119866119896 (Δ119905)]
2
119866 (Δ119905)
rmse119877-119866 (Δ119905) =
radicsum119899minus1
119896=0[119877119896 (Δ119905) minus 119866119896 (Δ119905)]
2
119899
rmse119877-119866 (Δ119905) = radic
sum119899minus1
119896=0[119877119896 (Δ119905) minus 119866119896 (Δ119905)]
2
sum119899minus1
119896=0[119866119896 (Δ119905)]
2
corr119877-119866 (Δ119905)
=
sum119899minus1
119896=0[119877119894119896 (Δ119905) minus 119877119894 (Δ119905)] [119866119895119896 (Δ119905) minus 119866119895 (Δ119905)]
radicsum119899minus1
119896=0[119877119894119896 (Δ119905) minus 119877119894 (Δ119905)]
2radicsum119899minus1
119896=0[119866119895119896 (Δ119905) minus 119866119895 (Δ119905)]
2
(6)
where 119899 is the number of 119877 and 119866 pairs at a given temporalintegrationΔ119905 and the overbar indicates the temporal average
33 Spatial Correlation and Variability of the Rain FieldsRain gauges are widely used to measure rainfall amountsdue tomany advantages including easy installation mobilitylow price and ease of data processing The rain gauge datais used as a reference for the ground rainfall amount forthe evaluation of rain retrieval algorithms adjustment ofother instruments data assimilation and other hydrologicalapplications However it is affected by instrumental errorsand representativeness errors due to the variability of rainfields
In a TB rain gauge the instrumental errors are due toimproper calibration the time delay of the tips leakage dueto its measuring principle and so forthThis error is exploredin the Appendix
The representativeness errors are subdivided into spatialand temporal representativeness They are mainly caused byhaving a small sampling area and a limited sampling time dueto the gauge measuring principle The actual rain fields varywith different spatiotemporal scales that are not equivalentto the spatial and temporal resolution of the rain gaugeTherefore the rain gauge data represent the actual rain fieldsin a very limited manner
These errors should be considered and minimized priorto using rain gauge data as ground truth Furthermore thedegree of these errors has to be understood quantitativelysince there can be other error sources in various applicationsIn this section the spatial and temporal variability of the rainfields are investigated using the rain data collected by the tenrain gauges for the month of July 2010 at Muan Korea
First the Spearman correlation coefficient 119903119894119895(Δ119905)
between a pair of 119866 values from the 119894th and 119895th individualrain gauge with integration time Δ119905 was calculated fromthe following equation to obtain the Spearman correlationcoefficient of the rain fields
119903119894119895 (Δ119905)
=
sum119899minus1
119896=0[119866119894119896 (Δ119905) minus 119866119894 (Δ119905)] [119866119895119896 (Δ119905) minus 119866119895 (Δ119905)]
radicsum119899minus1
119896=0[119866119894119896 (Δ119905) minus 119866119894 (Δ119905)]
2radicsum119899minus1
119896=0[119866119895119896 (Δ119905) minus 119866119895 (Δ119905)]
2
(7)
where 119899 is the number of gauge data pairs at given Δ119905 and theoverbar indicates the temporal average of the rainfall rate Itwas derived withΔ119905 to research the spatial and temporal scaleof the variation for the rain field
In addition the effect of the natural variability of the rainfields within two different areas (038 km2 and 6 km2) wasexplored by analyzing the bias and random error between thespatially averaged rainfall rate ⟨119866⟩ and 119866 from the individual
Advances in Meteorology 7
32km
21 km
11 km
20km40km60km
minus180 18020 30 40 50 60 70 80 90 100120140160100(a)
32 km
21 km
11 km
20km40km60km
minus180 18020 30 40 50 60 70 80 90 100120140160100(b)
Figure 5 PPIs of measured (a) and filtered (b) ΨDP of the NIMS-XPOL at 2101UTC July 10 2010
gauge The NB and NSD of 119866119894with ⟨119866⟩ were calculated as
follows
NB (119894) =(sum119899minus1
119896=0[119866119894119896 (Δ119905) minus ⟨119866119896 (Δ119905)⟩]) 119899
⟨119866 (Δ119905)⟩
NSD (119894) =(1119899)radic(sum
119899minus1
119896=0[119866119894119896 (Δ119905) minus ⟨119866119896 (Δ119905)⟩]
2)
⟨119866 (Δ119905)⟩
(8)
where 119894 indicates the 119894th gauge and the overbar indicates thetemporal average
This analysis provided a reasonable spatial average andtemporal accumulation resolution for the calculation ofmoreaccurate and realistic rainfall amounts from the gauge dataand a quantitative guideline regarding the errors in rainestimation due to the variability of the rain fields
4 Results
41 Attenuation Correction of Reflectivity and DifferentialReflectivity The attenuation correction algorithm in Sec-tion 31 was applied to the NIMS-XPOL radar data ΦDPwas estimated by eliminating 120575 and the observational noisefrom measured ΨDP using an iterative FIR filter The offsetof ΦDP was then adjusted Figure 5 shows the PPIs ofmeasured ΨDP and filtered ΦDP After applying the FIR filterthe small scale variation due to 120575 and observational noisein measured ΨDP was reduced and filtered ΨDP (estimatedΦDP) monotonically increased with an increase in rangeThe maximum value of estimated ΦDP reached to 140∘ Inaddition estimated ΦDP at the azimuth angles of 150∘ and200∘ sharply increased with the loss of signal beyond thesesharp gradient areas This reflected that strong convective
cells existed in these directions and that these cells caused thesevere signal attenuation EstimatedΦDP values between 250
∘
and 360∘ azimuth angle decreased after filtering since ΦDPoffsets were positive values in these directions Figure 6 showsthe range profile of measured and filtered ΨDP for estimationof ΦDP at 71
∘ 160∘ and 334∘ azimuth angles The unexpectedlarge values near the radarwere removed by the120588HV threshold(red line)The small fluctuation inmeasuredΨDP due to 120575 andthe observational noise was effectively removed after filteringThe value of estimated ΦDP was adjusted to 0 by the offset(blue line)
Estimated ΦDP was used to calculate the path integrated119860119867 and the measured119885
119867values were corrected as shown in
Figure 7 Figure 7(a) represents measured-attenuated119885119867and
relatively small values of 119885119867at far ranges This may indicate
that the 119885119867values at far ranges were affected by the strong
attenuation due to strong convective cells near the radarThisspatial pattern of 119885
119867was the typical feature of the attenuated
radar signal due to the rain After correcting the attenuationthe value of the 119885
119867field became larger over the entire area
and the radially attenuated 119885119867
was corrected by showinga typical rain cell pattern in 119885
119867(Figure 7(b)) The largest
difference between the uncorrected and the corrected 119885119867at
2101UTCwas over 50 dB However the white area behind thestrong echoes was not detected as there was a complete loss ofthe signal powerThese regions could not be corrected by thisalgorithmThe circle pattern of119885
119867fields within about a 4 km
range from theNIMS-XPOL could not present the actual rainechoes due to the instrumental issues (the same was true for119885DR) These results are also presented in the range profiles inFigure 8
Similarly measured 119885DR were corrected with the pathintegrated 119860DP derived from the path integrated 119860
119867using
(4) (see Figure 9) Before applying attenuation correction
8 Advances in Meteorology
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
0
50
100
150
200
ΦDP (∘) azimuth angle = 71∘
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
0
50
100
150
200
ΦDP (∘) azimuth angle = 160∘
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
0
50
100
150
200
ΦDP (∘) azimuth angle = 334∘
(c)
Figure 6 Range profiles of measured (black) 120588HV threshold isolated point removing (red) and FIR filter and offset adjusted (blue) ΨDP forestimation of ΦDP at 71
∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
32km
21 km
11 km
20km40km60km
minus10 10 15 20 25 30 35 40 45 50 55 60 65 700(a)
32 km
21 km
11 km
20km40km60km
minus10 10 15 20 25 30 35 40 45 50 55 60 65 700(b)
Figure 7 PPIs of measured (a) and attenuation corrected (b) 119885119867of the NIMS-XPOL for the case in Figure 5
Advances in Meteorology 9
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 71∘
minus20
0
20
40
60
80
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 160∘
minus20
0
20
40
60
80
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 334∘
minus20
0
20
40
60
80
(c)
Figure 8 Range profiles of measured (line) calibration bias 120588HV threshold and isolated point removing (dot) and attenuation corrected(dash) 119885
119867at 71∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
the attenuation signature in119885DR was determined Ideally the119885DR value for rain should be positive since rain drops areoblate due to aerodynamic force However themeasured119885DRfield in Figure 9 showed a negative value over wide areasAfter correcting the attenuation the119885DR field was larger than0 dB and showed a realistic rain patternThis is also presentedin the range profile of Figure 10The largest difference of119885DRbetween before and after attenuation correction at 2101UTCwas about 8 dB The 119885DR values around the 220∘ azimuthangle beyond 40 km were the largest They did not matchwith the area of the largest 119885DR values (around the 210∘azimuth angle with a range of 20 km and 40 km) This wasan unexpected result because 119885DR is proportional to 119885
119867in
theory This mismatch between 119885119867
and 119885DR at far rangemay be due to contamination due to beam broadening andaccumulated errors due to uncertainty in the attenuationcorrection with different ranges
Figures 11 and 12 show the time series of 119885119867
and119885DR at the ten gauge sites before and after applying theattenuation correction The differences between uncorrectedand corrected values existed after 1800UTC and were at amaximumaround 2100UTCThis attenuation correctionwas
evaluated through comparison with the theoretical values of119885119867and 119885DR In Figure 13 the uncorrected 119885
119867-119885DR scatter
plots are apart from the theoretical119885119867-119885DR values [23] After
attenuation correction the119885119867-119885DR scatter plots were close to
the calculated valuesThis indicated that119885119867and119885DR over the
gauge sites were corrected with reliable values
42 Variability of the Rain Fields The Spearman correlationcoefficient of the rainfall rate from the 10 spatially dis-tributed rain gauges was calculated with accumulation timeand separation distance (Figure 14) The overall Spearmancorrelation coefficient decreased with increasing separationdistance at all accumulation timesThis tendency was similarto the result from Habib and Krajewski [24] However thecorrelation value was higher in this study Note that the gaugetip resolution in Habib and Krajewski [24] was 0254mmwhile it was 02mm in this study In addition the rainfall ratesfrom the gauge data used in this study mainly consisted ofmoderate rainfall rates while Habib and Krajewski [24] useddata containing heavy rainfall rates
For accumulation times of less than 10min the corre-lation at the same distance was dramatically increased with
10 Advances in Meteorology
32km
21 km
11 km
20km40km60km
18
14
00
02
04
06
26
40
minus10
22
minus20
30
10
50
minus30
(a)
32 km
21 km
11 km
20km40km60km
18
14
00
02
04
06
26
40
minus10
22
minus20
30
10
50
minus30
(b)
Figure 9 PPIs of measured (a) and attenuation corrected (b) 119885DR of NIMS-XPOL for the case in Figure 5
increasing accumulation time and the slope of the correlationfunction was steep This was due to the reduction of theinstrumental errors of the gauge data by time accumulationThis indicated that rainfall accumulated for less than 10minfrom single gauge data cannot represent the rainfall over itssurrounding area due to instrumental errors For accumula-tion times longer than 10min the slope became less steepThis reflected the remaining variability of the rain field Inaddition the accumulated rainfall rate of about 10min wassuitable for minimizing the instrumental uncertainty whilepreserving the variation in the rain field The correlationreached about 098 at 05 km in 10min of accumulationand this approached 098 at 10 km at 60min accumulationTherefore a single rain gauge mostly represented the rainfallrate within 05 km and 10 km with 10min and 60minaccumulation respectively
The NB between an areal averaged rainfall rate ⟨119866⟩ anda single gauge rainfall rate 119866 is shown with accumulationtime in Figures 15(a) and 16(a) Where the average area was038 km2 (corresponding to 9 pixels of NIMS-XPOL radarat 16 km range) the NB reached a 6ndash8 maximum Wherethe averaged area is 6 km2 (corresponding to the coverage forall of the gauges) the NB reached about a 10 maximumFurthermore the NB hardly changed with accumulationtime Note that this statistic was derived after removinginstrumental bias This indicated that the rainfall rate froma single gauge was biased due to the spatial variability of therain fields and this bias was maintained with accumulationtime
The NSD of a single gauge for a mean area of 038 km2was over 40 and it reached about 70ndash90 for a mean areaof 6 km2 at an accumulation time of 1min (see Figures 15(b)and 16(b)) This statistic included instrumental errors (about
40ndash60 at 16mmhminus1 and 60ndash80 at 10mmhminus1 as shownin the Appendix) as well as the spatial variability of the rainfields This indicated that the variance between the point andarea values from the gauge data at short accumulation timeswas similar to the errors due to the instrumental uncertaintyof the rain gaugeTheNSD in 60min accumulation decreasedto over 10 and about 15ndash45 with a mean area of 038 km2and 6 km2 respectively including the NSD of 6ndash10 due toinstrumental uncertainty This indicated that the rainfall ratefrom single gauges may differ significantly from the actualrainfall rate due to the variability of the rain fields and theinstrumental uncertainty in short accumulation times
43 Verification of Rainfall Retrieval The rainfall rate fromthe NIMS-XPOL radar measurements was retrieved withtwo rainfall rate estimators 119877(119885
119867) and 119877(119885
119867 119885DR) The
estimated rainfall rate was verified through comparison withthe rainfall rate (119866) from the rain gauge data These twoestimators were applied before and after the attenuation cor-rection First 119877(038 km2) was compared with 119877(007 km2)at 10min accumulation (Figure 17) Using 119877(119885
119867) the NB
NSD rmse and r rmse without the attenuation correctionwere significantly large at 70 127 127 (mmhr) and78 respectively However they were reduced to minus152 521 (mmhr) and 32 respectively after applying theattenuation correction Using 119877(119885
119867 119885DR) the NB was just
3 NSD was 49 rmse was 496 (mmhr) and r rmsewas 31 before applying attenuation correction This wassimilar to the result from 119877(119885
119867) after attenuation correction
However there was no significant change in119877(119885119867 119885DR)with
attenuation correction This means that 119877(119885119867 119885DR) was less
sensitive to rain attenuation than was 119877(119885119867) because 119885
119867is
proportional to 119877 and 119885DR is inversely proportional to 119877
Advances in Meteorology 11
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 71∘
minus10
minus5
0
5
10
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 160∘
minus10
minus5
0
5
10
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 334∘
minus10
minus5
0
5
10
(c)
Figure 10 Range profiles of measured (line) calibration bias 120588HV threshold and isolated point removing (dot) and attenuation corrected(dash) 119885DR at 71∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
in this rain estimator (see Table 3) Furthermore comparingthe analysis of variance results between areal (038 km2) andpoint (007m2) values from the gauge data (Section 42)half of the errors after attenuation correction were producedby the small scale variability of the rain fields and theinstrumental uncertainty of the rain gauge
A similar comparison is shown in Figure 18 for a 60minaccumulation Using 119877(119885
119867) the NB NSD rmse and r rmse
were decreased from minus65 86 71 (mmhr) and 70to minus5 34 28 (mmhr) and 28 by the attenuationcorrection respectively Using 119877(119885
119867 119885DR) the NB NSD
rmse and r rmse were only slightly decreased after theattenuation correction However the NSD rmse and r rmsewere reduced when compared at 10min accumulation In thesame accumulation time the errors in the gauge data reachedto over 15
This comparison was affected by different sampling areasof the gauges and the radar Therefore the areal averagevalues were compared (Figure 19) Using119877(119885
119867)with the areal
averaged radar measurements the NB was reduced fromminus69 to 7 NSD was reduced from 122 to 42 rmsewas reduced from 39 (mmhr) to 13 (mmhr) and r rmse
was reduced from 77 to 26 at an accumulation time of10min Using 119877(119885
119867 119885DR) the NB NSD rmse and r rmse
were decreased fromminus24 67 21 (mmhr) and 42 to 539 12 (mmhr) and 24 respectively
TheNB and r rmsewithmatched sampling areas betweenthe gauge and the radar are shown in Figures 20 and 21 Thereduction of NB due to temporal accumulation was smallin both of the rain estimators The NB was reduced to lessthan 7 by attenuation correction for all accumulation timeswith both119877(119885
119867) and119877(119885
119867 119885DR)The attenuation correction
effect was significantly greater in119877(119885119867) Similarly the r rmse
was reduced significantly in 119877(119885119867) by attenuation correction
(from 68 to 25 at a 60min accumulation) The reductionof the r rmse due to temporal accumulation was larger in119877(119885119867 119885DR) because the significant observational noise of
119885DR was smoothed The r rmse was decreased from 18to 15 in 119877(119885
119867 119885DR) at a 60min accumulation time after
attenuation correction
5 Conclusions
In this study an attenuation correction algorithm based onΦDP was developed to improve the accuracy of the rainfall
12 Advances in Meteorology
1st gauge site 2nd gauge site
3rd gauge site 4th gauge site
5th gauge site 6th gauge site
7th gauge site 8th gauge site
9th gauge site 10th gauge site
17 18 19 20 2116Time (h)
17 18 19 20 2116Time (h)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
UncorrectedCorrected
UncorrectedCorrected
Figure 11 Time series of measured (red circle) and attenuation corrected (blue circle) 119885119867at the 10 rain gauge sites
Advances in Meteorology 13
UncorrectedCorrected
UncorrectedCorrected
17 18 19 20 2116Time (h)
17 18 19 20 2116Time (h)
1st gauge site 2nd gauge site
3rd gauge site 4th gauge site
5th gauge site 6th gauge site
7th gauge site 8th gauge site
9th gauge site 10th gauge site
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
Figure 12 Time series of measured (red circle) and attenuation corrected (blue circle) 119885DR at the 10 rain gauge sites
14 Advances in Meteorology
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Uncorrected(a)
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Corrected Simulated from DSD(b)
Figure 13 Scatter plots of 119885119867and 119885DR before (a) and after (b) attenuation correction at the 10 rain gauge sites DSD observation results (dot
[21])
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
04
05
06
07
08
09
10
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
Accum time = 1min
Accum time = 10min
Accum time = 2min
Accum time = 20min
Accum time = 5min
Accum time = 60min
Figure 14 The Spearman correlation coefficient 119903 between data from two rain gauges at different accumulation times
estimation using X-band radar This algorithm was evalu-ated with radar measurements from NIMS-XPOL Howevermeasured ΨDP contained 120575 and observation noise as wellas ΦDP Furthermore the ΨDP offset varied with azimuthangles and decreased near the radar when rain cells were
above the radar In this study ΦDP was estimated by aniterative filtering technique and by adjusting with an offsetdetermined as the minimum value ofΦDP within 3 km of theradar As a result estimated ΦDP monotonically increasedThe path integrated value of119860
119867and119860DP was calculated from
Advances in Meteorology 15
2nd 6th 9th gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(038km2) versus G
(a) Normalized bias
⟨G⟩(038km2) versus G
2nd 6th 9th gaugesMean NSD
Mean R
0
20
40
60
80
100
NSD
()
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 15 (a) The NB and (b) NSD between ⟨119866⟩(038 km2) and 119866 from three rain gauges with accumulation time ⟨119866⟩ is the temporallyaveraged rainfall rate of ⟨119866⟩
All gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(6km2) versus G
(a) Normalized bias
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
0
20
40
60
80
100
NSD
()
All gaugesMean NSD
Mean R
⟨G⟩(6km2) versus G
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 16 (a) The NB and (b) NSD between ⟨119866⟩(6 km2) and 119866 from 10 rain gauges with accumulation time ⟨119866⟩ is the temporally averaged⟨119866⟩ with accumulation time
estimated ΦDP using 119860119867-ΦDP and 119860DP-ΦDP relationships
derived from the DSDs data Measured 119885119867
and 119885DR werecorrected with the calculated path integrated 119860
119867and 119860DP
by each gate The attenuated signal patterns of 119885119867and 119885DR
were restored to a typical rain echo pattern However theattenuation corrected 119885
119867and 119885DR fields were not matched
at distant ranges This was likely because of contaminationdue to beam broadening and the accumulation of errorswith increasing range Comparing the 119885
119867-119885DR scatter plots
over the 10 rain gauge sites with the results obtained froma scattering simulation with the DSD data the attenuation
corrected 119885119867-119885DR scatter plots are in good agreement with
the 119885119867-119885DR scatter plots from the DSDs
The validation for the improvement of accuracy in therainfall estimation through attenuation correction was per-formed by comparing with dense rain gauge network data119877(119885119867) from the attenuated (observed)119885
119867produced a severe
underestimation with an NB NSD rmse and r rmse of70 127 127 (mmhr) and 78 at 10min accumulationrespectively while 119877(119885
119867) from the attenuation corrected 119885
119867
showed better agreement with the gauge measurements withan NB NSD rmse and r rmse of minus1 52 521 (mmhr)
16 Advances in Meteorology
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus070
NSD = 127
rmse = 1279
corr = 0858
Corrected
NB = minus001
NSD = 052
rmse = 521
corr = 0927
r_rmse = 0786 r_rmse = 0320
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = 003
NSD = 049
rmse = 496
r_rmse = 0305
corr = 0936
Corrected
NB = 001
NSD = 050
rmse = 502
r_rmse = 0308
corr = 0936
(b) 119877(119885119867119885DR)
Figure 17 The scatter plots of 119877(038 km2 10min) versus 119866(007m2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
CorrectedUncorrected
R(m
mh
)
0
10
20
30
40
10 20 30 400G (mmh)
Uncorrected
NB = minus065
NSD = 086
rmse = 706
r_rmse = 0702
corr = 0878
Corrected
NB = minus005
NSD = 034
rmse = 281
r_rmse = 0279
corr = 0866
(a) 119877(119885119867)
CorrectedUncorrected
R(m
mh
)
10 20 30 400G (mmh)
0
10
20
30
40
Uncorrected
NB = 008
NSD = 023
rmse = 189
r_rmse = 0188
corr = 0942
Corrected
NB = 007
NSD = 025
rmse = 205
r_rmse = 0203
corr = 0940
(b) 119877(119885119867119885DR)
Figure 18 The scatter plots of 119877(038 km2 60min) versus 119866(007m2 60min) (a) using 119885119867and (b) using 119885
119867and 119885DR
and 32 at 10min accumulation respectively At 60minaccumulation theNBNSD rmse and r rmsewere decreasedfrom minus65 86 71 (mmhr) and 70 to minus5 3428 (mmhr) and 28 respectively Using 119877(119885
119867 119885DR) the
NB NSD rmse and r rmse were only slightly reduced even if
attenuation was corrected but were smaller than when using119877(119885119867) This indicated that 119877(119885
119867 119885DR) was less sensitive to
the attenuation than 119877(119885119867) In addition half of the errors
that remained after attenuation correction were caused by thesmall scale variability of the rain fields and the instrumental
Advances in Meteorology 17
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus069
NSD = 122
rmse = 391
r_rmse = 0770
corr = 0893
Corrected
NB = 007
NSD = 042
rmse = 133
r_rmse = 0262
corr = 0960
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = minus024
NSD = 067
rmse = 215
r_rmse = 0424
corr = 0718
Corrected
NB = 005
NSD = 039
rmse = 124
r_rmse = 0245
corr = 0891
(b) 119877(119885119867119885DR)
Figure 19 The scatter plots of 119877(6 km2 10min) versus 119866(6 km2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
R(ZH)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(a) 119877(119885119867)
R(ZH ZDR)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(b) 119877(119885119867119885DR)
Figure 20 The normalized bias of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and 119885DR The red
point-line and the blue point-line indicate without and with the attenuation correction respectively
uncertainty of the rain gauges Using an areal averagedvalue the NB was reduced to less than 7 with attenua-tion correction using either rainfall estimator regardless ofaccumulation timeThe r rmse were significantly reduced byattenuation correction in 119877(119885
119867) and the reduction of r rmse
with 119877(119885119867 119885DR) was even more dramatic
As summarized above the rain attenuation for 119885119867
and 119885DR in the X-band was corrected by ΦDP and theresults showed a good agreement with the value obtainedby scattering simulations with the DSD datasets The resultsof comparison with rain gauge data also showed a signifi-cant improvement in the accuracy of rainfall estimation by
18 Advances in Meteorology
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10r_
rmse
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH)
(a) 119877(119885119867)
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10
r_rm
se
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH ZDR)
(b) 119877(119885119867119885DR)
Figure 21 The relative root-mean square error of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and
119885DR The red point-line and the blue point-line indicate without and with the attenuation correction respectively
attenuation correction which supports the quantitative useof the X-band radar in rain data retrieval
Appendix
A Instrumental Uncertainty of the Rain Gauge
The instrumental uncertainty of the TB rain gauge is asso-ciated with improper calibration leakage and precipitationmissing the gauge due to the measurement principle andwetting and evaporation losses In addition external sourcessuch as wind turbulence and installed position also con-tribute to the instrumental uncertainty Ciach [17] reportedthat the TB gauge data suffers a significant uncertainty asso-ciated with random differences between closely collocatedTB gauges through the analysis of the error as a functionof the rainfall intensity for three timescales In addition theimproper calibration induces a bias in the rainfall measure-ment and this bias cannot be removed by increasing accu-mulation timeTherefore the proper correction or processingof TB gauge data is necessary to obtain accurate and reliablereading of the rainfall amount
In this appendix instrumental biases were calculatedto reduce the instrumental uncertainty through an idealexperiment in the laboratory and a field intercomparisonwith 10 TB rain gauges In addition the random errors werequantitatively calculated as a function of rainfall rate andaccumulation time
A1 Correction of Instrumental Bias The instrumental bias119861119868(119894) of the 119894th gauge was defined as a combination
of the absolute bias 119861119860(ref) of the reference gauge and
the relative bias 119861119877(119894) of the 119894th gauge according to the
following equation
119861119868 (119894) = 119861119860 (ref) times 119861119877 (119894) (A1)
119861119860(ref) signified the ratio of the calculated value from
the instrumental resolution of the reference gauge to the truevalue measured by other instruments
119861119860 (ref) =
rainfalltruthrainfallgauge (ref)
(A2)
In this study 119861119860(ref) was derived by an ideal experiment
using an electronic scale First the reference gauge was puton the scale and the scale was initialized The water dropsflowed out into the receiving orifice of the reference gaugeuntil the tipping number became 60Themeasured weight of60 tips was 2144 g while the theoretical (or expected) valuewas 2244 g for the reference gauge (receiving orifice size of1545 cm 60 tips and 02mm tip resolution) Therefore theabsolute bias of the reference gauge from (A2) was about0955 This indicates that the rainfall amount measured bythis reference gauge was overestimated by about 45 andthis bias should be corrected prior to any interpretation
119861119877(119894) represented the ratio of the observed (recorded)
value from the 119894th gauge to the observed (recorded) valuefrom the reference gauge
119861119877 (119894) =
rainfallgauge (ref)rainfallgauge (119894)
(A3)
119861119877(119894) was calculated from ten collocated rain gauges
including the reference gauge located at an observation field
Advances in Meteorology 19
Table 4 The instrumental and relative biases of the 10 tipping bucket rain gauges
Gauge ID RG1 RG2 RG3 RG4 RG5 RG6 RG7 RG8 RG9 RG10Instrumental bias 094 108 099 100 096 104 098 102 108 098Relative bias 098 113 103 105 100 109 102 107 113 102
013610152230406080120150200250300400
()
Rain
rate
(mm
h)
100101
Accumulation time (min)
10
100NSD ()
Figure 22 Two-dimensional normalized standard deviationbetween the standard rainfall and each rain gauge
at KNUKoreaThe observational period for intercomparisonwas from October 2009 to June 2010 (14 rainy days) andthe total accumulated rainfall for this period was about300mm The results are presented in Table 4 All relativebiases were mainly larger than 10 This indicated that themost of gauges underestimated the rainfall amount comparedto the reference rain gauge The relative bias of the referencegauge (119894 = 5) was 10
Finally 119861119860(119894) was calculated from (A1) (see Table 4) All
instrumental biases were between 090 and 11 and the meanvalue of the biases was nearly 10 This indicated that therainfall rate from a single gauge data had a maximum 10bias These biases were corrected for each gauge to removeexisting instrumental errors
A2 Quantification of Random Errors To quantify the ran-dom error due to instrumental uncertainty the NSD of theten gauges was calculated with accumulation time and rainrate using data from an intercomparison The NSD of the 119894thgauge compared to the reference gauge was derived for 20rain rate classes (Δ119877) and 17 accumulation time classes (Δ119905)as follows
NSD119894 (Δ119905 Δ119877)
=
(1119899)radicsum119899minus1
119896=0[1198661015840
119870(Δ119905 Δ119877) minus 119866119894119896 (Δ119905 Δ119877)]
2
1198661015840
(A4)
where 1198661015840 and 119866 indicate the reference and individual gaugesrespectively The overbar means the temporal average ofthe rainfall rate The NSDs of nine different gauges (exceptthe reference gauge) were averaged to obtain the two-dimensional average NSD (Figure 22) The average NSD
tended to decrease with increasing accumulation time andwith increasing rain rate The NSD was about 10ndash15when accumulation time was 10min and the rain rate was10mmhminus1 The instrumental errors were quite significant forlonger accumulation times and higher rainfall rates Theseerrors should be used as a guideline when a single gauge isused as a reference for any validation study
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper has been revised and extended fromMs Young-AOhrsquos MS thesis [25] of Kyungpook National University Thisresearch is supported by Development and Application ofCross Governmental Dual-Pol Radar Harmonization (WRC-2013-A-1) project of theWeather Radar Center KoreaMeteo-rological AdministrationThis work was funded by the KoreaMeteorological Administration Research and DevelopmentProgram under Grant KMIPA2015-1010
References
[1] V N Bringi V Chandrasekar N Balakrishnan and D SZrnic ldquoAn examination of propagation effects in rainfall onradar measurements at microwave frequenciesrdquo Journal ofAtmospheric and Oceanic Technology vol 7 no 6 pp 829ndash8401990
[2] A Huggel W Schmid and AWaldvogel ldquoRaindrop size distri-butions and the radar bright bandrdquo Journal of AppliedMeteorol-ogy vol 35 no 10 pp 1688ndash1701 1996
[3] G W Lee Errors in rain measurement by radar effect of vari-ability of drop size distributions [PhD thesis]McGill University2003
[4] J D Tuttle and R E Rinehart ldquoAttenuation correction in dual-wavelength analysesrdquo Journal of ClimateampAppliedMeteorologyvol 22 no 11 pp 1914ndash1921 1983
[5] S-G Park V N Bringi V Chandrasekar M Maki andK Iwanami ldquoCorrection of radar reflectivity and differentialreflectivity for rain attenuation at X band Part I theoretical andempirical basisrdquo Journal of Atmospheric and Oceanic Technol-ogy vol 22 no 11 pp 1621ndash1632 2005
[6] G Scarchilli E Gorgucci V Chandrasekar and T A SeligaldquoRainfall estimation using polarimetric techniques at C-bandfrequenciesrdquo Journal of Applied Meteorology vol 32 no 6 pp1150ndash1160 1993
[7] M N Anagnostou E N Anagnostou and J VivekanandanldquoCorrection for rain path specific and differential attenuation ofX-band dual-polarization observationsrdquo IEEE Transactions onGeoscience and Remote Sensing vol 44 no 9 pp 2470ndash24802006
20 Advances in Meteorology
[8] J Kalogiros M N Anagnostou E N Anagnostou M Mon-topoli E Picciotti and F S Marzano ldquoEvaluation of a newpolarimetric algorithm for rain-path attenuation correction ofX-band radar observations against disdrometerrdquo IEEE Transac-tions on Geoscience and Remote Sensing vol 52 no 2 pp 1369ndash1380 2014
[9] W-Y Chang J Vivekanandan and T-C C Wang ldquoEstimationof X-band polarimetric radar attenuation and measurementuncertainty using a variational methodrdquo Journal of AppliedMeteorology and Climatology vol 53 no 4 pp 1099ndash1119 2014
[10] J Hubbert and V N Bringi ldquoAn iterative filtering technique forthe analysis of copolar differential phase and dual-frequencyradar measurementsrdquo Journal of Atmospheric and OceanicTechnology vol 12 no 3 pp 643ndash648 1995
[11] K V Mishra W F Krajewski R Goska et al ldquoDeploy-ment and performance analyses of high-resolution iowa XPOLradar system during the NASA IFloodS campaignrdquo Journal ofHydrometeorology vol 17 no 2 pp 455ndash479 2016
[12] M Thurai G J Huang V N Bringi W L Randeu and MSchonhuber ldquoDrop shapes model comparisons and calcu-lations of polarimetric radar parameters in rainrdquo Journal ofAtmospheric and Oceanic Technology vol 24 no 6 pp 1019ndash1032 2007
[13] M D Humphrey J D Istok J Y Lee J A Hevesi and AL Flint ldquoA new method for automated dynamic calibration oftipping-bucket rain gaugesrdquo Journal of Atmospheric andOceanicTechnology vol 14 no 6 pp 1513ndash1519 1997
[14] E HabibW F Krajewski V Nespor andA Kruger ldquoNumericalsimulation studies of rain gage data correction due to windeffectrdquo Journal of Geophysical ResearchAtmospheres vol 104 no16 Article ID 1999JD900228 pp 19723ndash19733 1999
[15] G J Ciach and W F Krajewski ldquoOn the estimation of radarrainfall error variancerdquoAdvances inWater Resources vol 22 no6 pp 585ndash595 1999
[16] E Habib W F Krajewski and A Kruger ldquoSampling errors oftipping-bucket rain gaugemeasurementsrdquo Journal of HydrologicEngineering vol 6 no 2 pp 159ndash166 2001
[17] G J Ciach ldquoLocal random errors in tipping-bucket rain gaugemeasurementsrdquo Journal of Atmospheric andOceanic Technologyvol 20 no 5 pp 752ndash759 2003
[18] EHabibW F Krajewski andG J Ciach ldquoEstimation of rainfallinterstation correlationrdquo Journal ofHydrometeorology vol 2 no6 pp 621ndash629 2001
[19] NIMS ldquoStudy on estimation of rainfall amount and classifi-cation of hydrometeor species using dual-polarimetric radar(III)rdquo pp 1ndash20 2010
[20] E Gorgucci V Chandrasekar and L Baldini ldquoCorrection ofX-band radar observation for propagation effects based on theself-consistency principlerdquo Journal of Atmospheric and OceanicTechnology vol 23 pp 1668ndash1681 2006
[21] NIMS ldquoStudy on estimating raindrop size distribution forquantitative estimation of radar rainfall amountrdquo pp 1ndash55 2009
[22] NIMS ldquoQuantitative precipitation estimation and hydrometeoridentification using dual-polarization radar-Phase IIrdquo pp 1ndash742009
[23] S-G Park M Maki K Iwanami V N Bringi and V Chan-drasekar ldquoCorrection of radar reflectivity and differential reflec-tivity for rain attenuation at X band Part II Evaluation andapplicationrdquo Journal of Atmospheric and Oceanic Technologyvol 22 no 11 pp 1633ndash1655 2005
[24] E Habib and W F Krajewski ldquoUncertainty analysis of theTRMM ground-validation radar-rainfall products applicationto the TEFLUN-B field campaignrdquo Journal of Applied Meteorol-ogy vol 41 no 5 pp 558ndash572 2002
[25] Y-A Oh Rainfall estimation from X-band dual-polarizationradar measurements effects of attenuation correction and evalu-ation with a dense rain gauge network [MS thesis] KyungpookNational University 2011
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom
Applied ampEnvironmentalSoil Science
Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal of
Geophysics
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
Advances in Meteorology 5
3510
3506
3502
3498
3494 Unnam
12628 12636 12644
1 234 96
8 75
10
R
Figure 3 Layout of NIMS-XPOL and the 10 rain gauges nearUnnam-myeon Muan-gun Jeollanam-do Korea R shows thelocation of the NIMS-XPOL the numbered points indicate theobservational sites of the 10 rain gauges
respectively119870DP(119903) is the specific differential one-way phaseshift calculated using the following equation
119870DP (119903) =ΦDP (1199032) minus ΦDP (1199031)
2 (1199032minus 1199031)
(2)
where 119903 is between 1199031and 1199032 The coefficients 120590 (= 03445)
and 120576 (= 01705) were derived from 119860119867 119860DP and 119870DP
obtained by scattering simulations with the DSDs data col-lected by the PrecipitationOccurrence Sensor System (POSS)from March to September of 2001 in Pusan Korea [21] Thetotal path integrated attenuation and differential attenuationwere calculated withΦDP by integrating the following
2int
119903
0
119860119867(1199031015840) 1198891199031015840= 120590ΦDP (119903) (3)
2int
119903
0
119860DP (1199031015840) 1198891199031015840= 120576 times 2int
119903
0
119860119867(1199031015840) 1198891199031015840 (4)
Number 2 in the equations indicates that the values are two-way values
Finally the measured reflectivity 119885119867-meas(119903) [dBZ] and
the differential reflectivity 119885DR-meas(119903) [dB] at 119903 were cor-rected with the calculated two-way path integrated attenua-tion and differential attenuation
119885119867-corr (119903) = 119885119867-meas (119903) + 2int
119903
0
119860119867(1199031015840) 1198891199031015840
119885DR-corr (119903) = 119885DR-meas (119903) + 2int119903
0
119860DP (1199031015840) 1198891199031015840
(5)
32 Radar Rain Estimation and Comparison with Rain GaugeData A comparison between estimated 119877 from the radar
Measured dual-polarization parameters(ZH ZDR ΦDP 120588HV)
Quality control
(ii) Speckling echo removal
(i) Calibration for instrumental bias of ZH ZDR
(iii) 120588HV threshold application
(ii) Adjustment of its offset(i) 20th FIR filter application to ΨDP
Estimation of ΦDP
Calculation of AH and ADP
Correction of ZH and ZDR
from DSDs120590 and 120576
Figure 4 The schematic diagram of the attenuation correctionalgorithm
parameters and the rain rate (119866) from the rain gauge data wasperformed to evaluate the attenuation correction algorithmIn addition to the rain attenuation the difference between119877 and 119866 was caused by the errors due to the instrumentaluncertainty and representativeness errors of the rain gaugedata the radar observational noise and the errors in the radarrainfall estimation The bias due to the instrumental uncer-tainty in gauge data was removed by properly calibrating thegauges through laboratory tests and a gauge intercomparisonThe representativeness errors were due to the mismatch ofthe sampling area (volume) between the gauge and the radarThe random error in the gauge data could be minimized bytime and areal integrationThe errors due to themeasurementnoise in the radar data could be reduced in a similar mannerThe time offset between the rain gauge and the radar samplingarea was ignored due to the close distance (about 16 km) andthe low elevation angle (3∘)
Therefore 119877 and 119866 were derived with two different radarsampling areas and with increasing accumulation time
(1)119877(038 km2) versus119866(0071198982)119877(038 km2)was estimated
from the average radar measurement with 9 pixels (038 km2at 16 km range) The gauge rainfall rate 119866(007m2) wasderived from the rainfall accumulation from an individualrain gauge that had a sampling area of 007m2
(2) 119877(6 1198961198982) versus 119866(6 1198961198982
) 119877(6 km2) was estimated fromthe areal averaged radar parameters at a 16 km times 31 km radar
6 Advances in Meteorology
Table 3 Two rain rate retrieval algorithms and boundary conditions [22]
119877 (119885119867) = 120572119885
120573
119867119877 (119885DR) = 120572119885
120573
1198671001120574119885DR
Z gt 35 dBZ Z lt 35 dBZ Z gt 35 dBZ 119885DR gt 03 dB Z lt 35 dBZ 119885DR gt 03 dB120572 353 times 10
minus2383 times 10
minus2103 times 10
minus2145 times 10
minus2
120573 621 times 10minus2
608 times 10minus2
985 times 10minus1
890 times 10minus1
120574 mdash mdash minus6479 minus5295
measurement resolution 119866(6 km2) was the average value of119866(007m2) from the ten gauges within the same area
Estimated 119877 from radar parameters was calculated withtwo rain estimators as shown in Table 3These rain estimatorswere derived by a least squares fit with the theoretical 119885
119867
119885DR and 119877 obtained by scattering simulations with theDSD data collected by POSS from March to Septemberof 2001 in Pusan Korea [21] To improve the estimationaccuracy each estimator was derived for two categoriesThe reflectivity threshold of 35 dBZ was used as a stan-dard stratiformconvective rain classification [22] The 119885DRthreshold of 03 dB was used to minimize the effect of the119885DR measurement noiseThePPI data with an elevation angleof 3∘ was used to suppress the beam shielding and groundclutterThe119885
119867and119885DR fields from the radar were smoothed
by the linear average of values at 3 times 3 pixels to reduce theobservational noise
The bias in 119866 from the ten rain gauges was corrected bythe laboratory test and the intercomparison with field dataThe detail is presented in Appendix To evaluate the atten-uation correction algorithm and verify the rainfall retrievalaccuracy the comparisons of119877 and119866were represented by thegeneral statistics such as normalized bias (NB) normalizedstandard deviation (NSD) root-mean square error (rmse)relative rms error (r rmse) and Spearman correlation (corr)as follows
NB119877-119866 (Δ119905) =
(1119899)sum119899minus1
119896=0[119877119896 (Δ119905) minus 119866119896 (Δ119905)]
119866 (Δ119905)
NSD119877-119866 (Δ119905) =
radic(1119899)sum119899minus1
119896=0[119877119896 (Δ119905) minus 119866119896 (Δ119905)]
2
119866 (Δ119905)
rmse119877-119866 (Δ119905) =
radicsum119899minus1
119896=0[119877119896 (Δ119905) minus 119866119896 (Δ119905)]
2
119899
rmse119877-119866 (Δ119905) = radic
sum119899minus1
119896=0[119877119896 (Δ119905) minus 119866119896 (Δ119905)]
2
sum119899minus1
119896=0[119866119896 (Δ119905)]
2
corr119877-119866 (Δ119905)
=
sum119899minus1
119896=0[119877119894119896 (Δ119905) minus 119877119894 (Δ119905)] [119866119895119896 (Δ119905) minus 119866119895 (Δ119905)]
radicsum119899minus1
119896=0[119877119894119896 (Δ119905) minus 119877119894 (Δ119905)]
2radicsum119899minus1
119896=0[119866119895119896 (Δ119905) minus 119866119895 (Δ119905)]
2
(6)
where 119899 is the number of 119877 and 119866 pairs at a given temporalintegrationΔ119905 and the overbar indicates the temporal average
33 Spatial Correlation and Variability of the Rain FieldsRain gauges are widely used to measure rainfall amountsdue tomany advantages including easy installation mobilitylow price and ease of data processing The rain gauge datais used as a reference for the ground rainfall amount forthe evaluation of rain retrieval algorithms adjustment ofother instruments data assimilation and other hydrologicalapplications However it is affected by instrumental errorsand representativeness errors due to the variability of rainfields
In a TB rain gauge the instrumental errors are due toimproper calibration the time delay of the tips leakage dueto its measuring principle and so forthThis error is exploredin the Appendix
The representativeness errors are subdivided into spatialand temporal representativeness They are mainly caused byhaving a small sampling area and a limited sampling time dueto the gauge measuring principle The actual rain fields varywith different spatiotemporal scales that are not equivalentto the spatial and temporal resolution of the rain gaugeTherefore the rain gauge data represent the actual rain fieldsin a very limited manner
These errors should be considered and minimized priorto using rain gauge data as ground truth Furthermore thedegree of these errors has to be understood quantitativelysince there can be other error sources in various applicationsIn this section the spatial and temporal variability of the rainfields are investigated using the rain data collected by the tenrain gauges for the month of July 2010 at Muan Korea
First the Spearman correlation coefficient 119903119894119895(Δ119905)
between a pair of 119866 values from the 119894th and 119895th individualrain gauge with integration time Δ119905 was calculated fromthe following equation to obtain the Spearman correlationcoefficient of the rain fields
119903119894119895 (Δ119905)
=
sum119899minus1
119896=0[119866119894119896 (Δ119905) minus 119866119894 (Δ119905)] [119866119895119896 (Δ119905) minus 119866119895 (Δ119905)]
radicsum119899minus1
119896=0[119866119894119896 (Δ119905) minus 119866119894 (Δ119905)]
2radicsum119899minus1
119896=0[119866119895119896 (Δ119905) minus 119866119895 (Δ119905)]
2
(7)
where 119899 is the number of gauge data pairs at given Δ119905 and theoverbar indicates the temporal average of the rainfall rate Itwas derived withΔ119905 to research the spatial and temporal scaleof the variation for the rain field
In addition the effect of the natural variability of the rainfields within two different areas (038 km2 and 6 km2) wasexplored by analyzing the bias and random error between thespatially averaged rainfall rate ⟨119866⟩ and 119866 from the individual
Advances in Meteorology 7
32km
21 km
11 km
20km40km60km
minus180 18020 30 40 50 60 70 80 90 100120140160100(a)
32 km
21 km
11 km
20km40km60km
minus180 18020 30 40 50 60 70 80 90 100120140160100(b)
Figure 5 PPIs of measured (a) and filtered (b) ΨDP of the NIMS-XPOL at 2101UTC July 10 2010
gauge The NB and NSD of 119866119894with ⟨119866⟩ were calculated as
follows
NB (119894) =(sum119899minus1
119896=0[119866119894119896 (Δ119905) minus ⟨119866119896 (Δ119905)⟩]) 119899
⟨119866 (Δ119905)⟩
NSD (119894) =(1119899)radic(sum
119899minus1
119896=0[119866119894119896 (Δ119905) minus ⟨119866119896 (Δ119905)⟩]
2)
⟨119866 (Δ119905)⟩
(8)
where 119894 indicates the 119894th gauge and the overbar indicates thetemporal average
This analysis provided a reasonable spatial average andtemporal accumulation resolution for the calculation ofmoreaccurate and realistic rainfall amounts from the gauge dataand a quantitative guideline regarding the errors in rainestimation due to the variability of the rain fields
4 Results
41 Attenuation Correction of Reflectivity and DifferentialReflectivity The attenuation correction algorithm in Sec-tion 31 was applied to the NIMS-XPOL radar data ΦDPwas estimated by eliminating 120575 and the observational noisefrom measured ΨDP using an iterative FIR filter The offsetof ΦDP was then adjusted Figure 5 shows the PPIs ofmeasured ΨDP and filtered ΦDP After applying the FIR filterthe small scale variation due to 120575 and observational noisein measured ΨDP was reduced and filtered ΨDP (estimatedΦDP) monotonically increased with an increase in rangeThe maximum value of estimated ΦDP reached to 140∘ Inaddition estimated ΦDP at the azimuth angles of 150∘ and200∘ sharply increased with the loss of signal beyond thesesharp gradient areas This reflected that strong convective
cells existed in these directions and that these cells caused thesevere signal attenuation EstimatedΦDP values between 250
∘
and 360∘ azimuth angle decreased after filtering since ΦDPoffsets were positive values in these directions Figure 6 showsthe range profile of measured and filtered ΨDP for estimationof ΦDP at 71
∘ 160∘ and 334∘ azimuth angles The unexpectedlarge values near the radarwere removed by the120588HV threshold(red line)The small fluctuation inmeasuredΨDP due to 120575 andthe observational noise was effectively removed after filteringThe value of estimated ΦDP was adjusted to 0 by the offset(blue line)
Estimated ΦDP was used to calculate the path integrated119860119867 and the measured119885
119867values were corrected as shown in
Figure 7 Figure 7(a) represents measured-attenuated119885119867and
relatively small values of 119885119867at far ranges This may indicate
that the 119885119867values at far ranges were affected by the strong
attenuation due to strong convective cells near the radarThisspatial pattern of 119885
119867was the typical feature of the attenuated
radar signal due to the rain After correcting the attenuationthe value of the 119885
119867field became larger over the entire area
and the radially attenuated 119885119867
was corrected by showinga typical rain cell pattern in 119885
119867(Figure 7(b)) The largest
difference between the uncorrected and the corrected 119885119867at
2101UTCwas over 50 dB However the white area behind thestrong echoes was not detected as there was a complete loss ofthe signal powerThese regions could not be corrected by thisalgorithmThe circle pattern of119885
119867fields within about a 4 km
range from theNIMS-XPOL could not present the actual rainechoes due to the instrumental issues (the same was true for119885DR) These results are also presented in the range profiles inFigure 8
Similarly measured 119885DR were corrected with the pathintegrated 119860DP derived from the path integrated 119860
119867using
(4) (see Figure 9) Before applying attenuation correction
8 Advances in Meteorology
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
0
50
100
150
200
ΦDP (∘) azimuth angle = 71∘
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
0
50
100
150
200
ΦDP (∘) azimuth angle = 160∘
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
0
50
100
150
200
ΦDP (∘) azimuth angle = 334∘
(c)
Figure 6 Range profiles of measured (black) 120588HV threshold isolated point removing (red) and FIR filter and offset adjusted (blue) ΨDP forestimation of ΦDP at 71
∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
32km
21 km
11 km
20km40km60km
minus10 10 15 20 25 30 35 40 45 50 55 60 65 700(a)
32 km
21 km
11 km
20km40km60km
minus10 10 15 20 25 30 35 40 45 50 55 60 65 700(b)
Figure 7 PPIs of measured (a) and attenuation corrected (b) 119885119867of the NIMS-XPOL for the case in Figure 5
Advances in Meteorology 9
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 71∘
minus20
0
20
40
60
80
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 160∘
minus20
0
20
40
60
80
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 334∘
minus20
0
20
40
60
80
(c)
Figure 8 Range profiles of measured (line) calibration bias 120588HV threshold and isolated point removing (dot) and attenuation corrected(dash) 119885
119867at 71∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
the attenuation signature in119885DR was determined Ideally the119885DR value for rain should be positive since rain drops areoblate due to aerodynamic force However themeasured119885DRfield in Figure 9 showed a negative value over wide areasAfter correcting the attenuation the119885DR field was larger than0 dB and showed a realistic rain patternThis is also presentedin the range profile of Figure 10The largest difference of119885DRbetween before and after attenuation correction at 2101UTCwas about 8 dB The 119885DR values around the 220∘ azimuthangle beyond 40 km were the largest They did not matchwith the area of the largest 119885DR values (around the 210∘azimuth angle with a range of 20 km and 40 km) This wasan unexpected result because 119885DR is proportional to 119885
119867in
theory This mismatch between 119885119867
and 119885DR at far rangemay be due to contamination due to beam broadening andaccumulated errors due to uncertainty in the attenuationcorrection with different ranges
Figures 11 and 12 show the time series of 119885119867
and119885DR at the ten gauge sites before and after applying theattenuation correction The differences between uncorrectedand corrected values existed after 1800UTC and were at amaximumaround 2100UTCThis attenuation correctionwas
evaluated through comparison with the theoretical values of119885119867and 119885DR In Figure 13 the uncorrected 119885
119867-119885DR scatter
plots are apart from the theoretical119885119867-119885DR values [23] After
attenuation correction the119885119867-119885DR scatter plots were close to
the calculated valuesThis indicated that119885119867and119885DR over the
gauge sites were corrected with reliable values
42 Variability of the Rain Fields The Spearman correlationcoefficient of the rainfall rate from the 10 spatially dis-tributed rain gauges was calculated with accumulation timeand separation distance (Figure 14) The overall Spearmancorrelation coefficient decreased with increasing separationdistance at all accumulation timesThis tendency was similarto the result from Habib and Krajewski [24] However thecorrelation value was higher in this study Note that the gaugetip resolution in Habib and Krajewski [24] was 0254mmwhile it was 02mm in this study In addition the rainfall ratesfrom the gauge data used in this study mainly consisted ofmoderate rainfall rates while Habib and Krajewski [24] useddata containing heavy rainfall rates
For accumulation times of less than 10min the corre-lation at the same distance was dramatically increased with
10 Advances in Meteorology
32km
21 km
11 km
20km40km60km
18
14
00
02
04
06
26
40
minus10
22
minus20
30
10
50
minus30
(a)
32 km
21 km
11 km
20km40km60km
18
14
00
02
04
06
26
40
minus10
22
minus20
30
10
50
minus30
(b)
Figure 9 PPIs of measured (a) and attenuation corrected (b) 119885DR of NIMS-XPOL for the case in Figure 5
increasing accumulation time and the slope of the correlationfunction was steep This was due to the reduction of theinstrumental errors of the gauge data by time accumulationThis indicated that rainfall accumulated for less than 10minfrom single gauge data cannot represent the rainfall over itssurrounding area due to instrumental errors For accumula-tion times longer than 10min the slope became less steepThis reflected the remaining variability of the rain field Inaddition the accumulated rainfall rate of about 10min wassuitable for minimizing the instrumental uncertainty whilepreserving the variation in the rain field The correlationreached about 098 at 05 km in 10min of accumulationand this approached 098 at 10 km at 60min accumulationTherefore a single rain gauge mostly represented the rainfallrate within 05 km and 10 km with 10min and 60minaccumulation respectively
The NB between an areal averaged rainfall rate ⟨119866⟩ anda single gauge rainfall rate 119866 is shown with accumulationtime in Figures 15(a) and 16(a) Where the average area was038 km2 (corresponding to 9 pixels of NIMS-XPOL radarat 16 km range) the NB reached a 6ndash8 maximum Wherethe averaged area is 6 km2 (corresponding to the coverage forall of the gauges) the NB reached about a 10 maximumFurthermore the NB hardly changed with accumulationtime Note that this statistic was derived after removinginstrumental bias This indicated that the rainfall rate froma single gauge was biased due to the spatial variability of therain fields and this bias was maintained with accumulationtime
The NSD of a single gauge for a mean area of 038 km2was over 40 and it reached about 70ndash90 for a mean areaof 6 km2 at an accumulation time of 1min (see Figures 15(b)and 16(b)) This statistic included instrumental errors (about
40ndash60 at 16mmhminus1 and 60ndash80 at 10mmhminus1 as shownin the Appendix) as well as the spatial variability of the rainfields This indicated that the variance between the point andarea values from the gauge data at short accumulation timeswas similar to the errors due to the instrumental uncertaintyof the rain gaugeTheNSD in 60min accumulation decreasedto over 10 and about 15ndash45 with a mean area of 038 km2and 6 km2 respectively including the NSD of 6ndash10 due toinstrumental uncertainty This indicated that the rainfall ratefrom single gauges may differ significantly from the actualrainfall rate due to the variability of the rain fields and theinstrumental uncertainty in short accumulation times
43 Verification of Rainfall Retrieval The rainfall rate fromthe NIMS-XPOL radar measurements was retrieved withtwo rainfall rate estimators 119877(119885
119867) and 119877(119885
119867 119885DR) The
estimated rainfall rate was verified through comparison withthe rainfall rate (119866) from the rain gauge data These twoestimators were applied before and after the attenuation cor-rection First 119877(038 km2) was compared with 119877(007 km2)at 10min accumulation (Figure 17) Using 119877(119885
119867) the NB
NSD rmse and r rmse without the attenuation correctionwere significantly large at 70 127 127 (mmhr) and78 respectively However they were reduced to minus152 521 (mmhr) and 32 respectively after applying theattenuation correction Using 119877(119885
119867 119885DR) the NB was just
3 NSD was 49 rmse was 496 (mmhr) and r rmsewas 31 before applying attenuation correction This wassimilar to the result from 119877(119885
119867) after attenuation correction
However there was no significant change in119877(119885119867 119885DR)with
attenuation correction This means that 119877(119885119867 119885DR) was less
sensitive to rain attenuation than was 119877(119885119867) because 119885
119867is
proportional to 119877 and 119885DR is inversely proportional to 119877
Advances in Meteorology 11
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 71∘
minus10
minus5
0
5
10
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 160∘
minus10
minus5
0
5
10
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 334∘
minus10
minus5
0
5
10
(c)
Figure 10 Range profiles of measured (line) calibration bias 120588HV threshold and isolated point removing (dot) and attenuation corrected(dash) 119885DR at 71∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
in this rain estimator (see Table 3) Furthermore comparingthe analysis of variance results between areal (038 km2) andpoint (007m2) values from the gauge data (Section 42)half of the errors after attenuation correction were producedby the small scale variability of the rain fields and theinstrumental uncertainty of the rain gauge
A similar comparison is shown in Figure 18 for a 60minaccumulation Using 119877(119885
119867) the NB NSD rmse and r rmse
were decreased from minus65 86 71 (mmhr) and 70to minus5 34 28 (mmhr) and 28 by the attenuationcorrection respectively Using 119877(119885
119867 119885DR) the NB NSD
rmse and r rmse were only slightly decreased after theattenuation correction However the NSD rmse and r rmsewere reduced when compared at 10min accumulation In thesame accumulation time the errors in the gauge data reachedto over 15
This comparison was affected by different sampling areasof the gauges and the radar Therefore the areal averagevalues were compared (Figure 19) Using119877(119885
119867)with the areal
averaged radar measurements the NB was reduced fromminus69 to 7 NSD was reduced from 122 to 42 rmsewas reduced from 39 (mmhr) to 13 (mmhr) and r rmse
was reduced from 77 to 26 at an accumulation time of10min Using 119877(119885
119867 119885DR) the NB NSD rmse and r rmse
were decreased fromminus24 67 21 (mmhr) and 42 to 539 12 (mmhr) and 24 respectively
TheNB and r rmsewithmatched sampling areas betweenthe gauge and the radar are shown in Figures 20 and 21 Thereduction of NB due to temporal accumulation was smallin both of the rain estimators The NB was reduced to lessthan 7 by attenuation correction for all accumulation timeswith both119877(119885
119867) and119877(119885
119867 119885DR)The attenuation correction
effect was significantly greater in119877(119885119867) Similarly the r rmse
was reduced significantly in 119877(119885119867) by attenuation correction
(from 68 to 25 at a 60min accumulation) The reductionof the r rmse due to temporal accumulation was larger in119877(119885119867 119885DR) because the significant observational noise of
119885DR was smoothed The r rmse was decreased from 18to 15 in 119877(119885
119867 119885DR) at a 60min accumulation time after
attenuation correction
5 Conclusions
In this study an attenuation correction algorithm based onΦDP was developed to improve the accuracy of the rainfall
12 Advances in Meteorology
1st gauge site 2nd gauge site
3rd gauge site 4th gauge site
5th gauge site 6th gauge site
7th gauge site 8th gauge site
9th gauge site 10th gauge site
17 18 19 20 2116Time (h)
17 18 19 20 2116Time (h)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
UncorrectedCorrected
UncorrectedCorrected
Figure 11 Time series of measured (red circle) and attenuation corrected (blue circle) 119885119867at the 10 rain gauge sites
Advances in Meteorology 13
UncorrectedCorrected
UncorrectedCorrected
17 18 19 20 2116Time (h)
17 18 19 20 2116Time (h)
1st gauge site 2nd gauge site
3rd gauge site 4th gauge site
5th gauge site 6th gauge site
7th gauge site 8th gauge site
9th gauge site 10th gauge site
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
Figure 12 Time series of measured (red circle) and attenuation corrected (blue circle) 119885DR at the 10 rain gauge sites
14 Advances in Meteorology
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Uncorrected(a)
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Corrected Simulated from DSD(b)
Figure 13 Scatter plots of 119885119867and 119885DR before (a) and after (b) attenuation correction at the 10 rain gauge sites DSD observation results (dot
[21])
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
04
05
06
07
08
09
10
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
Accum time = 1min
Accum time = 10min
Accum time = 2min
Accum time = 20min
Accum time = 5min
Accum time = 60min
Figure 14 The Spearman correlation coefficient 119903 between data from two rain gauges at different accumulation times
estimation using X-band radar This algorithm was evalu-ated with radar measurements from NIMS-XPOL Howevermeasured ΨDP contained 120575 and observation noise as wellas ΦDP Furthermore the ΨDP offset varied with azimuthangles and decreased near the radar when rain cells were
above the radar In this study ΦDP was estimated by aniterative filtering technique and by adjusting with an offsetdetermined as the minimum value ofΦDP within 3 km of theradar As a result estimated ΦDP monotonically increasedThe path integrated value of119860
119867and119860DP was calculated from
Advances in Meteorology 15
2nd 6th 9th gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(038km2) versus G
(a) Normalized bias
⟨G⟩(038km2) versus G
2nd 6th 9th gaugesMean NSD
Mean R
0
20
40
60
80
100
NSD
()
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 15 (a) The NB and (b) NSD between ⟨119866⟩(038 km2) and 119866 from three rain gauges with accumulation time ⟨119866⟩ is the temporallyaveraged rainfall rate of ⟨119866⟩
All gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(6km2) versus G
(a) Normalized bias
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
0
20
40
60
80
100
NSD
()
All gaugesMean NSD
Mean R
⟨G⟩(6km2) versus G
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 16 (a) The NB and (b) NSD between ⟨119866⟩(6 km2) and 119866 from 10 rain gauges with accumulation time ⟨119866⟩ is the temporally averaged⟨119866⟩ with accumulation time
estimated ΦDP using 119860119867-ΦDP and 119860DP-ΦDP relationships
derived from the DSDs data Measured 119885119867
and 119885DR werecorrected with the calculated path integrated 119860
119867and 119860DP
by each gate The attenuated signal patterns of 119885119867and 119885DR
were restored to a typical rain echo pattern However theattenuation corrected 119885
119867and 119885DR fields were not matched
at distant ranges This was likely because of contaminationdue to beam broadening and the accumulation of errorswith increasing range Comparing the 119885
119867-119885DR scatter plots
over the 10 rain gauge sites with the results obtained froma scattering simulation with the DSD data the attenuation
corrected 119885119867-119885DR scatter plots are in good agreement with
the 119885119867-119885DR scatter plots from the DSDs
The validation for the improvement of accuracy in therainfall estimation through attenuation correction was per-formed by comparing with dense rain gauge network data119877(119885119867) from the attenuated (observed)119885
119867produced a severe
underestimation with an NB NSD rmse and r rmse of70 127 127 (mmhr) and 78 at 10min accumulationrespectively while 119877(119885
119867) from the attenuation corrected 119885
119867
showed better agreement with the gauge measurements withan NB NSD rmse and r rmse of minus1 52 521 (mmhr)
16 Advances in Meteorology
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus070
NSD = 127
rmse = 1279
corr = 0858
Corrected
NB = minus001
NSD = 052
rmse = 521
corr = 0927
r_rmse = 0786 r_rmse = 0320
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = 003
NSD = 049
rmse = 496
r_rmse = 0305
corr = 0936
Corrected
NB = 001
NSD = 050
rmse = 502
r_rmse = 0308
corr = 0936
(b) 119877(119885119867119885DR)
Figure 17 The scatter plots of 119877(038 km2 10min) versus 119866(007m2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
CorrectedUncorrected
R(m
mh
)
0
10
20
30
40
10 20 30 400G (mmh)
Uncorrected
NB = minus065
NSD = 086
rmse = 706
r_rmse = 0702
corr = 0878
Corrected
NB = minus005
NSD = 034
rmse = 281
r_rmse = 0279
corr = 0866
(a) 119877(119885119867)
CorrectedUncorrected
R(m
mh
)
10 20 30 400G (mmh)
0
10
20
30
40
Uncorrected
NB = 008
NSD = 023
rmse = 189
r_rmse = 0188
corr = 0942
Corrected
NB = 007
NSD = 025
rmse = 205
r_rmse = 0203
corr = 0940
(b) 119877(119885119867119885DR)
Figure 18 The scatter plots of 119877(038 km2 60min) versus 119866(007m2 60min) (a) using 119885119867and (b) using 119885
119867and 119885DR
and 32 at 10min accumulation respectively At 60minaccumulation theNBNSD rmse and r rmsewere decreasedfrom minus65 86 71 (mmhr) and 70 to minus5 3428 (mmhr) and 28 respectively Using 119877(119885
119867 119885DR) the
NB NSD rmse and r rmse were only slightly reduced even if
attenuation was corrected but were smaller than when using119877(119885119867) This indicated that 119877(119885
119867 119885DR) was less sensitive to
the attenuation than 119877(119885119867) In addition half of the errors
that remained after attenuation correction were caused by thesmall scale variability of the rain fields and the instrumental
Advances in Meteorology 17
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus069
NSD = 122
rmse = 391
r_rmse = 0770
corr = 0893
Corrected
NB = 007
NSD = 042
rmse = 133
r_rmse = 0262
corr = 0960
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = minus024
NSD = 067
rmse = 215
r_rmse = 0424
corr = 0718
Corrected
NB = 005
NSD = 039
rmse = 124
r_rmse = 0245
corr = 0891
(b) 119877(119885119867119885DR)
Figure 19 The scatter plots of 119877(6 km2 10min) versus 119866(6 km2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
R(ZH)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(a) 119877(119885119867)
R(ZH ZDR)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(b) 119877(119885119867119885DR)
Figure 20 The normalized bias of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and 119885DR The red
point-line and the blue point-line indicate without and with the attenuation correction respectively
uncertainty of the rain gauges Using an areal averagedvalue the NB was reduced to less than 7 with attenua-tion correction using either rainfall estimator regardless ofaccumulation timeThe r rmse were significantly reduced byattenuation correction in 119877(119885
119867) and the reduction of r rmse
with 119877(119885119867 119885DR) was even more dramatic
As summarized above the rain attenuation for 119885119867
and 119885DR in the X-band was corrected by ΦDP and theresults showed a good agreement with the value obtainedby scattering simulations with the DSD datasets The resultsof comparison with rain gauge data also showed a signifi-cant improvement in the accuracy of rainfall estimation by
18 Advances in Meteorology
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10r_
rmse
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH)
(a) 119877(119885119867)
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10
r_rm
se
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH ZDR)
(b) 119877(119885119867119885DR)
Figure 21 The relative root-mean square error of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and
119885DR The red point-line and the blue point-line indicate without and with the attenuation correction respectively
attenuation correction which supports the quantitative useof the X-band radar in rain data retrieval
Appendix
A Instrumental Uncertainty of the Rain Gauge
The instrumental uncertainty of the TB rain gauge is asso-ciated with improper calibration leakage and precipitationmissing the gauge due to the measurement principle andwetting and evaporation losses In addition external sourcessuch as wind turbulence and installed position also con-tribute to the instrumental uncertainty Ciach [17] reportedthat the TB gauge data suffers a significant uncertainty asso-ciated with random differences between closely collocatedTB gauges through the analysis of the error as a functionof the rainfall intensity for three timescales In addition theimproper calibration induces a bias in the rainfall measure-ment and this bias cannot be removed by increasing accu-mulation timeTherefore the proper correction or processingof TB gauge data is necessary to obtain accurate and reliablereading of the rainfall amount
In this appendix instrumental biases were calculatedto reduce the instrumental uncertainty through an idealexperiment in the laboratory and a field intercomparisonwith 10 TB rain gauges In addition the random errors werequantitatively calculated as a function of rainfall rate andaccumulation time
A1 Correction of Instrumental Bias The instrumental bias119861119868(119894) of the 119894th gauge was defined as a combination
of the absolute bias 119861119860(ref) of the reference gauge and
the relative bias 119861119877(119894) of the 119894th gauge according to the
following equation
119861119868 (119894) = 119861119860 (ref) times 119861119877 (119894) (A1)
119861119860(ref) signified the ratio of the calculated value from
the instrumental resolution of the reference gauge to the truevalue measured by other instruments
119861119860 (ref) =
rainfalltruthrainfallgauge (ref)
(A2)
In this study 119861119860(ref) was derived by an ideal experiment
using an electronic scale First the reference gauge was puton the scale and the scale was initialized The water dropsflowed out into the receiving orifice of the reference gaugeuntil the tipping number became 60Themeasured weight of60 tips was 2144 g while the theoretical (or expected) valuewas 2244 g for the reference gauge (receiving orifice size of1545 cm 60 tips and 02mm tip resolution) Therefore theabsolute bias of the reference gauge from (A2) was about0955 This indicates that the rainfall amount measured bythis reference gauge was overestimated by about 45 andthis bias should be corrected prior to any interpretation
119861119877(119894) represented the ratio of the observed (recorded)
value from the 119894th gauge to the observed (recorded) valuefrom the reference gauge
119861119877 (119894) =
rainfallgauge (ref)rainfallgauge (119894)
(A3)
119861119877(119894) was calculated from ten collocated rain gauges
including the reference gauge located at an observation field
Advances in Meteorology 19
Table 4 The instrumental and relative biases of the 10 tipping bucket rain gauges
Gauge ID RG1 RG2 RG3 RG4 RG5 RG6 RG7 RG8 RG9 RG10Instrumental bias 094 108 099 100 096 104 098 102 108 098Relative bias 098 113 103 105 100 109 102 107 113 102
013610152230406080120150200250300400
()
Rain
rate
(mm
h)
100101
Accumulation time (min)
10
100NSD ()
Figure 22 Two-dimensional normalized standard deviationbetween the standard rainfall and each rain gauge
at KNUKoreaThe observational period for intercomparisonwas from October 2009 to June 2010 (14 rainy days) andthe total accumulated rainfall for this period was about300mm The results are presented in Table 4 All relativebiases were mainly larger than 10 This indicated that themost of gauges underestimated the rainfall amount comparedto the reference rain gauge The relative bias of the referencegauge (119894 = 5) was 10
Finally 119861119860(119894) was calculated from (A1) (see Table 4) All
instrumental biases were between 090 and 11 and the meanvalue of the biases was nearly 10 This indicated that therainfall rate from a single gauge data had a maximum 10bias These biases were corrected for each gauge to removeexisting instrumental errors
A2 Quantification of Random Errors To quantify the ran-dom error due to instrumental uncertainty the NSD of theten gauges was calculated with accumulation time and rainrate using data from an intercomparison The NSD of the 119894thgauge compared to the reference gauge was derived for 20rain rate classes (Δ119877) and 17 accumulation time classes (Δ119905)as follows
NSD119894 (Δ119905 Δ119877)
=
(1119899)radicsum119899minus1
119896=0[1198661015840
119870(Δ119905 Δ119877) minus 119866119894119896 (Δ119905 Δ119877)]
2
1198661015840
(A4)
where 1198661015840 and 119866 indicate the reference and individual gaugesrespectively The overbar means the temporal average ofthe rainfall rate The NSDs of nine different gauges (exceptthe reference gauge) were averaged to obtain the two-dimensional average NSD (Figure 22) The average NSD
tended to decrease with increasing accumulation time andwith increasing rain rate The NSD was about 10ndash15when accumulation time was 10min and the rain rate was10mmhminus1 The instrumental errors were quite significant forlonger accumulation times and higher rainfall rates Theseerrors should be used as a guideline when a single gauge isused as a reference for any validation study
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper has been revised and extended fromMs Young-AOhrsquos MS thesis [25] of Kyungpook National University Thisresearch is supported by Development and Application ofCross Governmental Dual-Pol Radar Harmonization (WRC-2013-A-1) project of theWeather Radar Center KoreaMeteo-rological AdministrationThis work was funded by the KoreaMeteorological Administration Research and DevelopmentProgram under Grant KMIPA2015-1010
References
[1] V N Bringi V Chandrasekar N Balakrishnan and D SZrnic ldquoAn examination of propagation effects in rainfall onradar measurements at microwave frequenciesrdquo Journal ofAtmospheric and Oceanic Technology vol 7 no 6 pp 829ndash8401990
[2] A Huggel W Schmid and AWaldvogel ldquoRaindrop size distri-butions and the radar bright bandrdquo Journal of AppliedMeteorol-ogy vol 35 no 10 pp 1688ndash1701 1996
[3] G W Lee Errors in rain measurement by radar effect of vari-ability of drop size distributions [PhD thesis]McGill University2003
[4] J D Tuttle and R E Rinehart ldquoAttenuation correction in dual-wavelength analysesrdquo Journal of ClimateampAppliedMeteorologyvol 22 no 11 pp 1914ndash1921 1983
[5] S-G Park V N Bringi V Chandrasekar M Maki andK Iwanami ldquoCorrection of radar reflectivity and differentialreflectivity for rain attenuation at X band Part I theoretical andempirical basisrdquo Journal of Atmospheric and Oceanic Technol-ogy vol 22 no 11 pp 1621ndash1632 2005
[6] G Scarchilli E Gorgucci V Chandrasekar and T A SeligaldquoRainfall estimation using polarimetric techniques at C-bandfrequenciesrdquo Journal of Applied Meteorology vol 32 no 6 pp1150ndash1160 1993
[7] M N Anagnostou E N Anagnostou and J VivekanandanldquoCorrection for rain path specific and differential attenuation ofX-band dual-polarization observationsrdquo IEEE Transactions onGeoscience and Remote Sensing vol 44 no 9 pp 2470ndash24802006
20 Advances in Meteorology
[8] J Kalogiros M N Anagnostou E N Anagnostou M Mon-topoli E Picciotti and F S Marzano ldquoEvaluation of a newpolarimetric algorithm for rain-path attenuation correction ofX-band radar observations against disdrometerrdquo IEEE Transac-tions on Geoscience and Remote Sensing vol 52 no 2 pp 1369ndash1380 2014
[9] W-Y Chang J Vivekanandan and T-C C Wang ldquoEstimationof X-band polarimetric radar attenuation and measurementuncertainty using a variational methodrdquo Journal of AppliedMeteorology and Climatology vol 53 no 4 pp 1099ndash1119 2014
[10] J Hubbert and V N Bringi ldquoAn iterative filtering technique forthe analysis of copolar differential phase and dual-frequencyradar measurementsrdquo Journal of Atmospheric and OceanicTechnology vol 12 no 3 pp 643ndash648 1995
[11] K V Mishra W F Krajewski R Goska et al ldquoDeploy-ment and performance analyses of high-resolution iowa XPOLradar system during the NASA IFloodS campaignrdquo Journal ofHydrometeorology vol 17 no 2 pp 455ndash479 2016
[12] M Thurai G J Huang V N Bringi W L Randeu and MSchonhuber ldquoDrop shapes model comparisons and calcu-lations of polarimetric radar parameters in rainrdquo Journal ofAtmospheric and Oceanic Technology vol 24 no 6 pp 1019ndash1032 2007
[13] M D Humphrey J D Istok J Y Lee J A Hevesi and AL Flint ldquoA new method for automated dynamic calibration oftipping-bucket rain gaugesrdquo Journal of Atmospheric andOceanicTechnology vol 14 no 6 pp 1513ndash1519 1997
[14] E HabibW F Krajewski V Nespor andA Kruger ldquoNumericalsimulation studies of rain gage data correction due to windeffectrdquo Journal of Geophysical ResearchAtmospheres vol 104 no16 Article ID 1999JD900228 pp 19723ndash19733 1999
[15] G J Ciach and W F Krajewski ldquoOn the estimation of radarrainfall error variancerdquoAdvances inWater Resources vol 22 no6 pp 585ndash595 1999
[16] E Habib W F Krajewski and A Kruger ldquoSampling errors oftipping-bucket rain gaugemeasurementsrdquo Journal of HydrologicEngineering vol 6 no 2 pp 159ndash166 2001
[17] G J Ciach ldquoLocal random errors in tipping-bucket rain gaugemeasurementsrdquo Journal of Atmospheric andOceanic Technologyvol 20 no 5 pp 752ndash759 2003
[18] EHabibW F Krajewski andG J Ciach ldquoEstimation of rainfallinterstation correlationrdquo Journal ofHydrometeorology vol 2 no6 pp 621ndash629 2001
[19] NIMS ldquoStudy on estimation of rainfall amount and classifi-cation of hydrometeor species using dual-polarimetric radar(III)rdquo pp 1ndash20 2010
[20] E Gorgucci V Chandrasekar and L Baldini ldquoCorrection ofX-band radar observation for propagation effects based on theself-consistency principlerdquo Journal of Atmospheric and OceanicTechnology vol 23 pp 1668ndash1681 2006
[21] NIMS ldquoStudy on estimating raindrop size distribution forquantitative estimation of radar rainfall amountrdquo pp 1ndash55 2009
[22] NIMS ldquoQuantitative precipitation estimation and hydrometeoridentification using dual-polarization radar-Phase IIrdquo pp 1ndash742009
[23] S-G Park M Maki K Iwanami V N Bringi and V Chan-drasekar ldquoCorrection of radar reflectivity and differential reflec-tivity for rain attenuation at X band Part II Evaluation andapplicationrdquo Journal of Atmospheric and Oceanic Technologyvol 22 no 11 pp 1633ndash1655 2005
[24] E Habib and W F Krajewski ldquoUncertainty analysis of theTRMM ground-validation radar-rainfall products applicationto the TEFLUN-B field campaignrdquo Journal of Applied Meteorol-ogy vol 41 no 5 pp 558ndash572 2002
[25] Y-A Oh Rainfall estimation from X-band dual-polarizationradar measurements effects of attenuation correction and evalu-ation with a dense rain gauge network [MS thesis] KyungpookNational University 2011
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom
Applied ampEnvironmentalSoil Science
Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal of
Geophysics
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
6 Advances in Meteorology
Table 3 Two rain rate retrieval algorithms and boundary conditions [22]
119877 (119885119867) = 120572119885
120573
119867119877 (119885DR) = 120572119885
120573
1198671001120574119885DR
Z gt 35 dBZ Z lt 35 dBZ Z gt 35 dBZ 119885DR gt 03 dB Z lt 35 dBZ 119885DR gt 03 dB120572 353 times 10
minus2383 times 10
minus2103 times 10
minus2145 times 10
minus2
120573 621 times 10minus2
608 times 10minus2
985 times 10minus1
890 times 10minus1
120574 mdash mdash minus6479 minus5295
measurement resolution 119866(6 km2) was the average value of119866(007m2) from the ten gauges within the same area
Estimated 119877 from radar parameters was calculated withtwo rain estimators as shown in Table 3These rain estimatorswere derived by a least squares fit with the theoretical 119885
119867
119885DR and 119877 obtained by scattering simulations with theDSD data collected by POSS from March to Septemberof 2001 in Pusan Korea [21] To improve the estimationaccuracy each estimator was derived for two categoriesThe reflectivity threshold of 35 dBZ was used as a stan-dard stratiformconvective rain classification [22] The 119885DRthreshold of 03 dB was used to minimize the effect of the119885DR measurement noiseThePPI data with an elevation angleof 3∘ was used to suppress the beam shielding and groundclutterThe119885
119867and119885DR fields from the radar were smoothed
by the linear average of values at 3 times 3 pixels to reduce theobservational noise
The bias in 119866 from the ten rain gauges was corrected bythe laboratory test and the intercomparison with field dataThe detail is presented in Appendix To evaluate the atten-uation correction algorithm and verify the rainfall retrievalaccuracy the comparisons of119877 and119866were represented by thegeneral statistics such as normalized bias (NB) normalizedstandard deviation (NSD) root-mean square error (rmse)relative rms error (r rmse) and Spearman correlation (corr)as follows
NB119877-119866 (Δ119905) =
(1119899)sum119899minus1
119896=0[119877119896 (Δ119905) minus 119866119896 (Δ119905)]
119866 (Δ119905)
NSD119877-119866 (Δ119905) =
radic(1119899)sum119899minus1
119896=0[119877119896 (Δ119905) minus 119866119896 (Δ119905)]
2
119866 (Δ119905)
rmse119877-119866 (Δ119905) =
radicsum119899minus1
119896=0[119877119896 (Δ119905) minus 119866119896 (Δ119905)]
2
119899
rmse119877-119866 (Δ119905) = radic
sum119899minus1
119896=0[119877119896 (Δ119905) minus 119866119896 (Δ119905)]
2
sum119899minus1
119896=0[119866119896 (Δ119905)]
2
corr119877-119866 (Δ119905)
=
sum119899minus1
119896=0[119877119894119896 (Δ119905) minus 119877119894 (Δ119905)] [119866119895119896 (Δ119905) minus 119866119895 (Δ119905)]
radicsum119899minus1
119896=0[119877119894119896 (Δ119905) minus 119877119894 (Δ119905)]
2radicsum119899minus1
119896=0[119866119895119896 (Δ119905) minus 119866119895 (Δ119905)]
2
(6)
where 119899 is the number of 119877 and 119866 pairs at a given temporalintegrationΔ119905 and the overbar indicates the temporal average
33 Spatial Correlation and Variability of the Rain FieldsRain gauges are widely used to measure rainfall amountsdue tomany advantages including easy installation mobilitylow price and ease of data processing The rain gauge datais used as a reference for the ground rainfall amount forthe evaluation of rain retrieval algorithms adjustment ofother instruments data assimilation and other hydrologicalapplications However it is affected by instrumental errorsand representativeness errors due to the variability of rainfields
In a TB rain gauge the instrumental errors are due toimproper calibration the time delay of the tips leakage dueto its measuring principle and so forthThis error is exploredin the Appendix
The representativeness errors are subdivided into spatialand temporal representativeness They are mainly caused byhaving a small sampling area and a limited sampling time dueto the gauge measuring principle The actual rain fields varywith different spatiotemporal scales that are not equivalentto the spatial and temporal resolution of the rain gaugeTherefore the rain gauge data represent the actual rain fieldsin a very limited manner
These errors should be considered and minimized priorto using rain gauge data as ground truth Furthermore thedegree of these errors has to be understood quantitativelysince there can be other error sources in various applicationsIn this section the spatial and temporal variability of the rainfields are investigated using the rain data collected by the tenrain gauges for the month of July 2010 at Muan Korea
First the Spearman correlation coefficient 119903119894119895(Δ119905)
between a pair of 119866 values from the 119894th and 119895th individualrain gauge with integration time Δ119905 was calculated fromthe following equation to obtain the Spearman correlationcoefficient of the rain fields
119903119894119895 (Δ119905)
=
sum119899minus1
119896=0[119866119894119896 (Δ119905) minus 119866119894 (Δ119905)] [119866119895119896 (Δ119905) minus 119866119895 (Δ119905)]
radicsum119899minus1
119896=0[119866119894119896 (Δ119905) minus 119866119894 (Δ119905)]
2radicsum119899minus1
119896=0[119866119895119896 (Δ119905) minus 119866119895 (Δ119905)]
2
(7)
where 119899 is the number of gauge data pairs at given Δ119905 and theoverbar indicates the temporal average of the rainfall rate Itwas derived withΔ119905 to research the spatial and temporal scaleof the variation for the rain field
In addition the effect of the natural variability of the rainfields within two different areas (038 km2 and 6 km2) wasexplored by analyzing the bias and random error between thespatially averaged rainfall rate ⟨119866⟩ and 119866 from the individual
Advances in Meteorology 7
32km
21 km
11 km
20km40km60km
minus180 18020 30 40 50 60 70 80 90 100120140160100(a)
32 km
21 km
11 km
20km40km60km
minus180 18020 30 40 50 60 70 80 90 100120140160100(b)
Figure 5 PPIs of measured (a) and filtered (b) ΨDP of the NIMS-XPOL at 2101UTC July 10 2010
gauge The NB and NSD of 119866119894with ⟨119866⟩ were calculated as
follows
NB (119894) =(sum119899minus1
119896=0[119866119894119896 (Δ119905) minus ⟨119866119896 (Δ119905)⟩]) 119899
⟨119866 (Δ119905)⟩
NSD (119894) =(1119899)radic(sum
119899minus1
119896=0[119866119894119896 (Δ119905) minus ⟨119866119896 (Δ119905)⟩]
2)
⟨119866 (Δ119905)⟩
(8)
where 119894 indicates the 119894th gauge and the overbar indicates thetemporal average
This analysis provided a reasonable spatial average andtemporal accumulation resolution for the calculation ofmoreaccurate and realistic rainfall amounts from the gauge dataand a quantitative guideline regarding the errors in rainestimation due to the variability of the rain fields
4 Results
41 Attenuation Correction of Reflectivity and DifferentialReflectivity The attenuation correction algorithm in Sec-tion 31 was applied to the NIMS-XPOL radar data ΦDPwas estimated by eliminating 120575 and the observational noisefrom measured ΨDP using an iterative FIR filter The offsetof ΦDP was then adjusted Figure 5 shows the PPIs ofmeasured ΨDP and filtered ΦDP After applying the FIR filterthe small scale variation due to 120575 and observational noisein measured ΨDP was reduced and filtered ΨDP (estimatedΦDP) monotonically increased with an increase in rangeThe maximum value of estimated ΦDP reached to 140∘ Inaddition estimated ΦDP at the azimuth angles of 150∘ and200∘ sharply increased with the loss of signal beyond thesesharp gradient areas This reflected that strong convective
cells existed in these directions and that these cells caused thesevere signal attenuation EstimatedΦDP values between 250
∘
and 360∘ azimuth angle decreased after filtering since ΦDPoffsets were positive values in these directions Figure 6 showsthe range profile of measured and filtered ΨDP for estimationof ΦDP at 71
∘ 160∘ and 334∘ azimuth angles The unexpectedlarge values near the radarwere removed by the120588HV threshold(red line)The small fluctuation inmeasuredΨDP due to 120575 andthe observational noise was effectively removed after filteringThe value of estimated ΦDP was adjusted to 0 by the offset(blue line)
Estimated ΦDP was used to calculate the path integrated119860119867 and the measured119885
119867values were corrected as shown in
Figure 7 Figure 7(a) represents measured-attenuated119885119867and
relatively small values of 119885119867at far ranges This may indicate
that the 119885119867values at far ranges were affected by the strong
attenuation due to strong convective cells near the radarThisspatial pattern of 119885
119867was the typical feature of the attenuated
radar signal due to the rain After correcting the attenuationthe value of the 119885
119867field became larger over the entire area
and the radially attenuated 119885119867
was corrected by showinga typical rain cell pattern in 119885
119867(Figure 7(b)) The largest
difference between the uncorrected and the corrected 119885119867at
2101UTCwas over 50 dB However the white area behind thestrong echoes was not detected as there was a complete loss ofthe signal powerThese regions could not be corrected by thisalgorithmThe circle pattern of119885
119867fields within about a 4 km
range from theNIMS-XPOL could not present the actual rainechoes due to the instrumental issues (the same was true for119885DR) These results are also presented in the range profiles inFigure 8
Similarly measured 119885DR were corrected with the pathintegrated 119860DP derived from the path integrated 119860
119867using
(4) (see Figure 9) Before applying attenuation correction
8 Advances in Meteorology
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
0
50
100
150
200
ΦDP (∘) azimuth angle = 71∘
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
0
50
100
150
200
ΦDP (∘) azimuth angle = 160∘
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
0
50
100
150
200
ΦDP (∘) azimuth angle = 334∘
(c)
Figure 6 Range profiles of measured (black) 120588HV threshold isolated point removing (red) and FIR filter and offset adjusted (blue) ΨDP forestimation of ΦDP at 71
∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
32km
21 km
11 km
20km40km60km
minus10 10 15 20 25 30 35 40 45 50 55 60 65 700(a)
32 km
21 km
11 km
20km40km60km
minus10 10 15 20 25 30 35 40 45 50 55 60 65 700(b)
Figure 7 PPIs of measured (a) and attenuation corrected (b) 119885119867of the NIMS-XPOL for the case in Figure 5
Advances in Meteorology 9
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 71∘
minus20
0
20
40
60
80
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 160∘
minus20
0
20
40
60
80
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 334∘
minus20
0
20
40
60
80
(c)
Figure 8 Range profiles of measured (line) calibration bias 120588HV threshold and isolated point removing (dot) and attenuation corrected(dash) 119885
119867at 71∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
the attenuation signature in119885DR was determined Ideally the119885DR value for rain should be positive since rain drops areoblate due to aerodynamic force However themeasured119885DRfield in Figure 9 showed a negative value over wide areasAfter correcting the attenuation the119885DR field was larger than0 dB and showed a realistic rain patternThis is also presentedin the range profile of Figure 10The largest difference of119885DRbetween before and after attenuation correction at 2101UTCwas about 8 dB The 119885DR values around the 220∘ azimuthangle beyond 40 km were the largest They did not matchwith the area of the largest 119885DR values (around the 210∘azimuth angle with a range of 20 km and 40 km) This wasan unexpected result because 119885DR is proportional to 119885
119867in
theory This mismatch between 119885119867
and 119885DR at far rangemay be due to contamination due to beam broadening andaccumulated errors due to uncertainty in the attenuationcorrection with different ranges
Figures 11 and 12 show the time series of 119885119867
and119885DR at the ten gauge sites before and after applying theattenuation correction The differences between uncorrectedand corrected values existed after 1800UTC and were at amaximumaround 2100UTCThis attenuation correctionwas
evaluated through comparison with the theoretical values of119885119867and 119885DR In Figure 13 the uncorrected 119885
119867-119885DR scatter
plots are apart from the theoretical119885119867-119885DR values [23] After
attenuation correction the119885119867-119885DR scatter plots were close to
the calculated valuesThis indicated that119885119867and119885DR over the
gauge sites were corrected with reliable values
42 Variability of the Rain Fields The Spearman correlationcoefficient of the rainfall rate from the 10 spatially dis-tributed rain gauges was calculated with accumulation timeand separation distance (Figure 14) The overall Spearmancorrelation coefficient decreased with increasing separationdistance at all accumulation timesThis tendency was similarto the result from Habib and Krajewski [24] However thecorrelation value was higher in this study Note that the gaugetip resolution in Habib and Krajewski [24] was 0254mmwhile it was 02mm in this study In addition the rainfall ratesfrom the gauge data used in this study mainly consisted ofmoderate rainfall rates while Habib and Krajewski [24] useddata containing heavy rainfall rates
For accumulation times of less than 10min the corre-lation at the same distance was dramatically increased with
10 Advances in Meteorology
32km
21 km
11 km
20km40km60km
18
14
00
02
04
06
26
40
minus10
22
minus20
30
10
50
minus30
(a)
32 km
21 km
11 km
20km40km60km
18
14
00
02
04
06
26
40
minus10
22
minus20
30
10
50
minus30
(b)
Figure 9 PPIs of measured (a) and attenuation corrected (b) 119885DR of NIMS-XPOL for the case in Figure 5
increasing accumulation time and the slope of the correlationfunction was steep This was due to the reduction of theinstrumental errors of the gauge data by time accumulationThis indicated that rainfall accumulated for less than 10minfrom single gauge data cannot represent the rainfall over itssurrounding area due to instrumental errors For accumula-tion times longer than 10min the slope became less steepThis reflected the remaining variability of the rain field Inaddition the accumulated rainfall rate of about 10min wassuitable for minimizing the instrumental uncertainty whilepreserving the variation in the rain field The correlationreached about 098 at 05 km in 10min of accumulationand this approached 098 at 10 km at 60min accumulationTherefore a single rain gauge mostly represented the rainfallrate within 05 km and 10 km with 10min and 60minaccumulation respectively
The NB between an areal averaged rainfall rate ⟨119866⟩ anda single gauge rainfall rate 119866 is shown with accumulationtime in Figures 15(a) and 16(a) Where the average area was038 km2 (corresponding to 9 pixels of NIMS-XPOL radarat 16 km range) the NB reached a 6ndash8 maximum Wherethe averaged area is 6 km2 (corresponding to the coverage forall of the gauges) the NB reached about a 10 maximumFurthermore the NB hardly changed with accumulationtime Note that this statistic was derived after removinginstrumental bias This indicated that the rainfall rate froma single gauge was biased due to the spatial variability of therain fields and this bias was maintained with accumulationtime
The NSD of a single gauge for a mean area of 038 km2was over 40 and it reached about 70ndash90 for a mean areaof 6 km2 at an accumulation time of 1min (see Figures 15(b)and 16(b)) This statistic included instrumental errors (about
40ndash60 at 16mmhminus1 and 60ndash80 at 10mmhminus1 as shownin the Appendix) as well as the spatial variability of the rainfields This indicated that the variance between the point andarea values from the gauge data at short accumulation timeswas similar to the errors due to the instrumental uncertaintyof the rain gaugeTheNSD in 60min accumulation decreasedto over 10 and about 15ndash45 with a mean area of 038 km2and 6 km2 respectively including the NSD of 6ndash10 due toinstrumental uncertainty This indicated that the rainfall ratefrom single gauges may differ significantly from the actualrainfall rate due to the variability of the rain fields and theinstrumental uncertainty in short accumulation times
43 Verification of Rainfall Retrieval The rainfall rate fromthe NIMS-XPOL radar measurements was retrieved withtwo rainfall rate estimators 119877(119885
119867) and 119877(119885
119867 119885DR) The
estimated rainfall rate was verified through comparison withthe rainfall rate (119866) from the rain gauge data These twoestimators were applied before and after the attenuation cor-rection First 119877(038 km2) was compared with 119877(007 km2)at 10min accumulation (Figure 17) Using 119877(119885
119867) the NB
NSD rmse and r rmse without the attenuation correctionwere significantly large at 70 127 127 (mmhr) and78 respectively However they were reduced to minus152 521 (mmhr) and 32 respectively after applying theattenuation correction Using 119877(119885
119867 119885DR) the NB was just
3 NSD was 49 rmse was 496 (mmhr) and r rmsewas 31 before applying attenuation correction This wassimilar to the result from 119877(119885
119867) after attenuation correction
However there was no significant change in119877(119885119867 119885DR)with
attenuation correction This means that 119877(119885119867 119885DR) was less
sensitive to rain attenuation than was 119877(119885119867) because 119885
119867is
proportional to 119877 and 119885DR is inversely proportional to 119877
Advances in Meteorology 11
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 71∘
minus10
minus5
0
5
10
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 160∘
minus10
minus5
0
5
10
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 334∘
minus10
minus5
0
5
10
(c)
Figure 10 Range profiles of measured (line) calibration bias 120588HV threshold and isolated point removing (dot) and attenuation corrected(dash) 119885DR at 71∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
in this rain estimator (see Table 3) Furthermore comparingthe analysis of variance results between areal (038 km2) andpoint (007m2) values from the gauge data (Section 42)half of the errors after attenuation correction were producedby the small scale variability of the rain fields and theinstrumental uncertainty of the rain gauge
A similar comparison is shown in Figure 18 for a 60minaccumulation Using 119877(119885
119867) the NB NSD rmse and r rmse
were decreased from minus65 86 71 (mmhr) and 70to minus5 34 28 (mmhr) and 28 by the attenuationcorrection respectively Using 119877(119885
119867 119885DR) the NB NSD
rmse and r rmse were only slightly decreased after theattenuation correction However the NSD rmse and r rmsewere reduced when compared at 10min accumulation In thesame accumulation time the errors in the gauge data reachedto over 15
This comparison was affected by different sampling areasof the gauges and the radar Therefore the areal averagevalues were compared (Figure 19) Using119877(119885
119867)with the areal
averaged radar measurements the NB was reduced fromminus69 to 7 NSD was reduced from 122 to 42 rmsewas reduced from 39 (mmhr) to 13 (mmhr) and r rmse
was reduced from 77 to 26 at an accumulation time of10min Using 119877(119885
119867 119885DR) the NB NSD rmse and r rmse
were decreased fromminus24 67 21 (mmhr) and 42 to 539 12 (mmhr) and 24 respectively
TheNB and r rmsewithmatched sampling areas betweenthe gauge and the radar are shown in Figures 20 and 21 Thereduction of NB due to temporal accumulation was smallin both of the rain estimators The NB was reduced to lessthan 7 by attenuation correction for all accumulation timeswith both119877(119885
119867) and119877(119885
119867 119885DR)The attenuation correction
effect was significantly greater in119877(119885119867) Similarly the r rmse
was reduced significantly in 119877(119885119867) by attenuation correction
(from 68 to 25 at a 60min accumulation) The reductionof the r rmse due to temporal accumulation was larger in119877(119885119867 119885DR) because the significant observational noise of
119885DR was smoothed The r rmse was decreased from 18to 15 in 119877(119885
119867 119885DR) at a 60min accumulation time after
attenuation correction
5 Conclusions
In this study an attenuation correction algorithm based onΦDP was developed to improve the accuracy of the rainfall
12 Advances in Meteorology
1st gauge site 2nd gauge site
3rd gauge site 4th gauge site
5th gauge site 6th gauge site
7th gauge site 8th gauge site
9th gauge site 10th gauge site
17 18 19 20 2116Time (h)
17 18 19 20 2116Time (h)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
UncorrectedCorrected
UncorrectedCorrected
Figure 11 Time series of measured (red circle) and attenuation corrected (blue circle) 119885119867at the 10 rain gauge sites
Advances in Meteorology 13
UncorrectedCorrected
UncorrectedCorrected
17 18 19 20 2116Time (h)
17 18 19 20 2116Time (h)
1st gauge site 2nd gauge site
3rd gauge site 4th gauge site
5th gauge site 6th gauge site
7th gauge site 8th gauge site
9th gauge site 10th gauge site
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
Figure 12 Time series of measured (red circle) and attenuation corrected (blue circle) 119885DR at the 10 rain gauge sites
14 Advances in Meteorology
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Uncorrected(a)
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Corrected Simulated from DSD(b)
Figure 13 Scatter plots of 119885119867and 119885DR before (a) and after (b) attenuation correction at the 10 rain gauge sites DSD observation results (dot
[21])
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
04
05
06
07
08
09
10
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
Accum time = 1min
Accum time = 10min
Accum time = 2min
Accum time = 20min
Accum time = 5min
Accum time = 60min
Figure 14 The Spearman correlation coefficient 119903 between data from two rain gauges at different accumulation times
estimation using X-band radar This algorithm was evalu-ated with radar measurements from NIMS-XPOL Howevermeasured ΨDP contained 120575 and observation noise as wellas ΦDP Furthermore the ΨDP offset varied with azimuthangles and decreased near the radar when rain cells were
above the radar In this study ΦDP was estimated by aniterative filtering technique and by adjusting with an offsetdetermined as the minimum value ofΦDP within 3 km of theradar As a result estimated ΦDP monotonically increasedThe path integrated value of119860
119867and119860DP was calculated from
Advances in Meteorology 15
2nd 6th 9th gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(038km2) versus G
(a) Normalized bias
⟨G⟩(038km2) versus G
2nd 6th 9th gaugesMean NSD
Mean R
0
20
40
60
80
100
NSD
()
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 15 (a) The NB and (b) NSD between ⟨119866⟩(038 km2) and 119866 from three rain gauges with accumulation time ⟨119866⟩ is the temporallyaveraged rainfall rate of ⟨119866⟩
All gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(6km2) versus G
(a) Normalized bias
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
0
20
40
60
80
100
NSD
()
All gaugesMean NSD
Mean R
⟨G⟩(6km2) versus G
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 16 (a) The NB and (b) NSD between ⟨119866⟩(6 km2) and 119866 from 10 rain gauges with accumulation time ⟨119866⟩ is the temporally averaged⟨119866⟩ with accumulation time
estimated ΦDP using 119860119867-ΦDP and 119860DP-ΦDP relationships
derived from the DSDs data Measured 119885119867
and 119885DR werecorrected with the calculated path integrated 119860
119867and 119860DP
by each gate The attenuated signal patterns of 119885119867and 119885DR
were restored to a typical rain echo pattern However theattenuation corrected 119885
119867and 119885DR fields were not matched
at distant ranges This was likely because of contaminationdue to beam broadening and the accumulation of errorswith increasing range Comparing the 119885
119867-119885DR scatter plots
over the 10 rain gauge sites with the results obtained froma scattering simulation with the DSD data the attenuation
corrected 119885119867-119885DR scatter plots are in good agreement with
the 119885119867-119885DR scatter plots from the DSDs
The validation for the improvement of accuracy in therainfall estimation through attenuation correction was per-formed by comparing with dense rain gauge network data119877(119885119867) from the attenuated (observed)119885
119867produced a severe
underestimation with an NB NSD rmse and r rmse of70 127 127 (mmhr) and 78 at 10min accumulationrespectively while 119877(119885
119867) from the attenuation corrected 119885
119867
showed better agreement with the gauge measurements withan NB NSD rmse and r rmse of minus1 52 521 (mmhr)
16 Advances in Meteorology
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus070
NSD = 127
rmse = 1279
corr = 0858
Corrected
NB = minus001
NSD = 052
rmse = 521
corr = 0927
r_rmse = 0786 r_rmse = 0320
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = 003
NSD = 049
rmse = 496
r_rmse = 0305
corr = 0936
Corrected
NB = 001
NSD = 050
rmse = 502
r_rmse = 0308
corr = 0936
(b) 119877(119885119867119885DR)
Figure 17 The scatter plots of 119877(038 km2 10min) versus 119866(007m2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
CorrectedUncorrected
R(m
mh
)
0
10
20
30
40
10 20 30 400G (mmh)
Uncorrected
NB = minus065
NSD = 086
rmse = 706
r_rmse = 0702
corr = 0878
Corrected
NB = minus005
NSD = 034
rmse = 281
r_rmse = 0279
corr = 0866
(a) 119877(119885119867)
CorrectedUncorrected
R(m
mh
)
10 20 30 400G (mmh)
0
10
20
30
40
Uncorrected
NB = 008
NSD = 023
rmse = 189
r_rmse = 0188
corr = 0942
Corrected
NB = 007
NSD = 025
rmse = 205
r_rmse = 0203
corr = 0940
(b) 119877(119885119867119885DR)
Figure 18 The scatter plots of 119877(038 km2 60min) versus 119866(007m2 60min) (a) using 119885119867and (b) using 119885
119867and 119885DR
and 32 at 10min accumulation respectively At 60minaccumulation theNBNSD rmse and r rmsewere decreasedfrom minus65 86 71 (mmhr) and 70 to minus5 3428 (mmhr) and 28 respectively Using 119877(119885
119867 119885DR) the
NB NSD rmse and r rmse were only slightly reduced even if
attenuation was corrected but were smaller than when using119877(119885119867) This indicated that 119877(119885
119867 119885DR) was less sensitive to
the attenuation than 119877(119885119867) In addition half of the errors
that remained after attenuation correction were caused by thesmall scale variability of the rain fields and the instrumental
Advances in Meteorology 17
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus069
NSD = 122
rmse = 391
r_rmse = 0770
corr = 0893
Corrected
NB = 007
NSD = 042
rmse = 133
r_rmse = 0262
corr = 0960
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = minus024
NSD = 067
rmse = 215
r_rmse = 0424
corr = 0718
Corrected
NB = 005
NSD = 039
rmse = 124
r_rmse = 0245
corr = 0891
(b) 119877(119885119867119885DR)
Figure 19 The scatter plots of 119877(6 km2 10min) versus 119866(6 km2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
R(ZH)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(a) 119877(119885119867)
R(ZH ZDR)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(b) 119877(119885119867119885DR)
Figure 20 The normalized bias of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and 119885DR The red
point-line and the blue point-line indicate without and with the attenuation correction respectively
uncertainty of the rain gauges Using an areal averagedvalue the NB was reduced to less than 7 with attenua-tion correction using either rainfall estimator regardless ofaccumulation timeThe r rmse were significantly reduced byattenuation correction in 119877(119885
119867) and the reduction of r rmse
with 119877(119885119867 119885DR) was even more dramatic
As summarized above the rain attenuation for 119885119867
and 119885DR in the X-band was corrected by ΦDP and theresults showed a good agreement with the value obtainedby scattering simulations with the DSD datasets The resultsof comparison with rain gauge data also showed a signifi-cant improvement in the accuracy of rainfall estimation by
18 Advances in Meteorology
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10r_
rmse
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH)
(a) 119877(119885119867)
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10
r_rm
se
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH ZDR)
(b) 119877(119885119867119885DR)
Figure 21 The relative root-mean square error of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and
119885DR The red point-line and the blue point-line indicate without and with the attenuation correction respectively
attenuation correction which supports the quantitative useof the X-band radar in rain data retrieval
Appendix
A Instrumental Uncertainty of the Rain Gauge
The instrumental uncertainty of the TB rain gauge is asso-ciated with improper calibration leakage and precipitationmissing the gauge due to the measurement principle andwetting and evaporation losses In addition external sourcessuch as wind turbulence and installed position also con-tribute to the instrumental uncertainty Ciach [17] reportedthat the TB gauge data suffers a significant uncertainty asso-ciated with random differences between closely collocatedTB gauges through the analysis of the error as a functionof the rainfall intensity for three timescales In addition theimproper calibration induces a bias in the rainfall measure-ment and this bias cannot be removed by increasing accu-mulation timeTherefore the proper correction or processingof TB gauge data is necessary to obtain accurate and reliablereading of the rainfall amount
In this appendix instrumental biases were calculatedto reduce the instrumental uncertainty through an idealexperiment in the laboratory and a field intercomparisonwith 10 TB rain gauges In addition the random errors werequantitatively calculated as a function of rainfall rate andaccumulation time
A1 Correction of Instrumental Bias The instrumental bias119861119868(119894) of the 119894th gauge was defined as a combination
of the absolute bias 119861119860(ref) of the reference gauge and
the relative bias 119861119877(119894) of the 119894th gauge according to the
following equation
119861119868 (119894) = 119861119860 (ref) times 119861119877 (119894) (A1)
119861119860(ref) signified the ratio of the calculated value from
the instrumental resolution of the reference gauge to the truevalue measured by other instruments
119861119860 (ref) =
rainfalltruthrainfallgauge (ref)
(A2)
In this study 119861119860(ref) was derived by an ideal experiment
using an electronic scale First the reference gauge was puton the scale and the scale was initialized The water dropsflowed out into the receiving orifice of the reference gaugeuntil the tipping number became 60Themeasured weight of60 tips was 2144 g while the theoretical (or expected) valuewas 2244 g for the reference gauge (receiving orifice size of1545 cm 60 tips and 02mm tip resolution) Therefore theabsolute bias of the reference gauge from (A2) was about0955 This indicates that the rainfall amount measured bythis reference gauge was overestimated by about 45 andthis bias should be corrected prior to any interpretation
119861119877(119894) represented the ratio of the observed (recorded)
value from the 119894th gauge to the observed (recorded) valuefrom the reference gauge
119861119877 (119894) =
rainfallgauge (ref)rainfallgauge (119894)
(A3)
119861119877(119894) was calculated from ten collocated rain gauges
including the reference gauge located at an observation field
Advances in Meteorology 19
Table 4 The instrumental and relative biases of the 10 tipping bucket rain gauges
Gauge ID RG1 RG2 RG3 RG4 RG5 RG6 RG7 RG8 RG9 RG10Instrumental bias 094 108 099 100 096 104 098 102 108 098Relative bias 098 113 103 105 100 109 102 107 113 102
013610152230406080120150200250300400
()
Rain
rate
(mm
h)
100101
Accumulation time (min)
10
100NSD ()
Figure 22 Two-dimensional normalized standard deviationbetween the standard rainfall and each rain gauge
at KNUKoreaThe observational period for intercomparisonwas from October 2009 to June 2010 (14 rainy days) andthe total accumulated rainfall for this period was about300mm The results are presented in Table 4 All relativebiases were mainly larger than 10 This indicated that themost of gauges underestimated the rainfall amount comparedto the reference rain gauge The relative bias of the referencegauge (119894 = 5) was 10
Finally 119861119860(119894) was calculated from (A1) (see Table 4) All
instrumental biases were between 090 and 11 and the meanvalue of the biases was nearly 10 This indicated that therainfall rate from a single gauge data had a maximum 10bias These biases were corrected for each gauge to removeexisting instrumental errors
A2 Quantification of Random Errors To quantify the ran-dom error due to instrumental uncertainty the NSD of theten gauges was calculated with accumulation time and rainrate using data from an intercomparison The NSD of the 119894thgauge compared to the reference gauge was derived for 20rain rate classes (Δ119877) and 17 accumulation time classes (Δ119905)as follows
NSD119894 (Δ119905 Δ119877)
=
(1119899)radicsum119899minus1
119896=0[1198661015840
119870(Δ119905 Δ119877) minus 119866119894119896 (Δ119905 Δ119877)]
2
1198661015840
(A4)
where 1198661015840 and 119866 indicate the reference and individual gaugesrespectively The overbar means the temporal average ofthe rainfall rate The NSDs of nine different gauges (exceptthe reference gauge) were averaged to obtain the two-dimensional average NSD (Figure 22) The average NSD
tended to decrease with increasing accumulation time andwith increasing rain rate The NSD was about 10ndash15when accumulation time was 10min and the rain rate was10mmhminus1 The instrumental errors were quite significant forlonger accumulation times and higher rainfall rates Theseerrors should be used as a guideline when a single gauge isused as a reference for any validation study
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper has been revised and extended fromMs Young-AOhrsquos MS thesis [25] of Kyungpook National University Thisresearch is supported by Development and Application ofCross Governmental Dual-Pol Radar Harmonization (WRC-2013-A-1) project of theWeather Radar Center KoreaMeteo-rological AdministrationThis work was funded by the KoreaMeteorological Administration Research and DevelopmentProgram under Grant KMIPA2015-1010
References
[1] V N Bringi V Chandrasekar N Balakrishnan and D SZrnic ldquoAn examination of propagation effects in rainfall onradar measurements at microwave frequenciesrdquo Journal ofAtmospheric and Oceanic Technology vol 7 no 6 pp 829ndash8401990
[2] A Huggel W Schmid and AWaldvogel ldquoRaindrop size distri-butions and the radar bright bandrdquo Journal of AppliedMeteorol-ogy vol 35 no 10 pp 1688ndash1701 1996
[3] G W Lee Errors in rain measurement by radar effect of vari-ability of drop size distributions [PhD thesis]McGill University2003
[4] J D Tuttle and R E Rinehart ldquoAttenuation correction in dual-wavelength analysesrdquo Journal of ClimateampAppliedMeteorologyvol 22 no 11 pp 1914ndash1921 1983
[5] S-G Park V N Bringi V Chandrasekar M Maki andK Iwanami ldquoCorrection of radar reflectivity and differentialreflectivity for rain attenuation at X band Part I theoretical andempirical basisrdquo Journal of Atmospheric and Oceanic Technol-ogy vol 22 no 11 pp 1621ndash1632 2005
[6] G Scarchilli E Gorgucci V Chandrasekar and T A SeligaldquoRainfall estimation using polarimetric techniques at C-bandfrequenciesrdquo Journal of Applied Meteorology vol 32 no 6 pp1150ndash1160 1993
[7] M N Anagnostou E N Anagnostou and J VivekanandanldquoCorrection for rain path specific and differential attenuation ofX-band dual-polarization observationsrdquo IEEE Transactions onGeoscience and Remote Sensing vol 44 no 9 pp 2470ndash24802006
20 Advances in Meteorology
[8] J Kalogiros M N Anagnostou E N Anagnostou M Mon-topoli E Picciotti and F S Marzano ldquoEvaluation of a newpolarimetric algorithm for rain-path attenuation correction ofX-band radar observations against disdrometerrdquo IEEE Transac-tions on Geoscience and Remote Sensing vol 52 no 2 pp 1369ndash1380 2014
[9] W-Y Chang J Vivekanandan and T-C C Wang ldquoEstimationof X-band polarimetric radar attenuation and measurementuncertainty using a variational methodrdquo Journal of AppliedMeteorology and Climatology vol 53 no 4 pp 1099ndash1119 2014
[10] J Hubbert and V N Bringi ldquoAn iterative filtering technique forthe analysis of copolar differential phase and dual-frequencyradar measurementsrdquo Journal of Atmospheric and OceanicTechnology vol 12 no 3 pp 643ndash648 1995
[11] K V Mishra W F Krajewski R Goska et al ldquoDeploy-ment and performance analyses of high-resolution iowa XPOLradar system during the NASA IFloodS campaignrdquo Journal ofHydrometeorology vol 17 no 2 pp 455ndash479 2016
[12] M Thurai G J Huang V N Bringi W L Randeu and MSchonhuber ldquoDrop shapes model comparisons and calcu-lations of polarimetric radar parameters in rainrdquo Journal ofAtmospheric and Oceanic Technology vol 24 no 6 pp 1019ndash1032 2007
[13] M D Humphrey J D Istok J Y Lee J A Hevesi and AL Flint ldquoA new method for automated dynamic calibration oftipping-bucket rain gaugesrdquo Journal of Atmospheric andOceanicTechnology vol 14 no 6 pp 1513ndash1519 1997
[14] E HabibW F Krajewski V Nespor andA Kruger ldquoNumericalsimulation studies of rain gage data correction due to windeffectrdquo Journal of Geophysical ResearchAtmospheres vol 104 no16 Article ID 1999JD900228 pp 19723ndash19733 1999
[15] G J Ciach and W F Krajewski ldquoOn the estimation of radarrainfall error variancerdquoAdvances inWater Resources vol 22 no6 pp 585ndash595 1999
[16] E Habib W F Krajewski and A Kruger ldquoSampling errors oftipping-bucket rain gaugemeasurementsrdquo Journal of HydrologicEngineering vol 6 no 2 pp 159ndash166 2001
[17] G J Ciach ldquoLocal random errors in tipping-bucket rain gaugemeasurementsrdquo Journal of Atmospheric andOceanic Technologyvol 20 no 5 pp 752ndash759 2003
[18] EHabibW F Krajewski andG J Ciach ldquoEstimation of rainfallinterstation correlationrdquo Journal ofHydrometeorology vol 2 no6 pp 621ndash629 2001
[19] NIMS ldquoStudy on estimation of rainfall amount and classifi-cation of hydrometeor species using dual-polarimetric radar(III)rdquo pp 1ndash20 2010
[20] E Gorgucci V Chandrasekar and L Baldini ldquoCorrection ofX-band radar observation for propagation effects based on theself-consistency principlerdquo Journal of Atmospheric and OceanicTechnology vol 23 pp 1668ndash1681 2006
[21] NIMS ldquoStudy on estimating raindrop size distribution forquantitative estimation of radar rainfall amountrdquo pp 1ndash55 2009
[22] NIMS ldquoQuantitative precipitation estimation and hydrometeoridentification using dual-polarization radar-Phase IIrdquo pp 1ndash742009
[23] S-G Park M Maki K Iwanami V N Bringi and V Chan-drasekar ldquoCorrection of radar reflectivity and differential reflec-tivity for rain attenuation at X band Part II Evaluation andapplicationrdquo Journal of Atmospheric and Oceanic Technologyvol 22 no 11 pp 1633ndash1655 2005
[24] E Habib and W F Krajewski ldquoUncertainty analysis of theTRMM ground-validation radar-rainfall products applicationto the TEFLUN-B field campaignrdquo Journal of Applied Meteorol-ogy vol 41 no 5 pp 558ndash572 2002
[25] Y-A Oh Rainfall estimation from X-band dual-polarizationradar measurements effects of attenuation correction and evalu-ation with a dense rain gauge network [MS thesis] KyungpookNational University 2011
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom
Applied ampEnvironmentalSoil Science
Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal of
Geophysics
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
Advances in Meteorology 7
32km
21 km
11 km
20km40km60km
minus180 18020 30 40 50 60 70 80 90 100120140160100(a)
32 km
21 km
11 km
20km40km60km
minus180 18020 30 40 50 60 70 80 90 100120140160100(b)
Figure 5 PPIs of measured (a) and filtered (b) ΨDP of the NIMS-XPOL at 2101UTC July 10 2010
gauge The NB and NSD of 119866119894with ⟨119866⟩ were calculated as
follows
NB (119894) =(sum119899minus1
119896=0[119866119894119896 (Δ119905) minus ⟨119866119896 (Δ119905)⟩]) 119899
⟨119866 (Δ119905)⟩
NSD (119894) =(1119899)radic(sum
119899minus1
119896=0[119866119894119896 (Δ119905) minus ⟨119866119896 (Δ119905)⟩]
2)
⟨119866 (Δ119905)⟩
(8)
where 119894 indicates the 119894th gauge and the overbar indicates thetemporal average
This analysis provided a reasonable spatial average andtemporal accumulation resolution for the calculation ofmoreaccurate and realistic rainfall amounts from the gauge dataand a quantitative guideline regarding the errors in rainestimation due to the variability of the rain fields
4 Results
41 Attenuation Correction of Reflectivity and DifferentialReflectivity The attenuation correction algorithm in Sec-tion 31 was applied to the NIMS-XPOL radar data ΦDPwas estimated by eliminating 120575 and the observational noisefrom measured ΨDP using an iterative FIR filter The offsetof ΦDP was then adjusted Figure 5 shows the PPIs ofmeasured ΨDP and filtered ΦDP After applying the FIR filterthe small scale variation due to 120575 and observational noisein measured ΨDP was reduced and filtered ΨDP (estimatedΦDP) monotonically increased with an increase in rangeThe maximum value of estimated ΦDP reached to 140∘ Inaddition estimated ΦDP at the azimuth angles of 150∘ and200∘ sharply increased with the loss of signal beyond thesesharp gradient areas This reflected that strong convective
cells existed in these directions and that these cells caused thesevere signal attenuation EstimatedΦDP values between 250
∘
and 360∘ azimuth angle decreased after filtering since ΦDPoffsets were positive values in these directions Figure 6 showsthe range profile of measured and filtered ΨDP for estimationof ΦDP at 71
∘ 160∘ and 334∘ azimuth angles The unexpectedlarge values near the radarwere removed by the120588HV threshold(red line)The small fluctuation inmeasuredΨDP due to 120575 andthe observational noise was effectively removed after filteringThe value of estimated ΦDP was adjusted to 0 by the offset(blue line)
Estimated ΦDP was used to calculate the path integrated119860119867 and the measured119885
119867values were corrected as shown in
Figure 7 Figure 7(a) represents measured-attenuated119885119867and
relatively small values of 119885119867at far ranges This may indicate
that the 119885119867values at far ranges were affected by the strong
attenuation due to strong convective cells near the radarThisspatial pattern of 119885
119867was the typical feature of the attenuated
radar signal due to the rain After correcting the attenuationthe value of the 119885
119867field became larger over the entire area
and the radially attenuated 119885119867
was corrected by showinga typical rain cell pattern in 119885
119867(Figure 7(b)) The largest
difference between the uncorrected and the corrected 119885119867at
2101UTCwas over 50 dB However the white area behind thestrong echoes was not detected as there was a complete loss ofthe signal powerThese regions could not be corrected by thisalgorithmThe circle pattern of119885
119867fields within about a 4 km
range from theNIMS-XPOL could not present the actual rainechoes due to the instrumental issues (the same was true for119885DR) These results are also presented in the range profiles inFigure 8
Similarly measured 119885DR were corrected with the pathintegrated 119860DP derived from the path integrated 119860
119867using
(4) (see Figure 9) Before applying attenuation correction
8 Advances in Meteorology
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
0
50
100
150
200
ΦDP (∘) azimuth angle = 71∘
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
0
50
100
150
200
ΦDP (∘) azimuth angle = 160∘
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
0
50
100
150
200
ΦDP (∘) azimuth angle = 334∘
(c)
Figure 6 Range profiles of measured (black) 120588HV threshold isolated point removing (red) and FIR filter and offset adjusted (blue) ΨDP forestimation of ΦDP at 71
∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
32km
21 km
11 km
20km40km60km
minus10 10 15 20 25 30 35 40 45 50 55 60 65 700(a)
32 km
21 km
11 km
20km40km60km
minus10 10 15 20 25 30 35 40 45 50 55 60 65 700(b)
Figure 7 PPIs of measured (a) and attenuation corrected (b) 119885119867of the NIMS-XPOL for the case in Figure 5
Advances in Meteorology 9
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 71∘
minus20
0
20
40
60
80
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 160∘
minus20
0
20
40
60
80
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 334∘
minus20
0
20
40
60
80
(c)
Figure 8 Range profiles of measured (line) calibration bias 120588HV threshold and isolated point removing (dot) and attenuation corrected(dash) 119885
119867at 71∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
the attenuation signature in119885DR was determined Ideally the119885DR value for rain should be positive since rain drops areoblate due to aerodynamic force However themeasured119885DRfield in Figure 9 showed a negative value over wide areasAfter correcting the attenuation the119885DR field was larger than0 dB and showed a realistic rain patternThis is also presentedin the range profile of Figure 10The largest difference of119885DRbetween before and after attenuation correction at 2101UTCwas about 8 dB The 119885DR values around the 220∘ azimuthangle beyond 40 km were the largest They did not matchwith the area of the largest 119885DR values (around the 210∘azimuth angle with a range of 20 km and 40 km) This wasan unexpected result because 119885DR is proportional to 119885
119867in
theory This mismatch between 119885119867
and 119885DR at far rangemay be due to contamination due to beam broadening andaccumulated errors due to uncertainty in the attenuationcorrection with different ranges
Figures 11 and 12 show the time series of 119885119867
and119885DR at the ten gauge sites before and after applying theattenuation correction The differences between uncorrectedand corrected values existed after 1800UTC and were at amaximumaround 2100UTCThis attenuation correctionwas
evaluated through comparison with the theoretical values of119885119867and 119885DR In Figure 13 the uncorrected 119885
119867-119885DR scatter
plots are apart from the theoretical119885119867-119885DR values [23] After
attenuation correction the119885119867-119885DR scatter plots were close to
the calculated valuesThis indicated that119885119867and119885DR over the
gauge sites were corrected with reliable values
42 Variability of the Rain Fields The Spearman correlationcoefficient of the rainfall rate from the 10 spatially dis-tributed rain gauges was calculated with accumulation timeand separation distance (Figure 14) The overall Spearmancorrelation coefficient decreased with increasing separationdistance at all accumulation timesThis tendency was similarto the result from Habib and Krajewski [24] However thecorrelation value was higher in this study Note that the gaugetip resolution in Habib and Krajewski [24] was 0254mmwhile it was 02mm in this study In addition the rainfall ratesfrom the gauge data used in this study mainly consisted ofmoderate rainfall rates while Habib and Krajewski [24] useddata containing heavy rainfall rates
For accumulation times of less than 10min the corre-lation at the same distance was dramatically increased with
10 Advances in Meteorology
32km
21 km
11 km
20km40km60km
18
14
00
02
04
06
26
40
minus10
22
minus20
30
10
50
minus30
(a)
32 km
21 km
11 km
20km40km60km
18
14
00
02
04
06
26
40
minus10
22
minus20
30
10
50
minus30
(b)
Figure 9 PPIs of measured (a) and attenuation corrected (b) 119885DR of NIMS-XPOL for the case in Figure 5
increasing accumulation time and the slope of the correlationfunction was steep This was due to the reduction of theinstrumental errors of the gauge data by time accumulationThis indicated that rainfall accumulated for less than 10minfrom single gauge data cannot represent the rainfall over itssurrounding area due to instrumental errors For accumula-tion times longer than 10min the slope became less steepThis reflected the remaining variability of the rain field Inaddition the accumulated rainfall rate of about 10min wassuitable for minimizing the instrumental uncertainty whilepreserving the variation in the rain field The correlationreached about 098 at 05 km in 10min of accumulationand this approached 098 at 10 km at 60min accumulationTherefore a single rain gauge mostly represented the rainfallrate within 05 km and 10 km with 10min and 60minaccumulation respectively
The NB between an areal averaged rainfall rate ⟨119866⟩ anda single gauge rainfall rate 119866 is shown with accumulationtime in Figures 15(a) and 16(a) Where the average area was038 km2 (corresponding to 9 pixels of NIMS-XPOL radarat 16 km range) the NB reached a 6ndash8 maximum Wherethe averaged area is 6 km2 (corresponding to the coverage forall of the gauges) the NB reached about a 10 maximumFurthermore the NB hardly changed with accumulationtime Note that this statistic was derived after removinginstrumental bias This indicated that the rainfall rate froma single gauge was biased due to the spatial variability of therain fields and this bias was maintained with accumulationtime
The NSD of a single gauge for a mean area of 038 km2was over 40 and it reached about 70ndash90 for a mean areaof 6 km2 at an accumulation time of 1min (see Figures 15(b)and 16(b)) This statistic included instrumental errors (about
40ndash60 at 16mmhminus1 and 60ndash80 at 10mmhminus1 as shownin the Appendix) as well as the spatial variability of the rainfields This indicated that the variance between the point andarea values from the gauge data at short accumulation timeswas similar to the errors due to the instrumental uncertaintyof the rain gaugeTheNSD in 60min accumulation decreasedto over 10 and about 15ndash45 with a mean area of 038 km2and 6 km2 respectively including the NSD of 6ndash10 due toinstrumental uncertainty This indicated that the rainfall ratefrom single gauges may differ significantly from the actualrainfall rate due to the variability of the rain fields and theinstrumental uncertainty in short accumulation times
43 Verification of Rainfall Retrieval The rainfall rate fromthe NIMS-XPOL radar measurements was retrieved withtwo rainfall rate estimators 119877(119885
119867) and 119877(119885
119867 119885DR) The
estimated rainfall rate was verified through comparison withthe rainfall rate (119866) from the rain gauge data These twoestimators were applied before and after the attenuation cor-rection First 119877(038 km2) was compared with 119877(007 km2)at 10min accumulation (Figure 17) Using 119877(119885
119867) the NB
NSD rmse and r rmse without the attenuation correctionwere significantly large at 70 127 127 (mmhr) and78 respectively However they were reduced to minus152 521 (mmhr) and 32 respectively after applying theattenuation correction Using 119877(119885
119867 119885DR) the NB was just
3 NSD was 49 rmse was 496 (mmhr) and r rmsewas 31 before applying attenuation correction This wassimilar to the result from 119877(119885
119867) after attenuation correction
However there was no significant change in119877(119885119867 119885DR)with
attenuation correction This means that 119877(119885119867 119885DR) was less
sensitive to rain attenuation than was 119877(119885119867) because 119885
119867is
proportional to 119877 and 119885DR is inversely proportional to 119877
Advances in Meteorology 11
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 71∘
minus10
minus5
0
5
10
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 160∘
minus10
minus5
0
5
10
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 334∘
minus10
minus5
0
5
10
(c)
Figure 10 Range profiles of measured (line) calibration bias 120588HV threshold and isolated point removing (dot) and attenuation corrected(dash) 119885DR at 71∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
in this rain estimator (see Table 3) Furthermore comparingthe analysis of variance results between areal (038 km2) andpoint (007m2) values from the gauge data (Section 42)half of the errors after attenuation correction were producedby the small scale variability of the rain fields and theinstrumental uncertainty of the rain gauge
A similar comparison is shown in Figure 18 for a 60minaccumulation Using 119877(119885
119867) the NB NSD rmse and r rmse
were decreased from minus65 86 71 (mmhr) and 70to minus5 34 28 (mmhr) and 28 by the attenuationcorrection respectively Using 119877(119885
119867 119885DR) the NB NSD
rmse and r rmse were only slightly decreased after theattenuation correction However the NSD rmse and r rmsewere reduced when compared at 10min accumulation In thesame accumulation time the errors in the gauge data reachedto over 15
This comparison was affected by different sampling areasof the gauges and the radar Therefore the areal averagevalues were compared (Figure 19) Using119877(119885
119867)with the areal
averaged radar measurements the NB was reduced fromminus69 to 7 NSD was reduced from 122 to 42 rmsewas reduced from 39 (mmhr) to 13 (mmhr) and r rmse
was reduced from 77 to 26 at an accumulation time of10min Using 119877(119885
119867 119885DR) the NB NSD rmse and r rmse
were decreased fromminus24 67 21 (mmhr) and 42 to 539 12 (mmhr) and 24 respectively
TheNB and r rmsewithmatched sampling areas betweenthe gauge and the radar are shown in Figures 20 and 21 Thereduction of NB due to temporal accumulation was smallin both of the rain estimators The NB was reduced to lessthan 7 by attenuation correction for all accumulation timeswith both119877(119885
119867) and119877(119885
119867 119885DR)The attenuation correction
effect was significantly greater in119877(119885119867) Similarly the r rmse
was reduced significantly in 119877(119885119867) by attenuation correction
(from 68 to 25 at a 60min accumulation) The reductionof the r rmse due to temporal accumulation was larger in119877(119885119867 119885DR) because the significant observational noise of
119885DR was smoothed The r rmse was decreased from 18to 15 in 119877(119885
119867 119885DR) at a 60min accumulation time after
attenuation correction
5 Conclusions
In this study an attenuation correction algorithm based onΦDP was developed to improve the accuracy of the rainfall
12 Advances in Meteorology
1st gauge site 2nd gauge site
3rd gauge site 4th gauge site
5th gauge site 6th gauge site
7th gauge site 8th gauge site
9th gauge site 10th gauge site
17 18 19 20 2116Time (h)
17 18 19 20 2116Time (h)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
UncorrectedCorrected
UncorrectedCorrected
Figure 11 Time series of measured (red circle) and attenuation corrected (blue circle) 119885119867at the 10 rain gauge sites
Advances in Meteorology 13
UncorrectedCorrected
UncorrectedCorrected
17 18 19 20 2116Time (h)
17 18 19 20 2116Time (h)
1st gauge site 2nd gauge site
3rd gauge site 4th gauge site
5th gauge site 6th gauge site
7th gauge site 8th gauge site
9th gauge site 10th gauge site
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
Figure 12 Time series of measured (red circle) and attenuation corrected (blue circle) 119885DR at the 10 rain gauge sites
14 Advances in Meteorology
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Uncorrected(a)
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Corrected Simulated from DSD(b)
Figure 13 Scatter plots of 119885119867and 119885DR before (a) and after (b) attenuation correction at the 10 rain gauge sites DSD observation results (dot
[21])
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
04
05
06
07
08
09
10
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
Accum time = 1min
Accum time = 10min
Accum time = 2min
Accum time = 20min
Accum time = 5min
Accum time = 60min
Figure 14 The Spearman correlation coefficient 119903 between data from two rain gauges at different accumulation times
estimation using X-band radar This algorithm was evalu-ated with radar measurements from NIMS-XPOL Howevermeasured ΨDP contained 120575 and observation noise as wellas ΦDP Furthermore the ΨDP offset varied with azimuthangles and decreased near the radar when rain cells were
above the radar In this study ΦDP was estimated by aniterative filtering technique and by adjusting with an offsetdetermined as the minimum value ofΦDP within 3 km of theradar As a result estimated ΦDP monotonically increasedThe path integrated value of119860
119867and119860DP was calculated from
Advances in Meteorology 15
2nd 6th 9th gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(038km2) versus G
(a) Normalized bias
⟨G⟩(038km2) versus G
2nd 6th 9th gaugesMean NSD
Mean R
0
20
40
60
80
100
NSD
()
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 15 (a) The NB and (b) NSD between ⟨119866⟩(038 km2) and 119866 from three rain gauges with accumulation time ⟨119866⟩ is the temporallyaveraged rainfall rate of ⟨119866⟩
All gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(6km2) versus G
(a) Normalized bias
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
0
20
40
60
80
100
NSD
()
All gaugesMean NSD
Mean R
⟨G⟩(6km2) versus G
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 16 (a) The NB and (b) NSD between ⟨119866⟩(6 km2) and 119866 from 10 rain gauges with accumulation time ⟨119866⟩ is the temporally averaged⟨119866⟩ with accumulation time
estimated ΦDP using 119860119867-ΦDP and 119860DP-ΦDP relationships
derived from the DSDs data Measured 119885119867
and 119885DR werecorrected with the calculated path integrated 119860
119867and 119860DP
by each gate The attenuated signal patterns of 119885119867and 119885DR
were restored to a typical rain echo pattern However theattenuation corrected 119885
119867and 119885DR fields were not matched
at distant ranges This was likely because of contaminationdue to beam broadening and the accumulation of errorswith increasing range Comparing the 119885
119867-119885DR scatter plots
over the 10 rain gauge sites with the results obtained froma scattering simulation with the DSD data the attenuation
corrected 119885119867-119885DR scatter plots are in good agreement with
the 119885119867-119885DR scatter plots from the DSDs
The validation for the improvement of accuracy in therainfall estimation through attenuation correction was per-formed by comparing with dense rain gauge network data119877(119885119867) from the attenuated (observed)119885
119867produced a severe
underestimation with an NB NSD rmse and r rmse of70 127 127 (mmhr) and 78 at 10min accumulationrespectively while 119877(119885
119867) from the attenuation corrected 119885
119867
showed better agreement with the gauge measurements withan NB NSD rmse and r rmse of minus1 52 521 (mmhr)
16 Advances in Meteorology
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus070
NSD = 127
rmse = 1279
corr = 0858
Corrected
NB = minus001
NSD = 052
rmse = 521
corr = 0927
r_rmse = 0786 r_rmse = 0320
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = 003
NSD = 049
rmse = 496
r_rmse = 0305
corr = 0936
Corrected
NB = 001
NSD = 050
rmse = 502
r_rmse = 0308
corr = 0936
(b) 119877(119885119867119885DR)
Figure 17 The scatter plots of 119877(038 km2 10min) versus 119866(007m2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
CorrectedUncorrected
R(m
mh
)
0
10
20
30
40
10 20 30 400G (mmh)
Uncorrected
NB = minus065
NSD = 086
rmse = 706
r_rmse = 0702
corr = 0878
Corrected
NB = minus005
NSD = 034
rmse = 281
r_rmse = 0279
corr = 0866
(a) 119877(119885119867)
CorrectedUncorrected
R(m
mh
)
10 20 30 400G (mmh)
0
10
20
30
40
Uncorrected
NB = 008
NSD = 023
rmse = 189
r_rmse = 0188
corr = 0942
Corrected
NB = 007
NSD = 025
rmse = 205
r_rmse = 0203
corr = 0940
(b) 119877(119885119867119885DR)
Figure 18 The scatter plots of 119877(038 km2 60min) versus 119866(007m2 60min) (a) using 119885119867and (b) using 119885
119867and 119885DR
and 32 at 10min accumulation respectively At 60minaccumulation theNBNSD rmse and r rmsewere decreasedfrom minus65 86 71 (mmhr) and 70 to minus5 3428 (mmhr) and 28 respectively Using 119877(119885
119867 119885DR) the
NB NSD rmse and r rmse were only slightly reduced even if
attenuation was corrected but were smaller than when using119877(119885119867) This indicated that 119877(119885
119867 119885DR) was less sensitive to
the attenuation than 119877(119885119867) In addition half of the errors
that remained after attenuation correction were caused by thesmall scale variability of the rain fields and the instrumental
Advances in Meteorology 17
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus069
NSD = 122
rmse = 391
r_rmse = 0770
corr = 0893
Corrected
NB = 007
NSD = 042
rmse = 133
r_rmse = 0262
corr = 0960
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = minus024
NSD = 067
rmse = 215
r_rmse = 0424
corr = 0718
Corrected
NB = 005
NSD = 039
rmse = 124
r_rmse = 0245
corr = 0891
(b) 119877(119885119867119885DR)
Figure 19 The scatter plots of 119877(6 km2 10min) versus 119866(6 km2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
R(ZH)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(a) 119877(119885119867)
R(ZH ZDR)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(b) 119877(119885119867119885DR)
Figure 20 The normalized bias of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and 119885DR The red
point-line and the blue point-line indicate without and with the attenuation correction respectively
uncertainty of the rain gauges Using an areal averagedvalue the NB was reduced to less than 7 with attenua-tion correction using either rainfall estimator regardless ofaccumulation timeThe r rmse were significantly reduced byattenuation correction in 119877(119885
119867) and the reduction of r rmse
with 119877(119885119867 119885DR) was even more dramatic
As summarized above the rain attenuation for 119885119867
and 119885DR in the X-band was corrected by ΦDP and theresults showed a good agreement with the value obtainedby scattering simulations with the DSD datasets The resultsof comparison with rain gauge data also showed a signifi-cant improvement in the accuracy of rainfall estimation by
18 Advances in Meteorology
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10r_
rmse
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH)
(a) 119877(119885119867)
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10
r_rm
se
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH ZDR)
(b) 119877(119885119867119885DR)
Figure 21 The relative root-mean square error of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and
119885DR The red point-line and the blue point-line indicate without and with the attenuation correction respectively
attenuation correction which supports the quantitative useof the X-band radar in rain data retrieval
Appendix
A Instrumental Uncertainty of the Rain Gauge
The instrumental uncertainty of the TB rain gauge is asso-ciated with improper calibration leakage and precipitationmissing the gauge due to the measurement principle andwetting and evaporation losses In addition external sourcessuch as wind turbulence and installed position also con-tribute to the instrumental uncertainty Ciach [17] reportedthat the TB gauge data suffers a significant uncertainty asso-ciated with random differences between closely collocatedTB gauges through the analysis of the error as a functionof the rainfall intensity for three timescales In addition theimproper calibration induces a bias in the rainfall measure-ment and this bias cannot be removed by increasing accu-mulation timeTherefore the proper correction or processingof TB gauge data is necessary to obtain accurate and reliablereading of the rainfall amount
In this appendix instrumental biases were calculatedto reduce the instrumental uncertainty through an idealexperiment in the laboratory and a field intercomparisonwith 10 TB rain gauges In addition the random errors werequantitatively calculated as a function of rainfall rate andaccumulation time
A1 Correction of Instrumental Bias The instrumental bias119861119868(119894) of the 119894th gauge was defined as a combination
of the absolute bias 119861119860(ref) of the reference gauge and
the relative bias 119861119877(119894) of the 119894th gauge according to the
following equation
119861119868 (119894) = 119861119860 (ref) times 119861119877 (119894) (A1)
119861119860(ref) signified the ratio of the calculated value from
the instrumental resolution of the reference gauge to the truevalue measured by other instruments
119861119860 (ref) =
rainfalltruthrainfallgauge (ref)
(A2)
In this study 119861119860(ref) was derived by an ideal experiment
using an electronic scale First the reference gauge was puton the scale and the scale was initialized The water dropsflowed out into the receiving orifice of the reference gaugeuntil the tipping number became 60Themeasured weight of60 tips was 2144 g while the theoretical (or expected) valuewas 2244 g for the reference gauge (receiving orifice size of1545 cm 60 tips and 02mm tip resolution) Therefore theabsolute bias of the reference gauge from (A2) was about0955 This indicates that the rainfall amount measured bythis reference gauge was overestimated by about 45 andthis bias should be corrected prior to any interpretation
119861119877(119894) represented the ratio of the observed (recorded)
value from the 119894th gauge to the observed (recorded) valuefrom the reference gauge
119861119877 (119894) =
rainfallgauge (ref)rainfallgauge (119894)
(A3)
119861119877(119894) was calculated from ten collocated rain gauges
including the reference gauge located at an observation field
Advances in Meteorology 19
Table 4 The instrumental and relative biases of the 10 tipping bucket rain gauges
Gauge ID RG1 RG2 RG3 RG4 RG5 RG6 RG7 RG8 RG9 RG10Instrumental bias 094 108 099 100 096 104 098 102 108 098Relative bias 098 113 103 105 100 109 102 107 113 102
013610152230406080120150200250300400
()
Rain
rate
(mm
h)
100101
Accumulation time (min)
10
100NSD ()
Figure 22 Two-dimensional normalized standard deviationbetween the standard rainfall and each rain gauge
at KNUKoreaThe observational period for intercomparisonwas from October 2009 to June 2010 (14 rainy days) andthe total accumulated rainfall for this period was about300mm The results are presented in Table 4 All relativebiases were mainly larger than 10 This indicated that themost of gauges underestimated the rainfall amount comparedto the reference rain gauge The relative bias of the referencegauge (119894 = 5) was 10
Finally 119861119860(119894) was calculated from (A1) (see Table 4) All
instrumental biases were between 090 and 11 and the meanvalue of the biases was nearly 10 This indicated that therainfall rate from a single gauge data had a maximum 10bias These biases were corrected for each gauge to removeexisting instrumental errors
A2 Quantification of Random Errors To quantify the ran-dom error due to instrumental uncertainty the NSD of theten gauges was calculated with accumulation time and rainrate using data from an intercomparison The NSD of the 119894thgauge compared to the reference gauge was derived for 20rain rate classes (Δ119877) and 17 accumulation time classes (Δ119905)as follows
NSD119894 (Δ119905 Δ119877)
=
(1119899)radicsum119899minus1
119896=0[1198661015840
119870(Δ119905 Δ119877) minus 119866119894119896 (Δ119905 Δ119877)]
2
1198661015840
(A4)
where 1198661015840 and 119866 indicate the reference and individual gaugesrespectively The overbar means the temporal average ofthe rainfall rate The NSDs of nine different gauges (exceptthe reference gauge) were averaged to obtain the two-dimensional average NSD (Figure 22) The average NSD
tended to decrease with increasing accumulation time andwith increasing rain rate The NSD was about 10ndash15when accumulation time was 10min and the rain rate was10mmhminus1 The instrumental errors were quite significant forlonger accumulation times and higher rainfall rates Theseerrors should be used as a guideline when a single gauge isused as a reference for any validation study
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper has been revised and extended fromMs Young-AOhrsquos MS thesis [25] of Kyungpook National University Thisresearch is supported by Development and Application ofCross Governmental Dual-Pol Radar Harmonization (WRC-2013-A-1) project of theWeather Radar Center KoreaMeteo-rological AdministrationThis work was funded by the KoreaMeteorological Administration Research and DevelopmentProgram under Grant KMIPA2015-1010
References
[1] V N Bringi V Chandrasekar N Balakrishnan and D SZrnic ldquoAn examination of propagation effects in rainfall onradar measurements at microwave frequenciesrdquo Journal ofAtmospheric and Oceanic Technology vol 7 no 6 pp 829ndash8401990
[2] A Huggel W Schmid and AWaldvogel ldquoRaindrop size distri-butions and the radar bright bandrdquo Journal of AppliedMeteorol-ogy vol 35 no 10 pp 1688ndash1701 1996
[3] G W Lee Errors in rain measurement by radar effect of vari-ability of drop size distributions [PhD thesis]McGill University2003
[4] J D Tuttle and R E Rinehart ldquoAttenuation correction in dual-wavelength analysesrdquo Journal of ClimateampAppliedMeteorologyvol 22 no 11 pp 1914ndash1921 1983
[5] S-G Park V N Bringi V Chandrasekar M Maki andK Iwanami ldquoCorrection of radar reflectivity and differentialreflectivity for rain attenuation at X band Part I theoretical andempirical basisrdquo Journal of Atmospheric and Oceanic Technol-ogy vol 22 no 11 pp 1621ndash1632 2005
[6] G Scarchilli E Gorgucci V Chandrasekar and T A SeligaldquoRainfall estimation using polarimetric techniques at C-bandfrequenciesrdquo Journal of Applied Meteorology vol 32 no 6 pp1150ndash1160 1993
[7] M N Anagnostou E N Anagnostou and J VivekanandanldquoCorrection for rain path specific and differential attenuation ofX-band dual-polarization observationsrdquo IEEE Transactions onGeoscience and Remote Sensing vol 44 no 9 pp 2470ndash24802006
20 Advances in Meteorology
[8] J Kalogiros M N Anagnostou E N Anagnostou M Mon-topoli E Picciotti and F S Marzano ldquoEvaluation of a newpolarimetric algorithm for rain-path attenuation correction ofX-band radar observations against disdrometerrdquo IEEE Transac-tions on Geoscience and Remote Sensing vol 52 no 2 pp 1369ndash1380 2014
[9] W-Y Chang J Vivekanandan and T-C C Wang ldquoEstimationof X-band polarimetric radar attenuation and measurementuncertainty using a variational methodrdquo Journal of AppliedMeteorology and Climatology vol 53 no 4 pp 1099ndash1119 2014
[10] J Hubbert and V N Bringi ldquoAn iterative filtering technique forthe analysis of copolar differential phase and dual-frequencyradar measurementsrdquo Journal of Atmospheric and OceanicTechnology vol 12 no 3 pp 643ndash648 1995
[11] K V Mishra W F Krajewski R Goska et al ldquoDeploy-ment and performance analyses of high-resolution iowa XPOLradar system during the NASA IFloodS campaignrdquo Journal ofHydrometeorology vol 17 no 2 pp 455ndash479 2016
[12] M Thurai G J Huang V N Bringi W L Randeu and MSchonhuber ldquoDrop shapes model comparisons and calcu-lations of polarimetric radar parameters in rainrdquo Journal ofAtmospheric and Oceanic Technology vol 24 no 6 pp 1019ndash1032 2007
[13] M D Humphrey J D Istok J Y Lee J A Hevesi and AL Flint ldquoA new method for automated dynamic calibration oftipping-bucket rain gaugesrdquo Journal of Atmospheric andOceanicTechnology vol 14 no 6 pp 1513ndash1519 1997
[14] E HabibW F Krajewski V Nespor andA Kruger ldquoNumericalsimulation studies of rain gage data correction due to windeffectrdquo Journal of Geophysical ResearchAtmospheres vol 104 no16 Article ID 1999JD900228 pp 19723ndash19733 1999
[15] G J Ciach and W F Krajewski ldquoOn the estimation of radarrainfall error variancerdquoAdvances inWater Resources vol 22 no6 pp 585ndash595 1999
[16] E Habib W F Krajewski and A Kruger ldquoSampling errors oftipping-bucket rain gaugemeasurementsrdquo Journal of HydrologicEngineering vol 6 no 2 pp 159ndash166 2001
[17] G J Ciach ldquoLocal random errors in tipping-bucket rain gaugemeasurementsrdquo Journal of Atmospheric andOceanic Technologyvol 20 no 5 pp 752ndash759 2003
[18] EHabibW F Krajewski andG J Ciach ldquoEstimation of rainfallinterstation correlationrdquo Journal ofHydrometeorology vol 2 no6 pp 621ndash629 2001
[19] NIMS ldquoStudy on estimation of rainfall amount and classifi-cation of hydrometeor species using dual-polarimetric radar(III)rdquo pp 1ndash20 2010
[20] E Gorgucci V Chandrasekar and L Baldini ldquoCorrection ofX-band radar observation for propagation effects based on theself-consistency principlerdquo Journal of Atmospheric and OceanicTechnology vol 23 pp 1668ndash1681 2006
[21] NIMS ldquoStudy on estimating raindrop size distribution forquantitative estimation of radar rainfall amountrdquo pp 1ndash55 2009
[22] NIMS ldquoQuantitative precipitation estimation and hydrometeoridentification using dual-polarization radar-Phase IIrdquo pp 1ndash742009
[23] S-G Park M Maki K Iwanami V N Bringi and V Chan-drasekar ldquoCorrection of radar reflectivity and differential reflec-tivity for rain attenuation at X band Part II Evaluation andapplicationrdquo Journal of Atmospheric and Oceanic Technologyvol 22 no 11 pp 1633ndash1655 2005
[24] E Habib and W F Krajewski ldquoUncertainty analysis of theTRMM ground-validation radar-rainfall products applicationto the TEFLUN-B field campaignrdquo Journal of Applied Meteorol-ogy vol 41 no 5 pp 558ndash572 2002
[25] Y-A Oh Rainfall estimation from X-band dual-polarizationradar measurements effects of attenuation correction and evalu-ation with a dense rain gauge network [MS thesis] KyungpookNational University 2011
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom
Applied ampEnvironmentalSoil Science
Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal of
Geophysics
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
8 Advances in Meteorology
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
0
50
100
150
200
ΦDP (∘) azimuth angle = 71∘
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
0
50
100
150
200
ΦDP (∘) azimuth angle = 160∘
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
0
50
100
150
200
ΦDP (∘) azimuth angle = 334∘
(c)
Figure 6 Range profiles of measured (black) 120588HV threshold isolated point removing (red) and FIR filter and offset adjusted (blue) ΨDP forestimation of ΦDP at 71
∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
32km
21 km
11 km
20km40km60km
minus10 10 15 20 25 30 35 40 45 50 55 60 65 700(a)
32 km
21 km
11 km
20km40km60km
minus10 10 15 20 25 30 35 40 45 50 55 60 65 700(b)
Figure 7 PPIs of measured (a) and attenuation corrected (b) 119885119867of the NIMS-XPOL for the case in Figure 5
Advances in Meteorology 9
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 71∘
minus20
0
20
40
60
80
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 160∘
minus20
0
20
40
60
80
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 334∘
minus20
0
20
40
60
80
(c)
Figure 8 Range profiles of measured (line) calibration bias 120588HV threshold and isolated point removing (dot) and attenuation corrected(dash) 119885
119867at 71∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
the attenuation signature in119885DR was determined Ideally the119885DR value for rain should be positive since rain drops areoblate due to aerodynamic force However themeasured119885DRfield in Figure 9 showed a negative value over wide areasAfter correcting the attenuation the119885DR field was larger than0 dB and showed a realistic rain patternThis is also presentedin the range profile of Figure 10The largest difference of119885DRbetween before and after attenuation correction at 2101UTCwas about 8 dB The 119885DR values around the 220∘ azimuthangle beyond 40 km were the largest They did not matchwith the area of the largest 119885DR values (around the 210∘azimuth angle with a range of 20 km and 40 km) This wasan unexpected result because 119885DR is proportional to 119885
119867in
theory This mismatch between 119885119867
and 119885DR at far rangemay be due to contamination due to beam broadening andaccumulated errors due to uncertainty in the attenuationcorrection with different ranges
Figures 11 and 12 show the time series of 119885119867
and119885DR at the ten gauge sites before and after applying theattenuation correction The differences between uncorrectedand corrected values existed after 1800UTC and were at amaximumaround 2100UTCThis attenuation correctionwas
evaluated through comparison with the theoretical values of119885119867and 119885DR In Figure 13 the uncorrected 119885
119867-119885DR scatter
plots are apart from the theoretical119885119867-119885DR values [23] After
attenuation correction the119885119867-119885DR scatter plots were close to
the calculated valuesThis indicated that119885119867and119885DR over the
gauge sites were corrected with reliable values
42 Variability of the Rain Fields The Spearman correlationcoefficient of the rainfall rate from the 10 spatially dis-tributed rain gauges was calculated with accumulation timeand separation distance (Figure 14) The overall Spearmancorrelation coefficient decreased with increasing separationdistance at all accumulation timesThis tendency was similarto the result from Habib and Krajewski [24] However thecorrelation value was higher in this study Note that the gaugetip resolution in Habib and Krajewski [24] was 0254mmwhile it was 02mm in this study In addition the rainfall ratesfrom the gauge data used in this study mainly consisted ofmoderate rainfall rates while Habib and Krajewski [24] useddata containing heavy rainfall rates
For accumulation times of less than 10min the corre-lation at the same distance was dramatically increased with
10 Advances in Meteorology
32km
21 km
11 km
20km40km60km
18
14
00
02
04
06
26
40
minus10
22
minus20
30
10
50
minus30
(a)
32 km
21 km
11 km
20km40km60km
18
14
00
02
04
06
26
40
minus10
22
minus20
30
10
50
minus30
(b)
Figure 9 PPIs of measured (a) and attenuation corrected (b) 119885DR of NIMS-XPOL for the case in Figure 5
increasing accumulation time and the slope of the correlationfunction was steep This was due to the reduction of theinstrumental errors of the gauge data by time accumulationThis indicated that rainfall accumulated for less than 10minfrom single gauge data cannot represent the rainfall over itssurrounding area due to instrumental errors For accumula-tion times longer than 10min the slope became less steepThis reflected the remaining variability of the rain field Inaddition the accumulated rainfall rate of about 10min wassuitable for minimizing the instrumental uncertainty whilepreserving the variation in the rain field The correlationreached about 098 at 05 km in 10min of accumulationand this approached 098 at 10 km at 60min accumulationTherefore a single rain gauge mostly represented the rainfallrate within 05 km and 10 km with 10min and 60minaccumulation respectively
The NB between an areal averaged rainfall rate ⟨119866⟩ anda single gauge rainfall rate 119866 is shown with accumulationtime in Figures 15(a) and 16(a) Where the average area was038 km2 (corresponding to 9 pixels of NIMS-XPOL radarat 16 km range) the NB reached a 6ndash8 maximum Wherethe averaged area is 6 km2 (corresponding to the coverage forall of the gauges) the NB reached about a 10 maximumFurthermore the NB hardly changed with accumulationtime Note that this statistic was derived after removinginstrumental bias This indicated that the rainfall rate froma single gauge was biased due to the spatial variability of therain fields and this bias was maintained with accumulationtime
The NSD of a single gauge for a mean area of 038 km2was over 40 and it reached about 70ndash90 for a mean areaof 6 km2 at an accumulation time of 1min (see Figures 15(b)and 16(b)) This statistic included instrumental errors (about
40ndash60 at 16mmhminus1 and 60ndash80 at 10mmhminus1 as shownin the Appendix) as well as the spatial variability of the rainfields This indicated that the variance between the point andarea values from the gauge data at short accumulation timeswas similar to the errors due to the instrumental uncertaintyof the rain gaugeTheNSD in 60min accumulation decreasedto over 10 and about 15ndash45 with a mean area of 038 km2and 6 km2 respectively including the NSD of 6ndash10 due toinstrumental uncertainty This indicated that the rainfall ratefrom single gauges may differ significantly from the actualrainfall rate due to the variability of the rain fields and theinstrumental uncertainty in short accumulation times
43 Verification of Rainfall Retrieval The rainfall rate fromthe NIMS-XPOL radar measurements was retrieved withtwo rainfall rate estimators 119877(119885
119867) and 119877(119885
119867 119885DR) The
estimated rainfall rate was verified through comparison withthe rainfall rate (119866) from the rain gauge data These twoestimators were applied before and after the attenuation cor-rection First 119877(038 km2) was compared with 119877(007 km2)at 10min accumulation (Figure 17) Using 119877(119885
119867) the NB
NSD rmse and r rmse without the attenuation correctionwere significantly large at 70 127 127 (mmhr) and78 respectively However they were reduced to minus152 521 (mmhr) and 32 respectively after applying theattenuation correction Using 119877(119885
119867 119885DR) the NB was just
3 NSD was 49 rmse was 496 (mmhr) and r rmsewas 31 before applying attenuation correction This wassimilar to the result from 119877(119885
119867) after attenuation correction
However there was no significant change in119877(119885119867 119885DR)with
attenuation correction This means that 119877(119885119867 119885DR) was less
sensitive to rain attenuation than was 119877(119885119867) because 119885
119867is
proportional to 119877 and 119885DR is inversely proportional to 119877
Advances in Meteorology 11
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 71∘
minus10
minus5
0
5
10
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 160∘
minus10
minus5
0
5
10
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 334∘
minus10
minus5
0
5
10
(c)
Figure 10 Range profiles of measured (line) calibration bias 120588HV threshold and isolated point removing (dot) and attenuation corrected(dash) 119885DR at 71∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
in this rain estimator (see Table 3) Furthermore comparingthe analysis of variance results between areal (038 km2) andpoint (007m2) values from the gauge data (Section 42)half of the errors after attenuation correction were producedby the small scale variability of the rain fields and theinstrumental uncertainty of the rain gauge
A similar comparison is shown in Figure 18 for a 60minaccumulation Using 119877(119885
119867) the NB NSD rmse and r rmse
were decreased from minus65 86 71 (mmhr) and 70to minus5 34 28 (mmhr) and 28 by the attenuationcorrection respectively Using 119877(119885
119867 119885DR) the NB NSD
rmse and r rmse were only slightly decreased after theattenuation correction However the NSD rmse and r rmsewere reduced when compared at 10min accumulation In thesame accumulation time the errors in the gauge data reachedto over 15
This comparison was affected by different sampling areasof the gauges and the radar Therefore the areal averagevalues were compared (Figure 19) Using119877(119885
119867)with the areal
averaged radar measurements the NB was reduced fromminus69 to 7 NSD was reduced from 122 to 42 rmsewas reduced from 39 (mmhr) to 13 (mmhr) and r rmse
was reduced from 77 to 26 at an accumulation time of10min Using 119877(119885
119867 119885DR) the NB NSD rmse and r rmse
were decreased fromminus24 67 21 (mmhr) and 42 to 539 12 (mmhr) and 24 respectively
TheNB and r rmsewithmatched sampling areas betweenthe gauge and the radar are shown in Figures 20 and 21 Thereduction of NB due to temporal accumulation was smallin both of the rain estimators The NB was reduced to lessthan 7 by attenuation correction for all accumulation timeswith both119877(119885
119867) and119877(119885
119867 119885DR)The attenuation correction
effect was significantly greater in119877(119885119867) Similarly the r rmse
was reduced significantly in 119877(119885119867) by attenuation correction
(from 68 to 25 at a 60min accumulation) The reductionof the r rmse due to temporal accumulation was larger in119877(119885119867 119885DR) because the significant observational noise of
119885DR was smoothed The r rmse was decreased from 18to 15 in 119877(119885
119867 119885DR) at a 60min accumulation time after
attenuation correction
5 Conclusions
In this study an attenuation correction algorithm based onΦDP was developed to improve the accuracy of the rainfall
12 Advances in Meteorology
1st gauge site 2nd gauge site
3rd gauge site 4th gauge site
5th gauge site 6th gauge site
7th gauge site 8th gauge site
9th gauge site 10th gauge site
17 18 19 20 2116Time (h)
17 18 19 20 2116Time (h)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
UncorrectedCorrected
UncorrectedCorrected
Figure 11 Time series of measured (red circle) and attenuation corrected (blue circle) 119885119867at the 10 rain gauge sites
Advances in Meteorology 13
UncorrectedCorrected
UncorrectedCorrected
17 18 19 20 2116Time (h)
17 18 19 20 2116Time (h)
1st gauge site 2nd gauge site
3rd gauge site 4th gauge site
5th gauge site 6th gauge site
7th gauge site 8th gauge site
9th gauge site 10th gauge site
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
Figure 12 Time series of measured (red circle) and attenuation corrected (blue circle) 119885DR at the 10 rain gauge sites
14 Advances in Meteorology
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Uncorrected(a)
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Corrected Simulated from DSD(b)
Figure 13 Scatter plots of 119885119867and 119885DR before (a) and after (b) attenuation correction at the 10 rain gauge sites DSD observation results (dot
[21])
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
04
05
06
07
08
09
10
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
Accum time = 1min
Accum time = 10min
Accum time = 2min
Accum time = 20min
Accum time = 5min
Accum time = 60min
Figure 14 The Spearman correlation coefficient 119903 between data from two rain gauges at different accumulation times
estimation using X-band radar This algorithm was evalu-ated with radar measurements from NIMS-XPOL Howevermeasured ΨDP contained 120575 and observation noise as wellas ΦDP Furthermore the ΨDP offset varied with azimuthangles and decreased near the radar when rain cells were
above the radar In this study ΦDP was estimated by aniterative filtering technique and by adjusting with an offsetdetermined as the minimum value ofΦDP within 3 km of theradar As a result estimated ΦDP monotonically increasedThe path integrated value of119860
119867and119860DP was calculated from
Advances in Meteorology 15
2nd 6th 9th gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(038km2) versus G
(a) Normalized bias
⟨G⟩(038km2) versus G
2nd 6th 9th gaugesMean NSD
Mean R
0
20
40
60
80
100
NSD
()
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 15 (a) The NB and (b) NSD between ⟨119866⟩(038 km2) and 119866 from three rain gauges with accumulation time ⟨119866⟩ is the temporallyaveraged rainfall rate of ⟨119866⟩
All gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(6km2) versus G
(a) Normalized bias
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
0
20
40
60
80
100
NSD
()
All gaugesMean NSD
Mean R
⟨G⟩(6km2) versus G
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 16 (a) The NB and (b) NSD between ⟨119866⟩(6 km2) and 119866 from 10 rain gauges with accumulation time ⟨119866⟩ is the temporally averaged⟨119866⟩ with accumulation time
estimated ΦDP using 119860119867-ΦDP and 119860DP-ΦDP relationships
derived from the DSDs data Measured 119885119867
and 119885DR werecorrected with the calculated path integrated 119860
119867and 119860DP
by each gate The attenuated signal patterns of 119885119867and 119885DR
were restored to a typical rain echo pattern However theattenuation corrected 119885
119867and 119885DR fields were not matched
at distant ranges This was likely because of contaminationdue to beam broadening and the accumulation of errorswith increasing range Comparing the 119885
119867-119885DR scatter plots
over the 10 rain gauge sites with the results obtained froma scattering simulation with the DSD data the attenuation
corrected 119885119867-119885DR scatter plots are in good agreement with
the 119885119867-119885DR scatter plots from the DSDs
The validation for the improvement of accuracy in therainfall estimation through attenuation correction was per-formed by comparing with dense rain gauge network data119877(119885119867) from the attenuated (observed)119885
119867produced a severe
underestimation with an NB NSD rmse and r rmse of70 127 127 (mmhr) and 78 at 10min accumulationrespectively while 119877(119885
119867) from the attenuation corrected 119885
119867
showed better agreement with the gauge measurements withan NB NSD rmse and r rmse of minus1 52 521 (mmhr)
16 Advances in Meteorology
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus070
NSD = 127
rmse = 1279
corr = 0858
Corrected
NB = minus001
NSD = 052
rmse = 521
corr = 0927
r_rmse = 0786 r_rmse = 0320
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = 003
NSD = 049
rmse = 496
r_rmse = 0305
corr = 0936
Corrected
NB = 001
NSD = 050
rmse = 502
r_rmse = 0308
corr = 0936
(b) 119877(119885119867119885DR)
Figure 17 The scatter plots of 119877(038 km2 10min) versus 119866(007m2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
CorrectedUncorrected
R(m
mh
)
0
10
20
30
40
10 20 30 400G (mmh)
Uncorrected
NB = minus065
NSD = 086
rmse = 706
r_rmse = 0702
corr = 0878
Corrected
NB = minus005
NSD = 034
rmse = 281
r_rmse = 0279
corr = 0866
(a) 119877(119885119867)
CorrectedUncorrected
R(m
mh
)
10 20 30 400G (mmh)
0
10
20
30
40
Uncorrected
NB = 008
NSD = 023
rmse = 189
r_rmse = 0188
corr = 0942
Corrected
NB = 007
NSD = 025
rmse = 205
r_rmse = 0203
corr = 0940
(b) 119877(119885119867119885DR)
Figure 18 The scatter plots of 119877(038 km2 60min) versus 119866(007m2 60min) (a) using 119885119867and (b) using 119885
119867and 119885DR
and 32 at 10min accumulation respectively At 60minaccumulation theNBNSD rmse and r rmsewere decreasedfrom minus65 86 71 (mmhr) and 70 to minus5 3428 (mmhr) and 28 respectively Using 119877(119885
119867 119885DR) the
NB NSD rmse and r rmse were only slightly reduced even if
attenuation was corrected but were smaller than when using119877(119885119867) This indicated that 119877(119885
119867 119885DR) was less sensitive to
the attenuation than 119877(119885119867) In addition half of the errors
that remained after attenuation correction were caused by thesmall scale variability of the rain fields and the instrumental
Advances in Meteorology 17
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus069
NSD = 122
rmse = 391
r_rmse = 0770
corr = 0893
Corrected
NB = 007
NSD = 042
rmse = 133
r_rmse = 0262
corr = 0960
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = minus024
NSD = 067
rmse = 215
r_rmse = 0424
corr = 0718
Corrected
NB = 005
NSD = 039
rmse = 124
r_rmse = 0245
corr = 0891
(b) 119877(119885119867119885DR)
Figure 19 The scatter plots of 119877(6 km2 10min) versus 119866(6 km2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
R(ZH)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(a) 119877(119885119867)
R(ZH ZDR)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(b) 119877(119885119867119885DR)
Figure 20 The normalized bias of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and 119885DR The red
point-line and the blue point-line indicate without and with the attenuation correction respectively
uncertainty of the rain gauges Using an areal averagedvalue the NB was reduced to less than 7 with attenua-tion correction using either rainfall estimator regardless ofaccumulation timeThe r rmse were significantly reduced byattenuation correction in 119877(119885
119867) and the reduction of r rmse
with 119877(119885119867 119885DR) was even more dramatic
As summarized above the rain attenuation for 119885119867
and 119885DR in the X-band was corrected by ΦDP and theresults showed a good agreement with the value obtainedby scattering simulations with the DSD datasets The resultsof comparison with rain gauge data also showed a signifi-cant improvement in the accuracy of rainfall estimation by
18 Advances in Meteorology
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10r_
rmse
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH)
(a) 119877(119885119867)
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10
r_rm
se
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH ZDR)
(b) 119877(119885119867119885DR)
Figure 21 The relative root-mean square error of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and
119885DR The red point-line and the blue point-line indicate without and with the attenuation correction respectively
attenuation correction which supports the quantitative useof the X-band radar in rain data retrieval
Appendix
A Instrumental Uncertainty of the Rain Gauge
The instrumental uncertainty of the TB rain gauge is asso-ciated with improper calibration leakage and precipitationmissing the gauge due to the measurement principle andwetting and evaporation losses In addition external sourcessuch as wind turbulence and installed position also con-tribute to the instrumental uncertainty Ciach [17] reportedthat the TB gauge data suffers a significant uncertainty asso-ciated with random differences between closely collocatedTB gauges through the analysis of the error as a functionof the rainfall intensity for three timescales In addition theimproper calibration induces a bias in the rainfall measure-ment and this bias cannot be removed by increasing accu-mulation timeTherefore the proper correction or processingof TB gauge data is necessary to obtain accurate and reliablereading of the rainfall amount
In this appendix instrumental biases were calculatedto reduce the instrumental uncertainty through an idealexperiment in the laboratory and a field intercomparisonwith 10 TB rain gauges In addition the random errors werequantitatively calculated as a function of rainfall rate andaccumulation time
A1 Correction of Instrumental Bias The instrumental bias119861119868(119894) of the 119894th gauge was defined as a combination
of the absolute bias 119861119860(ref) of the reference gauge and
the relative bias 119861119877(119894) of the 119894th gauge according to the
following equation
119861119868 (119894) = 119861119860 (ref) times 119861119877 (119894) (A1)
119861119860(ref) signified the ratio of the calculated value from
the instrumental resolution of the reference gauge to the truevalue measured by other instruments
119861119860 (ref) =
rainfalltruthrainfallgauge (ref)
(A2)
In this study 119861119860(ref) was derived by an ideal experiment
using an electronic scale First the reference gauge was puton the scale and the scale was initialized The water dropsflowed out into the receiving orifice of the reference gaugeuntil the tipping number became 60Themeasured weight of60 tips was 2144 g while the theoretical (or expected) valuewas 2244 g for the reference gauge (receiving orifice size of1545 cm 60 tips and 02mm tip resolution) Therefore theabsolute bias of the reference gauge from (A2) was about0955 This indicates that the rainfall amount measured bythis reference gauge was overestimated by about 45 andthis bias should be corrected prior to any interpretation
119861119877(119894) represented the ratio of the observed (recorded)
value from the 119894th gauge to the observed (recorded) valuefrom the reference gauge
119861119877 (119894) =
rainfallgauge (ref)rainfallgauge (119894)
(A3)
119861119877(119894) was calculated from ten collocated rain gauges
including the reference gauge located at an observation field
Advances in Meteorology 19
Table 4 The instrumental and relative biases of the 10 tipping bucket rain gauges
Gauge ID RG1 RG2 RG3 RG4 RG5 RG6 RG7 RG8 RG9 RG10Instrumental bias 094 108 099 100 096 104 098 102 108 098Relative bias 098 113 103 105 100 109 102 107 113 102
013610152230406080120150200250300400
()
Rain
rate
(mm
h)
100101
Accumulation time (min)
10
100NSD ()
Figure 22 Two-dimensional normalized standard deviationbetween the standard rainfall and each rain gauge
at KNUKoreaThe observational period for intercomparisonwas from October 2009 to June 2010 (14 rainy days) andthe total accumulated rainfall for this period was about300mm The results are presented in Table 4 All relativebiases were mainly larger than 10 This indicated that themost of gauges underestimated the rainfall amount comparedto the reference rain gauge The relative bias of the referencegauge (119894 = 5) was 10
Finally 119861119860(119894) was calculated from (A1) (see Table 4) All
instrumental biases were between 090 and 11 and the meanvalue of the biases was nearly 10 This indicated that therainfall rate from a single gauge data had a maximum 10bias These biases were corrected for each gauge to removeexisting instrumental errors
A2 Quantification of Random Errors To quantify the ran-dom error due to instrumental uncertainty the NSD of theten gauges was calculated with accumulation time and rainrate using data from an intercomparison The NSD of the 119894thgauge compared to the reference gauge was derived for 20rain rate classes (Δ119877) and 17 accumulation time classes (Δ119905)as follows
NSD119894 (Δ119905 Δ119877)
=
(1119899)radicsum119899minus1
119896=0[1198661015840
119870(Δ119905 Δ119877) minus 119866119894119896 (Δ119905 Δ119877)]
2
1198661015840
(A4)
where 1198661015840 and 119866 indicate the reference and individual gaugesrespectively The overbar means the temporal average ofthe rainfall rate The NSDs of nine different gauges (exceptthe reference gauge) were averaged to obtain the two-dimensional average NSD (Figure 22) The average NSD
tended to decrease with increasing accumulation time andwith increasing rain rate The NSD was about 10ndash15when accumulation time was 10min and the rain rate was10mmhminus1 The instrumental errors were quite significant forlonger accumulation times and higher rainfall rates Theseerrors should be used as a guideline when a single gauge isused as a reference for any validation study
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper has been revised and extended fromMs Young-AOhrsquos MS thesis [25] of Kyungpook National University Thisresearch is supported by Development and Application ofCross Governmental Dual-Pol Radar Harmonization (WRC-2013-A-1) project of theWeather Radar Center KoreaMeteo-rological AdministrationThis work was funded by the KoreaMeteorological Administration Research and DevelopmentProgram under Grant KMIPA2015-1010
References
[1] V N Bringi V Chandrasekar N Balakrishnan and D SZrnic ldquoAn examination of propagation effects in rainfall onradar measurements at microwave frequenciesrdquo Journal ofAtmospheric and Oceanic Technology vol 7 no 6 pp 829ndash8401990
[2] A Huggel W Schmid and AWaldvogel ldquoRaindrop size distri-butions and the radar bright bandrdquo Journal of AppliedMeteorol-ogy vol 35 no 10 pp 1688ndash1701 1996
[3] G W Lee Errors in rain measurement by radar effect of vari-ability of drop size distributions [PhD thesis]McGill University2003
[4] J D Tuttle and R E Rinehart ldquoAttenuation correction in dual-wavelength analysesrdquo Journal of ClimateampAppliedMeteorologyvol 22 no 11 pp 1914ndash1921 1983
[5] S-G Park V N Bringi V Chandrasekar M Maki andK Iwanami ldquoCorrection of radar reflectivity and differentialreflectivity for rain attenuation at X band Part I theoretical andempirical basisrdquo Journal of Atmospheric and Oceanic Technol-ogy vol 22 no 11 pp 1621ndash1632 2005
[6] G Scarchilli E Gorgucci V Chandrasekar and T A SeligaldquoRainfall estimation using polarimetric techniques at C-bandfrequenciesrdquo Journal of Applied Meteorology vol 32 no 6 pp1150ndash1160 1993
[7] M N Anagnostou E N Anagnostou and J VivekanandanldquoCorrection for rain path specific and differential attenuation ofX-band dual-polarization observationsrdquo IEEE Transactions onGeoscience and Remote Sensing vol 44 no 9 pp 2470ndash24802006
20 Advances in Meteorology
[8] J Kalogiros M N Anagnostou E N Anagnostou M Mon-topoli E Picciotti and F S Marzano ldquoEvaluation of a newpolarimetric algorithm for rain-path attenuation correction ofX-band radar observations against disdrometerrdquo IEEE Transac-tions on Geoscience and Remote Sensing vol 52 no 2 pp 1369ndash1380 2014
[9] W-Y Chang J Vivekanandan and T-C C Wang ldquoEstimationof X-band polarimetric radar attenuation and measurementuncertainty using a variational methodrdquo Journal of AppliedMeteorology and Climatology vol 53 no 4 pp 1099ndash1119 2014
[10] J Hubbert and V N Bringi ldquoAn iterative filtering technique forthe analysis of copolar differential phase and dual-frequencyradar measurementsrdquo Journal of Atmospheric and OceanicTechnology vol 12 no 3 pp 643ndash648 1995
[11] K V Mishra W F Krajewski R Goska et al ldquoDeploy-ment and performance analyses of high-resolution iowa XPOLradar system during the NASA IFloodS campaignrdquo Journal ofHydrometeorology vol 17 no 2 pp 455ndash479 2016
[12] M Thurai G J Huang V N Bringi W L Randeu and MSchonhuber ldquoDrop shapes model comparisons and calcu-lations of polarimetric radar parameters in rainrdquo Journal ofAtmospheric and Oceanic Technology vol 24 no 6 pp 1019ndash1032 2007
[13] M D Humphrey J D Istok J Y Lee J A Hevesi and AL Flint ldquoA new method for automated dynamic calibration oftipping-bucket rain gaugesrdquo Journal of Atmospheric andOceanicTechnology vol 14 no 6 pp 1513ndash1519 1997
[14] E HabibW F Krajewski V Nespor andA Kruger ldquoNumericalsimulation studies of rain gage data correction due to windeffectrdquo Journal of Geophysical ResearchAtmospheres vol 104 no16 Article ID 1999JD900228 pp 19723ndash19733 1999
[15] G J Ciach and W F Krajewski ldquoOn the estimation of radarrainfall error variancerdquoAdvances inWater Resources vol 22 no6 pp 585ndash595 1999
[16] E Habib W F Krajewski and A Kruger ldquoSampling errors oftipping-bucket rain gaugemeasurementsrdquo Journal of HydrologicEngineering vol 6 no 2 pp 159ndash166 2001
[17] G J Ciach ldquoLocal random errors in tipping-bucket rain gaugemeasurementsrdquo Journal of Atmospheric andOceanic Technologyvol 20 no 5 pp 752ndash759 2003
[18] EHabibW F Krajewski andG J Ciach ldquoEstimation of rainfallinterstation correlationrdquo Journal ofHydrometeorology vol 2 no6 pp 621ndash629 2001
[19] NIMS ldquoStudy on estimation of rainfall amount and classifi-cation of hydrometeor species using dual-polarimetric radar(III)rdquo pp 1ndash20 2010
[20] E Gorgucci V Chandrasekar and L Baldini ldquoCorrection ofX-band radar observation for propagation effects based on theself-consistency principlerdquo Journal of Atmospheric and OceanicTechnology vol 23 pp 1668ndash1681 2006
[21] NIMS ldquoStudy on estimating raindrop size distribution forquantitative estimation of radar rainfall amountrdquo pp 1ndash55 2009
[22] NIMS ldquoQuantitative precipitation estimation and hydrometeoridentification using dual-polarization radar-Phase IIrdquo pp 1ndash742009
[23] S-G Park M Maki K Iwanami V N Bringi and V Chan-drasekar ldquoCorrection of radar reflectivity and differential reflec-tivity for rain attenuation at X band Part II Evaluation andapplicationrdquo Journal of Atmospheric and Oceanic Technologyvol 22 no 11 pp 1633ndash1655 2005
[24] E Habib and W F Krajewski ldquoUncertainty analysis of theTRMM ground-validation radar-rainfall products applicationto the TEFLUN-B field campaignrdquo Journal of Applied Meteorol-ogy vol 41 no 5 pp 558ndash572 2002
[25] Y-A Oh Rainfall estimation from X-band dual-polarizationradar measurements effects of attenuation correction and evalu-ation with a dense rain gauge network [MS thesis] KyungpookNational University 2011
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom
Applied ampEnvironmentalSoil Science
Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal of
Geophysics
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
Advances in Meteorology 9
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 71∘
minus20
0
20
40
60
80
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 160∘
minus20
0
20
40
60
80
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZH (dBZ) azimuth angle = 334∘
minus20
0
20
40
60
80
(c)
Figure 8 Range profiles of measured (line) calibration bias 120588HV threshold and isolated point removing (dot) and attenuation corrected(dash) 119885
119867at 71∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
the attenuation signature in119885DR was determined Ideally the119885DR value for rain should be positive since rain drops areoblate due to aerodynamic force However themeasured119885DRfield in Figure 9 showed a negative value over wide areasAfter correcting the attenuation the119885DR field was larger than0 dB and showed a realistic rain patternThis is also presentedin the range profile of Figure 10The largest difference of119885DRbetween before and after attenuation correction at 2101UTCwas about 8 dB The 119885DR values around the 220∘ azimuthangle beyond 40 km were the largest They did not matchwith the area of the largest 119885DR values (around the 210∘azimuth angle with a range of 20 km and 40 km) This wasan unexpected result because 119885DR is proportional to 119885
119867in
theory This mismatch between 119885119867
and 119885DR at far rangemay be due to contamination due to beam broadening andaccumulated errors due to uncertainty in the attenuationcorrection with different ranges
Figures 11 and 12 show the time series of 119885119867
and119885DR at the ten gauge sites before and after applying theattenuation correction The differences between uncorrectedand corrected values existed after 1800UTC and were at amaximumaround 2100UTCThis attenuation correctionwas
evaluated through comparison with the theoretical values of119885119867and 119885DR In Figure 13 the uncorrected 119885
119867-119885DR scatter
plots are apart from the theoretical119885119867-119885DR values [23] After
attenuation correction the119885119867-119885DR scatter plots were close to
the calculated valuesThis indicated that119885119867and119885DR over the
gauge sites were corrected with reliable values
42 Variability of the Rain Fields The Spearman correlationcoefficient of the rainfall rate from the 10 spatially dis-tributed rain gauges was calculated with accumulation timeand separation distance (Figure 14) The overall Spearmancorrelation coefficient decreased with increasing separationdistance at all accumulation timesThis tendency was similarto the result from Habib and Krajewski [24] However thecorrelation value was higher in this study Note that the gaugetip resolution in Habib and Krajewski [24] was 0254mmwhile it was 02mm in this study In addition the rainfall ratesfrom the gauge data used in this study mainly consisted ofmoderate rainfall rates while Habib and Krajewski [24] useddata containing heavy rainfall rates
For accumulation times of less than 10min the corre-lation at the same distance was dramatically increased with
10 Advances in Meteorology
32km
21 km
11 km
20km40km60km
18
14
00
02
04
06
26
40
minus10
22
minus20
30
10
50
minus30
(a)
32 km
21 km
11 km
20km40km60km
18
14
00
02
04
06
26
40
minus10
22
minus20
30
10
50
minus30
(b)
Figure 9 PPIs of measured (a) and attenuation corrected (b) 119885DR of NIMS-XPOL for the case in Figure 5
increasing accumulation time and the slope of the correlationfunction was steep This was due to the reduction of theinstrumental errors of the gauge data by time accumulationThis indicated that rainfall accumulated for less than 10minfrom single gauge data cannot represent the rainfall over itssurrounding area due to instrumental errors For accumula-tion times longer than 10min the slope became less steepThis reflected the remaining variability of the rain field Inaddition the accumulated rainfall rate of about 10min wassuitable for minimizing the instrumental uncertainty whilepreserving the variation in the rain field The correlationreached about 098 at 05 km in 10min of accumulationand this approached 098 at 10 km at 60min accumulationTherefore a single rain gauge mostly represented the rainfallrate within 05 km and 10 km with 10min and 60minaccumulation respectively
The NB between an areal averaged rainfall rate ⟨119866⟩ anda single gauge rainfall rate 119866 is shown with accumulationtime in Figures 15(a) and 16(a) Where the average area was038 km2 (corresponding to 9 pixels of NIMS-XPOL radarat 16 km range) the NB reached a 6ndash8 maximum Wherethe averaged area is 6 km2 (corresponding to the coverage forall of the gauges) the NB reached about a 10 maximumFurthermore the NB hardly changed with accumulationtime Note that this statistic was derived after removinginstrumental bias This indicated that the rainfall rate froma single gauge was biased due to the spatial variability of therain fields and this bias was maintained with accumulationtime
The NSD of a single gauge for a mean area of 038 km2was over 40 and it reached about 70ndash90 for a mean areaof 6 km2 at an accumulation time of 1min (see Figures 15(b)and 16(b)) This statistic included instrumental errors (about
40ndash60 at 16mmhminus1 and 60ndash80 at 10mmhminus1 as shownin the Appendix) as well as the spatial variability of the rainfields This indicated that the variance between the point andarea values from the gauge data at short accumulation timeswas similar to the errors due to the instrumental uncertaintyof the rain gaugeTheNSD in 60min accumulation decreasedto over 10 and about 15ndash45 with a mean area of 038 km2and 6 km2 respectively including the NSD of 6ndash10 due toinstrumental uncertainty This indicated that the rainfall ratefrom single gauges may differ significantly from the actualrainfall rate due to the variability of the rain fields and theinstrumental uncertainty in short accumulation times
43 Verification of Rainfall Retrieval The rainfall rate fromthe NIMS-XPOL radar measurements was retrieved withtwo rainfall rate estimators 119877(119885
119867) and 119877(119885
119867 119885DR) The
estimated rainfall rate was verified through comparison withthe rainfall rate (119866) from the rain gauge data These twoestimators were applied before and after the attenuation cor-rection First 119877(038 km2) was compared with 119877(007 km2)at 10min accumulation (Figure 17) Using 119877(119885
119867) the NB
NSD rmse and r rmse without the attenuation correctionwere significantly large at 70 127 127 (mmhr) and78 respectively However they were reduced to minus152 521 (mmhr) and 32 respectively after applying theattenuation correction Using 119877(119885
119867 119885DR) the NB was just
3 NSD was 49 rmse was 496 (mmhr) and r rmsewas 31 before applying attenuation correction This wassimilar to the result from 119877(119885
119867) after attenuation correction
However there was no significant change in119877(119885119867 119885DR)with
attenuation correction This means that 119877(119885119867 119885DR) was less
sensitive to rain attenuation than was 119877(119885119867) because 119885
119867is
proportional to 119877 and 119885DR is inversely proportional to 119877
Advances in Meteorology 11
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 71∘
minus10
minus5
0
5
10
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 160∘
minus10
minus5
0
5
10
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 334∘
minus10
minus5
0
5
10
(c)
Figure 10 Range profiles of measured (line) calibration bias 120588HV threshold and isolated point removing (dot) and attenuation corrected(dash) 119885DR at 71∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
in this rain estimator (see Table 3) Furthermore comparingthe analysis of variance results between areal (038 km2) andpoint (007m2) values from the gauge data (Section 42)half of the errors after attenuation correction were producedby the small scale variability of the rain fields and theinstrumental uncertainty of the rain gauge
A similar comparison is shown in Figure 18 for a 60minaccumulation Using 119877(119885
119867) the NB NSD rmse and r rmse
were decreased from minus65 86 71 (mmhr) and 70to minus5 34 28 (mmhr) and 28 by the attenuationcorrection respectively Using 119877(119885
119867 119885DR) the NB NSD
rmse and r rmse were only slightly decreased after theattenuation correction However the NSD rmse and r rmsewere reduced when compared at 10min accumulation In thesame accumulation time the errors in the gauge data reachedto over 15
This comparison was affected by different sampling areasof the gauges and the radar Therefore the areal averagevalues were compared (Figure 19) Using119877(119885
119867)with the areal
averaged radar measurements the NB was reduced fromminus69 to 7 NSD was reduced from 122 to 42 rmsewas reduced from 39 (mmhr) to 13 (mmhr) and r rmse
was reduced from 77 to 26 at an accumulation time of10min Using 119877(119885
119867 119885DR) the NB NSD rmse and r rmse
were decreased fromminus24 67 21 (mmhr) and 42 to 539 12 (mmhr) and 24 respectively
TheNB and r rmsewithmatched sampling areas betweenthe gauge and the radar are shown in Figures 20 and 21 Thereduction of NB due to temporal accumulation was smallin both of the rain estimators The NB was reduced to lessthan 7 by attenuation correction for all accumulation timeswith both119877(119885
119867) and119877(119885
119867 119885DR)The attenuation correction
effect was significantly greater in119877(119885119867) Similarly the r rmse
was reduced significantly in 119877(119885119867) by attenuation correction
(from 68 to 25 at a 60min accumulation) The reductionof the r rmse due to temporal accumulation was larger in119877(119885119867 119885DR) because the significant observational noise of
119885DR was smoothed The r rmse was decreased from 18to 15 in 119877(119885
119867 119885DR) at a 60min accumulation time after
attenuation correction
5 Conclusions
In this study an attenuation correction algorithm based onΦDP was developed to improve the accuracy of the rainfall
12 Advances in Meteorology
1st gauge site 2nd gauge site
3rd gauge site 4th gauge site
5th gauge site 6th gauge site
7th gauge site 8th gauge site
9th gauge site 10th gauge site
17 18 19 20 2116Time (h)
17 18 19 20 2116Time (h)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
UncorrectedCorrected
UncorrectedCorrected
Figure 11 Time series of measured (red circle) and attenuation corrected (blue circle) 119885119867at the 10 rain gauge sites
Advances in Meteorology 13
UncorrectedCorrected
UncorrectedCorrected
17 18 19 20 2116Time (h)
17 18 19 20 2116Time (h)
1st gauge site 2nd gauge site
3rd gauge site 4th gauge site
5th gauge site 6th gauge site
7th gauge site 8th gauge site
9th gauge site 10th gauge site
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
Figure 12 Time series of measured (red circle) and attenuation corrected (blue circle) 119885DR at the 10 rain gauge sites
14 Advances in Meteorology
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Uncorrected(a)
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Corrected Simulated from DSD(b)
Figure 13 Scatter plots of 119885119867and 119885DR before (a) and after (b) attenuation correction at the 10 rain gauge sites DSD observation results (dot
[21])
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
04
05
06
07
08
09
10
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
Accum time = 1min
Accum time = 10min
Accum time = 2min
Accum time = 20min
Accum time = 5min
Accum time = 60min
Figure 14 The Spearman correlation coefficient 119903 between data from two rain gauges at different accumulation times
estimation using X-band radar This algorithm was evalu-ated with radar measurements from NIMS-XPOL Howevermeasured ΨDP contained 120575 and observation noise as wellas ΦDP Furthermore the ΨDP offset varied with azimuthangles and decreased near the radar when rain cells were
above the radar In this study ΦDP was estimated by aniterative filtering technique and by adjusting with an offsetdetermined as the minimum value ofΦDP within 3 km of theradar As a result estimated ΦDP monotonically increasedThe path integrated value of119860
119867and119860DP was calculated from
Advances in Meteorology 15
2nd 6th 9th gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(038km2) versus G
(a) Normalized bias
⟨G⟩(038km2) versus G
2nd 6th 9th gaugesMean NSD
Mean R
0
20
40
60
80
100
NSD
()
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 15 (a) The NB and (b) NSD between ⟨119866⟩(038 km2) and 119866 from three rain gauges with accumulation time ⟨119866⟩ is the temporallyaveraged rainfall rate of ⟨119866⟩
All gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(6km2) versus G
(a) Normalized bias
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
0
20
40
60
80
100
NSD
()
All gaugesMean NSD
Mean R
⟨G⟩(6km2) versus G
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 16 (a) The NB and (b) NSD between ⟨119866⟩(6 km2) and 119866 from 10 rain gauges with accumulation time ⟨119866⟩ is the temporally averaged⟨119866⟩ with accumulation time
estimated ΦDP using 119860119867-ΦDP and 119860DP-ΦDP relationships
derived from the DSDs data Measured 119885119867
and 119885DR werecorrected with the calculated path integrated 119860
119867and 119860DP
by each gate The attenuated signal patterns of 119885119867and 119885DR
were restored to a typical rain echo pattern However theattenuation corrected 119885
119867and 119885DR fields were not matched
at distant ranges This was likely because of contaminationdue to beam broadening and the accumulation of errorswith increasing range Comparing the 119885
119867-119885DR scatter plots
over the 10 rain gauge sites with the results obtained froma scattering simulation with the DSD data the attenuation
corrected 119885119867-119885DR scatter plots are in good agreement with
the 119885119867-119885DR scatter plots from the DSDs
The validation for the improvement of accuracy in therainfall estimation through attenuation correction was per-formed by comparing with dense rain gauge network data119877(119885119867) from the attenuated (observed)119885
119867produced a severe
underestimation with an NB NSD rmse and r rmse of70 127 127 (mmhr) and 78 at 10min accumulationrespectively while 119877(119885
119867) from the attenuation corrected 119885
119867
showed better agreement with the gauge measurements withan NB NSD rmse and r rmse of minus1 52 521 (mmhr)
16 Advances in Meteorology
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus070
NSD = 127
rmse = 1279
corr = 0858
Corrected
NB = minus001
NSD = 052
rmse = 521
corr = 0927
r_rmse = 0786 r_rmse = 0320
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = 003
NSD = 049
rmse = 496
r_rmse = 0305
corr = 0936
Corrected
NB = 001
NSD = 050
rmse = 502
r_rmse = 0308
corr = 0936
(b) 119877(119885119867119885DR)
Figure 17 The scatter plots of 119877(038 km2 10min) versus 119866(007m2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
CorrectedUncorrected
R(m
mh
)
0
10
20
30
40
10 20 30 400G (mmh)
Uncorrected
NB = minus065
NSD = 086
rmse = 706
r_rmse = 0702
corr = 0878
Corrected
NB = minus005
NSD = 034
rmse = 281
r_rmse = 0279
corr = 0866
(a) 119877(119885119867)
CorrectedUncorrected
R(m
mh
)
10 20 30 400G (mmh)
0
10
20
30
40
Uncorrected
NB = 008
NSD = 023
rmse = 189
r_rmse = 0188
corr = 0942
Corrected
NB = 007
NSD = 025
rmse = 205
r_rmse = 0203
corr = 0940
(b) 119877(119885119867119885DR)
Figure 18 The scatter plots of 119877(038 km2 60min) versus 119866(007m2 60min) (a) using 119885119867and (b) using 119885
119867and 119885DR
and 32 at 10min accumulation respectively At 60minaccumulation theNBNSD rmse and r rmsewere decreasedfrom minus65 86 71 (mmhr) and 70 to minus5 3428 (mmhr) and 28 respectively Using 119877(119885
119867 119885DR) the
NB NSD rmse and r rmse were only slightly reduced even if
attenuation was corrected but were smaller than when using119877(119885119867) This indicated that 119877(119885
119867 119885DR) was less sensitive to
the attenuation than 119877(119885119867) In addition half of the errors
that remained after attenuation correction were caused by thesmall scale variability of the rain fields and the instrumental
Advances in Meteorology 17
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus069
NSD = 122
rmse = 391
r_rmse = 0770
corr = 0893
Corrected
NB = 007
NSD = 042
rmse = 133
r_rmse = 0262
corr = 0960
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = minus024
NSD = 067
rmse = 215
r_rmse = 0424
corr = 0718
Corrected
NB = 005
NSD = 039
rmse = 124
r_rmse = 0245
corr = 0891
(b) 119877(119885119867119885DR)
Figure 19 The scatter plots of 119877(6 km2 10min) versus 119866(6 km2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
R(ZH)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(a) 119877(119885119867)
R(ZH ZDR)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(b) 119877(119885119867119885DR)
Figure 20 The normalized bias of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and 119885DR The red
point-line and the blue point-line indicate without and with the attenuation correction respectively
uncertainty of the rain gauges Using an areal averagedvalue the NB was reduced to less than 7 with attenua-tion correction using either rainfall estimator regardless ofaccumulation timeThe r rmse were significantly reduced byattenuation correction in 119877(119885
119867) and the reduction of r rmse
with 119877(119885119867 119885DR) was even more dramatic
As summarized above the rain attenuation for 119885119867
and 119885DR in the X-band was corrected by ΦDP and theresults showed a good agreement with the value obtainedby scattering simulations with the DSD datasets The resultsof comparison with rain gauge data also showed a signifi-cant improvement in the accuracy of rainfall estimation by
18 Advances in Meteorology
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10r_
rmse
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH)
(a) 119877(119885119867)
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10
r_rm
se
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH ZDR)
(b) 119877(119885119867119885DR)
Figure 21 The relative root-mean square error of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and
119885DR The red point-line and the blue point-line indicate without and with the attenuation correction respectively
attenuation correction which supports the quantitative useof the X-band radar in rain data retrieval
Appendix
A Instrumental Uncertainty of the Rain Gauge
The instrumental uncertainty of the TB rain gauge is asso-ciated with improper calibration leakage and precipitationmissing the gauge due to the measurement principle andwetting and evaporation losses In addition external sourcessuch as wind turbulence and installed position also con-tribute to the instrumental uncertainty Ciach [17] reportedthat the TB gauge data suffers a significant uncertainty asso-ciated with random differences between closely collocatedTB gauges through the analysis of the error as a functionof the rainfall intensity for three timescales In addition theimproper calibration induces a bias in the rainfall measure-ment and this bias cannot be removed by increasing accu-mulation timeTherefore the proper correction or processingof TB gauge data is necessary to obtain accurate and reliablereading of the rainfall amount
In this appendix instrumental biases were calculatedto reduce the instrumental uncertainty through an idealexperiment in the laboratory and a field intercomparisonwith 10 TB rain gauges In addition the random errors werequantitatively calculated as a function of rainfall rate andaccumulation time
A1 Correction of Instrumental Bias The instrumental bias119861119868(119894) of the 119894th gauge was defined as a combination
of the absolute bias 119861119860(ref) of the reference gauge and
the relative bias 119861119877(119894) of the 119894th gauge according to the
following equation
119861119868 (119894) = 119861119860 (ref) times 119861119877 (119894) (A1)
119861119860(ref) signified the ratio of the calculated value from
the instrumental resolution of the reference gauge to the truevalue measured by other instruments
119861119860 (ref) =
rainfalltruthrainfallgauge (ref)
(A2)
In this study 119861119860(ref) was derived by an ideal experiment
using an electronic scale First the reference gauge was puton the scale and the scale was initialized The water dropsflowed out into the receiving orifice of the reference gaugeuntil the tipping number became 60Themeasured weight of60 tips was 2144 g while the theoretical (or expected) valuewas 2244 g for the reference gauge (receiving orifice size of1545 cm 60 tips and 02mm tip resolution) Therefore theabsolute bias of the reference gauge from (A2) was about0955 This indicates that the rainfall amount measured bythis reference gauge was overestimated by about 45 andthis bias should be corrected prior to any interpretation
119861119877(119894) represented the ratio of the observed (recorded)
value from the 119894th gauge to the observed (recorded) valuefrom the reference gauge
119861119877 (119894) =
rainfallgauge (ref)rainfallgauge (119894)
(A3)
119861119877(119894) was calculated from ten collocated rain gauges
including the reference gauge located at an observation field
Advances in Meteorology 19
Table 4 The instrumental and relative biases of the 10 tipping bucket rain gauges
Gauge ID RG1 RG2 RG3 RG4 RG5 RG6 RG7 RG8 RG9 RG10Instrumental bias 094 108 099 100 096 104 098 102 108 098Relative bias 098 113 103 105 100 109 102 107 113 102
013610152230406080120150200250300400
()
Rain
rate
(mm
h)
100101
Accumulation time (min)
10
100NSD ()
Figure 22 Two-dimensional normalized standard deviationbetween the standard rainfall and each rain gauge
at KNUKoreaThe observational period for intercomparisonwas from October 2009 to June 2010 (14 rainy days) andthe total accumulated rainfall for this period was about300mm The results are presented in Table 4 All relativebiases were mainly larger than 10 This indicated that themost of gauges underestimated the rainfall amount comparedto the reference rain gauge The relative bias of the referencegauge (119894 = 5) was 10
Finally 119861119860(119894) was calculated from (A1) (see Table 4) All
instrumental biases were between 090 and 11 and the meanvalue of the biases was nearly 10 This indicated that therainfall rate from a single gauge data had a maximum 10bias These biases were corrected for each gauge to removeexisting instrumental errors
A2 Quantification of Random Errors To quantify the ran-dom error due to instrumental uncertainty the NSD of theten gauges was calculated with accumulation time and rainrate using data from an intercomparison The NSD of the 119894thgauge compared to the reference gauge was derived for 20rain rate classes (Δ119877) and 17 accumulation time classes (Δ119905)as follows
NSD119894 (Δ119905 Δ119877)
=
(1119899)radicsum119899minus1
119896=0[1198661015840
119870(Δ119905 Δ119877) minus 119866119894119896 (Δ119905 Δ119877)]
2
1198661015840
(A4)
where 1198661015840 and 119866 indicate the reference and individual gaugesrespectively The overbar means the temporal average ofthe rainfall rate The NSDs of nine different gauges (exceptthe reference gauge) were averaged to obtain the two-dimensional average NSD (Figure 22) The average NSD
tended to decrease with increasing accumulation time andwith increasing rain rate The NSD was about 10ndash15when accumulation time was 10min and the rain rate was10mmhminus1 The instrumental errors were quite significant forlonger accumulation times and higher rainfall rates Theseerrors should be used as a guideline when a single gauge isused as a reference for any validation study
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper has been revised and extended fromMs Young-AOhrsquos MS thesis [25] of Kyungpook National University Thisresearch is supported by Development and Application ofCross Governmental Dual-Pol Radar Harmonization (WRC-2013-A-1) project of theWeather Radar Center KoreaMeteo-rological AdministrationThis work was funded by the KoreaMeteorological Administration Research and DevelopmentProgram under Grant KMIPA2015-1010
References
[1] V N Bringi V Chandrasekar N Balakrishnan and D SZrnic ldquoAn examination of propagation effects in rainfall onradar measurements at microwave frequenciesrdquo Journal ofAtmospheric and Oceanic Technology vol 7 no 6 pp 829ndash8401990
[2] A Huggel W Schmid and AWaldvogel ldquoRaindrop size distri-butions and the radar bright bandrdquo Journal of AppliedMeteorol-ogy vol 35 no 10 pp 1688ndash1701 1996
[3] G W Lee Errors in rain measurement by radar effect of vari-ability of drop size distributions [PhD thesis]McGill University2003
[4] J D Tuttle and R E Rinehart ldquoAttenuation correction in dual-wavelength analysesrdquo Journal of ClimateampAppliedMeteorologyvol 22 no 11 pp 1914ndash1921 1983
[5] S-G Park V N Bringi V Chandrasekar M Maki andK Iwanami ldquoCorrection of radar reflectivity and differentialreflectivity for rain attenuation at X band Part I theoretical andempirical basisrdquo Journal of Atmospheric and Oceanic Technol-ogy vol 22 no 11 pp 1621ndash1632 2005
[6] G Scarchilli E Gorgucci V Chandrasekar and T A SeligaldquoRainfall estimation using polarimetric techniques at C-bandfrequenciesrdquo Journal of Applied Meteorology vol 32 no 6 pp1150ndash1160 1993
[7] M N Anagnostou E N Anagnostou and J VivekanandanldquoCorrection for rain path specific and differential attenuation ofX-band dual-polarization observationsrdquo IEEE Transactions onGeoscience and Remote Sensing vol 44 no 9 pp 2470ndash24802006
20 Advances in Meteorology
[8] J Kalogiros M N Anagnostou E N Anagnostou M Mon-topoli E Picciotti and F S Marzano ldquoEvaluation of a newpolarimetric algorithm for rain-path attenuation correction ofX-band radar observations against disdrometerrdquo IEEE Transac-tions on Geoscience and Remote Sensing vol 52 no 2 pp 1369ndash1380 2014
[9] W-Y Chang J Vivekanandan and T-C C Wang ldquoEstimationof X-band polarimetric radar attenuation and measurementuncertainty using a variational methodrdquo Journal of AppliedMeteorology and Climatology vol 53 no 4 pp 1099ndash1119 2014
[10] J Hubbert and V N Bringi ldquoAn iterative filtering technique forthe analysis of copolar differential phase and dual-frequencyradar measurementsrdquo Journal of Atmospheric and OceanicTechnology vol 12 no 3 pp 643ndash648 1995
[11] K V Mishra W F Krajewski R Goska et al ldquoDeploy-ment and performance analyses of high-resolution iowa XPOLradar system during the NASA IFloodS campaignrdquo Journal ofHydrometeorology vol 17 no 2 pp 455ndash479 2016
[12] M Thurai G J Huang V N Bringi W L Randeu and MSchonhuber ldquoDrop shapes model comparisons and calcu-lations of polarimetric radar parameters in rainrdquo Journal ofAtmospheric and Oceanic Technology vol 24 no 6 pp 1019ndash1032 2007
[13] M D Humphrey J D Istok J Y Lee J A Hevesi and AL Flint ldquoA new method for automated dynamic calibration oftipping-bucket rain gaugesrdquo Journal of Atmospheric andOceanicTechnology vol 14 no 6 pp 1513ndash1519 1997
[14] E HabibW F Krajewski V Nespor andA Kruger ldquoNumericalsimulation studies of rain gage data correction due to windeffectrdquo Journal of Geophysical ResearchAtmospheres vol 104 no16 Article ID 1999JD900228 pp 19723ndash19733 1999
[15] G J Ciach and W F Krajewski ldquoOn the estimation of radarrainfall error variancerdquoAdvances inWater Resources vol 22 no6 pp 585ndash595 1999
[16] E Habib W F Krajewski and A Kruger ldquoSampling errors oftipping-bucket rain gaugemeasurementsrdquo Journal of HydrologicEngineering vol 6 no 2 pp 159ndash166 2001
[17] G J Ciach ldquoLocal random errors in tipping-bucket rain gaugemeasurementsrdquo Journal of Atmospheric andOceanic Technologyvol 20 no 5 pp 752ndash759 2003
[18] EHabibW F Krajewski andG J Ciach ldquoEstimation of rainfallinterstation correlationrdquo Journal ofHydrometeorology vol 2 no6 pp 621ndash629 2001
[19] NIMS ldquoStudy on estimation of rainfall amount and classifi-cation of hydrometeor species using dual-polarimetric radar(III)rdquo pp 1ndash20 2010
[20] E Gorgucci V Chandrasekar and L Baldini ldquoCorrection ofX-band radar observation for propagation effects based on theself-consistency principlerdquo Journal of Atmospheric and OceanicTechnology vol 23 pp 1668ndash1681 2006
[21] NIMS ldquoStudy on estimating raindrop size distribution forquantitative estimation of radar rainfall amountrdquo pp 1ndash55 2009
[22] NIMS ldquoQuantitative precipitation estimation and hydrometeoridentification using dual-polarization radar-Phase IIrdquo pp 1ndash742009
[23] S-G Park M Maki K Iwanami V N Bringi and V Chan-drasekar ldquoCorrection of radar reflectivity and differential reflec-tivity for rain attenuation at X band Part II Evaluation andapplicationrdquo Journal of Atmospheric and Oceanic Technologyvol 22 no 11 pp 1633ndash1655 2005
[24] E Habib and W F Krajewski ldquoUncertainty analysis of theTRMM ground-validation radar-rainfall products applicationto the TEFLUN-B field campaignrdquo Journal of Applied Meteorol-ogy vol 41 no 5 pp 558ndash572 2002
[25] Y-A Oh Rainfall estimation from X-band dual-polarizationradar measurements effects of attenuation correction and evalu-ation with a dense rain gauge network [MS thesis] KyungpookNational University 2011
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom
Applied ampEnvironmentalSoil Science
Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal of
Geophysics
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
10 Advances in Meteorology
32km
21 km
11 km
20km40km60km
18
14
00
02
04
06
26
40
minus10
22
minus20
30
10
50
minus30
(a)
32 km
21 km
11 km
20km40km60km
18
14
00
02
04
06
26
40
minus10
22
minus20
30
10
50
minus30
(b)
Figure 9 PPIs of measured (a) and attenuation corrected (b) 119885DR of NIMS-XPOL for the case in Figure 5
increasing accumulation time and the slope of the correlationfunction was steep This was due to the reduction of theinstrumental errors of the gauge data by time accumulationThis indicated that rainfall accumulated for less than 10minfrom single gauge data cannot represent the rainfall over itssurrounding area due to instrumental errors For accumula-tion times longer than 10min the slope became less steepThis reflected the remaining variability of the rain field Inaddition the accumulated rainfall rate of about 10min wassuitable for minimizing the instrumental uncertainty whilepreserving the variation in the rain field The correlationreached about 098 at 05 km in 10min of accumulationand this approached 098 at 10 km at 60min accumulationTherefore a single rain gauge mostly represented the rainfallrate within 05 km and 10 km with 10min and 60minaccumulation respectively
The NB between an areal averaged rainfall rate ⟨119866⟩ anda single gauge rainfall rate 119866 is shown with accumulationtime in Figures 15(a) and 16(a) Where the average area was038 km2 (corresponding to 9 pixels of NIMS-XPOL radarat 16 km range) the NB reached a 6ndash8 maximum Wherethe averaged area is 6 km2 (corresponding to the coverage forall of the gauges) the NB reached about a 10 maximumFurthermore the NB hardly changed with accumulationtime Note that this statistic was derived after removinginstrumental bias This indicated that the rainfall rate froma single gauge was biased due to the spatial variability of therain fields and this bias was maintained with accumulationtime
The NSD of a single gauge for a mean area of 038 km2was over 40 and it reached about 70ndash90 for a mean areaof 6 km2 at an accumulation time of 1min (see Figures 15(b)and 16(b)) This statistic included instrumental errors (about
40ndash60 at 16mmhminus1 and 60ndash80 at 10mmhminus1 as shownin the Appendix) as well as the spatial variability of the rainfields This indicated that the variance between the point andarea values from the gauge data at short accumulation timeswas similar to the errors due to the instrumental uncertaintyof the rain gaugeTheNSD in 60min accumulation decreasedto over 10 and about 15ndash45 with a mean area of 038 km2and 6 km2 respectively including the NSD of 6ndash10 due toinstrumental uncertainty This indicated that the rainfall ratefrom single gauges may differ significantly from the actualrainfall rate due to the variability of the rain fields and theinstrumental uncertainty in short accumulation times
43 Verification of Rainfall Retrieval The rainfall rate fromthe NIMS-XPOL radar measurements was retrieved withtwo rainfall rate estimators 119877(119885
119867) and 119877(119885
119867 119885DR) The
estimated rainfall rate was verified through comparison withthe rainfall rate (119866) from the rain gauge data These twoestimators were applied before and after the attenuation cor-rection First 119877(038 km2) was compared with 119877(007 km2)at 10min accumulation (Figure 17) Using 119877(119885
119867) the NB
NSD rmse and r rmse without the attenuation correctionwere significantly large at 70 127 127 (mmhr) and78 respectively However they were reduced to minus152 521 (mmhr) and 32 respectively after applying theattenuation correction Using 119877(119885
119867 119885DR) the NB was just
3 NSD was 49 rmse was 496 (mmhr) and r rmsewas 31 before applying attenuation correction This wassimilar to the result from 119877(119885
119867) after attenuation correction
However there was no significant change in119877(119885119867 119885DR)with
attenuation correction This means that 119877(119885119867 119885DR) was less
sensitive to rain attenuation than was 119877(119885119867) because 119885
119867is
proportional to 119877 and 119885DR is inversely proportional to 119877
Advances in Meteorology 11
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 71∘
minus10
minus5
0
5
10
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 160∘
minus10
minus5
0
5
10
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 334∘
minus10
minus5
0
5
10
(c)
Figure 10 Range profiles of measured (line) calibration bias 120588HV threshold and isolated point removing (dot) and attenuation corrected(dash) 119885DR at 71∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
in this rain estimator (see Table 3) Furthermore comparingthe analysis of variance results between areal (038 km2) andpoint (007m2) values from the gauge data (Section 42)half of the errors after attenuation correction were producedby the small scale variability of the rain fields and theinstrumental uncertainty of the rain gauge
A similar comparison is shown in Figure 18 for a 60minaccumulation Using 119877(119885
119867) the NB NSD rmse and r rmse
were decreased from minus65 86 71 (mmhr) and 70to minus5 34 28 (mmhr) and 28 by the attenuationcorrection respectively Using 119877(119885
119867 119885DR) the NB NSD
rmse and r rmse were only slightly decreased after theattenuation correction However the NSD rmse and r rmsewere reduced when compared at 10min accumulation In thesame accumulation time the errors in the gauge data reachedto over 15
This comparison was affected by different sampling areasof the gauges and the radar Therefore the areal averagevalues were compared (Figure 19) Using119877(119885
119867)with the areal
averaged radar measurements the NB was reduced fromminus69 to 7 NSD was reduced from 122 to 42 rmsewas reduced from 39 (mmhr) to 13 (mmhr) and r rmse
was reduced from 77 to 26 at an accumulation time of10min Using 119877(119885
119867 119885DR) the NB NSD rmse and r rmse
were decreased fromminus24 67 21 (mmhr) and 42 to 539 12 (mmhr) and 24 respectively
TheNB and r rmsewithmatched sampling areas betweenthe gauge and the radar are shown in Figures 20 and 21 Thereduction of NB due to temporal accumulation was smallin both of the rain estimators The NB was reduced to lessthan 7 by attenuation correction for all accumulation timeswith both119877(119885
119867) and119877(119885
119867 119885DR)The attenuation correction
effect was significantly greater in119877(119885119867) Similarly the r rmse
was reduced significantly in 119877(119885119867) by attenuation correction
(from 68 to 25 at a 60min accumulation) The reductionof the r rmse due to temporal accumulation was larger in119877(119885119867 119885DR) because the significant observational noise of
119885DR was smoothed The r rmse was decreased from 18to 15 in 119877(119885
119867 119885DR) at a 60min accumulation time after
attenuation correction
5 Conclusions
In this study an attenuation correction algorithm based onΦDP was developed to improve the accuracy of the rainfall
12 Advances in Meteorology
1st gauge site 2nd gauge site
3rd gauge site 4th gauge site
5th gauge site 6th gauge site
7th gauge site 8th gauge site
9th gauge site 10th gauge site
17 18 19 20 2116Time (h)
17 18 19 20 2116Time (h)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
UncorrectedCorrected
UncorrectedCorrected
Figure 11 Time series of measured (red circle) and attenuation corrected (blue circle) 119885119867at the 10 rain gauge sites
Advances in Meteorology 13
UncorrectedCorrected
UncorrectedCorrected
17 18 19 20 2116Time (h)
17 18 19 20 2116Time (h)
1st gauge site 2nd gauge site
3rd gauge site 4th gauge site
5th gauge site 6th gauge site
7th gauge site 8th gauge site
9th gauge site 10th gauge site
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
Figure 12 Time series of measured (red circle) and attenuation corrected (blue circle) 119885DR at the 10 rain gauge sites
14 Advances in Meteorology
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Uncorrected(a)
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Corrected Simulated from DSD(b)
Figure 13 Scatter plots of 119885119867and 119885DR before (a) and after (b) attenuation correction at the 10 rain gauge sites DSD observation results (dot
[21])
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
04
05
06
07
08
09
10
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
Accum time = 1min
Accum time = 10min
Accum time = 2min
Accum time = 20min
Accum time = 5min
Accum time = 60min
Figure 14 The Spearman correlation coefficient 119903 between data from two rain gauges at different accumulation times
estimation using X-band radar This algorithm was evalu-ated with radar measurements from NIMS-XPOL Howevermeasured ΨDP contained 120575 and observation noise as wellas ΦDP Furthermore the ΨDP offset varied with azimuthangles and decreased near the radar when rain cells were
above the radar In this study ΦDP was estimated by aniterative filtering technique and by adjusting with an offsetdetermined as the minimum value ofΦDP within 3 km of theradar As a result estimated ΦDP monotonically increasedThe path integrated value of119860
119867and119860DP was calculated from
Advances in Meteorology 15
2nd 6th 9th gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(038km2) versus G
(a) Normalized bias
⟨G⟩(038km2) versus G
2nd 6th 9th gaugesMean NSD
Mean R
0
20
40
60
80
100
NSD
()
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 15 (a) The NB and (b) NSD between ⟨119866⟩(038 km2) and 119866 from three rain gauges with accumulation time ⟨119866⟩ is the temporallyaveraged rainfall rate of ⟨119866⟩
All gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(6km2) versus G
(a) Normalized bias
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
0
20
40
60
80
100
NSD
()
All gaugesMean NSD
Mean R
⟨G⟩(6km2) versus G
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 16 (a) The NB and (b) NSD between ⟨119866⟩(6 km2) and 119866 from 10 rain gauges with accumulation time ⟨119866⟩ is the temporally averaged⟨119866⟩ with accumulation time
estimated ΦDP using 119860119867-ΦDP and 119860DP-ΦDP relationships
derived from the DSDs data Measured 119885119867
and 119885DR werecorrected with the calculated path integrated 119860
119867and 119860DP
by each gate The attenuated signal patterns of 119885119867and 119885DR
were restored to a typical rain echo pattern However theattenuation corrected 119885
119867and 119885DR fields were not matched
at distant ranges This was likely because of contaminationdue to beam broadening and the accumulation of errorswith increasing range Comparing the 119885
119867-119885DR scatter plots
over the 10 rain gauge sites with the results obtained froma scattering simulation with the DSD data the attenuation
corrected 119885119867-119885DR scatter plots are in good agreement with
the 119885119867-119885DR scatter plots from the DSDs
The validation for the improvement of accuracy in therainfall estimation through attenuation correction was per-formed by comparing with dense rain gauge network data119877(119885119867) from the attenuated (observed)119885
119867produced a severe
underestimation with an NB NSD rmse and r rmse of70 127 127 (mmhr) and 78 at 10min accumulationrespectively while 119877(119885
119867) from the attenuation corrected 119885
119867
showed better agreement with the gauge measurements withan NB NSD rmse and r rmse of minus1 52 521 (mmhr)
16 Advances in Meteorology
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus070
NSD = 127
rmse = 1279
corr = 0858
Corrected
NB = minus001
NSD = 052
rmse = 521
corr = 0927
r_rmse = 0786 r_rmse = 0320
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = 003
NSD = 049
rmse = 496
r_rmse = 0305
corr = 0936
Corrected
NB = 001
NSD = 050
rmse = 502
r_rmse = 0308
corr = 0936
(b) 119877(119885119867119885DR)
Figure 17 The scatter plots of 119877(038 km2 10min) versus 119866(007m2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
CorrectedUncorrected
R(m
mh
)
0
10
20
30
40
10 20 30 400G (mmh)
Uncorrected
NB = minus065
NSD = 086
rmse = 706
r_rmse = 0702
corr = 0878
Corrected
NB = minus005
NSD = 034
rmse = 281
r_rmse = 0279
corr = 0866
(a) 119877(119885119867)
CorrectedUncorrected
R(m
mh
)
10 20 30 400G (mmh)
0
10
20
30
40
Uncorrected
NB = 008
NSD = 023
rmse = 189
r_rmse = 0188
corr = 0942
Corrected
NB = 007
NSD = 025
rmse = 205
r_rmse = 0203
corr = 0940
(b) 119877(119885119867119885DR)
Figure 18 The scatter plots of 119877(038 km2 60min) versus 119866(007m2 60min) (a) using 119885119867and (b) using 119885
119867and 119885DR
and 32 at 10min accumulation respectively At 60minaccumulation theNBNSD rmse and r rmsewere decreasedfrom minus65 86 71 (mmhr) and 70 to minus5 3428 (mmhr) and 28 respectively Using 119877(119885
119867 119885DR) the
NB NSD rmse and r rmse were only slightly reduced even if
attenuation was corrected but were smaller than when using119877(119885119867) This indicated that 119877(119885
119867 119885DR) was less sensitive to
the attenuation than 119877(119885119867) In addition half of the errors
that remained after attenuation correction were caused by thesmall scale variability of the rain fields and the instrumental
Advances in Meteorology 17
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus069
NSD = 122
rmse = 391
r_rmse = 0770
corr = 0893
Corrected
NB = 007
NSD = 042
rmse = 133
r_rmse = 0262
corr = 0960
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = minus024
NSD = 067
rmse = 215
r_rmse = 0424
corr = 0718
Corrected
NB = 005
NSD = 039
rmse = 124
r_rmse = 0245
corr = 0891
(b) 119877(119885119867119885DR)
Figure 19 The scatter plots of 119877(6 km2 10min) versus 119866(6 km2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
R(ZH)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(a) 119877(119885119867)
R(ZH ZDR)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(b) 119877(119885119867119885DR)
Figure 20 The normalized bias of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and 119885DR The red
point-line and the blue point-line indicate without and with the attenuation correction respectively
uncertainty of the rain gauges Using an areal averagedvalue the NB was reduced to less than 7 with attenua-tion correction using either rainfall estimator regardless ofaccumulation timeThe r rmse were significantly reduced byattenuation correction in 119877(119885
119867) and the reduction of r rmse
with 119877(119885119867 119885DR) was even more dramatic
As summarized above the rain attenuation for 119885119867
and 119885DR in the X-band was corrected by ΦDP and theresults showed a good agreement with the value obtainedby scattering simulations with the DSD datasets The resultsof comparison with rain gauge data also showed a signifi-cant improvement in the accuracy of rainfall estimation by
18 Advances in Meteorology
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10r_
rmse
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH)
(a) 119877(119885119867)
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10
r_rm
se
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH ZDR)
(b) 119877(119885119867119885DR)
Figure 21 The relative root-mean square error of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and
119885DR The red point-line and the blue point-line indicate without and with the attenuation correction respectively
attenuation correction which supports the quantitative useof the X-band radar in rain data retrieval
Appendix
A Instrumental Uncertainty of the Rain Gauge
The instrumental uncertainty of the TB rain gauge is asso-ciated with improper calibration leakage and precipitationmissing the gauge due to the measurement principle andwetting and evaporation losses In addition external sourcessuch as wind turbulence and installed position also con-tribute to the instrumental uncertainty Ciach [17] reportedthat the TB gauge data suffers a significant uncertainty asso-ciated with random differences between closely collocatedTB gauges through the analysis of the error as a functionof the rainfall intensity for three timescales In addition theimproper calibration induces a bias in the rainfall measure-ment and this bias cannot be removed by increasing accu-mulation timeTherefore the proper correction or processingof TB gauge data is necessary to obtain accurate and reliablereading of the rainfall amount
In this appendix instrumental biases were calculatedto reduce the instrumental uncertainty through an idealexperiment in the laboratory and a field intercomparisonwith 10 TB rain gauges In addition the random errors werequantitatively calculated as a function of rainfall rate andaccumulation time
A1 Correction of Instrumental Bias The instrumental bias119861119868(119894) of the 119894th gauge was defined as a combination
of the absolute bias 119861119860(ref) of the reference gauge and
the relative bias 119861119877(119894) of the 119894th gauge according to the
following equation
119861119868 (119894) = 119861119860 (ref) times 119861119877 (119894) (A1)
119861119860(ref) signified the ratio of the calculated value from
the instrumental resolution of the reference gauge to the truevalue measured by other instruments
119861119860 (ref) =
rainfalltruthrainfallgauge (ref)
(A2)
In this study 119861119860(ref) was derived by an ideal experiment
using an electronic scale First the reference gauge was puton the scale and the scale was initialized The water dropsflowed out into the receiving orifice of the reference gaugeuntil the tipping number became 60Themeasured weight of60 tips was 2144 g while the theoretical (or expected) valuewas 2244 g for the reference gauge (receiving orifice size of1545 cm 60 tips and 02mm tip resolution) Therefore theabsolute bias of the reference gauge from (A2) was about0955 This indicates that the rainfall amount measured bythis reference gauge was overestimated by about 45 andthis bias should be corrected prior to any interpretation
119861119877(119894) represented the ratio of the observed (recorded)
value from the 119894th gauge to the observed (recorded) valuefrom the reference gauge
119861119877 (119894) =
rainfallgauge (ref)rainfallgauge (119894)
(A3)
119861119877(119894) was calculated from ten collocated rain gauges
including the reference gauge located at an observation field
Advances in Meteorology 19
Table 4 The instrumental and relative biases of the 10 tipping bucket rain gauges
Gauge ID RG1 RG2 RG3 RG4 RG5 RG6 RG7 RG8 RG9 RG10Instrumental bias 094 108 099 100 096 104 098 102 108 098Relative bias 098 113 103 105 100 109 102 107 113 102
013610152230406080120150200250300400
()
Rain
rate
(mm
h)
100101
Accumulation time (min)
10
100NSD ()
Figure 22 Two-dimensional normalized standard deviationbetween the standard rainfall and each rain gauge
at KNUKoreaThe observational period for intercomparisonwas from October 2009 to June 2010 (14 rainy days) andthe total accumulated rainfall for this period was about300mm The results are presented in Table 4 All relativebiases were mainly larger than 10 This indicated that themost of gauges underestimated the rainfall amount comparedto the reference rain gauge The relative bias of the referencegauge (119894 = 5) was 10
Finally 119861119860(119894) was calculated from (A1) (see Table 4) All
instrumental biases were between 090 and 11 and the meanvalue of the biases was nearly 10 This indicated that therainfall rate from a single gauge data had a maximum 10bias These biases were corrected for each gauge to removeexisting instrumental errors
A2 Quantification of Random Errors To quantify the ran-dom error due to instrumental uncertainty the NSD of theten gauges was calculated with accumulation time and rainrate using data from an intercomparison The NSD of the 119894thgauge compared to the reference gauge was derived for 20rain rate classes (Δ119877) and 17 accumulation time classes (Δ119905)as follows
NSD119894 (Δ119905 Δ119877)
=
(1119899)radicsum119899minus1
119896=0[1198661015840
119870(Δ119905 Δ119877) minus 119866119894119896 (Δ119905 Δ119877)]
2
1198661015840
(A4)
where 1198661015840 and 119866 indicate the reference and individual gaugesrespectively The overbar means the temporal average ofthe rainfall rate The NSDs of nine different gauges (exceptthe reference gauge) were averaged to obtain the two-dimensional average NSD (Figure 22) The average NSD
tended to decrease with increasing accumulation time andwith increasing rain rate The NSD was about 10ndash15when accumulation time was 10min and the rain rate was10mmhminus1 The instrumental errors were quite significant forlonger accumulation times and higher rainfall rates Theseerrors should be used as a guideline when a single gauge isused as a reference for any validation study
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper has been revised and extended fromMs Young-AOhrsquos MS thesis [25] of Kyungpook National University Thisresearch is supported by Development and Application ofCross Governmental Dual-Pol Radar Harmonization (WRC-2013-A-1) project of theWeather Radar Center KoreaMeteo-rological AdministrationThis work was funded by the KoreaMeteorological Administration Research and DevelopmentProgram under Grant KMIPA2015-1010
References
[1] V N Bringi V Chandrasekar N Balakrishnan and D SZrnic ldquoAn examination of propagation effects in rainfall onradar measurements at microwave frequenciesrdquo Journal ofAtmospheric and Oceanic Technology vol 7 no 6 pp 829ndash8401990
[2] A Huggel W Schmid and AWaldvogel ldquoRaindrop size distri-butions and the radar bright bandrdquo Journal of AppliedMeteorol-ogy vol 35 no 10 pp 1688ndash1701 1996
[3] G W Lee Errors in rain measurement by radar effect of vari-ability of drop size distributions [PhD thesis]McGill University2003
[4] J D Tuttle and R E Rinehart ldquoAttenuation correction in dual-wavelength analysesrdquo Journal of ClimateampAppliedMeteorologyvol 22 no 11 pp 1914ndash1921 1983
[5] S-G Park V N Bringi V Chandrasekar M Maki andK Iwanami ldquoCorrection of radar reflectivity and differentialreflectivity for rain attenuation at X band Part I theoretical andempirical basisrdquo Journal of Atmospheric and Oceanic Technol-ogy vol 22 no 11 pp 1621ndash1632 2005
[6] G Scarchilli E Gorgucci V Chandrasekar and T A SeligaldquoRainfall estimation using polarimetric techniques at C-bandfrequenciesrdquo Journal of Applied Meteorology vol 32 no 6 pp1150ndash1160 1993
[7] M N Anagnostou E N Anagnostou and J VivekanandanldquoCorrection for rain path specific and differential attenuation ofX-band dual-polarization observationsrdquo IEEE Transactions onGeoscience and Remote Sensing vol 44 no 9 pp 2470ndash24802006
20 Advances in Meteorology
[8] J Kalogiros M N Anagnostou E N Anagnostou M Mon-topoli E Picciotti and F S Marzano ldquoEvaluation of a newpolarimetric algorithm for rain-path attenuation correction ofX-band radar observations against disdrometerrdquo IEEE Transac-tions on Geoscience and Remote Sensing vol 52 no 2 pp 1369ndash1380 2014
[9] W-Y Chang J Vivekanandan and T-C C Wang ldquoEstimationof X-band polarimetric radar attenuation and measurementuncertainty using a variational methodrdquo Journal of AppliedMeteorology and Climatology vol 53 no 4 pp 1099ndash1119 2014
[10] J Hubbert and V N Bringi ldquoAn iterative filtering technique forthe analysis of copolar differential phase and dual-frequencyradar measurementsrdquo Journal of Atmospheric and OceanicTechnology vol 12 no 3 pp 643ndash648 1995
[11] K V Mishra W F Krajewski R Goska et al ldquoDeploy-ment and performance analyses of high-resolution iowa XPOLradar system during the NASA IFloodS campaignrdquo Journal ofHydrometeorology vol 17 no 2 pp 455ndash479 2016
[12] M Thurai G J Huang V N Bringi W L Randeu and MSchonhuber ldquoDrop shapes model comparisons and calcu-lations of polarimetric radar parameters in rainrdquo Journal ofAtmospheric and Oceanic Technology vol 24 no 6 pp 1019ndash1032 2007
[13] M D Humphrey J D Istok J Y Lee J A Hevesi and AL Flint ldquoA new method for automated dynamic calibration oftipping-bucket rain gaugesrdquo Journal of Atmospheric andOceanicTechnology vol 14 no 6 pp 1513ndash1519 1997
[14] E HabibW F Krajewski V Nespor andA Kruger ldquoNumericalsimulation studies of rain gage data correction due to windeffectrdquo Journal of Geophysical ResearchAtmospheres vol 104 no16 Article ID 1999JD900228 pp 19723ndash19733 1999
[15] G J Ciach and W F Krajewski ldquoOn the estimation of radarrainfall error variancerdquoAdvances inWater Resources vol 22 no6 pp 585ndash595 1999
[16] E Habib W F Krajewski and A Kruger ldquoSampling errors oftipping-bucket rain gaugemeasurementsrdquo Journal of HydrologicEngineering vol 6 no 2 pp 159ndash166 2001
[17] G J Ciach ldquoLocal random errors in tipping-bucket rain gaugemeasurementsrdquo Journal of Atmospheric andOceanic Technologyvol 20 no 5 pp 752ndash759 2003
[18] EHabibW F Krajewski andG J Ciach ldquoEstimation of rainfallinterstation correlationrdquo Journal ofHydrometeorology vol 2 no6 pp 621ndash629 2001
[19] NIMS ldquoStudy on estimation of rainfall amount and classifi-cation of hydrometeor species using dual-polarimetric radar(III)rdquo pp 1ndash20 2010
[20] E Gorgucci V Chandrasekar and L Baldini ldquoCorrection ofX-band radar observation for propagation effects based on theself-consistency principlerdquo Journal of Atmospheric and OceanicTechnology vol 23 pp 1668ndash1681 2006
[21] NIMS ldquoStudy on estimating raindrop size distribution forquantitative estimation of radar rainfall amountrdquo pp 1ndash55 2009
[22] NIMS ldquoQuantitative precipitation estimation and hydrometeoridentification using dual-polarization radar-Phase IIrdquo pp 1ndash742009
[23] S-G Park M Maki K Iwanami V N Bringi and V Chan-drasekar ldquoCorrection of radar reflectivity and differential reflec-tivity for rain attenuation at X band Part II Evaluation andapplicationrdquo Journal of Atmospheric and Oceanic Technologyvol 22 no 11 pp 1633ndash1655 2005
[24] E Habib and W F Krajewski ldquoUncertainty analysis of theTRMM ground-validation radar-rainfall products applicationto the TEFLUN-B field campaignrdquo Journal of Applied Meteorol-ogy vol 41 no 5 pp 558ndash572 2002
[25] Y-A Oh Rainfall estimation from X-band dual-polarizationradar measurements effects of attenuation correction and evalu-ation with a dense rain gauge network [MS thesis] KyungpookNational University 2011
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom
Applied ampEnvironmentalSoil Science
Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal of
Geophysics
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
Advances in Meteorology 11
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 71∘
minus10
minus5
0
5
10
(a)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 160∘
minus10
minus5
0
5
10
(b)
Measured dataPreprocessed dataCorrected data
10 20 30 40 50 600Slant range (km)
ZDR (dB) azimuth angle = 334∘
minus10
minus5
0
5
10
(c)
Figure 10 Range profiles of measured (line) calibration bias 120588HV threshold and isolated point removing (dot) and attenuation corrected(dash) 119885DR at 71∘ (a) 160∘ (b) and 334∘ (c) azimuth angles of NIMS-XPOL for the case in Figure 5
in this rain estimator (see Table 3) Furthermore comparingthe analysis of variance results between areal (038 km2) andpoint (007m2) values from the gauge data (Section 42)half of the errors after attenuation correction were producedby the small scale variability of the rain fields and theinstrumental uncertainty of the rain gauge
A similar comparison is shown in Figure 18 for a 60minaccumulation Using 119877(119885
119867) the NB NSD rmse and r rmse
were decreased from minus65 86 71 (mmhr) and 70to minus5 34 28 (mmhr) and 28 by the attenuationcorrection respectively Using 119877(119885
119867 119885DR) the NB NSD
rmse and r rmse were only slightly decreased after theattenuation correction However the NSD rmse and r rmsewere reduced when compared at 10min accumulation In thesame accumulation time the errors in the gauge data reachedto over 15
This comparison was affected by different sampling areasof the gauges and the radar Therefore the areal averagevalues were compared (Figure 19) Using119877(119885
119867)with the areal
averaged radar measurements the NB was reduced fromminus69 to 7 NSD was reduced from 122 to 42 rmsewas reduced from 39 (mmhr) to 13 (mmhr) and r rmse
was reduced from 77 to 26 at an accumulation time of10min Using 119877(119885
119867 119885DR) the NB NSD rmse and r rmse
were decreased fromminus24 67 21 (mmhr) and 42 to 539 12 (mmhr) and 24 respectively
TheNB and r rmsewithmatched sampling areas betweenthe gauge and the radar are shown in Figures 20 and 21 Thereduction of NB due to temporal accumulation was smallin both of the rain estimators The NB was reduced to lessthan 7 by attenuation correction for all accumulation timeswith both119877(119885
119867) and119877(119885
119867 119885DR)The attenuation correction
effect was significantly greater in119877(119885119867) Similarly the r rmse
was reduced significantly in 119877(119885119867) by attenuation correction
(from 68 to 25 at a 60min accumulation) The reductionof the r rmse due to temporal accumulation was larger in119877(119885119867 119885DR) because the significant observational noise of
119885DR was smoothed The r rmse was decreased from 18to 15 in 119877(119885
119867 119885DR) at a 60min accumulation time after
attenuation correction
5 Conclusions
In this study an attenuation correction algorithm based onΦDP was developed to improve the accuracy of the rainfall
12 Advances in Meteorology
1st gauge site 2nd gauge site
3rd gauge site 4th gauge site
5th gauge site 6th gauge site
7th gauge site 8th gauge site
9th gauge site 10th gauge site
17 18 19 20 2116Time (h)
17 18 19 20 2116Time (h)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
UncorrectedCorrected
UncorrectedCorrected
Figure 11 Time series of measured (red circle) and attenuation corrected (blue circle) 119885119867at the 10 rain gauge sites
Advances in Meteorology 13
UncorrectedCorrected
UncorrectedCorrected
17 18 19 20 2116Time (h)
17 18 19 20 2116Time (h)
1st gauge site 2nd gauge site
3rd gauge site 4th gauge site
5th gauge site 6th gauge site
7th gauge site 8th gauge site
9th gauge site 10th gauge site
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
Figure 12 Time series of measured (red circle) and attenuation corrected (blue circle) 119885DR at the 10 rain gauge sites
14 Advances in Meteorology
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Uncorrected(a)
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Corrected Simulated from DSD(b)
Figure 13 Scatter plots of 119885119867and 119885DR before (a) and after (b) attenuation correction at the 10 rain gauge sites DSD observation results (dot
[21])
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
04
05
06
07
08
09
10
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
Accum time = 1min
Accum time = 10min
Accum time = 2min
Accum time = 20min
Accum time = 5min
Accum time = 60min
Figure 14 The Spearman correlation coefficient 119903 between data from two rain gauges at different accumulation times
estimation using X-band radar This algorithm was evalu-ated with radar measurements from NIMS-XPOL Howevermeasured ΨDP contained 120575 and observation noise as wellas ΦDP Furthermore the ΨDP offset varied with azimuthangles and decreased near the radar when rain cells were
above the radar In this study ΦDP was estimated by aniterative filtering technique and by adjusting with an offsetdetermined as the minimum value ofΦDP within 3 km of theradar As a result estimated ΦDP monotonically increasedThe path integrated value of119860
119867and119860DP was calculated from
Advances in Meteorology 15
2nd 6th 9th gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(038km2) versus G
(a) Normalized bias
⟨G⟩(038km2) versus G
2nd 6th 9th gaugesMean NSD
Mean R
0
20
40
60
80
100
NSD
()
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 15 (a) The NB and (b) NSD between ⟨119866⟩(038 km2) and 119866 from three rain gauges with accumulation time ⟨119866⟩ is the temporallyaveraged rainfall rate of ⟨119866⟩
All gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(6km2) versus G
(a) Normalized bias
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
0
20
40
60
80
100
NSD
()
All gaugesMean NSD
Mean R
⟨G⟩(6km2) versus G
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 16 (a) The NB and (b) NSD between ⟨119866⟩(6 km2) and 119866 from 10 rain gauges with accumulation time ⟨119866⟩ is the temporally averaged⟨119866⟩ with accumulation time
estimated ΦDP using 119860119867-ΦDP and 119860DP-ΦDP relationships
derived from the DSDs data Measured 119885119867
and 119885DR werecorrected with the calculated path integrated 119860
119867and 119860DP
by each gate The attenuated signal patterns of 119885119867and 119885DR
were restored to a typical rain echo pattern However theattenuation corrected 119885
119867and 119885DR fields were not matched
at distant ranges This was likely because of contaminationdue to beam broadening and the accumulation of errorswith increasing range Comparing the 119885
119867-119885DR scatter plots
over the 10 rain gauge sites with the results obtained froma scattering simulation with the DSD data the attenuation
corrected 119885119867-119885DR scatter plots are in good agreement with
the 119885119867-119885DR scatter plots from the DSDs
The validation for the improvement of accuracy in therainfall estimation through attenuation correction was per-formed by comparing with dense rain gauge network data119877(119885119867) from the attenuated (observed)119885
119867produced a severe
underestimation with an NB NSD rmse and r rmse of70 127 127 (mmhr) and 78 at 10min accumulationrespectively while 119877(119885
119867) from the attenuation corrected 119885
119867
showed better agreement with the gauge measurements withan NB NSD rmse and r rmse of minus1 52 521 (mmhr)
16 Advances in Meteorology
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus070
NSD = 127
rmse = 1279
corr = 0858
Corrected
NB = minus001
NSD = 052
rmse = 521
corr = 0927
r_rmse = 0786 r_rmse = 0320
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = 003
NSD = 049
rmse = 496
r_rmse = 0305
corr = 0936
Corrected
NB = 001
NSD = 050
rmse = 502
r_rmse = 0308
corr = 0936
(b) 119877(119885119867119885DR)
Figure 17 The scatter plots of 119877(038 km2 10min) versus 119866(007m2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
CorrectedUncorrected
R(m
mh
)
0
10
20
30
40
10 20 30 400G (mmh)
Uncorrected
NB = minus065
NSD = 086
rmse = 706
r_rmse = 0702
corr = 0878
Corrected
NB = minus005
NSD = 034
rmse = 281
r_rmse = 0279
corr = 0866
(a) 119877(119885119867)
CorrectedUncorrected
R(m
mh
)
10 20 30 400G (mmh)
0
10
20
30
40
Uncorrected
NB = 008
NSD = 023
rmse = 189
r_rmse = 0188
corr = 0942
Corrected
NB = 007
NSD = 025
rmse = 205
r_rmse = 0203
corr = 0940
(b) 119877(119885119867119885DR)
Figure 18 The scatter plots of 119877(038 km2 60min) versus 119866(007m2 60min) (a) using 119885119867and (b) using 119885
119867and 119885DR
and 32 at 10min accumulation respectively At 60minaccumulation theNBNSD rmse and r rmsewere decreasedfrom minus65 86 71 (mmhr) and 70 to minus5 3428 (mmhr) and 28 respectively Using 119877(119885
119867 119885DR) the
NB NSD rmse and r rmse were only slightly reduced even if
attenuation was corrected but were smaller than when using119877(119885119867) This indicated that 119877(119885
119867 119885DR) was less sensitive to
the attenuation than 119877(119885119867) In addition half of the errors
that remained after attenuation correction were caused by thesmall scale variability of the rain fields and the instrumental
Advances in Meteorology 17
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus069
NSD = 122
rmse = 391
r_rmse = 0770
corr = 0893
Corrected
NB = 007
NSD = 042
rmse = 133
r_rmse = 0262
corr = 0960
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = minus024
NSD = 067
rmse = 215
r_rmse = 0424
corr = 0718
Corrected
NB = 005
NSD = 039
rmse = 124
r_rmse = 0245
corr = 0891
(b) 119877(119885119867119885DR)
Figure 19 The scatter plots of 119877(6 km2 10min) versus 119866(6 km2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
R(ZH)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(a) 119877(119885119867)
R(ZH ZDR)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(b) 119877(119885119867119885DR)
Figure 20 The normalized bias of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and 119885DR The red
point-line and the blue point-line indicate without and with the attenuation correction respectively
uncertainty of the rain gauges Using an areal averagedvalue the NB was reduced to less than 7 with attenua-tion correction using either rainfall estimator regardless ofaccumulation timeThe r rmse were significantly reduced byattenuation correction in 119877(119885
119867) and the reduction of r rmse
with 119877(119885119867 119885DR) was even more dramatic
As summarized above the rain attenuation for 119885119867
and 119885DR in the X-band was corrected by ΦDP and theresults showed a good agreement with the value obtainedby scattering simulations with the DSD datasets The resultsof comparison with rain gauge data also showed a signifi-cant improvement in the accuracy of rainfall estimation by
18 Advances in Meteorology
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10r_
rmse
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH)
(a) 119877(119885119867)
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10
r_rm
se
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH ZDR)
(b) 119877(119885119867119885DR)
Figure 21 The relative root-mean square error of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and
119885DR The red point-line and the blue point-line indicate without and with the attenuation correction respectively
attenuation correction which supports the quantitative useof the X-band radar in rain data retrieval
Appendix
A Instrumental Uncertainty of the Rain Gauge
The instrumental uncertainty of the TB rain gauge is asso-ciated with improper calibration leakage and precipitationmissing the gauge due to the measurement principle andwetting and evaporation losses In addition external sourcessuch as wind turbulence and installed position also con-tribute to the instrumental uncertainty Ciach [17] reportedthat the TB gauge data suffers a significant uncertainty asso-ciated with random differences between closely collocatedTB gauges through the analysis of the error as a functionof the rainfall intensity for three timescales In addition theimproper calibration induces a bias in the rainfall measure-ment and this bias cannot be removed by increasing accu-mulation timeTherefore the proper correction or processingof TB gauge data is necessary to obtain accurate and reliablereading of the rainfall amount
In this appendix instrumental biases were calculatedto reduce the instrumental uncertainty through an idealexperiment in the laboratory and a field intercomparisonwith 10 TB rain gauges In addition the random errors werequantitatively calculated as a function of rainfall rate andaccumulation time
A1 Correction of Instrumental Bias The instrumental bias119861119868(119894) of the 119894th gauge was defined as a combination
of the absolute bias 119861119860(ref) of the reference gauge and
the relative bias 119861119877(119894) of the 119894th gauge according to the
following equation
119861119868 (119894) = 119861119860 (ref) times 119861119877 (119894) (A1)
119861119860(ref) signified the ratio of the calculated value from
the instrumental resolution of the reference gauge to the truevalue measured by other instruments
119861119860 (ref) =
rainfalltruthrainfallgauge (ref)
(A2)
In this study 119861119860(ref) was derived by an ideal experiment
using an electronic scale First the reference gauge was puton the scale and the scale was initialized The water dropsflowed out into the receiving orifice of the reference gaugeuntil the tipping number became 60Themeasured weight of60 tips was 2144 g while the theoretical (or expected) valuewas 2244 g for the reference gauge (receiving orifice size of1545 cm 60 tips and 02mm tip resolution) Therefore theabsolute bias of the reference gauge from (A2) was about0955 This indicates that the rainfall amount measured bythis reference gauge was overestimated by about 45 andthis bias should be corrected prior to any interpretation
119861119877(119894) represented the ratio of the observed (recorded)
value from the 119894th gauge to the observed (recorded) valuefrom the reference gauge
119861119877 (119894) =
rainfallgauge (ref)rainfallgauge (119894)
(A3)
119861119877(119894) was calculated from ten collocated rain gauges
including the reference gauge located at an observation field
Advances in Meteorology 19
Table 4 The instrumental and relative biases of the 10 tipping bucket rain gauges
Gauge ID RG1 RG2 RG3 RG4 RG5 RG6 RG7 RG8 RG9 RG10Instrumental bias 094 108 099 100 096 104 098 102 108 098Relative bias 098 113 103 105 100 109 102 107 113 102
013610152230406080120150200250300400
()
Rain
rate
(mm
h)
100101
Accumulation time (min)
10
100NSD ()
Figure 22 Two-dimensional normalized standard deviationbetween the standard rainfall and each rain gauge
at KNUKoreaThe observational period for intercomparisonwas from October 2009 to June 2010 (14 rainy days) andthe total accumulated rainfall for this period was about300mm The results are presented in Table 4 All relativebiases were mainly larger than 10 This indicated that themost of gauges underestimated the rainfall amount comparedto the reference rain gauge The relative bias of the referencegauge (119894 = 5) was 10
Finally 119861119860(119894) was calculated from (A1) (see Table 4) All
instrumental biases were between 090 and 11 and the meanvalue of the biases was nearly 10 This indicated that therainfall rate from a single gauge data had a maximum 10bias These biases were corrected for each gauge to removeexisting instrumental errors
A2 Quantification of Random Errors To quantify the ran-dom error due to instrumental uncertainty the NSD of theten gauges was calculated with accumulation time and rainrate using data from an intercomparison The NSD of the 119894thgauge compared to the reference gauge was derived for 20rain rate classes (Δ119877) and 17 accumulation time classes (Δ119905)as follows
NSD119894 (Δ119905 Δ119877)
=
(1119899)radicsum119899minus1
119896=0[1198661015840
119870(Δ119905 Δ119877) minus 119866119894119896 (Δ119905 Δ119877)]
2
1198661015840
(A4)
where 1198661015840 and 119866 indicate the reference and individual gaugesrespectively The overbar means the temporal average ofthe rainfall rate The NSDs of nine different gauges (exceptthe reference gauge) were averaged to obtain the two-dimensional average NSD (Figure 22) The average NSD
tended to decrease with increasing accumulation time andwith increasing rain rate The NSD was about 10ndash15when accumulation time was 10min and the rain rate was10mmhminus1 The instrumental errors were quite significant forlonger accumulation times and higher rainfall rates Theseerrors should be used as a guideline when a single gauge isused as a reference for any validation study
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper has been revised and extended fromMs Young-AOhrsquos MS thesis [25] of Kyungpook National University Thisresearch is supported by Development and Application ofCross Governmental Dual-Pol Radar Harmonization (WRC-2013-A-1) project of theWeather Radar Center KoreaMeteo-rological AdministrationThis work was funded by the KoreaMeteorological Administration Research and DevelopmentProgram under Grant KMIPA2015-1010
References
[1] V N Bringi V Chandrasekar N Balakrishnan and D SZrnic ldquoAn examination of propagation effects in rainfall onradar measurements at microwave frequenciesrdquo Journal ofAtmospheric and Oceanic Technology vol 7 no 6 pp 829ndash8401990
[2] A Huggel W Schmid and AWaldvogel ldquoRaindrop size distri-butions and the radar bright bandrdquo Journal of AppliedMeteorol-ogy vol 35 no 10 pp 1688ndash1701 1996
[3] G W Lee Errors in rain measurement by radar effect of vari-ability of drop size distributions [PhD thesis]McGill University2003
[4] J D Tuttle and R E Rinehart ldquoAttenuation correction in dual-wavelength analysesrdquo Journal of ClimateampAppliedMeteorologyvol 22 no 11 pp 1914ndash1921 1983
[5] S-G Park V N Bringi V Chandrasekar M Maki andK Iwanami ldquoCorrection of radar reflectivity and differentialreflectivity for rain attenuation at X band Part I theoretical andempirical basisrdquo Journal of Atmospheric and Oceanic Technol-ogy vol 22 no 11 pp 1621ndash1632 2005
[6] G Scarchilli E Gorgucci V Chandrasekar and T A SeligaldquoRainfall estimation using polarimetric techniques at C-bandfrequenciesrdquo Journal of Applied Meteorology vol 32 no 6 pp1150ndash1160 1993
[7] M N Anagnostou E N Anagnostou and J VivekanandanldquoCorrection for rain path specific and differential attenuation ofX-band dual-polarization observationsrdquo IEEE Transactions onGeoscience and Remote Sensing vol 44 no 9 pp 2470ndash24802006
20 Advances in Meteorology
[8] J Kalogiros M N Anagnostou E N Anagnostou M Mon-topoli E Picciotti and F S Marzano ldquoEvaluation of a newpolarimetric algorithm for rain-path attenuation correction ofX-band radar observations against disdrometerrdquo IEEE Transac-tions on Geoscience and Remote Sensing vol 52 no 2 pp 1369ndash1380 2014
[9] W-Y Chang J Vivekanandan and T-C C Wang ldquoEstimationof X-band polarimetric radar attenuation and measurementuncertainty using a variational methodrdquo Journal of AppliedMeteorology and Climatology vol 53 no 4 pp 1099ndash1119 2014
[10] J Hubbert and V N Bringi ldquoAn iterative filtering technique forthe analysis of copolar differential phase and dual-frequencyradar measurementsrdquo Journal of Atmospheric and OceanicTechnology vol 12 no 3 pp 643ndash648 1995
[11] K V Mishra W F Krajewski R Goska et al ldquoDeploy-ment and performance analyses of high-resolution iowa XPOLradar system during the NASA IFloodS campaignrdquo Journal ofHydrometeorology vol 17 no 2 pp 455ndash479 2016
[12] M Thurai G J Huang V N Bringi W L Randeu and MSchonhuber ldquoDrop shapes model comparisons and calcu-lations of polarimetric radar parameters in rainrdquo Journal ofAtmospheric and Oceanic Technology vol 24 no 6 pp 1019ndash1032 2007
[13] M D Humphrey J D Istok J Y Lee J A Hevesi and AL Flint ldquoA new method for automated dynamic calibration oftipping-bucket rain gaugesrdquo Journal of Atmospheric andOceanicTechnology vol 14 no 6 pp 1513ndash1519 1997
[14] E HabibW F Krajewski V Nespor andA Kruger ldquoNumericalsimulation studies of rain gage data correction due to windeffectrdquo Journal of Geophysical ResearchAtmospheres vol 104 no16 Article ID 1999JD900228 pp 19723ndash19733 1999
[15] G J Ciach and W F Krajewski ldquoOn the estimation of radarrainfall error variancerdquoAdvances inWater Resources vol 22 no6 pp 585ndash595 1999
[16] E Habib W F Krajewski and A Kruger ldquoSampling errors oftipping-bucket rain gaugemeasurementsrdquo Journal of HydrologicEngineering vol 6 no 2 pp 159ndash166 2001
[17] G J Ciach ldquoLocal random errors in tipping-bucket rain gaugemeasurementsrdquo Journal of Atmospheric andOceanic Technologyvol 20 no 5 pp 752ndash759 2003
[18] EHabibW F Krajewski andG J Ciach ldquoEstimation of rainfallinterstation correlationrdquo Journal ofHydrometeorology vol 2 no6 pp 621ndash629 2001
[19] NIMS ldquoStudy on estimation of rainfall amount and classifi-cation of hydrometeor species using dual-polarimetric radar(III)rdquo pp 1ndash20 2010
[20] E Gorgucci V Chandrasekar and L Baldini ldquoCorrection ofX-band radar observation for propagation effects based on theself-consistency principlerdquo Journal of Atmospheric and OceanicTechnology vol 23 pp 1668ndash1681 2006
[21] NIMS ldquoStudy on estimating raindrop size distribution forquantitative estimation of radar rainfall amountrdquo pp 1ndash55 2009
[22] NIMS ldquoQuantitative precipitation estimation and hydrometeoridentification using dual-polarization radar-Phase IIrdquo pp 1ndash742009
[23] S-G Park M Maki K Iwanami V N Bringi and V Chan-drasekar ldquoCorrection of radar reflectivity and differential reflec-tivity for rain attenuation at X band Part II Evaluation andapplicationrdquo Journal of Atmospheric and Oceanic Technologyvol 22 no 11 pp 1633ndash1655 2005
[24] E Habib and W F Krajewski ldquoUncertainty analysis of theTRMM ground-validation radar-rainfall products applicationto the TEFLUN-B field campaignrdquo Journal of Applied Meteorol-ogy vol 41 no 5 pp 558ndash572 2002
[25] Y-A Oh Rainfall estimation from X-band dual-polarizationradar measurements effects of attenuation correction and evalu-ation with a dense rain gauge network [MS thesis] KyungpookNational University 2011
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom
Applied ampEnvironmentalSoil Science
Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal of
Geophysics
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
12 Advances in Meteorology
1st gauge site 2nd gauge site
3rd gauge site 4th gauge site
5th gauge site 6th gauge site
7th gauge site 8th gauge site
9th gauge site 10th gauge site
17 18 19 20 2116Time (h)
17 18 19 20 2116Time (h)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
minus20
0
20
40
60
ZH
(dBZ
)
UncorrectedCorrected
UncorrectedCorrected
Figure 11 Time series of measured (red circle) and attenuation corrected (blue circle) 119885119867at the 10 rain gauge sites
Advances in Meteorology 13
UncorrectedCorrected
UncorrectedCorrected
17 18 19 20 2116Time (h)
17 18 19 20 2116Time (h)
1st gauge site 2nd gauge site
3rd gauge site 4th gauge site
5th gauge site 6th gauge site
7th gauge site 8th gauge site
9th gauge site 10th gauge site
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
Figure 12 Time series of measured (red circle) and attenuation corrected (blue circle) 119885DR at the 10 rain gauge sites
14 Advances in Meteorology
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Uncorrected(a)
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Corrected Simulated from DSD(b)
Figure 13 Scatter plots of 119885119867and 119885DR before (a) and after (b) attenuation correction at the 10 rain gauge sites DSD observation results (dot
[21])
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
04
05
06
07
08
09
10
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
Accum time = 1min
Accum time = 10min
Accum time = 2min
Accum time = 20min
Accum time = 5min
Accum time = 60min
Figure 14 The Spearman correlation coefficient 119903 between data from two rain gauges at different accumulation times
estimation using X-band radar This algorithm was evalu-ated with radar measurements from NIMS-XPOL Howevermeasured ΨDP contained 120575 and observation noise as wellas ΦDP Furthermore the ΨDP offset varied with azimuthangles and decreased near the radar when rain cells were
above the radar In this study ΦDP was estimated by aniterative filtering technique and by adjusting with an offsetdetermined as the minimum value ofΦDP within 3 km of theradar As a result estimated ΦDP monotonically increasedThe path integrated value of119860
119867and119860DP was calculated from
Advances in Meteorology 15
2nd 6th 9th gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(038km2) versus G
(a) Normalized bias
⟨G⟩(038km2) versus G
2nd 6th 9th gaugesMean NSD
Mean R
0
20
40
60
80
100
NSD
()
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 15 (a) The NB and (b) NSD between ⟨119866⟩(038 km2) and 119866 from three rain gauges with accumulation time ⟨119866⟩ is the temporallyaveraged rainfall rate of ⟨119866⟩
All gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(6km2) versus G
(a) Normalized bias
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
0
20
40
60
80
100
NSD
()
All gaugesMean NSD
Mean R
⟨G⟩(6km2) versus G
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 16 (a) The NB and (b) NSD between ⟨119866⟩(6 km2) and 119866 from 10 rain gauges with accumulation time ⟨119866⟩ is the temporally averaged⟨119866⟩ with accumulation time
estimated ΦDP using 119860119867-ΦDP and 119860DP-ΦDP relationships
derived from the DSDs data Measured 119885119867
and 119885DR werecorrected with the calculated path integrated 119860
119867and 119860DP
by each gate The attenuated signal patterns of 119885119867and 119885DR
were restored to a typical rain echo pattern However theattenuation corrected 119885
119867and 119885DR fields were not matched
at distant ranges This was likely because of contaminationdue to beam broadening and the accumulation of errorswith increasing range Comparing the 119885
119867-119885DR scatter plots
over the 10 rain gauge sites with the results obtained froma scattering simulation with the DSD data the attenuation
corrected 119885119867-119885DR scatter plots are in good agreement with
the 119885119867-119885DR scatter plots from the DSDs
The validation for the improvement of accuracy in therainfall estimation through attenuation correction was per-formed by comparing with dense rain gauge network data119877(119885119867) from the attenuated (observed)119885
119867produced a severe
underestimation with an NB NSD rmse and r rmse of70 127 127 (mmhr) and 78 at 10min accumulationrespectively while 119877(119885
119867) from the attenuation corrected 119885
119867
showed better agreement with the gauge measurements withan NB NSD rmse and r rmse of minus1 52 521 (mmhr)
16 Advances in Meteorology
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus070
NSD = 127
rmse = 1279
corr = 0858
Corrected
NB = minus001
NSD = 052
rmse = 521
corr = 0927
r_rmse = 0786 r_rmse = 0320
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = 003
NSD = 049
rmse = 496
r_rmse = 0305
corr = 0936
Corrected
NB = 001
NSD = 050
rmse = 502
r_rmse = 0308
corr = 0936
(b) 119877(119885119867119885DR)
Figure 17 The scatter plots of 119877(038 km2 10min) versus 119866(007m2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
CorrectedUncorrected
R(m
mh
)
0
10
20
30
40
10 20 30 400G (mmh)
Uncorrected
NB = minus065
NSD = 086
rmse = 706
r_rmse = 0702
corr = 0878
Corrected
NB = minus005
NSD = 034
rmse = 281
r_rmse = 0279
corr = 0866
(a) 119877(119885119867)
CorrectedUncorrected
R(m
mh
)
10 20 30 400G (mmh)
0
10
20
30
40
Uncorrected
NB = 008
NSD = 023
rmse = 189
r_rmse = 0188
corr = 0942
Corrected
NB = 007
NSD = 025
rmse = 205
r_rmse = 0203
corr = 0940
(b) 119877(119885119867119885DR)
Figure 18 The scatter plots of 119877(038 km2 60min) versus 119866(007m2 60min) (a) using 119885119867and (b) using 119885
119867and 119885DR
and 32 at 10min accumulation respectively At 60minaccumulation theNBNSD rmse and r rmsewere decreasedfrom minus65 86 71 (mmhr) and 70 to minus5 3428 (mmhr) and 28 respectively Using 119877(119885
119867 119885DR) the
NB NSD rmse and r rmse were only slightly reduced even if
attenuation was corrected but were smaller than when using119877(119885119867) This indicated that 119877(119885
119867 119885DR) was less sensitive to
the attenuation than 119877(119885119867) In addition half of the errors
that remained after attenuation correction were caused by thesmall scale variability of the rain fields and the instrumental
Advances in Meteorology 17
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus069
NSD = 122
rmse = 391
r_rmse = 0770
corr = 0893
Corrected
NB = 007
NSD = 042
rmse = 133
r_rmse = 0262
corr = 0960
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = minus024
NSD = 067
rmse = 215
r_rmse = 0424
corr = 0718
Corrected
NB = 005
NSD = 039
rmse = 124
r_rmse = 0245
corr = 0891
(b) 119877(119885119867119885DR)
Figure 19 The scatter plots of 119877(6 km2 10min) versus 119866(6 km2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
R(ZH)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(a) 119877(119885119867)
R(ZH ZDR)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(b) 119877(119885119867119885DR)
Figure 20 The normalized bias of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and 119885DR The red
point-line and the blue point-line indicate without and with the attenuation correction respectively
uncertainty of the rain gauges Using an areal averagedvalue the NB was reduced to less than 7 with attenua-tion correction using either rainfall estimator regardless ofaccumulation timeThe r rmse were significantly reduced byattenuation correction in 119877(119885
119867) and the reduction of r rmse
with 119877(119885119867 119885DR) was even more dramatic
As summarized above the rain attenuation for 119885119867
and 119885DR in the X-band was corrected by ΦDP and theresults showed a good agreement with the value obtainedby scattering simulations with the DSD datasets The resultsof comparison with rain gauge data also showed a signifi-cant improvement in the accuracy of rainfall estimation by
18 Advances in Meteorology
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10r_
rmse
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH)
(a) 119877(119885119867)
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10
r_rm
se
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH ZDR)
(b) 119877(119885119867119885DR)
Figure 21 The relative root-mean square error of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and
119885DR The red point-line and the blue point-line indicate without and with the attenuation correction respectively
attenuation correction which supports the quantitative useof the X-band radar in rain data retrieval
Appendix
A Instrumental Uncertainty of the Rain Gauge
The instrumental uncertainty of the TB rain gauge is asso-ciated with improper calibration leakage and precipitationmissing the gauge due to the measurement principle andwetting and evaporation losses In addition external sourcessuch as wind turbulence and installed position also con-tribute to the instrumental uncertainty Ciach [17] reportedthat the TB gauge data suffers a significant uncertainty asso-ciated with random differences between closely collocatedTB gauges through the analysis of the error as a functionof the rainfall intensity for three timescales In addition theimproper calibration induces a bias in the rainfall measure-ment and this bias cannot be removed by increasing accu-mulation timeTherefore the proper correction or processingof TB gauge data is necessary to obtain accurate and reliablereading of the rainfall amount
In this appendix instrumental biases were calculatedto reduce the instrumental uncertainty through an idealexperiment in the laboratory and a field intercomparisonwith 10 TB rain gauges In addition the random errors werequantitatively calculated as a function of rainfall rate andaccumulation time
A1 Correction of Instrumental Bias The instrumental bias119861119868(119894) of the 119894th gauge was defined as a combination
of the absolute bias 119861119860(ref) of the reference gauge and
the relative bias 119861119877(119894) of the 119894th gauge according to the
following equation
119861119868 (119894) = 119861119860 (ref) times 119861119877 (119894) (A1)
119861119860(ref) signified the ratio of the calculated value from
the instrumental resolution of the reference gauge to the truevalue measured by other instruments
119861119860 (ref) =
rainfalltruthrainfallgauge (ref)
(A2)
In this study 119861119860(ref) was derived by an ideal experiment
using an electronic scale First the reference gauge was puton the scale and the scale was initialized The water dropsflowed out into the receiving orifice of the reference gaugeuntil the tipping number became 60Themeasured weight of60 tips was 2144 g while the theoretical (or expected) valuewas 2244 g for the reference gauge (receiving orifice size of1545 cm 60 tips and 02mm tip resolution) Therefore theabsolute bias of the reference gauge from (A2) was about0955 This indicates that the rainfall amount measured bythis reference gauge was overestimated by about 45 andthis bias should be corrected prior to any interpretation
119861119877(119894) represented the ratio of the observed (recorded)
value from the 119894th gauge to the observed (recorded) valuefrom the reference gauge
119861119877 (119894) =
rainfallgauge (ref)rainfallgauge (119894)
(A3)
119861119877(119894) was calculated from ten collocated rain gauges
including the reference gauge located at an observation field
Advances in Meteorology 19
Table 4 The instrumental and relative biases of the 10 tipping bucket rain gauges
Gauge ID RG1 RG2 RG3 RG4 RG5 RG6 RG7 RG8 RG9 RG10Instrumental bias 094 108 099 100 096 104 098 102 108 098Relative bias 098 113 103 105 100 109 102 107 113 102
013610152230406080120150200250300400
()
Rain
rate
(mm
h)
100101
Accumulation time (min)
10
100NSD ()
Figure 22 Two-dimensional normalized standard deviationbetween the standard rainfall and each rain gauge
at KNUKoreaThe observational period for intercomparisonwas from October 2009 to June 2010 (14 rainy days) andthe total accumulated rainfall for this period was about300mm The results are presented in Table 4 All relativebiases were mainly larger than 10 This indicated that themost of gauges underestimated the rainfall amount comparedto the reference rain gauge The relative bias of the referencegauge (119894 = 5) was 10
Finally 119861119860(119894) was calculated from (A1) (see Table 4) All
instrumental biases were between 090 and 11 and the meanvalue of the biases was nearly 10 This indicated that therainfall rate from a single gauge data had a maximum 10bias These biases were corrected for each gauge to removeexisting instrumental errors
A2 Quantification of Random Errors To quantify the ran-dom error due to instrumental uncertainty the NSD of theten gauges was calculated with accumulation time and rainrate using data from an intercomparison The NSD of the 119894thgauge compared to the reference gauge was derived for 20rain rate classes (Δ119877) and 17 accumulation time classes (Δ119905)as follows
NSD119894 (Δ119905 Δ119877)
=
(1119899)radicsum119899minus1
119896=0[1198661015840
119870(Δ119905 Δ119877) minus 119866119894119896 (Δ119905 Δ119877)]
2
1198661015840
(A4)
where 1198661015840 and 119866 indicate the reference and individual gaugesrespectively The overbar means the temporal average ofthe rainfall rate The NSDs of nine different gauges (exceptthe reference gauge) were averaged to obtain the two-dimensional average NSD (Figure 22) The average NSD
tended to decrease with increasing accumulation time andwith increasing rain rate The NSD was about 10ndash15when accumulation time was 10min and the rain rate was10mmhminus1 The instrumental errors were quite significant forlonger accumulation times and higher rainfall rates Theseerrors should be used as a guideline when a single gauge isused as a reference for any validation study
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper has been revised and extended fromMs Young-AOhrsquos MS thesis [25] of Kyungpook National University Thisresearch is supported by Development and Application ofCross Governmental Dual-Pol Radar Harmonization (WRC-2013-A-1) project of theWeather Radar Center KoreaMeteo-rological AdministrationThis work was funded by the KoreaMeteorological Administration Research and DevelopmentProgram under Grant KMIPA2015-1010
References
[1] V N Bringi V Chandrasekar N Balakrishnan and D SZrnic ldquoAn examination of propagation effects in rainfall onradar measurements at microwave frequenciesrdquo Journal ofAtmospheric and Oceanic Technology vol 7 no 6 pp 829ndash8401990
[2] A Huggel W Schmid and AWaldvogel ldquoRaindrop size distri-butions and the radar bright bandrdquo Journal of AppliedMeteorol-ogy vol 35 no 10 pp 1688ndash1701 1996
[3] G W Lee Errors in rain measurement by radar effect of vari-ability of drop size distributions [PhD thesis]McGill University2003
[4] J D Tuttle and R E Rinehart ldquoAttenuation correction in dual-wavelength analysesrdquo Journal of ClimateampAppliedMeteorologyvol 22 no 11 pp 1914ndash1921 1983
[5] S-G Park V N Bringi V Chandrasekar M Maki andK Iwanami ldquoCorrection of radar reflectivity and differentialreflectivity for rain attenuation at X band Part I theoretical andempirical basisrdquo Journal of Atmospheric and Oceanic Technol-ogy vol 22 no 11 pp 1621ndash1632 2005
[6] G Scarchilli E Gorgucci V Chandrasekar and T A SeligaldquoRainfall estimation using polarimetric techniques at C-bandfrequenciesrdquo Journal of Applied Meteorology vol 32 no 6 pp1150ndash1160 1993
[7] M N Anagnostou E N Anagnostou and J VivekanandanldquoCorrection for rain path specific and differential attenuation ofX-band dual-polarization observationsrdquo IEEE Transactions onGeoscience and Remote Sensing vol 44 no 9 pp 2470ndash24802006
20 Advances in Meteorology
[8] J Kalogiros M N Anagnostou E N Anagnostou M Mon-topoli E Picciotti and F S Marzano ldquoEvaluation of a newpolarimetric algorithm for rain-path attenuation correction ofX-band radar observations against disdrometerrdquo IEEE Transac-tions on Geoscience and Remote Sensing vol 52 no 2 pp 1369ndash1380 2014
[9] W-Y Chang J Vivekanandan and T-C C Wang ldquoEstimationof X-band polarimetric radar attenuation and measurementuncertainty using a variational methodrdquo Journal of AppliedMeteorology and Climatology vol 53 no 4 pp 1099ndash1119 2014
[10] J Hubbert and V N Bringi ldquoAn iterative filtering technique forthe analysis of copolar differential phase and dual-frequencyradar measurementsrdquo Journal of Atmospheric and OceanicTechnology vol 12 no 3 pp 643ndash648 1995
[11] K V Mishra W F Krajewski R Goska et al ldquoDeploy-ment and performance analyses of high-resolution iowa XPOLradar system during the NASA IFloodS campaignrdquo Journal ofHydrometeorology vol 17 no 2 pp 455ndash479 2016
[12] M Thurai G J Huang V N Bringi W L Randeu and MSchonhuber ldquoDrop shapes model comparisons and calcu-lations of polarimetric radar parameters in rainrdquo Journal ofAtmospheric and Oceanic Technology vol 24 no 6 pp 1019ndash1032 2007
[13] M D Humphrey J D Istok J Y Lee J A Hevesi and AL Flint ldquoA new method for automated dynamic calibration oftipping-bucket rain gaugesrdquo Journal of Atmospheric andOceanicTechnology vol 14 no 6 pp 1513ndash1519 1997
[14] E HabibW F Krajewski V Nespor andA Kruger ldquoNumericalsimulation studies of rain gage data correction due to windeffectrdquo Journal of Geophysical ResearchAtmospheres vol 104 no16 Article ID 1999JD900228 pp 19723ndash19733 1999
[15] G J Ciach and W F Krajewski ldquoOn the estimation of radarrainfall error variancerdquoAdvances inWater Resources vol 22 no6 pp 585ndash595 1999
[16] E Habib W F Krajewski and A Kruger ldquoSampling errors oftipping-bucket rain gaugemeasurementsrdquo Journal of HydrologicEngineering vol 6 no 2 pp 159ndash166 2001
[17] G J Ciach ldquoLocal random errors in tipping-bucket rain gaugemeasurementsrdquo Journal of Atmospheric andOceanic Technologyvol 20 no 5 pp 752ndash759 2003
[18] EHabibW F Krajewski andG J Ciach ldquoEstimation of rainfallinterstation correlationrdquo Journal ofHydrometeorology vol 2 no6 pp 621ndash629 2001
[19] NIMS ldquoStudy on estimation of rainfall amount and classifi-cation of hydrometeor species using dual-polarimetric radar(III)rdquo pp 1ndash20 2010
[20] E Gorgucci V Chandrasekar and L Baldini ldquoCorrection ofX-band radar observation for propagation effects based on theself-consistency principlerdquo Journal of Atmospheric and OceanicTechnology vol 23 pp 1668ndash1681 2006
[21] NIMS ldquoStudy on estimating raindrop size distribution forquantitative estimation of radar rainfall amountrdquo pp 1ndash55 2009
[22] NIMS ldquoQuantitative precipitation estimation and hydrometeoridentification using dual-polarization radar-Phase IIrdquo pp 1ndash742009
[23] S-G Park M Maki K Iwanami V N Bringi and V Chan-drasekar ldquoCorrection of radar reflectivity and differential reflec-tivity for rain attenuation at X band Part II Evaluation andapplicationrdquo Journal of Atmospheric and Oceanic Technologyvol 22 no 11 pp 1633ndash1655 2005
[24] E Habib and W F Krajewski ldquoUncertainty analysis of theTRMM ground-validation radar-rainfall products applicationto the TEFLUN-B field campaignrdquo Journal of Applied Meteorol-ogy vol 41 no 5 pp 558ndash572 2002
[25] Y-A Oh Rainfall estimation from X-band dual-polarizationradar measurements effects of attenuation correction and evalu-ation with a dense rain gauge network [MS thesis] KyungpookNational University 2011
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom
Applied ampEnvironmentalSoil Science
Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal of
Geophysics
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
Advances in Meteorology 13
UncorrectedCorrected
UncorrectedCorrected
17 18 19 20 2116Time (h)
17 18 19 20 2116Time (h)
1st gauge site 2nd gauge site
3rd gauge site 4th gauge site
5th gauge site 6th gauge site
7th gauge site 8th gauge site
9th gauge site 10th gauge site
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
minus6
minus4
minus2
0
2
4
6
ZD
R(d
B)
Figure 12 Time series of measured (red circle) and attenuation corrected (blue circle) 119885DR at the 10 rain gauge sites
14 Advances in Meteorology
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Uncorrected(a)
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Corrected Simulated from DSD(b)
Figure 13 Scatter plots of 119885119867and 119885DR before (a) and after (b) attenuation correction at the 10 rain gauge sites DSD observation results (dot
[21])
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
04
05
06
07
08
09
10
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
Accum time = 1min
Accum time = 10min
Accum time = 2min
Accum time = 20min
Accum time = 5min
Accum time = 60min
Figure 14 The Spearman correlation coefficient 119903 between data from two rain gauges at different accumulation times
estimation using X-band radar This algorithm was evalu-ated with radar measurements from NIMS-XPOL Howevermeasured ΨDP contained 120575 and observation noise as wellas ΦDP Furthermore the ΨDP offset varied with azimuthangles and decreased near the radar when rain cells were
above the radar In this study ΦDP was estimated by aniterative filtering technique and by adjusting with an offsetdetermined as the minimum value ofΦDP within 3 km of theradar As a result estimated ΦDP monotonically increasedThe path integrated value of119860
119867and119860DP was calculated from
Advances in Meteorology 15
2nd 6th 9th gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(038km2) versus G
(a) Normalized bias
⟨G⟩(038km2) versus G
2nd 6th 9th gaugesMean NSD
Mean R
0
20
40
60
80
100
NSD
()
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 15 (a) The NB and (b) NSD between ⟨119866⟩(038 km2) and 119866 from three rain gauges with accumulation time ⟨119866⟩ is the temporallyaveraged rainfall rate of ⟨119866⟩
All gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(6km2) versus G
(a) Normalized bias
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
0
20
40
60
80
100
NSD
()
All gaugesMean NSD
Mean R
⟨G⟩(6km2) versus G
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 16 (a) The NB and (b) NSD between ⟨119866⟩(6 km2) and 119866 from 10 rain gauges with accumulation time ⟨119866⟩ is the temporally averaged⟨119866⟩ with accumulation time
estimated ΦDP using 119860119867-ΦDP and 119860DP-ΦDP relationships
derived from the DSDs data Measured 119885119867
and 119885DR werecorrected with the calculated path integrated 119860
119867and 119860DP
by each gate The attenuated signal patterns of 119885119867and 119885DR
were restored to a typical rain echo pattern However theattenuation corrected 119885
119867and 119885DR fields were not matched
at distant ranges This was likely because of contaminationdue to beam broadening and the accumulation of errorswith increasing range Comparing the 119885
119867-119885DR scatter plots
over the 10 rain gauge sites with the results obtained froma scattering simulation with the DSD data the attenuation
corrected 119885119867-119885DR scatter plots are in good agreement with
the 119885119867-119885DR scatter plots from the DSDs
The validation for the improvement of accuracy in therainfall estimation through attenuation correction was per-formed by comparing with dense rain gauge network data119877(119885119867) from the attenuated (observed)119885
119867produced a severe
underestimation with an NB NSD rmse and r rmse of70 127 127 (mmhr) and 78 at 10min accumulationrespectively while 119877(119885
119867) from the attenuation corrected 119885
119867
showed better agreement with the gauge measurements withan NB NSD rmse and r rmse of minus1 52 521 (mmhr)
16 Advances in Meteorology
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus070
NSD = 127
rmse = 1279
corr = 0858
Corrected
NB = minus001
NSD = 052
rmse = 521
corr = 0927
r_rmse = 0786 r_rmse = 0320
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = 003
NSD = 049
rmse = 496
r_rmse = 0305
corr = 0936
Corrected
NB = 001
NSD = 050
rmse = 502
r_rmse = 0308
corr = 0936
(b) 119877(119885119867119885DR)
Figure 17 The scatter plots of 119877(038 km2 10min) versus 119866(007m2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
CorrectedUncorrected
R(m
mh
)
0
10
20
30
40
10 20 30 400G (mmh)
Uncorrected
NB = minus065
NSD = 086
rmse = 706
r_rmse = 0702
corr = 0878
Corrected
NB = minus005
NSD = 034
rmse = 281
r_rmse = 0279
corr = 0866
(a) 119877(119885119867)
CorrectedUncorrected
R(m
mh
)
10 20 30 400G (mmh)
0
10
20
30
40
Uncorrected
NB = 008
NSD = 023
rmse = 189
r_rmse = 0188
corr = 0942
Corrected
NB = 007
NSD = 025
rmse = 205
r_rmse = 0203
corr = 0940
(b) 119877(119885119867119885DR)
Figure 18 The scatter plots of 119877(038 km2 60min) versus 119866(007m2 60min) (a) using 119885119867and (b) using 119885
119867and 119885DR
and 32 at 10min accumulation respectively At 60minaccumulation theNBNSD rmse and r rmsewere decreasedfrom minus65 86 71 (mmhr) and 70 to minus5 3428 (mmhr) and 28 respectively Using 119877(119885
119867 119885DR) the
NB NSD rmse and r rmse were only slightly reduced even if
attenuation was corrected but were smaller than when using119877(119885119867) This indicated that 119877(119885
119867 119885DR) was less sensitive to
the attenuation than 119877(119885119867) In addition half of the errors
that remained after attenuation correction were caused by thesmall scale variability of the rain fields and the instrumental
Advances in Meteorology 17
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus069
NSD = 122
rmse = 391
r_rmse = 0770
corr = 0893
Corrected
NB = 007
NSD = 042
rmse = 133
r_rmse = 0262
corr = 0960
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = minus024
NSD = 067
rmse = 215
r_rmse = 0424
corr = 0718
Corrected
NB = 005
NSD = 039
rmse = 124
r_rmse = 0245
corr = 0891
(b) 119877(119885119867119885DR)
Figure 19 The scatter plots of 119877(6 km2 10min) versus 119866(6 km2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
R(ZH)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(a) 119877(119885119867)
R(ZH ZDR)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(b) 119877(119885119867119885DR)
Figure 20 The normalized bias of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and 119885DR The red
point-line and the blue point-line indicate without and with the attenuation correction respectively
uncertainty of the rain gauges Using an areal averagedvalue the NB was reduced to less than 7 with attenua-tion correction using either rainfall estimator regardless ofaccumulation timeThe r rmse were significantly reduced byattenuation correction in 119877(119885
119867) and the reduction of r rmse
with 119877(119885119867 119885DR) was even more dramatic
As summarized above the rain attenuation for 119885119867
and 119885DR in the X-band was corrected by ΦDP and theresults showed a good agreement with the value obtainedby scattering simulations with the DSD datasets The resultsof comparison with rain gauge data also showed a signifi-cant improvement in the accuracy of rainfall estimation by
18 Advances in Meteorology
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10r_
rmse
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH)
(a) 119877(119885119867)
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10
r_rm
se
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH ZDR)
(b) 119877(119885119867119885DR)
Figure 21 The relative root-mean square error of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and
119885DR The red point-line and the blue point-line indicate without and with the attenuation correction respectively
attenuation correction which supports the quantitative useof the X-band radar in rain data retrieval
Appendix
A Instrumental Uncertainty of the Rain Gauge
The instrumental uncertainty of the TB rain gauge is asso-ciated with improper calibration leakage and precipitationmissing the gauge due to the measurement principle andwetting and evaporation losses In addition external sourcessuch as wind turbulence and installed position also con-tribute to the instrumental uncertainty Ciach [17] reportedthat the TB gauge data suffers a significant uncertainty asso-ciated with random differences between closely collocatedTB gauges through the analysis of the error as a functionof the rainfall intensity for three timescales In addition theimproper calibration induces a bias in the rainfall measure-ment and this bias cannot be removed by increasing accu-mulation timeTherefore the proper correction or processingof TB gauge data is necessary to obtain accurate and reliablereading of the rainfall amount
In this appendix instrumental biases were calculatedto reduce the instrumental uncertainty through an idealexperiment in the laboratory and a field intercomparisonwith 10 TB rain gauges In addition the random errors werequantitatively calculated as a function of rainfall rate andaccumulation time
A1 Correction of Instrumental Bias The instrumental bias119861119868(119894) of the 119894th gauge was defined as a combination
of the absolute bias 119861119860(ref) of the reference gauge and
the relative bias 119861119877(119894) of the 119894th gauge according to the
following equation
119861119868 (119894) = 119861119860 (ref) times 119861119877 (119894) (A1)
119861119860(ref) signified the ratio of the calculated value from
the instrumental resolution of the reference gauge to the truevalue measured by other instruments
119861119860 (ref) =
rainfalltruthrainfallgauge (ref)
(A2)
In this study 119861119860(ref) was derived by an ideal experiment
using an electronic scale First the reference gauge was puton the scale and the scale was initialized The water dropsflowed out into the receiving orifice of the reference gaugeuntil the tipping number became 60Themeasured weight of60 tips was 2144 g while the theoretical (or expected) valuewas 2244 g for the reference gauge (receiving orifice size of1545 cm 60 tips and 02mm tip resolution) Therefore theabsolute bias of the reference gauge from (A2) was about0955 This indicates that the rainfall amount measured bythis reference gauge was overestimated by about 45 andthis bias should be corrected prior to any interpretation
119861119877(119894) represented the ratio of the observed (recorded)
value from the 119894th gauge to the observed (recorded) valuefrom the reference gauge
119861119877 (119894) =
rainfallgauge (ref)rainfallgauge (119894)
(A3)
119861119877(119894) was calculated from ten collocated rain gauges
including the reference gauge located at an observation field
Advances in Meteorology 19
Table 4 The instrumental and relative biases of the 10 tipping bucket rain gauges
Gauge ID RG1 RG2 RG3 RG4 RG5 RG6 RG7 RG8 RG9 RG10Instrumental bias 094 108 099 100 096 104 098 102 108 098Relative bias 098 113 103 105 100 109 102 107 113 102
013610152230406080120150200250300400
()
Rain
rate
(mm
h)
100101
Accumulation time (min)
10
100NSD ()
Figure 22 Two-dimensional normalized standard deviationbetween the standard rainfall and each rain gauge
at KNUKoreaThe observational period for intercomparisonwas from October 2009 to June 2010 (14 rainy days) andthe total accumulated rainfall for this period was about300mm The results are presented in Table 4 All relativebiases were mainly larger than 10 This indicated that themost of gauges underestimated the rainfall amount comparedto the reference rain gauge The relative bias of the referencegauge (119894 = 5) was 10
Finally 119861119860(119894) was calculated from (A1) (see Table 4) All
instrumental biases were between 090 and 11 and the meanvalue of the biases was nearly 10 This indicated that therainfall rate from a single gauge data had a maximum 10bias These biases were corrected for each gauge to removeexisting instrumental errors
A2 Quantification of Random Errors To quantify the ran-dom error due to instrumental uncertainty the NSD of theten gauges was calculated with accumulation time and rainrate using data from an intercomparison The NSD of the 119894thgauge compared to the reference gauge was derived for 20rain rate classes (Δ119877) and 17 accumulation time classes (Δ119905)as follows
NSD119894 (Δ119905 Δ119877)
=
(1119899)radicsum119899minus1
119896=0[1198661015840
119870(Δ119905 Δ119877) minus 119866119894119896 (Δ119905 Δ119877)]
2
1198661015840
(A4)
where 1198661015840 and 119866 indicate the reference and individual gaugesrespectively The overbar means the temporal average ofthe rainfall rate The NSDs of nine different gauges (exceptthe reference gauge) were averaged to obtain the two-dimensional average NSD (Figure 22) The average NSD
tended to decrease with increasing accumulation time andwith increasing rain rate The NSD was about 10ndash15when accumulation time was 10min and the rain rate was10mmhminus1 The instrumental errors were quite significant forlonger accumulation times and higher rainfall rates Theseerrors should be used as a guideline when a single gauge isused as a reference for any validation study
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper has been revised and extended fromMs Young-AOhrsquos MS thesis [25] of Kyungpook National University Thisresearch is supported by Development and Application ofCross Governmental Dual-Pol Radar Harmonization (WRC-2013-A-1) project of theWeather Radar Center KoreaMeteo-rological AdministrationThis work was funded by the KoreaMeteorological Administration Research and DevelopmentProgram under Grant KMIPA2015-1010
References
[1] V N Bringi V Chandrasekar N Balakrishnan and D SZrnic ldquoAn examination of propagation effects in rainfall onradar measurements at microwave frequenciesrdquo Journal ofAtmospheric and Oceanic Technology vol 7 no 6 pp 829ndash8401990
[2] A Huggel W Schmid and AWaldvogel ldquoRaindrop size distri-butions and the radar bright bandrdquo Journal of AppliedMeteorol-ogy vol 35 no 10 pp 1688ndash1701 1996
[3] G W Lee Errors in rain measurement by radar effect of vari-ability of drop size distributions [PhD thesis]McGill University2003
[4] J D Tuttle and R E Rinehart ldquoAttenuation correction in dual-wavelength analysesrdquo Journal of ClimateampAppliedMeteorologyvol 22 no 11 pp 1914ndash1921 1983
[5] S-G Park V N Bringi V Chandrasekar M Maki andK Iwanami ldquoCorrection of radar reflectivity and differentialreflectivity for rain attenuation at X band Part I theoretical andempirical basisrdquo Journal of Atmospheric and Oceanic Technol-ogy vol 22 no 11 pp 1621ndash1632 2005
[6] G Scarchilli E Gorgucci V Chandrasekar and T A SeligaldquoRainfall estimation using polarimetric techniques at C-bandfrequenciesrdquo Journal of Applied Meteorology vol 32 no 6 pp1150ndash1160 1993
[7] M N Anagnostou E N Anagnostou and J VivekanandanldquoCorrection for rain path specific and differential attenuation ofX-band dual-polarization observationsrdquo IEEE Transactions onGeoscience and Remote Sensing vol 44 no 9 pp 2470ndash24802006
20 Advances in Meteorology
[8] J Kalogiros M N Anagnostou E N Anagnostou M Mon-topoli E Picciotti and F S Marzano ldquoEvaluation of a newpolarimetric algorithm for rain-path attenuation correction ofX-band radar observations against disdrometerrdquo IEEE Transac-tions on Geoscience and Remote Sensing vol 52 no 2 pp 1369ndash1380 2014
[9] W-Y Chang J Vivekanandan and T-C C Wang ldquoEstimationof X-band polarimetric radar attenuation and measurementuncertainty using a variational methodrdquo Journal of AppliedMeteorology and Climatology vol 53 no 4 pp 1099ndash1119 2014
[10] J Hubbert and V N Bringi ldquoAn iterative filtering technique forthe analysis of copolar differential phase and dual-frequencyradar measurementsrdquo Journal of Atmospheric and OceanicTechnology vol 12 no 3 pp 643ndash648 1995
[11] K V Mishra W F Krajewski R Goska et al ldquoDeploy-ment and performance analyses of high-resolution iowa XPOLradar system during the NASA IFloodS campaignrdquo Journal ofHydrometeorology vol 17 no 2 pp 455ndash479 2016
[12] M Thurai G J Huang V N Bringi W L Randeu and MSchonhuber ldquoDrop shapes model comparisons and calcu-lations of polarimetric radar parameters in rainrdquo Journal ofAtmospheric and Oceanic Technology vol 24 no 6 pp 1019ndash1032 2007
[13] M D Humphrey J D Istok J Y Lee J A Hevesi and AL Flint ldquoA new method for automated dynamic calibration oftipping-bucket rain gaugesrdquo Journal of Atmospheric andOceanicTechnology vol 14 no 6 pp 1513ndash1519 1997
[14] E HabibW F Krajewski V Nespor andA Kruger ldquoNumericalsimulation studies of rain gage data correction due to windeffectrdquo Journal of Geophysical ResearchAtmospheres vol 104 no16 Article ID 1999JD900228 pp 19723ndash19733 1999
[15] G J Ciach and W F Krajewski ldquoOn the estimation of radarrainfall error variancerdquoAdvances inWater Resources vol 22 no6 pp 585ndash595 1999
[16] E Habib W F Krajewski and A Kruger ldquoSampling errors oftipping-bucket rain gaugemeasurementsrdquo Journal of HydrologicEngineering vol 6 no 2 pp 159ndash166 2001
[17] G J Ciach ldquoLocal random errors in tipping-bucket rain gaugemeasurementsrdquo Journal of Atmospheric andOceanic Technologyvol 20 no 5 pp 752ndash759 2003
[18] EHabibW F Krajewski andG J Ciach ldquoEstimation of rainfallinterstation correlationrdquo Journal ofHydrometeorology vol 2 no6 pp 621ndash629 2001
[19] NIMS ldquoStudy on estimation of rainfall amount and classifi-cation of hydrometeor species using dual-polarimetric radar(III)rdquo pp 1ndash20 2010
[20] E Gorgucci V Chandrasekar and L Baldini ldquoCorrection ofX-band radar observation for propagation effects based on theself-consistency principlerdquo Journal of Atmospheric and OceanicTechnology vol 23 pp 1668ndash1681 2006
[21] NIMS ldquoStudy on estimating raindrop size distribution forquantitative estimation of radar rainfall amountrdquo pp 1ndash55 2009
[22] NIMS ldquoQuantitative precipitation estimation and hydrometeoridentification using dual-polarization radar-Phase IIrdquo pp 1ndash742009
[23] S-G Park M Maki K Iwanami V N Bringi and V Chan-drasekar ldquoCorrection of radar reflectivity and differential reflec-tivity for rain attenuation at X band Part II Evaluation andapplicationrdquo Journal of Atmospheric and Oceanic Technologyvol 22 no 11 pp 1633ndash1655 2005
[24] E Habib and W F Krajewski ldquoUncertainty analysis of theTRMM ground-validation radar-rainfall products applicationto the TEFLUN-B field campaignrdquo Journal of Applied Meteorol-ogy vol 41 no 5 pp 558ndash572 2002
[25] Y-A Oh Rainfall estimation from X-band dual-polarizationradar measurements effects of attenuation correction and evalu-ation with a dense rain gauge network [MS thesis] KyungpookNational University 2011
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom
Applied ampEnvironmentalSoil Science
Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal of
Geophysics
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
14 Advances in Meteorology
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Uncorrected(a)
minus6
minus4
minus2
0
2
4
ZD
R(d
B)
0 10 20 30 40 50 60 70minus10
ZH (dBZ)
Corrected Simulated from DSD(b)
Figure 13 Scatter plots of 119885119867and 119885DR before (a) and after (b) attenuation correction at the 10 rain gauge sites DSD observation results (dot
[21])
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
04
05
06
07
08
09
10
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
04
05
06
07
08
09
10
Cor
relat
ion
coeffi
cien
tr
0 2 3 41
Range between two rain gauges (km)
0 2 3 41
Range between two rain gauges (km)
Accum time = 1min
Accum time = 10min
Accum time = 2min
Accum time = 20min
Accum time = 5min
Accum time = 60min
Figure 14 The Spearman correlation coefficient 119903 between data from two rain gauges at different accumulation times
estimation using X-band radar This algorithm was evalu-ated with radar measurements from NIMS-XPOL Howevermeasured ΨDP contained 120575 and observation noise as wellas ΦDP Furthermore the ΨDP offset varied with azimuthangles and decreased near the radar when rain cells were
above the radar In this study ΦDP was estimated by aniterative filtering technique and by adjusting with an offsetdetermined as the minimum value ofΦDP within 3 km of theradar As a result estimated ΦDP monotonically increasedThe path integrated value of119860
119867and119860DP was calculated from
Advances in Meteorology 15
2nd 6th 9th gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(038km2) versus G
(a) Normalized bias
⟨G⟩(038km2) versus G
2nd 6th 9th gaugesMean NSD
Mean R
0
20
40
60
80
100
NSD
()
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 15 (a) The NB and (b) NSD between ⟨119866⟩(038 km2) and 119866 from three rain gauges with accumulation time ⟨119866⟩ is the temporallyaveraged rainfall rate of ⟨119866⟩
All gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(6km2) versus G
(a) Normalized bias
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
0
20
40
60
80
100
NSD
()
All gaugesMean NSD
Mean R
⟨G⟩(6km2) versus G
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 16 (a) The NB and (b) NSD between ⟨119866⟩(6 km2) and 119866 from 10 rain gauges with accumulation time ⟨119866⟩ is the temporally averaged⟨119866⟩ with accumulation time
estimated ΦDP using 119860119867-ΦDP and 119860DP-ΦDP relationships
derived from the DSDs data Measured 119885119867
and 119885DR werecorrected with the calculated path integrated 119860
119867and 119860DP
by each gate The attenuated signal patterns of 119885119867and 119885DR
were restored to a typical rain echo pattern However theattenuation corrected 119885
119867and 119885DR fields were not matched
at distant ranges This was likely because of contaminationdue to beam broadening and the accumulation of errorswith increasing range Comparing the 119885
119867-119885DR scatter plots
over the 10 rain gauge sites with the results obtained froma scattering simulation with the DSD data the attenuation
corrected 119885119867-119885DR scatter plots are in good agreement with
the 119885119867-119885DR scatter plots from the DSDs
The validation for the improvement of accuracy in therainfall estimation through attenuation correction was per-formed by comparing with dense rain gauge network data119877(119885119867) from the attenuated (observed)119885
119867produced a severe
underestimation with an NB NSD rmse and r rmse of70 127 127 (mmhr) and 78 at 10min accumulationrespectively while 119877(119885
119867) from the attenuation corrected 119885
119867
showed better agreement with the gauge measurements withan NB NSD rmse and r rmse of minus1 52 521 (mmhr)
16 Advances in Meteorology
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus070
NSD = 127
rmse = 1279
corr = 0858
Corrected
NB = minus001
NSD = 052
rmse = 521
corr = 0927
r_rmse = 0786 r_rmse = 0320
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = 003
NSD = 049
rmse = 496
r_rmse = 0305
corr = 0936
Corrected
NB = 001
NSD = 050
rmse = 502
r_rmse = 0308
corr = 0936
(b) 119877(119885119867119885DR)
Figure 17 The scatter plots of 119877(038 km2 10min) versus 119866(007m2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
CorrectedUncorrected
R(m
mh
)
0
10
20
30
40
10 20 30 400G (mmh)
Uncorrected
NB = minus065
NSD = 086
rmse = 706
r_rmse = 0702
corr = 0878
Corrected
NB = minus005
NSD = 034
rmse = 281
r_rmse = 0279
corr = 0866
(a) 119877(119885119867)
CorrectedUncorrected
R(m
mh
)
10 20 30 400G (mmh)
0
10
20
30
40
Uncorrected
NB = 008
NSD = 023
rmse = 189
r_rmse = 0188
corr = 0942
Corrected
NB = 007
NSD = 025
rmse = 205
r_rmse = 0203
corr = 0940
(b) 119877(119885119867119885DR)
Figure 18 The scatter plots of 119877(038 km2 60min) versus 119866(007m2 60min) (a) using 119885119867and (b) using 119885
119867and 119885DR
and 32 at 10min accumulation respectively At 60minaccumulation theNBNSD rmse and r rmsewere decreasedfrom minus65 86 71 (mmhr) and 70 to minus5 3428 (mmhr) and 28 respectively Using 119877(119885
119867 119885DR) the
NB NSD rmse and r rmse were only slightly reduced even if
attenuation was corrected but were smaller than when using119877(119885119867) This indicated that 119877(119885
119867 119885DR) was less sensitive to
the attenuation than 119877(119885119867) In addition half of the errors
that remained after attenuation correction were caused by thesmall scale variability of the rain fields and the instrumental
Advances in Meteorology 17
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus069
NSD = 122
rmse = 391
r_rmse = 0770
corr = 0893
Corrected
NB = 007
NSD = 042
rmse = 133
r_rmse = 0262
corr = 0960
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = minus024
NSD = 067
rmse = 215
r_rmse = 0424
corr = 0718
Corrected
NB = 005
NSD = 039
rmse = 124
r_rmse = 0245
corr = 0891
(b) 119877(119885119867119885DR)
Figure 19 The scatter plots of 119877(6 km2 10min) versus 119866(6 km2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
R(ZH)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(a) 119877(119885119867)
R(ZH ZDR)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(b) 119877(119885119867119885DR)
Figure 20 The normalized bias of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and 119885DR The red
point-line and the blue point-line indicate without and with the attenuation correction respectively
uncertainty of the rain gauges Using an areal averagedvalue the NB was reduced to less than 7 with attenua-tion correction using either rainfall estimator regardless ofaccumulation timeThe r rmse were significantly reduced byattenuation correction in 119877(119885
119867) and the reduction of r rmse
with 119877(119885119867 119885DR) was even more dramatic
As summarized above the rain attenuation for 119885119867
and 119885DR in the X-band was corrected by ΦDP and theresults showed a good agreement with the value obtainedby scattering simulations with the DSD datasets The resultsof comparison with rain gauge data also showed a signifi-cant improvement in the accuracy of rainfall estimation by
18 Advances in Meteorology
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10r_
rmse
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH)
(a) 119877(119885119867)
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10
r_rm
se
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH ZDR)
(b) 119877(119885119867119885DR)
Figure 21 The relative root-mean square error of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and
119885DR The red point-line and the blue point-line indicate without and with the attenuation correction respectively
attenuation correction which supports the quantitative useof the X-band radar in rain data retrieval
Appendix
A Instrumental Uncertainty of the Rain Gauge
The instrumental uncertainty of the TB rain gauge is asso-ciated with improper calibration leakage and precipitationmissing the gauge due to the measurement principle andwetting and evaporation losses In addition external sourcessuch as wind turbulence and installed position also con-tribute to the instrumental uncertainty Ciach [17] reportedthat the TB gauge data suffers a significant uncertainty asso-ciated with random differences between closely collocatedTB gauges through the analysis of the error as a functionof the rainfall intensity for three timescales In addition theimproper calibration induces a bias in the rainfall measure-ment and this bias cannot be removed by increasing accu-mulation timeTherefore the proper correction or processingof TB gauge data is necessary to obtain accurate and reliablereading of the rainfall amount
In this appendix instrumental biases were calculatedto reduce the instrumental uncertainty through an idealexperiment in the laboratory and a field intercomparisonwith 10 TB rain gauges In addition the random errors werequantitatively calculated as a function of rainfall rate andaccumulation time
A1 Correction of Instrumental Bias The instrumental bias119861119868(119894) of the 119894th gauge was defined as a combination
of the absolute bias 119861119860(ref) of the reference gauge and
the relative bias 119861119877(119894) of the 119894th gauge according to the
following equation
119861119868 (119894) = 119861119860 (ref) times 119861119877 (119894) (A1)
119861119860(ref) signified the ratio of the calculated value from
the instrumental resolution of the reference gauge to the truevalue measured by other instruments
119861119860 (ref) =
rainfalltruthrainfallgauge (ref)
(A2)
In this study 119861119860(ref) was derived by an ideal experiment
using an electronic scale First the reference gauge was puton the scale and the scale was initialized The water dropsflowed out into the receiving orifice of the reference gaugeuntil the tipping number became 60Themeasured weight of60 tips was 2144 g while the theoretical (or expected) valuewas 2244 g for the reference gauge (receiving orifice size of1545 cm 60 tips and 02mm tip resolution) Therefore theabsolute bias of the reference gauge from (A2) was about0955 This indicates that the rainfall amount measured bythis reference gauge was overestimated by about 45 andthis bias should be corrected prior to any interpretation
119861119877(119894) represented the ratio of the observed (recorded)
value from the 119894th gauge to the observed (recorded) valuefrom the reference gauge
119861119877 (119894) =
rainfallgauge (ref)rainfallgauge (119894)
(A3)
119861119877(119894) was calculated from ten collocated rain gauges
including the reference gauge located at an observation field
Advances in Meteorology 19
Table 4 The instrumental and relative biases of the 10 tipping bucket rain gauges
Gauge ID RG1 RG2 RG3 RG4 RG5 RG6 RG7 RG8 RG9 RG10Instrumental bias 094 108 099 100 096 104 098 102 108 098Relative bias 098 113 103 105 100 109 102 107 113 102
013610152230406080120150200250300400
()
Rain
rate
(mm
h)
100101
Accumulation time (min)
10
100NSD ()
Figure 22 Two-dimensional normalized standard deviationbetween the standard rainfall and each rain gauge
at KNUKoreaThe observational period for intercomparisonwas from October 2009 to June 2010 (14 rainy days) andthe total accumulated rainfall for this period was about300mm The results are presented in Table 4 All relativebiases were mainly larger than 10 This indicated that themost of gauges underestimated the rainfall amount comparedto the reference rain gauge The relative bias of the referencegauge (119894 = 5) was 10
Finally 119861119860(119894) was calculated from (A1) (see Table 4) All
instrumental biases were between 090 and 11 and the meanvalue of the biases was nearly 10 This indicated that therainfall rate from a single gauge data had a maximum 10bias These biases were corrected for each gauge to removeexisting instrumental errors
A2 Quantification of Random Errors To quantify the ran-dom error due to instrumental uncertainty the NSD of theten gauges was calculated with accumulation time and rainrate using data from an intercomparison The NSD of the 119894thgauge compared to the reference gauge was derived for 20rain rate classes (Δ119877) and 17 accumulation time classes (Δ119905)as follows
NSD119894 (Δ119905 Δ119877)
=
(1119899)radicsum119899minus1
119896=0[1198661015840
119870(Δ119905 Δ119877) minus 119866119894119896 (Δ119905 Δ119877)]
2
1198661015840
(A4)
where 1198661015840 and 119866 indicate the reference and individual gaugesrespectively The overbar means the temporal average ofthe rainfall rate The NSDs of nine different gauges (exceptthe reference gauge) were averaged to obtain the two-dimensional average NSD (Figure 22) The average NSD
tended to decrease with increasing accumulation time andwith increasing rain rate The NSD was about 10ndash15when accumulation time was 10min and the rain rate was10mmhminus1 The instrumental errors were quite significant forlonger accumulation times and higher rainfall rates Theseerrors should be used as a guideline when a single gauge isused as a reference for any validation study
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper has been revised and extended fromMs Young-AOhrsquos MS thesis [25] of Kyungpook National University Thisresearch is supported by Development and Application ofCross Governmental Dual-Pol Radar Harmonization (WRC-2013-A-1) project of theWeather Radar Center KoreaMeteo-rological AdministrationThis work was funded by the KoreaMeteorological Administration Research and DevelopmentProgram under Grant KMIPA2015-1010
References
[1] V N Bringi V Chandrasekar N Balakrishnan and D SZrnic ldquoAn examination of propagation effects in rainfall onradar measurements at microwave frequenciesrdquo Journal ofAtmospheric and Oceanic Technology vol 7 no 6 pp 829ndash8401990
[2] A Huggel W Schmid and AWaldvogel ldquoRaindrop size distri-butions and the radar bright bandrdquo Journal of AppliedMeteorol-ogy vol 35 no 10 pp 1688ndash1701 1996
[3] G W Lee Errors in rain measurement by radar effect of vari-ability of drop size distributions [PhD thesis]McGill University2003
[4] J D Tuttle and R E Rinehart ldquoAttenuation correction in dual-wavelength analysesrdquo Journal of ClimateampAppliedMeteorologyvol 22 no 11 pp 1914ndash1921 1983
[5] S-G Park V N Bringi V Chandrasekar M Maki andK Iwanami ldquoCorrection of radar reflectivity and differentialreflectivity for rain attenuation at X band Part I theoretical andempirical basisrdquo Journal of Atmospheric and Oceanic Technol-ogy vol 22 no 11 pp 1621ndash1632 2005
[6] G Scarchilli E Gorgucci V Chandrasekar and T A SeligaldquoRainfall estimation using polarimetric techniques at C-bandfrequenciesrdquo Journal of Applied Meteorology vol 32 no 6 pp1150ndash1160 1993
[7] M N Anagnostou E N Anagnostou and J VivekanandanldquoCorrection for rain path specific and differential attenuation ofX-band dual-polarization observationsrdquo IEEE Transactions onGeoscience and Remote Sensing vol 44 no 9 pp 2470ndash24802006
20 Advances in Meteorology
[8] J Kalogiros M N Anagnostou E N Anagnostou M Mon-topoli E Picciotti and F S Marzano ldquoEvaluation of a newpolarimetric algorithm for rain-path attenuation correction ofX-band radar observations against disdrometerrdquo IEEE Transac-tions on Geoscience and Remote Sensing vol 52 no 2 pp 1369ndash1380 2014
[9] W-Y Chang J Vivekanandan and T-C C Wang ldquoEstimationof X-band polarimetric radar attenuation and measurementuncertainty using a variational methodrdquo Journal of AppliedMeteorology and Climatology vol 53 no 4 pp 1099ndash1119 2014
[10] J Hubbert and V N Bringi ldquoAn iterative filtering technique forthe analysis of copolar differential phase and dual-frequencyradar measurementsrdquo Journal of Atmospheric and OceanicTechnology vol 12 no 3 pp 643ndash648 1995
[11] K V Mishra W F Krajewski R Goska et al ldquoDeploy-ment and performance analyses of high-resolution iowa XPOLradar system during the NASA IFloodS campaignrdquo Journal ofHydrometeorology vol 17 no 2 pp 455ndash479 2016
[12] M Thurai G J Huang V N Bringi W L Randeu and MSchonhuber ldquoDrop shapes model comparisons and calcu-lations of polarimetric radar parameters in rainrdquo Journal ofAtmospheric and Oceanic Technology vol 24 no 6 pp 1019ndash1032 2007
[13] M D Humphrey J D Istok J Y Lee J A Hevesi and AL Flint ldquoA new method for automated dynamic calibration oftipping-bucket rain gaugesrdquo Journal of Atmospheric andOceanicTechnology vol 14 no 6 pp 1513ndash1519 1997
[14] E HabibW F Krajewski V Nespor andA Kruger ldquoNumericalsimulation studies of rain gage data correction due to windeffectrdquo Journal of Geophysical ResearchAtmospheres vol 104 no16 Article ID 1999JD900228 pp 19723ndash19733 1999
[15] G J Ciach and W F Krajewski ldquoOn the estimation of radarrainfall error variancerdquoAdvances inWater Resources vol 22 no6 pp 585ndash595 1999
[16] E Habib W F Krajewski and A Kruger ldquoSampling errors oftipping-bucket rain gaugemeasurementsrdquo Journal of HydrologicEngineering vol 6 no 2 pp 159ndash166 2001
[17] G J Ciach ldquoLocal random errors in tipping-bucket rain gaugemeasurementsrdquo Journal of Atmospheric andOceanic Technologyvol 20 no 5 pp 752ndash759 2003
[18] EHabibW F Krajewski andG J Ciach ldquoEstimation of rainfallinterstation correlationrdquo Journal ofHydrometeorology vol 2 no6 pp 621ndash629 2001
[19] NIMS ldquoStudy on estimation of rainfall amount and classifi-cation of hydrometeor species using dual-polarimetric radar(III)rdquo pp 1ndash20 2010
[20] E Gorgucci V Chandrasekar and L Baldini ldquoCorrection ofX-band radar observation for propagation effects based on theself-consistency principlerdquo Journal of Atmospheric and OceanicTechnology vol 23 pp 1668ndash1681 2006
[21] NIMS ldquoStudy on estimating raindrop size distribution forquantitative estimation of radar rainfall amountrdquo pp 1ndash55 2009
[22] NIMS ldquoQuantitative precipitation estimation and hydrometeoridentification using dual-polarization radar-Phase IIrdquo pp 1ndash742009
[23] S-G Park M Maki K Iwanami V N Bringi and V Chan-drasekar ldquoCorrection of radar reflectivity and differential reflec-tivity for rain attenuation at X band Part II Evaluation andapplicationrdquo Journal of Atmospheric and Oceanic Technologyvol 22 no 11 pp 1633ndash1655 2005
[24] E Habib and W F Krajewski ldquoUncertainty analysis of theTRMM ground-validation radar-rainfall products applicationto the TEFLUN-B field campaignrdquo Journal of Applied Meteorol-ogy vol 41 no 5 pp 558ndash572 2002
[25] Y-A Oh Rainfall estimation from X-band dual-polarizationradar measurements effects of attenuation correction and evalu-ation with a dense rain gauge network [MS thesis] KyungpookNational University 2011
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom
Applied ampEnvironmentalSoil Science
Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal of
Geophysics
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
Advances in Meteorology 15
2nd 6th 9th gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(038km2) versus G
(a) Normalized bias
⟨G⟩(038km2) versus G
2nd 6th 9th gaugesMean NSD
Mean R
0
20
40
60
80
100
NSD
()
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 15 (a) The NB and (b) NSD between ⟨119866⟩(038 km2) and 119866 from three rain gauges with accumulation time ⟨119866⟩ is the temporallyaveraged rainfall rate of ⟨119866⟩
All gaugesMean NB
10 1001
Accumulation time (min)
minus20
minus10
0
10
20
NB
()
⟨G⟩(6km2) versus G
(a) Normalized bias
0
5
10
15
20
Mea
n ra
in ra
te (m
mh
)
0
20
40
60
80
100
NSD
()
All gaugesMean NSD
Mean R
⟨G⟩(6km2) versus G
10 1001
Accumulation time (min)
(b) Normalized standard deviation
Figure 16 (a) The NB and (b) NSD between ⟨119866⟩(6 km2) and 119866 from 10 rain gauges with accumulation time ⟨119866⟩ is the temporally averaged⟨119866⟩ with accumulation time
estimated ΦDP using 119860119867-ΦDP and 119860DP-ΦDP relationships
derived from the DSDs data Measured 119885119867
and 119885DR werecorrected with the calculated path integrated 119860
119867and 119860DP
by each gate The attenuated signal patterns of 119885119867and 119885DR
were restored to a typical rain echo pattern However theattenuation corrected 119885
119867and 119885DR fields were not matched
at distant ranges This was likely because of contaminationdue to beam broadening and the accumulation of errorswith increasing range Comparing the 119885
119867-119885DR scatter plots
over the 10 rain gauge sites with the results obtained froma scattering simulation with the DSD data the attenuation
corrected 119885119867-119885DR scatter plots are in good agreement with
the 119885119867-119885DR scatter plots from the DSDs
The validation for the improvement of accuracy in therainfall estimation through attenuation correction was per-formed by comparing with dense rain gauge network data119877(119885119867) from the attenuated (observed)119885
119867produced a severe
underestimation with an NB NSD rmse and r rmse of70 127 127 (mmhr) and 78 at 10min accumulationrespectively while 119877(119885
119867) from the attenuation corrected 119885
119867
showed better agreement with the gauge measurements withan NB NSD rmse and r rmse of minus1 52 521 (mmhr)
16 Advances in Meteorology
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus070
NSD = 127
rmse = 1279
corr = 0858
Corrected
NB = minus001
NSD = 052
rmse = 521
corr = 0927
r_rmse = 0786 r_rmse = 0320
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = 003
NSD = 049
rmse = 496
r_rmse = 0305
corr = 0936
Corrected
NB = 001
NSD = 050
rmse = 502
r_rmse = 0308
corr = 0936
(b) 119877(119885119867119885DR)
Figure 17 The scatter plots of 119877(038 km2 10min) versus 119866(007m2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
CorrectedUncorrected
R(m
mh
)
0
10
20
30
40
10 20 30 400G (mmh)
Uncorrected
NB = minus065
NSD = 086
rmse = 706
r_rmse = 0702
corr = 0878
Corrected
NB = minus005
NSD = 034
rmse = 281
r_rmse = 0279
corr = 0866
(a) 119877(119885119867)
CorrectedUncorrected
R(m
mh
)
10 20 30 400G (mmh)
0
10
20
30
40
Uncorrected
NB = 008
NSD = 023
rmse = 189
r_rmse = 0188
corr = 0942
Corrected
NB = 007
NSD = 025
rmse = 205
r_rmse = 0203
corr = 0940
(b) 119877(119885119867119885DR)
Figure 18 The scatter plots of 119877(038 km2 60min) versus 119866(007m2 60min) (a) using 119885119867and (b) using 119885
119867and 119885DR
and 32 at 10min accumulation respectively At 60minaccumulation theNBNSD rmse and r rmsewere decreasedfrom minus65 86 71 (mmhr) and 70 to minus5 3428 (mmhr) and 28 respectively Using 119877(119885
119867 119885DR) the
NB NSD rmse and r rmse were only slightly reduced even if
attenuation was corrected but were smaller than when using119877(119885119867) This indicated that 119877(119885
119867 119885DR) was less sensitive to
the attenuation than 119877(119885119867) In addition half of the errors
that remained after attenuation correction were caused by thesmall scale variability of the rain fields and the instrumental
Advances in Meteorology 17
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus069
NSD = 122
rmse = 391
r_rmse = 0770
corr = 0893
Corrected
NB = 007
NSD = 042
rmse = 133
r_rmse = 0262
corr = 0960
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = minus024
NSD = 067
rmse = 215
r_rmse = 0424
corr = 0718
Corrected
NB = 005
NSD = 039
rmse = 124
r_rmse = 0245
corr = 0891
(b) 119877(119885119867119885DR)
Figure 19 The scatter plots of 119877(6 km2 10min) versus 119866(6 km2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
R(ZH)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(a) 119877(119885119867)
R(ZH ZDR)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(b) 119877(119885119867119885DR)
Figure 20 The normalized bias of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and 119885DR The red
point-line and the blue point-line indicate without and with the attenuation correction respectively
uncertainty of the rain gauges Using an areal averagedvalue the NB was reduced to less than 7 with attenua-tion correction using either rainfall estimator regardless ofaccumulation timeThe r rmse were significantly reduced byattenuation correction in 119877(119885
119867) and the reduction of r rmse
with 119877(119885119867 119885DR) was even more dramatic
As summarized above the rain attenuation for 119885119867
and 119885DR in the X-band was corrected by ΦDP and theresults showed a good agreement with the value obtainedby scattering simulations with the DSD datasets The resultsof comparison with rain gauge data also showed a signifi-cant improvement in the accuracy of rainfall estimation by
18 Advances in Meteorology
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10r_
rmse
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH)
(a) 119877(119885119867)
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10
r_rm
se
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH ZDR)
(b) 119877(119885119867119885DR)
Figure 21 The relative root-mean square error of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and
119885DR The red point-line and the blue point-line indicate without and with the attenuation correction respectively
attenuation correction which supports the quantitative useof the X-band radar in rain data retrieval
Appendix
A Instrumental Uncertainty of the Rain Gauge
The instrumental uncertainty of the TB rain gauge is asso-ciated with improper calibration leakage and precipitationmissing the gauge due to the measurement principle andwetting and evaporation losses In addition external sourcessuch as wind turbulence and installed position also con-tribute to the instrumental uncertainty Ciach [17] reportedthat the TB gauge data suffers a significant uncertainty asso-ciated with random differences between closely collocatedTB gauges through the analysis of the error as a functionof the rainfall intensity for three timescales In addition theimproper calibration induces a bias in the rainfall measure-ment and this bias cannot be removed by increasing accu-mulation timeTherefore the proper correction or processingof TB gauge data is necessary to obtain accurate and reliablereading of the rainfall amount
In this appendix instrumental biases were calculatedto reduce the instrumental uncertainty through an idealexperiment in the laboratory and a field intercomparisonwith 10 TB rain gauges In addition the random errors werequantitatively calculated as a function of rainfall rate andaccumulation time
A1 Correction of Instrumental Bias The instrumental bias119861119868(119894) of the 119894th gauge was defined as a combination
of the absolute bias 119861119860(ref) of the reference gauge and
the relative bias 119861119877(119894) of the 119894th gauge according to the
following equation
119861119868 (119894) = 119861119860 (ref) times 119861119877 (119894) (A1)
119861119860(ref) signified the ratio of the calculated value from
the instrumental resolution of the reference gauge to the truevalue measured by other instruments
119861119860 (ref) =
rainfalltruthrainfallgauge (ref)
(A2)
In this study 119861119860(ref) was derived by an ideal experiment
using an electronic scale First the reference gauge was puton the scale and the scale was initialized The water dropsflowed out into the receiving orifice of the reference gaugeuntil the tipping number became 60Themeasured weight of60 tips was 2144 g while the theoretical (or expected) valuewas 2244 g for the reference gauge (receiving orifice size of1545 cm 60 tips and 02mm tip resolution) Therefore theabsolute bias of the reference gauge from (A2) was about0955 This indicates that the rainfall amount measured bythis reference gauge was overestimated by about 45 andthis bias should be corrected prior to any interpretation
119861119877(119894) represented the ratio of the observed (recorded)
value from the 119894th gauge to the observed (recorded) valuefrom the reference gauge
119861119877 (119894) =
rainfallgauge (ref)rainfallgauge (119894)
(A3)
119861119877(119894) was calculated from ten collocated rain gauges
including the reference gauge located at an observation field
Advances in Meteorology 19
Table 4 The instrumental and relative biases of the 10 tipping bucket rain gauges
Gauge ID RG1 RG2 RG3 RG4 RG5 RG6 RG7 RG8 RG9 RG10Instrumental bias 094 108 099 100 096 104 098 102 108 098Relative bias 098 113 103 105 100 109 102 107 113 102
013610152230406080120150200250300400
()
Rain
rate
(mm
h)
100101
Accumulation time (min)
10
100NSD ()
Figure 22 Two-dimensional normalized standard deviationbetween the standard rainfall and each rain gauge
at KNUKoreaThe observational period for intercomparisonwas from October 2009 to June 2010 (14 rainy days) andthe total accumulated rainfall for this period was about300mm The results are presented in Table 4 All relativebiases were mainly larger than 10 This indicated that themost of gauges underestimated the rainfall amount comparedto the reference rain gauge The relative bias of the referencegauge (119894 = 5) was 10
Finally 119861119860(119894) was calculated from (A1) (see Table 4) All
instrumental biases were between 090 and 11 and the meanvalue of the biases was nearly 10 This indicated that therainfall rate from a single gauge data had a maximum 10bias These biases were corrected for each gauge to removeexisting instrumental errors
A2 Quantification of Random Errors To quantify the ran-dom error due to instrumental uncertainty the NSD of theten gauges was calculated with accumulation time and rainrate using data from an intercomparison The NSD of the 119894thgauge compared to the reference gauge was derived for 20rain rate classes (Δ119877) and 17 accumulation time classes (Δ119905)as follows
NSD119894 (Δ119905 Δ119877)
=
(1119899)radicsum119899minus1
119896=0[1198661015840
119870(Δ119905 Δ119877) minus 119866119894119896 (Δ119905 Δ119877)]
2
1198661015840
(A4)
where 1198661015840 and 119866 indicate the reference and individual gaugesrespectively The overbar means the temporal average ofthe rainfall rate The NSDs of nine different gauges (exceptthe reference gauge) were averaged to obtain the two-dimensional average NSD (Figure 22) The average NSD
tended to decrease with increasing accumulation time andwith increasing rain rate The NSD was about 10ndash15when accumulation time was 10min and the rain rate was10mmhminus1 The instrumental errors were quite significant forlonger accumulation times and higher rainfall rates Theseerrors should be used as a guideline when a single gauge isused as a reference for any validation study
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper has been revised and extended fromMs Young-AOhrsquos MS thesis [25] of Kyungpook National University Thisresearch is supported by Development and Application ofCross Governmental Dual-Pol Radar Harmonization (WRC-2013-A-1) project of theWeather Radar Center KoreaMeteo-rological AdministrationThis work was funded by the KoreaMeteorological Administration Research and DevelopmentProgram under Grant KMIPA2015-1010
References
[1] V N Bringi V Chandrasekar N Balakrishnan and D SZrnic ldquoAn examination of propagation effects in rainfall onradar measurements at microwave frequenciesrdquo Journal ofAtmospheric and Oceanic Technology vol 7 no 6 pp 829ndash8401990
[2] A Huggel W Schmid and AWaldvogel ldquoRaindrop size distri-butions and the radar bright bandrdquo Journal of AppliedMeteorol-ogy vol 35 no 10 pp 1688ndash1701 1996
[3] G W Lee Errors in rain measurement by radar effect of vari-ability of drop size distributions [PhD thesis]McGill University2003
[4] J D Tuttle and R E Rinehart ldquoAttenuation correction in dual-wavelength analysesrdquo Journal of ClimateampAppliedMeteorologyvol 22 no 11 pp 1914ndash1921 1983
[5] S-G Park V N Bringi V Chandrasekar M Maki andK Iwanami ldquoCorrection of radar reflectivity and differentialreflectivity for rain attenuation at X band Part I theoretical andempirical basisrdquo Journal of Atmospheric and Oceanic Technol-ogy vol 22 no 11 pp 1621ndash1632 2005
[6] G Scarchilli E Gorgucci V Chandrasekar and T A SeligaldquoRainfall estimation using polarimetric techniques at C-bandfrequenciesrdquo Journal of Applied Meteorology vol 32 no 6 pp1150ndash1160 1993
[7] M N Anagnostou E N Anagnostou and J VivekanandanldquoCorrection for rain path specific and differential attenuation ofX-band dual-polarization observationsrdquo IEEE Transactions onGeoscience and Remote Sensing vol 44 no 9 pp 2470ndash24802006
20 Advances in Meteorology
[8] J Kalogiros M N Anagnostou E N Anagnostou M Mon-topoli E Picciotti and F S Marzano ldquoEvaluation of a newpolarimetric algorithm for rain-path attenuation correction ofX-band radar observations against disdrometerrdquo IEEE Transac-tions on Geoscience and Remote Sensing vol 52 no 2 pp 1369ndash1380 2014
[9] W-Y Chang J Vivekanandan and T-C C Wang ldquoEstimationof X-band polarimetric radar attenuation and measurementuncertainty using a variational methodrdquo Journal of AppliedMeteorology and Climatology vol 53 no 4 pp 1099ndash1119 2014
[10] J Hubbert and V N Bringi ldquoAn iterative filtering technique forthe analysis of copolar differential phase and dual-frequencyradar measurementsrdquo Journal of Atmospheric and OceanicTechnology vol 12 no 3 pp 643ndash648 1995
[11] K V Mishra W F Krajewski R Goska et al ldquoDeploy-ment and performance analyses of high-resolution iowa XPOLradar system during the NASA IFloodS campaignrdquo Journal ofHydrometeorology vol 17 no 2 pp 455ndash479 2016
[12] M Thurai G J Huang V N Bringi W L Randeu and MSchonhuber ldquoDrop shapes model comparisons and calcu-lations of polarimetric radar parameters in rainrdquo Journal ofAtmospheric and Oceanic Technology vol 24 no 6 pp 1019ndash1032 2007
[13] M D Humphrey J D Istok J Y Lee J A Hevesi and AL Flint ldquoA new method for automated dynamic calibration oftipping-bucket rain gaugesrdquo Journal of Atmospheric andOceanicTechnology vol 14 no 6 pp 1513ndash1519 1997
[14] E HabibW F Krajewski V Nespor andA Kruger ldquoNumericalsimulation studies of rain gage data correction due to windeffectrdquo Journal of Geophysical ResearchAtmospheres vol 104 no16 Article ID 1999JD900228 pp 19723ndash19733 1999
[15] G J Ciach and W F Krajewski ldquoOn the estimation of radarrainfall error variancerdquoAdvances inWater Resources vol 22 no6 pp 585ndash595 1999
[16] E Habib W F Krajewski and A Kruger ldquoSampling errors oftipping-bucket rain gaugemeasurementsrdquo Journal of HydrologicEngineering vol 6 no 2 pp 159ndash166 2001
[17] G J Ciach ldquoLocal random errors in tipping-bucket rain gaugemeasurementsrdquo Journal of Atmospheric andOceanic Technologyvol 20 no 5 pp 752ndash759 2003
[18] EHabibW F Krajewski andG J Ciach ldquoEstimation of rainfallinterstation correlationrdquo Journal ofHydrometeorology vol 2 no6 pp 621ndash629 2001
[19] NIMS ldquoStudy on estimation of rainfall amount and classifi-cation of hydrometeor species using dual-polarimetric radar(III)rdquo pp 1ndash20 2010
[20] E Gorgucci V Chandrasekar and L Baldini ldquoCorrection ofX-band radar observation for propagation effects based on theself-consistency principlerdquo Journal of Atmospheric and OceanicTechnology vol 23 pp 1668ndash1681 2006
[21] NIMS ldquoStudy on estimating raindrop size distribution forquantitative estimation of radar rainfall amountrdquo pp 1ndash55 2009
[22] NIMS ldquoQuantitative precipitation estimation and hydrometeoridentification using dual-polarization radar-Phase IIrdquo pp 1ndash742009
[23] S-G Park M Maki K Iwanami V N Bringi and V Chan-drasekar ldquoCorrection of radar reflectivity and differential reflec-tivity for rain attenuation at X band Part II Evaluation andapplicationrdquo Journal of Atmospheric and Oceanic Technologyvol 22 no 11 pp 1633ndash1655 2005
[24] E Habib and W F Krajewski ldquoUncertainty analysis of theTRMM ground-validation radar-rainfall products applicationto the TEFLUN-B field campaignrdquo Journal of Applied Meteorol-ogy vol 41 no 5 pp 558ndash572 2002
[25] Y-A Oh Rainfall estimation from X-band dual-polarizationradar measurements effects of attenuation correction and evalu-ation with a dense rain gauge network [MS thesis] KyungpookNational University 2011
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom
Applied ampEnvironmentalSoil Science
Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal of
Geophysics
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
16 Advances in Meteorology
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus070
NSD = 127
rmse = 1279
corr = 0858
Corrected
NB = minus001
NSD = 052
rmse = 521
corr = 0927
r_rmse = 0786 r_rmse = 0320
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = 003
NSD = 049
rmse = 496
r_rmse = 0305
corr = 0936
Corrected
NB = 001
NSD = 050
rmse = 502
r_rmse = 0308
corr = 0936
(b) 119877(119885119867119885DR)
Figure 17 The scatter plots of 119877(038 km2 10min) versus 119866(007m2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
CorrectedUncorrected
R(m
mh
)
0
10
20
30
40
10 20 30 400G (mmh)
Uncorrected
NB = minus065
NSD = 086
rmse = 706
r_rmse = 0702
corr = 0878
Corrected
NB = minus005
NSD = 034
rmse = 281
r_rmse = 0279
corr = 0866
(a) 119877(119885119867)
CorrectedUncorrected
R(m
mh
)
10 20 30 400G (mmh)
0
10
20
30
40
Uncorrected
NB = 008
NSD = 023
rmse = 189
r_rmse = 0188
corr = 0942
Corrected
NB = 007
NSD = 025
rmse = 205
r_rmse = 0203
corr = 0940
(b) 119877(119885119867119885DR)
Figure 18 The scatter plots of 119877(038 km2 60min) versus 119866(007m2 60min) (a) using 119885119867and (b) using 119885
119867and 119885DR
and 32 at 10min accumulation respectively At 60minaccumulation theNBNSD rmse and r rmsewere decreasedfrom minus65 86 71 (mmhr) and 70 to minus5 3428 (mmhr) and 28 respectively Using 119877(119885
119867 119885DR) the
NB NSD rmse and r rmse were only slightly reduced even if
attenuation was corrected but were smaller than when using119877(119885119867) This indicated that 119877(119885
119867 119885DR) was less sensitive to
the attenuation than 119877(119885119867) In addition half of the errors
that remained after attenuation correction were caused by thesmall scale variability of the rain fields and the instrumental
Advances in Meteorology 17
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus069
NSD = 122
rmse = 391
r_rmse = 0770
corr = 0893
Corrected
NB = 007
NSD = 042
rmse = 133
r_rmse = 0262
corr = 0960
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = minus024
NSD = 067
rmse = 215
r_rmse = 0424
corr = 0718
Corrected
NB = 005
NSD = 039
rmse = 124
r_rmse = 0245
corr = 0891
(b) 119877(119885119867119885DR)
Figure 19 The scatter plots of 119877(6 km2 10min) versus 119866(6 km2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
R(ZH)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(a) 119877(119885119867)
R(ZH ZDR)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(b) 119877(119885119867119885DR)
Figure 20 The normalized bias of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and 119885DR The red
point-line and the blue point-line indicate without and with the attenuation correction respectively
uncertainty of the rain gauges Using an areal averagedvalue the NB was reduced to less than 7 with attenua-tion correction using either rainfall estimator regardless ofaccumulation timeThe r rmse were significantly reduced byattenuation correction in 119877(119885
119867) and the reduction of r rmse
with 119877(119885119867 119885DR) was even more dramatic
As summarized above the rain attenuation for 119885119867
and 119885DR in the X-band was corrected by ΦDP and theresults showed a good agreement with the value obtainedby scattering simulations with the DSD datasets The resultsof comparison with rain gauge data also showed a signifi-cant improvement in the accuracy of rainfall estimation by
18 Advances in Meteorology
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10r_
rmse
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH)
(a) 119877(119885119867)
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10
r_rm
se
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH ZDR)
(b) 119877(119885119867119885DR)
Figure 21 The relative root-mean square error of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and
119885DR The red point-line and the blue point-line indicate without and with the attenuation correction respectively
attenuation correction which supports the quantitative useof the X-band radar in rain data retrieval
Appendix
A Instrumental Uncertainty of the Rain Gauge
The instrumental uncertainty of the TB rain gauge is asso-ciated with improper calibration leakage and precipitationmissing the gauge due to the measurement principle andwetting and evaporation losses In addition external sourcessuch as wind turbulence and installed position also con-tribute to the instrumental uncertainty Ciach [17] reportedthat the TB gauge data suffers a significant uncertainty asso-ciated with random differences between closely collocatedTB gauges through the analysis of the error as a functionof the rainfall intensity for three timescales In addition theimproper calibration induces a bias in the rainfall measure-ment and this bias cannot be removed by increasing accu-mulation timeTherefore the proper correction or processingof TB gauge data is necessary to obtain accurate and reliablereading of the rainfall amount
In this appendix instrumental biases were calculatedto reduce the instrumental uncertainty through an idealexperiment in the laboratory and a field intercomparisonwith 10 TB rain gauges In addition the random errors werequantitatively calculated as a function of rainfall rate andaccumulation time
A1 Correction of Instrumental Bias The instrumental bias119861119868(119894) of the 119894th gauge was defined as a combination
of the absolute bias 119861119860(ref) of the reference gauge and
the relative bias 119861119877(119894) of the 119894th gauge according to the
following equation
119861119868 (119894) = 119861119860 (ref) times 119861119877 (119894) (A1)
119861119860(ref) signified the ratio of the calculated value from
the instrumental resolution of the reference gauge to the truevalue measured by other instruments
119861119860 (ref) =
rainfalltruthrainfallgauge (ref)
(A2)
In this study 119861119860(ref) was derived by an ideal experiment
using an electronic scale First the reference gauge was puton the scale and the scale was initialized The water dropsflowed out into the receiving orifice of the reference gaugeuntil the tipping number became 60Themeasured weight of60 tips was 2144 g while the theoretical (or expected) valuewas 2244 g for the reference gauge (receiving orifice size of1545 cm 60 tips and 02mm tip resolution) Therefore theabsolute bias of the reference gauge from (A2) was about0955 This indicates that the rainfall amount measured bythis reference gauge was overestimated by about 45 andthis bias should be corrected prior to any interpretation
119861119877(119894) represented the ratio of the observed (recorded)
value from the 119894th gauge to the observed (recorded) valuefrom the reference gauge
119861119877 (119894) =
rainfallgauge (ref)rainfallgauge (119894)
(A3)
119861119877(119894) was calculated from ten collocated rain gauges
including the reference gauge located at an observation field
Advances in Meteorology 19
Table 4 The instrumental and relative biases of the 10 tipping bucket rain gauges
Gauge ID RG1 RG2 RG3 RG4 RG5 RG6 RG7 RG8 RG9 RG10Instrumental bias 094 108 099 100 096 104 098 102 108 098Relative bias 098 113 103 105 100 109 102 107 113 102
013610152230406080120150200250300400
()
Rain
rate
(mm
h)
100101
Accumulation time (min)
10
100NSD ()
Figure 22 Two-dimensional normalized standard deviationbetween the standard rainfall and each rain gauge
at KNUKoreaThe observational period for intercomparisonwas from October 2009 to June 2010 (14 rainy days) andthe total accumulated rainfall for this period was about300mm The results are presented in Table 4 All relativebiases were mainly larger than 10 This indicated that themost of gauges underestimated the rainfall amount comparedto the reference rain gauge The relative bias of the referencegauge (119894 = 5) was 10
Finally 119861119860(119894) was calculated from (A1) (see Table 4) All
instrumental biases were between 090 and 11 and the meanvalue of the biases was nearly 10 This indicated that therainfall rate from a single gauge data had a maximum 10bias These biases were corrected for each gauge to removeexisting instrumental errors
A2 Quantification of Random Errors To quantify the ran-dom error due to instrumental uncertainty the NSD of theten gauges was calculated with accumulation time and rainrate using data from an intercomparison The NSD of the 119894thgauge compared to the reference gauge was derived for 20rain rate classes (Δ119877) and 17 accumulation time classes (Δ119905)as follows
NSD119894 (Δ119905 Δ119877)
=
(1119899)radicsum119899minus1
119896=0[1198661015840
119870(Δ119905 Δ119877) minus 119866119894119896 (Δ119905 Δ119877)]
2
1198661015840
(A4)
where 1198661015840 and 119866 indicate the reference and individual gaugesrespectively The overbar means the temporal average ofthe rainfall rate The NSDs of nine different gauges (exceptthe reference gauge) were averaged to obtain the two-dimensional average NSD (Figure 22) The average NSD
tended to decrease with increasing accumulation time andwith increasing rain rate The NSD was about 10ndash15when accumulation time was 10min and the rain rate was10mmhminus1 The instrumental errors were quite significant forlonger accumulation times and higher rainfall rates Theseerrors should be used as a guideline when a single gauge isused as a reference for any validation study
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper has been revised and extended fromMs Young-AOhrsquos MS thesis [25] of Kyungpook National University Thisresearch is supported by Development and Application ofCross Governmental Dual-Pol Radar Harmonization (WRC-2013-A-1) project of theWeather Radar Center KoreaMeteo-rological AdministrationThis work was funded by the KoreaMeteorological Administration Research and DevelopmentProgram under Grant KMIPA2015-1010
References
[1] V N Bringi V Chandrasekar N Balakrishnan and D SZrnic ldquoAn examination of propagation effects in rainfall onradar measurements at microwave frequenciesrdquo Journal ofAtmospheric and Oceanic Technology vol 7 no 6 pp 829ndash8401990
[2] A Huggel W Schmid and AWaldvogel ldquoRaindrop size distri-butions and the radar bright bandrdquo Journal of AppliedMeteorol-ogy vol 35 no 10 pp 1688ndash1701 1996
[3] G W Lee Errors in rain measurement by radar effect of vari-ability of drop size distributions [PhD thesis]McGill University2003
[4] J D Tuttle and R E Rinehart ldquoAttenuation correction in dual-wavelength analysesrdquo Journal of ClimateampAppliedMeteorologyvol 22 no 11 pp 1914ndash1921 1983
[5] S-G Park V N Bringi V Chandrasekar M Maki andK Iwanami ldquoCorrection of radar reflectivity and differentialreflectivity for rain attenuation at X band Part I theoretical andempirical basisrdquo Journal of Atmospheric and Oceanic Technol-ogy vol 22 no 11 pp 1621ndash1632 2005
[6] G Scarchilli E Gorgucci V Chandrasekar and T A SeligaldquoRainfall estimation using polarimetric techniques at C-bandfrequenciesrdquo Journal of Applied Meteorology vol 32 no 6 pp1150ndash1160 1993
[7] M N Anagnostou E N Anagnostou and J VivekanandanldquoCorrection for rain path specific and differential attenuation ofX-band dual-polarization observationsrdquo IEEE Transactions onGeoscience and Remote Sensing vol 44 no 9 pp 2470ndash24802006
20 Advances in Meteorology
[8] J Kalogiros M N Anagnostou E N Anagnostou M Mon-topoli E Picciotti and F S Marzano ldquoEvaluation of a newpolarimetric algorithm for rain-path attenuation correction ofX-band radar observations against disdrometerrdquo IEEE Transac-tions on Geoscience and Remote Sensing vol 52 no 2 pp 1369ndash1380 2014
[9] W-Y Chang J Vivekanandan and T-C C Wang ldquoEstimationof X-band polarimetric radar attenuation and measurementuncertainty using a variational methodrdquo Journal of AppliedMeteorology and Climatology vol 53 no 4 pp 1099ndash1119 2014
[10] J Hubbert and V N Bringi ldquoAn iterative filtering technique forthe analysis of copolar differential phase and dual-frequencyradar measurementsrdquo Journal of Atmospheric and OceanicTechnology vol 12 no 3 pp 643ndash648 1995
[11] K V Mishra W F Krajewski R Goska et al ldquoDeploy-ment and performance analyses of high-resolution iowa XPOLradar system during the NASA IFloodS campaignrdquo Journal ofHydrometeorology vol 17 no 2 pp 455ndash479 2016
[12] M Thurai G J Huang V N Bringi W L Randeu and MSchonhuber ldquoDrop shapes model comparisons and calcu-lations of polarimetric radar parameters in rainrdquo Journal ofAtmospheric and Oceanic Technology vol 24 no 6 pp 1019ndash1032 2007
[13] M D Humphrey J D Istok J Y Lee J A Hevesi and AL Flint ldquoA new method for automated dynamic calibration oftipping-bucket rain gaugesrdquo Journal of Atmospheric andOceanicTechnology vol 14 no 6 pp 1513ndash1519 1997
[14] E HabibW F Krajewski V Nespor andA Kruger ldquoNumericalsimulation studies of rain gage data correction due to windeffectrdquo Journal of Geophysical ResearchAtmospheres vol 104 no16 Article ID 1999JD900228 pp 19723ndash19733 1999
[15] G J Ciach and W F Krajewski ldquoOn the estimation of radarrainfall error variancerdquoAdvances inWater Resources vol 22 no6 pp 585ndash595 1999
[16] E Habib W F Krajewski and A Kruger ldquoSampling errors oftipping-bucket rain gaugemeasurementsrdquo Journal of HydrologicEngineering vol 6 no 2 pp 159ndash166 2001
[17] G J Ciach ldquoLocal random errors in tipping-bucket rain gaugemeasurementsrdquo Journal of Atmospheric andOceanic Technologyvol 20 no 5 pp 752ndash759 2003
[18] EHabibW F Krajewski andG J Ciach ldquoEstimation of rainfallinterstation correlationrdquo Journal ofHydrometeorology vol 2 no6 pp 621ndash629 2001
[19] NIMS ldquoStudy on estimation of rainfall amount and classifi-cation of hydrometeor species using dual-polarimetric radar(III)rdquo pp 1ndash20 2010
[20] E Gorgucci V Chandrasekar and L Baldini ldquoCorrection ofX-band radar observation for propagation effects based on theself-consistency principlerdquo Journal of Atmospheric and OceanicTechnology vol 23 pp 1668ndash1681 2006
[21] NIMS ldquoStudy on estimating raindrop size distribution forquantitative estimation of radar rainfall amountrdquo pp 1ndash55 2009
[22] NIMS ldquoQuantitative precipitation estimation and hydrometeoridentification using dual-polarization radar-Phase IIrdquo pp 1ndash742009
[23] S-G Park M Maki K Iwanami V N Bringi and V Chan-drasekar ldquoCorrection of radar reflectivity and differential reflec-tivity for rain attenuation at X band Part II Evaluation andapplicationrdquo Journal of Atmospheric and Oceanic Technologyvol 22 no 11 pp 1633ndash1655 2005
[24] E Habib and W F Krajewski ldquoUncertainty analysis of theTRMM ground-validation radar-rainfall products applicationto the TEFLUN-B field campaignrdquo Journal of Applied Meteorol-ogy vol 41 no 5 pp 558ndash572 2002
[25] Y-A Oh Rainfall estimation from X-band dual-polarizationradar measurements effects of attenuation correction and evalu-ation with a dense rain gauge network [MS thesis] KyungpookNational University 2011
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom
Applied ampEnvironmentalSoil Science
Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal of
Geophysics
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
Advances in Meteorology 17
R(m
mh
)
CorrectedUncorrected
0
20
40
60
80
20 40 60 800G (mmh)
Uncorrected
NB = minus069
NSD = 122
rmse = 391
r_rmse = 0770
corr = 0893
Corrected
NB = 007
NSD = 042
rmse = 133
r_rmse = 0262
corr = 0960
(a) 119877(119885119867)
R(m
mh
)CorrectedUncorrected
20 40 60 800G (mmh)
0
20
40
60
80
Uncorrected
NB = minus024
NSD = 067
rmse = 215
r_rmse = 0424
corr = 0718
Corrected
NB = 005
NSD = 039
rmse = 124
r_rmse = 0245
corr = 0891
(b) 119877(119885119867119885DR)
Figure 19 The scatter plots of 119877(6 km2 10min) versus 119866(6 km2 10min) (a) using 119885119867and (b) using 119885
119867and 119885DR
R(ZH)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(a) 119877(119885119867)
R(ZH ZDR)
20 40 60 80 1000Accumulation time (min)
minus10
minus08
minus06
minus04
minus02
00
02
04
NB
CorrectedUncorrected
(b) 119877(119885119867119885DR)
Figure 20 The normalized bias of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and 119885DR The red
point-line and the blue point-line indicate without and with the attenuation correction respectively
uncertainty of the rain gauges Using an areal averagedvalue the NB was reduced to less than 7 with attenua-tion correction using either rainfall estimator regardless ofaccumulation timeThe r rmse were significantly reduced byattenuation correction in 119877(119885
119867) and the reduction of r rmse
with 119877(119885119867 119885DR) was even more dramatic
As summarized above the rain attenuation for 119885119867
and 119885DR in the X-band was corrected by ΦDP and theresults showed a good agreement with the value obtainedby scattering simulations with the DSD datasets The resultsof comparison with rain gauge data also showed a signifi-cant improvement in the accuracy of rainfall estimation by
18 Advances in Meteorology
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10r_
rmse
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH)
(a) 119877(119885119867)
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10
r_rm
se
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH ZDR)
(b) 119877(119885119867119885DR)
Figure 21 The relative root-mean square error of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and
119885DR The red point-line and the blue point-line indicate without and with the attenuation correction respectively
attenuation correction which supports the quantitative useof the X-band radar in rain data retrieval
Appendix
A Instrumental Uncertainty of the Rain Gauge
The instrumental uncertainty of the TB rain gauge is asso-ciated with improper calibration leakage and precipitationmissing the gauge due to the measurement principle andwetting and evaporation losses In addition external sourcessuch as wind turbulence and installed position also con-tribute to the instrumental uncertainty Ciach [17] reportedthat the TB gauge data suffers a significant uncertainty asso-ciated with random differences between closely collocatedTB gauges through the analysis of the error as a functionof the rainfall intensity for three timescales In addition theimproper calibration induces a bias in the rainfall measure-ment and this bias cannot be removed by increasing accu-mulation timeTherefore the proper correction or processingof TB gauge data is necessary to obtain accurate and reliablereading of the rainfall amount
In this appendix instrumental biases were calculatedto reduce the instrumental uncertainty through an idealexperiment in the laboratory and a field intercomparisonwith 10 TB rain gauges In addition the random errors werequantitatively calculated as a function of rainfall rate andaccumulation time
A1 Correction of Instrumental Bias The instrumental bias119861119868(119894) of the 119894th gauge was defined as a combination
of the absolute bias 119861119860(ref) of the reference gauge and
the relative bias 119861119877(119894) of the 119894th gauge according to the
following equation
119861119868 (119894) = 119861119860 (ref) times 119861119877 (119894) (A1)
119861119860(ref) signified the ratio of the calculated value from
the instrumental resolution of the reference gauge to the truevalue measured by other instruments
119861119860 (ref) =
rainfalltruthrainfallgauge (ref)
(A2)
In this study 119861119860(ref) was derived by an ideal experiment
using an electronic scale First the reference gauge was puton the scale and the scale was initialized The water dropsflowed out into the receiving orifice of the reference gaugeuntil the tipping number became 60Themeasured weight of60 tips was 2144 g while the theoretical (or expected) valuewas 2244 g for the reference gauge (receiving orifice size of1545 cm 60 tips and 02mm tip resolution) Therefore theabsolute bias of the reference gauge from (A2) was about0955 This indicates that the rainfall amount measured bythis reference gauge was overestimated by about 45 andthis bias should be corrected prior to any interpretation
119861119877(119894) represented the ratio of the observed (recorded)
value from the 119894th gauge to the observed (recorded) valuefrom the reference gauge
119861119877 (119894) =
rainfallgauge (ref)rainfallgauge (119894)
(A3)
119861119877(119894) was calculated from ten collocated rain gauges
including the reference gauge located at an observation field
Advances in Meteorology 19
Table 4 The instrumental and relative biases of the 10 tipping bucket rain gauges
Gauge ID RG1 RG2 RG3 RG4 RG5 RG6 RG7 RG8 RG9 RG10Instrumental bias 094 108 099 100 096 104 098 102 108 098Relative bias 098 113 103 105 100 109 102 107 113 102
013610152230406080120150200250300400
()
Rain
rate
(mm
h)
100101
Accumulation time (min)
10
100NSD ()
Figure 22 Two-dimensional normalized standard deviationbetween the standard rainfall and each rain gauge
at KNUKoreaThe observational period for intercomparisonwas from October 2009 to June 2010 (14 rainy days) andthe total accumulated rainfall for this period was about300mm The results are presented in Table 4 All relativebiases were mainly larger than 10 This indicated that themost of gauges underestimated the rainfall amount comparedto the reference rain gauge The relative bias of the referencegauge (119894 = 5) was 10
Finally 119861119860(119894) was calculated from (A1) (see Table 4) All
instrumental biases were between 090 and 11 and the meanvalue of the biases was nearly 10 This indicated that therainfall rate from a single gauge data had a maximum 10bias These biases were corrected for each gauge to removeexisting instrumental errors
A2 Quantification of Random Errors To quantify the ran-dom error due to instrumental uncertainty the NSD of theten gauges was calculated with accumulation time and rainrate using data from an intercomparison The NSD of the 119894thgauge compared to the reference gauge was derived for 20rain rate classes (Δ119877) and 17 accumulation time classes (Δ119905)as follows
NSD119894 (Δ119905 Δ119877)
=
(1119899)radicsum119899minus1
119896=0[1198661015840
119870(Δ119905 Δ119877) minus 119866119894119896 (Δ119905 Δ119877)]
2
1198661015840
(A4)
where 1198661015840 and 119866 indicate the reference and individual gaugesrespectively The overbar means the temporal average ofthe rainfall rate The NSDs of nine different gauges (exceptthe reference gauge) were averaged to obtain the two-dimensional average NSD (Figure 22) The average NSD
tended to decrease with increasing accumulation time andwith increasing rain rate The NSD was about 10ndash15when accumulation time was 10min and the rain rate was10mmhminus1 The instrumental errors were quite significant forlonger accumulation times and higher rainfall rates Theseerrors should be used as a guideline when a single gauge isused as a reference for any validation study
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper has been revised and extended fromMs Young-AOhrsquos MS thesis [25] of Kyungpook National University Thisresearch is supported by Development and Application ofCross Governmental Dual-Pol Radar Harmonization (WRC-2013-A-1) project of theWeather Radar Center KoreaMeteo-rological AdministrationThis work was funded by the KoreaMeteorological Administration Research and DevelopmentProgram under Grant KMIPA2015-1010
References
[1] V N Bringi V Chandrasekar N Balakrishnan and D SZrnic ldquoAn examination of propagation effects in rainfall onradar measurements at microwave frequenciesrdquo Journal ofAtmospheric and Oceanic Technology vol 7 no 6 pp 829ndash8401990
[2] A Huggel W Schmid and AWaldvogel ldquoRaindrop size distri-butions and the radar bright bandrdquo Journal of AppliedMeteorol-ogy vol 35 no 10 pp 1688ndash1701 1996
[3] G W Lee Errors in rain measurement by radar effect of vari-ability of drop size distributions [PhD thesis]McGill University2003
[4] J D Tuttle and R E Rinehart ldquoAttenuation correction in dual-wavelength analysesrdquo Journal of ClimateampAppliedMeteorologyvol 22 no 11 pp 1914ndash1921 1983
[5] S-G Park V N Bringi V Chandrasekar M Maki andK Iwanami ldquoCorrection of radar reflectivity and differentialreflectivity for rain attenuation at X band Part I theoretical andempirical basisrdquo Journal of Atmospheric and Oceanic Technol-ogy vol 22 no 11 pp 1621ndash1632 2005
[6] G Scarchilli E Gorgucci V Chandrasekar and T A SeligaldquoRainfall estimation using polarimetric techniques at C-bandfrequenciesrdquo Journal of Applied Meteorology vol 32 no 6 pp1150ndash1160 1993
[7] M N Anagnostou E N Anagnostou and J VivekanandanldquoCorrection for rain path specific and differential attenuation ofX-band dual-polarization observationsrdquo IEEE Transactions onGeoscience and Remote Sensing vol 44 no 9 pp 2470ndash24802006
20 Advances in Meteorology
[8] J Kalogiros M N Anagnostou E N Anagnostou M Mon-topoli E Picciotti and F S Marzano ldquoEvaluation of a newpolarimetric algorithm for rain-path attenuation correction ofX-band radar observations against disdrometerrdquo IEEE Transac-tions on Geoscience and Remote Sensing vol 52 no 2 pp 1369ndash1380 2014
[9] W-Y Chang J Vivekanandan and T-C C Wang ldquoEstimationof X-band polarimetric radar attenuation and measurementuncertainty using a variational methodrdquo Journal of AppliedMeteorology and Climatology vol 53 no 4 pp 1099ndash1119 2014
[10] J Hubbert and V N Bringi ldquoAn iterative filtering technique forthe analysis of copolar differential phase and dual-frequencyradar measurementsrdquo Journal of Atmospheric and OceanicTechnology vol 12 no 3 pp 643ndash648 1995
[11] K V Mishra W F Krajewski R Goska et al ldquoDeploy-ment and performance analyses of high-resolution iowa XPOLradar system during the NASA IFloodS campaignrdquo Journal ofHydrometeorology vol 17 no 2 pp 455ndash479 2016
[12] M Thurai G J Huang V N Bringi W L Randeu and MSchonhuber ldquoDrop shapes model comparisons and calcu-lations of polarimetric radar parameters in rainrdquo Journal ofAtmospheric and Oceanic Technology vol 24 no 6 pp 1019ndash1032 2007
[13] M D Humphrey J D Istok J Y Lee J A Hevesi and AL Flint ldquoA new method for automated dynamic calibration oftipping-bucket rain gaugesrdquo Journal of Atmospheric andOceanicTechnology vol 14 no 6 pp 1513ndash1519 1997
[14] E HabibW F Krajewski V Nespor andA Kruger ldquoNumericalsimulation studies of rain gage data correction due to windeffectrdquo Journal of Geophysical ResearchAtmospheres vol 104 no16 Article ID 1999JD900228 pp 19723ndash19733 1999
[15] G J Ciach and W F Krajewski ldquoOn the estimation of radarrainfall error variancerdquoAdvances inWater Resources vol 22 no6 pp 585ndash595 1999
[16] E Habib W F Krajewski and A Kruger ldquoSampling errors oftipping-bucket rain gaugemeasurementsrdquo Journal of HydrologicEngineering vol 6 no 2 pp 159ndash166 2001
[17] G J Ciach ldquoLocal random errors in tipping-bucket rain gaugemeasurementsrdquo Journal of Atmospheric andOceanic Technologyvol 20 no 5 pp 752ndash759 2003
[18] EHabibW F Krajewski andG J Ciach ldquoEstimation of rainfallinterstation correlationrdquo Journal ofHydrometeorology vol 2 no6 pp 621ndash629 2001
[19] NIMS ldquoStudy on estimation of rainfall amount and classifi-cation of hydrometeor species using dual-polarimetric radar(III)rdquo pp 1ndash20 2010
[20] E Gorgucci V Chandrasekar and L Baldini ldquoCorrection ofX-band radar observation for propagation effects based on theself-consistency principlerdquo Journal of Atmospheric and OceanicTechnology vol 23 pp 1668ndash1681 2006
[21] NIMS ldquoStudy on estimating raindrop size distribution forquantitative estimation of radar rainfall amountrdquo pp 1ndash55 2009
[22] NIMS ldquoQuantitative precipitation estimation and hydrometeoridentification using dual-polarization radar-Phase IIrdquo pp 1ndash742009
[23] S-G Park M Maki K Iwanami V N Bringi and V Chan-drasekar ldquoCorrection of radar reflectivity and differential reflec-tivity for rain attenuation at X band Part II Evaluation andapplicationrdquo Journal of Atmospheric and Oceanic Technologyvol 22 no 11 pp 1633ndash1655 2005
[24] E Habib and W F Krajewski ldquoUncertainty analysis of theTRMM ground-validation radar-rainfall products applicationto the TEFLUN-B field campaignrdquo Journal of Applied Meteorol-ogy vol 41 no 5 pp 558ndash572 2002
[25] Y-A Oh Rainfall estimation from X-band dual-polarizationradar measurements effects of attenuation correction and evalu-ation with a dense rain gauge network [MS thesis] KyungpookNational University 2011
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom
Applied ampEnvironmentalSoil Science
Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal of
Geophysics
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
18 Advances in Meteorology
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10r_
rmse
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH)
(a) 119877(119885119867)
CorrectedMean R
Uncorrected
20 40 60 80 1000Accumulation time (min)
00
02
04
06
08
10
r_rm
se
6
7
8
9
10
11
Mea
n ra
in ra
te (m
mh
)
R(ZH ZDR)
(b) 119877(119885119867119885DR)
Figure 21 The relative root-mean square error of 119877(6 km2) versus 119866(6 km2) with an accumulation time (a) using 119885119867and (b) using 119885
119867and
119885DR The red point-line and the blue point-line indicate without and with the attenuation correction respectively
attenuation correction which supports the quantitative useof the X-band radar in rain data retrieval
Appendix
A Instrumental Uncertainty of the Rain Gauge
The instrumental uncertainty of the TB rain gauge is asso-ciated with improper calibration leakage and precipitationmissing the gauge due to the measurement principle andwetting and evaporation losses In addition external sourcessuch as wind turbulence and installed position also con-tribute to the instrumental uncertainty Ciach [17] reportedthat the TB gauge data suffers a significant uncertainty asso-ciated with random differences between closely collocatedTB gauges through the analysis of the error as a functionof the rainfall intensity for three timescales In addition theimproper calibration induces a bias in the rainfall measure-ment and this bias cannot be removed by increasing accu-mulation timeTherefore the proper correction or processingof TB gauge data is necessary to obtain accurate and reliablereading of the rainfall amount
In this appendix instrumental biases were calculatedto reduce the instrumental uncertainty through an idealexperiment in the laboratory and a field intercomparisonwith 10 TB rain gauges In addition the random errors werequantitatively calculated as a function of rainfall rate andaccumulation time
A1 Correction of Instrumental Bias The instrumental bias119861119868(119894) of the 119894th gauge was defined as a combination
of the absolute bias 119861119860(ref) of the reference gauge and
the relative bias 119861119877(119894) of the 119894th gauge according to the
following equation
119861119868 (119894) = 119861119860 (ref) times 119861119877 (119894) (A1)
119861119860(ref) signified the ratio of the calculated value from
the instrumental resolution of the reference gauge to the truevalue measured by other instruments
119861119860 (ref) =
rainfalltruthrainfallgauge (ref)
(A2)
In this study 119861119860(ref) was derived by an ideal experiment
using an electronic scale First the reference gauge was puton the scale and the scale was initialized The water dropsflowed out into the receiving orifice of the reference gaugeuntil the tipping number became 60Themeasured weight of60 tips was 2144 g while the theoretical (or expected) valuewas 2244 g for the reference gauge (receiving orifice size of1545 cm 60 tips and 02mm tip resolution) Therefore theabsolute bias of the reference gauge from (A2) was about0955 This indicates that the rainfall amount measured bythis reference gauge was overestimated by about 45 andthis bias should be corrected prior to any interpretation
119861119877(119894) represented the ratio of the observed (recorded)
value from the 119894th gauge to the observed (recorded) valuefrom the reference gauge
119861119877 (119894) =
rainfallgauge (ref)rainfallgauge (119894)
(A3)
119861119877(119894) was calculated from ten collocated rain gauges
including the reference gauge located at an observation field
Advances in Meteorology 19
Table 4 The instrumental and relative biases of the 10 tipping bucket rain gauges
Gauge ID RG1 RG2 RG3 RG4 RG5 RG6 RG7 RG8 RG9 RG10Instrumental bias 094 108 099 100 096 104 098 102 108 098Relative bias 098 113 103 105 100 109 102 107 113 102
013610152230406080120150200250300400
()
Rain
rate
(mm
h)
100101
Accumulation time (min)
10
100NSD ()
Figure 22 Two-dimensional normalized standard deviationbetween the standard rainfall and each rain gauge
at KNUKoreaThe observational period for intercomparisonwas from October 2009 to June 2010 (14 rainy days) andthe total accumulated rainfall for this period was about300mm The results are presented in Table 4 All relativebiases were mainly larger than 10 This indicated that themost of gauges underestimated the rainfall amount comparedto the reference rain gauge The relative bias of the referencegauge (119894 = 5) was 10
Finally 119861119860(119894) was calculated from (A1) (see Table 4) All
instrumental biases were between 090 and 11 and the meanvalue of the biases was nearly 10 This indicated that therainfall rate from a single gauge data had a maximum 10bias These biases were corrected for each gauge to removeexisting instrumental errors
A2 Quantification of Random Errors To quantify the ran-dom error due to instrumental uncertainty the NSD of theten gauges was calculated with accumulation time and rainrate using data from an intercomparison The NSD of the 119894thgauge compared to the reference gauge was derived for 20rain rate classes (Δ119877) and 17 accumulation time classes (Δ119905)as follows
NSD119894 (Δ119905 Δ119877)
=
(1119899)radicsum119899minus1
119896=0[1198661015840
119870(Δ119905 Δ119877) minus 119866119894119896 (Δ119905 Δ119877)]
2
1198661015840
(A4)
where 1198661015840 and 119866 indicate the reference and individual gaugesrespectively The overbar means the temporal average ofthe rainfall rate The NSDs of nine different gauges (exceptthe reference gauge) were averaged to obtain the two-dimensional average NSD (Figure 22) The average NSD
tended to decrease with increasing accumulation time andwith increasing rain rate The NSD was about 10ndash15when accumulation time was 10min and the rain rate was10mmhminus1 The instrumental errors were quite significant forlonger accumulation times and higher rainfall rates Theseerrors should be used as a guideline when a single gauge isused as a reference for any validation study
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper has been revised and extended fromMs Young-AOhrsquos MS thesis [25] of Kyungpook National University Thisresearch is supported by Development and Application ofCross Governmental Dual-Pol Radar Harmonization (WRC-2013-A-1) project of theWeather Radar Center KoreaMeteo-rological AdministrationThis work was funded by the KoreaMeteorological Administration Research and DevelopmentProgram under Grant KMIPA2015-1010
References
[1] V N Bringi V Chandrasekar N Balakrishnan and D SZrnic ldquoAn examination of propagation effects in rainfall onradar measurements at microwave frequenciesrdquo Journal ofAtmospheric and Oceanic Technology vol 7 no 6 pp 829ndash8401990
[2] A Huggel W Schmid and AWaldvogel ldquoRaindrop size distri-butions and the radar bright bandrdquo Journal of AppliedMeteorol-ogy vol 35 no 10 pp 1688ndash1701 1996
[3] G W Lee Errors in rain measurement by radar effect of vari-ability of drop size distributions [PhD thesis]McGill University2003
[4] J D Tuttle and R E Rinehart ldquoAttenuation correction in dual-wavelength analysesrdquo Journal of ClimateampAppliedMeteorologyvol 22 no 11 pp 1914ndash1921 1983
[5] S-G Park V N Bringi V Chandrasekar M Maki andK Iwanami ldquoCorrection of radar reflectivity and differentialreflectivity for rain attenuation at X band Part I theoretical andempirical basisrdquo Journal of Atmospheric and Oceanic Technol-ogy vol 22 no 11 pp 1621ndash1632 2005
[6] G Scarchilli E Gorgucci V Chandrasekar and T A SeligaldquoRainfall estimation using polarimetric techniques at C-bandfrequenciesrdquo Journal of Applied Meteorology vol 32 no 6 pp1150ndash1160 1993
[7] M N Anagnostou E N Anagnostou and J VivekanandanldquoCorrection for rain path specific and differential attenuation ofX-band dual-polarization observationsrdquo IEEE Transactions onGeoscience and Remote Sensing vol 44 no 9 pp 2470ndash24802006
20 Advances in Meteorology
[8] J Kalogiros M N Anagnostou E N Anagnostou M Mon-topoli E Picciotti and F S Marzano ldquoEvaluation of a newpolarimetric algorithm for rain-path attenuation correction ofX-band radar observations against disdrometerrdquo IEEE Transac-tions on Geoscience and Remote Sensing vol 52 no 2 pp 1369ndash1380 2014
[9] W-Y Chang J Vivekanandan and T-C C Wang ldquoEstimationof X-band polarimetric radar attenuation and measurementuncertainty using a variational methodrdquo Journal of AppliedMeteorology and Climatology vol 53 no 4 pp 1099ndash1119 2014
[10] J Hubbert and V N Bringi ldquoAn iterative filtering technique forthe analysis of copolar differential phase and dual-frequencyradar measurementsrdquo Journal of Atmospheric and OceanicTechnology vol 12 no 3 pp 643ndash648 1995
[11] K V Mishra W F Krajewski R Goska et al ldquoDeploy-ment and performance analyses of high-resolution iowa XPOLradar system during the NASA IFloodS campaignrdquo Journal ofHydrometeorology vol 17 no 2 pp 455ndash479 2016
[12] M Thurai G J Huang V N Bringi W L Randeu and MSchonhuber ldquoDrop shapes model comparisons and calcu-lations of polarimetric radar parameters in rainrdquo Journal ofAtmospheric and Oceanic Technology vol 24 no 6 pp 1019ndash1032 2007
[13] M D Humphrey J D Istok J Y Lee J A Hevesi and AL Flint ldquoA new method for automated dynamic calibration oftipping-bucket rain gaugesrdquo Journal of Atmospheric andOceanicTechnology vol 14 no 6 pp 1513ndash1519 1997
[14] E HabibW F Krajewski V Nespor andA Kruger ldquoNumericalsimulation studies of rain gage data correction due to windeffectrdquo Journal of Geophysical ResearchAtmospheres vol 104 no16 Article ID 1999JD900228 pp 19723ndash19733 1999
[15] G J Ciach and W F Krajewski ldquoOn the estimation of radarrainfall error variancerdquoAdvances inWater Resources vol 22 no6 pp 585ndash595 1999
[16] E Habib W F Krajewski and A Kruger ldquoSampling errors oftipping-bucket rain gaugemeasurementsrdquo Journal of HydrologicEngineering vol 6 no 2 pp 159ndash166 2001
[17] G J Ciach ldquoLocal random errors in tipping-bucket rain gaugemeasurementsrdquo Journal of Atmospheric andOceanic Technologyvol 20 no 5 pp 752ndash759 2003
[18] EHabibW F Krajewski andG J Ciach ldquoEstimation of rainfallinterstation correlationrdquo Journal ofHydrometeorology vol 2 no6 pp 621ndash629 2001
[19] NIMS ldquoStudy on estimation of rainfall amount and classifi-cation of hydrometeor species using dual-polarimetric radar(III)rdquo pp 1ndash20 2010
[20] E Gorgucci V Chandrasekar and L Baldini ldquoCorrection ofX-band radar observation for propagation effects based on theself-consistency principlerdquo Journal of Atmospheric and OceanicTechnology vol 23 pp 1668ndash1681 2006
[21] NIMS ldquoStudy on estimating raindrop size distribution forquantitative estimation of radar rainfall amountrdquo pp 1ndash55 2009
[22] NIMS ldquoQuantitative precipitation estimation and hydrometeoridentification using dual-polarization radar-Phase IIrdquo pp 1ndash742009
[23] S-G Park M Maki K Iwanami V N Bringi and V Chan-drasekar ldquoCorrection of radar reflectivity and differential reflec-tivity for rain attenuation at X band Part II Evaluation andapplicationrdquo Journal of Atmospheric and Oceanic Technologyvol 22 no 11 pp 1633ndash1655 2005
[24] E Habib and W F Krajewski ldquoUncertainty analysis of theTRMM ground-validation radar-rainfall products applicationto the TEFLUN-B field campaignrdquo Journal of Applied Meteorol-ogy vol 41 no 5 pp 558ndash572 2002
[25] Y-A Oh Rainfall estimation from X-band dual-polarizationradar measurements effects of attenuation correction and evalu-ation with a dense rain gauge network [MS thesis] KyungpookNational University 2011
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom
Applied ampEnvironmentalSoil Science
Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal of
Geophysics
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
Advances in Meteorology 19
Table 4 The instrumental and relative biases of the 10 tipping bucket rain gauges
Gauge ID RG1 RG2 RG3 RG4 RG5 RG6 RG7 RG8 RG9 RG10Instrumental bias 094 108 099 100 096 104 098 102 108 098Relative bias 098 113 103 105 100 109 102 107 113 102
013610152230406080120150200250300400
()
Rain
rate
(mm
h)
100101
Accumulation time (min)
10
100NSD ()
Figure 22 Two-dimensional normalized standard deviationbetween the standard rainfall and each rain gauge
at KNUKoreaThe observational period for intercomparisonwas from October 2009 to June 2010 (14 rainy days) andthe total accumulated rainfall for this period was about300mm The results are presented in Table 4 All relativebiases were mainly larger than 10 This indicated that themost of gauges underestimated the rainfall amount comparedto the reference rain gauge The relative bias of the referencegauge (119894 = 5) was 10
Finally 119861119860(119894) was calculated from (A1) (see Table 4) All
instrumental biases were between 090 and 11 and the meanvalue of the biases was nearly 10 This indicated that therainfall rate from a single gauge data had a maximum 10bias These biases were corrected for each gauge to removeexisting instrumental errors
A2 Quantification of Random Errors To quantify the ran-dom error due to instrumental uncertainty the NSD of theten gauges was calculated with accumulation time and rainrate using data from an intercomparison The NSD of the 119894thgauge compared to the reference gauge was derived for 20rain rate classes (Δ119877) and 17 accumulation time classes (Δ119905)as follows
NSD119894 (Δ119905 Δ119877)
=
(1119899)radicsum119899minus1
119896=0[1198661015840
119870(Δ119905 Δ119877) minus 119866119894119896 (Δ119905 Δ119877)]
2
1198661015840
(A4)
where 1198661015840 and 119866 indicate the reference and individual gaugesrespectively The overbar means the temporal average ofthe rainfall rate The NSDs of nine different gauges (exceptthe reference gauge) were averaged to obtain the two-dimensional average NSD (Figure 22) The average NSD
tended to decrease with increasing accumulation time andwith increasing rain rate The NSD was about 10ndash15when accumulation time was 10min and the rain rate was10mmhminus1 The instrumental errors were quite significant forlonger accumulation times and higher rainfall rates Theseerrors should be used as a guideline when a single gauge isused as a reference for any validation study
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper has been revised and extended fromMs Young-AOhrsquos MS thesis [25] of Kyungpook National University Thisresearch is supported by Development and Application ofCross Governmental Dual-Pol Radar Harmonization (WRC-2013-A-1) project of theWeather Radar Center KoreaMeteo-rological AdministrationThis work was funded by the KoreaMeteorological Administration Research and DevelopmentProgram under Grant KMIPA2015-1010
References
[1] V N Bringi V Chandrasekar N Balakrishnan and D SZrnic ldquoAn examination of propagation effects in rainfall onradar measurements at microwave frequenciesrdquo Journal ofAtmospheric and Oceanic Technology vol 7 no 6 pp 829ndash8401990
[2] A Huggel W Schmid and AWaldvogel ldquoRaindrop size distri-butions and the radar bright bandrdquo Journal of AppliedMeteorol-ogy vol 35 no 10 pp 1688ndash1701 1996
[3] G W Lee Errors in rain measurement by radar effect of vari-ability of drop size distributions [PhD thesis]McGill University2003
[4] J D Tuttle and R E Rinehart ldquoAttenuation correction in dual-wavelength analysesrdquo Journal of ClimateampAppliedMeteorologyvol 22 no 11 pp 1914ndash1921 1983
[5] S-G Park V N Bringi V Chandrasekar M Maki andK Iwanami ldquoCorrection of radar reflectivity and differentialreflectivity for rain attenuation at X band Part I theoretical andempirical basisrdquo Journal of Atmospheric and Oceanic Technol-ogy vol 22 no 11 pp 1621ndash1632 2005
[6] G Scarchilli E Gorgucci V Chandrasekar and T A SeligaldquoRainfall estimation using polarimetric techniques at C-bandfrequenciesrdquo Journal of Applied Meteorology vol 32 no 6 pp1150ndash1160 1993
[7] M N Anagnostou E N Anagnostou and J VivekanandanldquoCorrection for rain path specific and differential attenuation ofX-band dual-polarization observationsrdquo IEEE Transactions onGeoscience and Remote Sensing vol 44 no 9 pp 2470ndash24802006
20 Advances in Meteorology
[8] J Kalogiros M N Anagnostou E N Anagnostou M Mon-topoli E Picciotti and F S Marzano ldquoEvaluation of a newpolarimetric algorithm for rain-path attenuation correction ofX-band radar observations against disdrometerrdquo IEEE Transac-tions on Geoscience and Remote Sensing vol 52 no 2 pp 1369ndash1380 2014
[9] W-Y Chang J Vivekanandan and T-C C Wang ldquoEstimationof X-band polarimetric radar attenuation and measurementuncertainty using a variational methodrdquo Journal of AppliedMeteorology and Climatology vol 53 no 4 pp 1099ndash1119 2014
[10] J Hubbert and V N Bringi ldquoAn iterative filtering technique forthe analysis of copolar differential phase and dual-frequencyradar measurementsrdquo Journal of Atmospheric and OceanicTechnology vol 12 no 3 pp 643ndash648 1995
[11] K V Mishra W F Krajewski R Goska et al ldquoDeploy-ment and performance analyses of high-resolution iowa XPOLradar system during the NASA IFloodS campaignrdquo Journal ofHydrometeorology vol 17 no 2 pp 455ndash479 2016
[12] M Thurai G J Huang V N Bringi W L Randeu and MSchonhuber ldquoDrop shapes model comparisons and calcu-lations of polarimetric radar parameters in rainrdquo Journal ofAtmospheric and Oceanic Technology vol 24 no 6 pp 1019ndash1032 2007
[13] M D Humphrey J D Istok J Y Lee J A Hevesi and AL Flint ldquoA new method for automated dynamic calibration oftipping-bucket rain gaugesrdquo Journal of Atmospheric andOceanicTechnology vol 14 no 6 pp 1513ndash1519 1997
[14] E HabibW F Krajewski V Nespor andA Kruger ldquoNumericalsimulation studies of rain gage data correction due to windeffectrdquo Journal of Geophysical ResearchAtmospheres vol 104 no16 Article ID 1999JD900228 pp 19723ndash19733 1999
[15] G J Ciach and W F Krajewski ldquoOn the estimation of radarrainfall error variancerdquoAdvances inWater Resources vol 22 no6 pp 585ndash595 1999
[16] E Habib W F Krajewski and A Kruger ldquoSampling errors oftipping-bucket rain gaugemeasurementsrdquo Journal of HydrologicEngineering vol 6 no 2 pp 159ndash166 2001
[17] G J Ciach ldquoLocal random errors in tipping-bucket rain gaugemeasurementsrdquo Journal of Atmospheric andOceanic Technologyvol 20 no 5 pp 752ndash759 2003
[18] EHabibW F Krajewski andG J Ciach ldquoEstimation of rainfallinterstation correlationrdquo Journal ofHydrometeorology vol 2 no6 pp 621ndash629 2001
[19] NIMS ldquoStudy on estimation of rainfall amount and classifi-cation of hydrometeor species using dual-polarimetric radar(III)rdquo pp 1ndash20 2010
[20] E Gorgucci V Chandrasekar and L Baldini ldquoCorrection ofX-band radar observation for propagation effects based on theself-consistency principlerdquo Journal of Atmospheric and OceanicTechnology vol 23 pp 1668ndash1681 2006
[21] NIMS ldquoStudy on estimating raindrop size distribution forquantitative estimation of radar rainfall amountrdquo pp 1ndash55 2009
[22] NIMS ldquoQuantitative precipitation estimation and hydrometeoridentification using dual-polarization radar-Phase IIrdquo pp 1ndash742009
[23] S-G Park M Maki K Iwanami V N Bringi and V Chan-drasekar ldquoCorrection of radar reflectivity and differential reflec-tivity for rain attenuation at X band Part II Evaluation andapplicationrdquo Journal of Atmospheric and Oceanic Technologyvol 22 no 11 pp 1633ndash1655 2005
[24] E Habib and W F Krajewski ldquoUncertainty analysis of theTRMM ground-validation radar-rainfall products applicationto the TEFLUN-B field campaignrdquo Journal of Applied Meteorol-ogy vol 41 no 5 pp 558ndash572 2002
[25] Y-A Oh Rainfall estimation from X-band dual-polarizationradar measurements effects of attenuation correction and evalu-ation with a dense rain gauge network [MS thesis] KyungpookNational University 2011
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom
Applied ampEnvironmentalSoil Science
Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal of
Geophysics
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
20 Advances in Meteorology
[8] J Kalogiros M N Anagnostou E N Anagnostou M Mon-topoli E Picciotti and F S Marzano ldquoEvaluation of a newpolarimetric algorithm for rain-path attenuation correction ofX-band radar observations against disdrometerrdquo IEEE Transac-tions on Geoscience and Remote Sensing vol 52 no 2 pp 1369ndash1380 2014
[9] W-Y Chang J Vivekanandan and T-C C Wang ldquoEstimationof X-band polarimetric radar attenuation and measurementuncertainty using a variational methodrdquo Journal of AppliedMeteorology and Climatology vol 53 no 4 pp 1099ndash1119 2014
[10] J Hubbert and V N Bringi ldquoAn iterative filtering technique forthe analysis of copolar differential phase and dual-frequencyradar measurementsrdquo Journal of Atmospheric and OceanicTechnology vol 12 no 3 pp 643ndash648 1995
[11] K V Mishra W F Krajewski R Goska et al ldquoDeploy-ment and performance analyses of high-resolution iowa XPOLradar system during the NASA IFloodS campaignrdquo Journal ofHydrometeorology vol 17 no 2 pp 455ndash479 2016
[12] M Thurai G J Huang V N Bringi W L Randeu and MSchonhuber ldquoDrop shapes model comparisons and calcu-lations of polarimetric radar parameters in rainrdquo Journal ofAtmospheric and Oceanic Technology vol 24 no 6 pp 1019ndash1032 2007
[13] M D Humphrey J D Istok J Y Lee J A Hevesi and AL Flint ldquoA new method for automated dynamic calibration oftipping-bucket rain gaugesrdquo Journal of Atmospheric andOceanicTechnology vol 14 no 6 pp 1513ndash1519 1997
[14] E HabibW F Krajewski V Nespor andA Kruger ldquoNumericalsimulation studies of rain gage data correction due to windeffectrdquo Journal of Geophysical ResearchAtmospheres vol 104 no16 Article ID 1999JD900228 pp 19723ndash19733 1999
[15] G J Ciach and W F Krajewski ldquoOn the estimation of radarrainfall error variancerdquoAdvances inWater Resources vol 22 no6 pp 585ndash595 1999
[16] E Habib W F Krajewski and A Kruger ldquoSampling errors oftipping-bucket rain gaugemeasurementsrdquo Journal of HydrologicEngineering vol 6 no 2 pp 159ndash166 2001
[17] G J Ciach ldquoLocal random errors in tipping-bucket rain gaugemeasurementsrdquo Journal of Atmospheric andOceanic Technologyvol 20 no 5 pp 752ndash759 2003
[18] EHabibW F Krajewski andG J Ciach ldquoEstimation of rainfallinterstation correlationrdquo Journal ofHydrometeorology vol 2 no6 pp 621ndash629 2001
[19] NIMS ldquoStudy on estimation of rainfall amount and classifi-cation of hydrometeor species using dual-polarimetric radar(III)rdquo pp 1ndash20 2010
[20] E Gorgucci V Chandrasekar and L Baldini ldquoCorrection ofX-band radar observation for propagation effects based on theself-consistency principlerdquo Journal of Atmospheric and OceanicTechnology vol 23 pp 1668ndash1681 2006
[21] NIMS ldquoStudy on estimating raindrop size distribution forquantitative estimation of radar rainfall amountrdquo pp 1ndash55 2009
[22] NIMS ldquoQuantitative precipitation estimation and hydrometeoridentification using dual-polarization radar-Phase IIrdquo pp 1ndash742009
[23] S-G Park M Maki K Iwanami V N Bringi and V Chan-drasekar ldquoCorrection of radar reflectivity and differential reflec-tivity for rain attenuation at X band Part II Evaluation andapplicationrdquo Journal of Atmospheric and Oceanic Technologyvol 22 no 11 pp 1633ndash1655 2005
[24] E Habib and W F Krajewski ldquoUncertainty analysis of theTRMM ground-validation radar-rainfall products applicationto the TEFLUN-B field campaignrdquo Journal of Applied Meteorol-ogy vol 41 no 5 pp 558ndash572 2002
[25] Y-A Oh Rainfall estimation from X-band dual-polarizationradar measurements effects of attenuation correction and evalu-ation with a dense rain gauge network [MS thesis] KyungpookNational University 2011
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom
Applied ampEnvironmentalSoil Science
Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal of
Geophysics
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom
Applied ampEnvironmentalSoil Science
Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
International Journal of
Geophysics
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in