atmospheric radiation measurement site atmospheric state best

Atmospheric Radiation Measurement site atmospheric state best

estimates for Atmospheric Infrared Sounder temperature and water

vapor retrieval validation

David C. Tobin,1 Henry E. Revercomb,1 Robert O. Knuteson,1 Barry M. Lesht,2

L. Larrabee Strow,3 Scott E. Hannon,3 Wayne F. Feltz,1 Leslie A. Moy,1 Eric J. Fetzer,4

and Ted S. Cress5

Received 20 April 2005; revised 14 June 2005; accepted 19 July 2005; published 16 March 2006.

[1] The Atmospheric Infrared Sounder (AIRS) is the first of a new generation of advancedsatellite-based atmospheric sounders with the capability of obtaining high–verticalresolution profiles of temperature and water vapor. The high-accuracy retrieval goals ofAIRS (e.g., 1 K RMS in 1 km layers below 100 mbar for air temperature, 10% RMS in 2 kmlayers below 100mbar for water vapor concentration), combinedwith the large temporal andspatial variability of the atmosphere and difficulties in making accurate measurements of theatmospheric state, necessitate careful and detailed validation using well-characterizedground-based sites. As part of ongoing AIRS Science Team efforts and a collaborative effortbetween the NASA Earth Observing System (EOS) project and the Department of EnergyAtmospheric Radiation Measurement (ARM) program, data from various ARM and otherobservations are used to create best estimates of the atmospheric state at the Aqua overpasstimes. The resulting validation data set is an ensemble of temperature and water vaporprofiles created from radiosondes launched at the approximate Aqua overpass times,interpolated to the exact overpass time using time continuous ground-based profiles,adjusted to account for spatial gradients within the Advanced Microwave Sounding Unit(AMSU) footprints, and supplemented with limited cloud observations. Estimates of thespectral surface infrared emissivity and local skin temperatures are also constructed. Relyingon the developed ARM infrastructure and previous and ongoing characterization studies ofthe ARMmeasurements, the data set provides a good combination of statistics and accuracywhich is essential for assessment of the advanced sounder products. Combined with thecollocated AIRS observations, the products are being used to study observed minuscalculated AIRS spectra, aimed at evaluation of the AIRS forward radiative transfer model,AIRS observed radiances, and temperature and water vapor profile retrievals. This paperprovides an introduction to the ARM site best estimate validation products and characterizesthe accuracy of the AIRS team version 4 atmospheric temperature and water vapor retrievalsusing the ARM products. The AIRS retrievals over tropical ocean are found to have verygood accuracy for both temperature and water vapor, with RMS errors approaching thetheoretical expectation for clear sky conditions, while retrievals over a midlatitude land sitehave poorer performance. The results demonstrate the importance of using specialized‘‘truth’’ sites for accurate assessment of the advanced sounder performance and motivate thecontinued refinement of theAIRS science team retrieval algorithm, particularly for retrievalsover land.

Citation: Tobin, D. C., H. E. Revercomb, R. O. Knuteson, B. M. Lesht, L. L. Strow, S. E. Hannon, W. F. Feltz, L. A. Moy, E. J.

Fetzer, and T. S. Cress (2006), Atmospheric Radiation Measurement site atmospheric state best estimates for Atmospheric Infrared

Sounder temperature and water vapor retrieval validation, J. Geophys. Res., 111, D09S14, doi:10.1029/2005JD006103.

1. Introduction

[2] Temperature and water vapor soundings from satel-lites have been available for more than 35 years, beginningwith the launch of two pioneering sensors on the Nimbus 3satellite in April 1969 [e.g., Wick, 1971]. By providingtwice daily atmospheric profiles on a global scale, thesatellite soundings were expected to provide significantimprovements in numerical weather prediction. Because

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 111, D09S14, doi:10.1029/2005JD006103, 2006

1Cooperative Institute for Meteorological Satellite Studies, SpaceScience and Engineering Center, University of Wisconsin, Madison,Wisconsin, USA.

2Argonne National Laboratory, Argonne, Illinois, USA.3Department of Physics, University of Maryland Baltimore County,

Baltimore, Maryland, USA.4NASA Jet Propulsion Laboratory, Pasadena, California, USA.5Pacific Northwest National Laboratory, Richland, Washington, USA.

Copyright 2006 by the American Geophysical Union.0148-0227/06/2005JD006103$09.00

D09S14 1 of 18

of improved data assimilation methods, use of the satellitedata for model initialization, and expanded computerresources, many significant improvements in forecast accu-racy have indeed been realized. However, the relatively lowvertical resolution provided by these broadband sensors,such as the Advanced TIROS Operational Vertical Sounder(ATOVS) on NOAA’s current Polar Operational Environ-mental Satellite, has limited the benefit of these satelliteobservations for improved weather forecasting. In manycircumstances, for example, the low vertical resolution isresponsible for large absolute errors as well as reducedhorizontal variance, and strong horizontal error correlationof the satellite soundings [e.g., Phillips, 1980]. In order toprovide further advances in numerical weather predictionand climate studies, the requirement for sensors and algo-rithms with improved vertical resolution and with an abilityto handle partially cloudy scenes was recognized. Begin-ning in the mid-1970s, extensive retrieval algorithm devel-opment, instrument design, and experimental field studieswere then conducted to determine the feasibility andrequirements of a future advanced sounder; the reader isreferred to Smith [1991] for an overview of the evolution ofthe satellite sounder to achieve high vertical resolution andimprove weather prediction. As a result of these efforts,specific requirements for the NOAA/NASA ‘‘InteragencySounder’’ were established in the late 1980s, providingmotivation for the Atmospheric Infrared Sounder (AIRS)as part of NASA’s Earth Observing System (EOS).[3] Launched into orbit on 4 May 2002 on the EOS Aqua

platform, AIRS is the first of a new generation of satellite-based advanced infrared sounders, designed to provide datawith higher vertical resolution and accuracy to numericalweather prediction centers for improved medium rangeweather forecasting. Aumann et al. [2003] present a thor-ough overview of the science objectives, data products,retrieval algorithms, and ground data processing concepts ofAIRS. Exploiting the higher vertical resolution made pos-sible by high spectral resolution and the combined use ofAdvanced Microwave Sounding Unit (AMSU) microwaveand infrared radiances in the presence of clouds, AIRS isexpected to produce retrievals of temperature and watervapor with high accuracy under clear and partly cloudyconditions (for scenes with cloud fraction up to 80%). TheAIRS retrieval performance requirements are stated in termsof RMS errors for vertically averaged layers in the tropo-sphere. The required performance is 1 K RMS in 1 kmvertical layers below 100 mbar for temperature profileretrievals and 20% (10% goal) RMS in 2 km vertical layersbelow 100 mbar for water vapor profile retrievals. There areno documented requirements for the yield or mean error(bias) of the retrieved profiles.[4] Detailed characterization of the accuracy of the satel-

lite products is required for many applications of the data.The high-accuracy retrieval goals of the advanced sounders,combined with the temporal and spatial variability of theatmosphere and difficulties in making accurate measure-ments of the atmospheric state, necessitates careful valida-tion using well-characterized ground-based sites. Starting inthe early 1990s the Department of Energy AtmosphericRadiation Measurement (ARM) program has developedseveral ground sites around the world to collect detailedmeasurements of the atmospheric state and coincident

radiation with an overall goal of improving the representa-tion of radiation and clouds in climate models [Stokes andSchwartz, 1994]. The ARM Climate Research Facilityground sites have matured and are now well suited forproviding data sets that are both accurately characterizedand statistically significant for performing satellite valida-tion [Ackerman and Stokes, 2003]. As part of an ongoingAIRS Science Team effort initiated under team member Dr.H. E. Revercomb and continuing under a collaborativeeffort between the NASA EOS project and the ARMprogram, data from the heavily instrumented ARM sitesare used to create ‘‘best estimates’’ of the atmospheric stateand surface properties at the Aqua overpass times. Com-bined with the colocated AIRS observations, the ARMproducts are being used for various studies, includingassessment of clear sky observed minus calculated radiancespectra, aimed at development and validation of the AIRSclear sky forward radiative transfer model. Validation of theAIRS clear sky radiative transfer algorithm using the ARMbest estimate and other data is presented by Strow et al.[2006].[5] The primary purpose of this paper is to provide a

description of ARM site best estimate validation products(section 2) and, using the ARM products, to provide anobjective and accurate characterization of the AIRS Teamversion 4 atmospheric temperature and water vapor retriev-als (section 3). Section 4 summarizes the results of thisstudy and plans for further development and applications ofthe ARM validation data.

2. ARM Site Atmospheric State and SurfaceBest Estimate Products

[6] Best estimates of vertical profiles of atmosphericpressure, air temperature, water vapor concentration, andsurface parameters are produced for Aqua overpasses of theARM sites using ARM and other data. This section includesan introduction to the ARM sites included in this effort, adescription of what data are used and how the best estimateproducts are created, and a discussion of the accuracy andstatistical significance of the products for AIRS temperatureand water vapor profile retrieval validation.[7] Three ARM sites, shown in Figure 1, are included in

this validation effort. These include the Tropical WesternPacific (TWP) site on the small equatorial island of Nauru,the Southern Great Plains (SGP) central facility site in northcentral Oklahoma, and the North Slope of Alaska (NSA)site near Barrow, Alaska. These sites were chosen by ARMto represent three of Earth’s primary climate regimes.Covering a range of conditions with varying degrees ofretrieval difficulty, the sites are also well suited for charac-terizing the AIRS retrieval performance. The TWP Naurusite has been very instrumental in AIRS validation becauseit is an ocean site with well-known surface emissivity, highwater vapor amounts, and low variability in the atmospherictemperature and water vapor. This provides a valuable dataset for validation of various components of the AIRSmolecular absorption and radiative transfer models andprovides the most straightforward retrieval conditions. Op-erational since 1991, the ARM SGP central facility site isone of the most heavily instrumented ground-based sites inthe world for atmospheric state and radiation measurements.

D09S14 TOBIN ET AL.: AIRS RETRIEVAL VALIDATION

2 of 18

D09S14

Numerous detailed experiments have been conducted at theSGP site in order to characterize and improve the accuracyof ARM’s temperature and water vapor profiling capabili-ties. The site is located within a large agricultural region andexperiences wide fluctuations on many timescales in theatmospheric temperature and water vapor, surface condi-tions, and cloud conditions. Having to deal with theseconditions, most notably the complex and unknown surfaceemission, satellite retrievals are more difficult at the SGPsite. The NSA site provides data of clouds, atmosphericstate, and radiative processes for cold and dry conditions. Itis located near the coastal town of Barrow on the ArcticOcean, with surrounding areas including open ocean, seaice, coastal inland tundra, and fresh water ecosystems.Passive remote sounding of arctic atmospheres is histori-cally very difficult because of low contrast of clouds and thevariable surface, and the NSA site serves as one of the mostdifficult sites for AIRS retrievals.[8] Each ARM site is instrumented with numerous

ground-based sensors to measure wind speed and direction,water vapor and liquid water, temperature, cloud properties,aerosols, and radiation. For AIRS temperature and watervapor retrieval validation, the primary profile measurementsare made with Vaisala RS-90 radiosondes [Paukkunen et al.,2001]. Depending on the particular site, radiosondes aretypically launched two to four times per day at the ARMsites. These normally scheduled launches were supple-mented to provide an optimal data set for AIRS validation.Three ‘‘phases’’ of dedicated radiosonde launches havebeen completed at each of the three ARM sites as ofOctober 2004. During these periods, two radiosondes arelaunched for Aqua overpasses of the ARM sites. The first

radiosonde is launched 45 to 90 min prior to the overpasstime and the second is launched just before (�5 min) theoverpass time, providing radiosonde measurements of theupper and lower troposphere at the AIRS overpass time. Atthe SGP and NSA sites, this is accomplished using tworadiosonde receivers, allowing two radiosondes to betracked simultaneously. At the TWP site, a single receiveris used and the first sounding is terminated prior to initiatingthe second. Each dedicated radiosonde launch phaseincludes approximately 90 overpasses (180 radiosondes)spanning several months and including both day andnighttime overpasses. Beginning and ending dates of eachphase are listed in Table 1. On a given day, the dedicatedlaunches are performed for the overpass with the smallestlocal zenith viewing angle. The launches are performed forall conditions except for totally overcast or precipitatingconditions (conditions also for which full infrared plusmicrowave AIRS retrievals are not performed) or for localzenith angles greater than 30 degrees. Additional dedicatedlaunch periods are scheduled for later in 2005.[9] Each Vaisala radiosonde water vapor profile is mul-

tiplied by a height-independent scale factor such that thetotal column precipitable water vapor (pwv) of the radio-sonde profile is equal to that measured simultaneously by aground-based two-channel microwave radiometer (MWR).This ‘‘microwave scaling’’ technique for the radiosondewater vapor profiles has several important advantages[Revercomb et al., 2003, and references therein] includingthe removal of significant radiosonde batch-to-batch cali-bration variations and diurnal biases in the radiosonde totalcolumn water vapor, and providing an absolute calibrationreference. The Goff Gratch formulation of vapor pressure

Figure 1. Locations of the ARM sites. Southern Great Plains (SGP) site in north central Oklahoma(97.5�W, 36.6�N), North Slope of Alaska (NSA) site in Barrow Alaska (156.6�W, 71.3�N), and TropicalWestern Pacific (TWP) site on the small equatorial island of Nauru (166.9�E, 0.5�S).

Table 1. Beginning and Ending Dates of Aqua Overpass Dedicated Radiosonde Launch Phases

Site Phase 1 Phase 2 Phase 3

SGP 2 September 2002 to 25 May 2003 9 September 2003 to 7 February 2004 2 April 2004 to 10 August 2004TWP 15 September 2002 to 29 May 2003 8 September 2003 to 20 March 2004 2 April 2004 to 29 September 2004NSA 1 September 2002 to 27 December 2002 10 September 2003 to 8 February 2004 23 April 2004 to 5 September 2004


3 of 18

D09S14

over liquid water [Smithsonian Institution, 1984] is used tocompute mass mixing ratio profiles from the radiosondetemperature and relative humidity measurements.[10] To account for temporal differences between the

AIRS overpass and the radiosonde measurements, timecontinuous profiles of temperature and water vapor at theARM site are used to interpolate the radiosonde profiles tothe overpass time. At the SGP site, AERI-plus [Feltz et al.,2003] profiles of temperature and water vapor are used forthis. The AERI-plus retrievals are derived from ground-based downwelling spectral radiance measurements by theAtmospheric Emitted Radiance Interferometer (AERI) andfrom coincident model analysis fields, providing verticaltemperature and water vapor profiles every 6 to 8 min withhigh vertical resolution in the boundary layer, and lower–vertical resolution profiles extending to �2 mbar. TheAERI-plus fields are obtained at the (altitude-dependent)radiosonde observation times and at the satellite overpasstime, and the relative changes in the fields are used tointerpolate the radiosonde profile to the overpass time. Thisis performed by adding AERI-plus temperature differencesto the radiosonde profile and multiplying the radiosondewater vapor profile by AERI-plus water vapor amountratios. This is done independently for each altitude bin ofthe AERI-plus fields. With this approach, the drift of theradiosonde position with height is not taken into consider-ation. The time interpolation is performed for the two

radiosondes with launch times bounding the overpass time,and the two profiles are then combined in a weighted mean,with the weights determined by the time difference and bythe magnitude of change in the AERI-plus fields betweenthe observation and overpass times. Only radiosondes thatreach a minimum air pressure of 200 mbar are used. AERI-plus time cross sections and radiosonde trajectories for anexample nighttime SGP site overpass on 25 July 2002 areshown in Figure 2. If the AERI-plus fields are not availablefor a given case, fields from the Rapid Update Cycle (RUC-2 [Benjamin et al., 2004]) are used. In the event that neitherAERI-plus nor RUC-2 fields are available, simple linearinterpolation between the two bounding radiosonde profilesis used to compute the profile at the overpass time. Note thatat the SGP site, continuous water vapor retrievals areavailable from a Raman Lidar. This sensor, however, wasnot fully operational during the past two years [Turner andGoldsmith, 2005] and it has therefore not been included inthe production of the best estimate profiles to date.[11] Figure 3 displays estimates of the short-term vari-

ability of the temperature and water vapor profiles at theSGP site at the Aqua overpass times, utilizing differencesbetween the pairs of microwave scaled radiosondeslaunched �60 min apart during the three Aqua overpassdedicated launch phases. Differences for individual high–vertical resolution profiles are shown. Analogous to how theAIRS retrieval validation statistics are computed (as de-

Figure 2. Time series of various SGP ARM data on 25 July 2002. (top) Cross section of AERI+temperature retrievals (with color scale and left-hand altitude scale) and time series of the near-surface airtemperature measured with a surface met station (dotted shaded curve) and the surface temperaturemeasured with a down-looking broadband IRT (dashed black curve), both with the right-hand y axisscale. (bottom) Cross section of AERI+ relative humidity retrievals (with color scale and left-handaltitude scale) and time series of total column precipitable water vapor measured by the MWR (dashedblack curve with right-hand y axis scale). In both panels the Aqua overpass time (black vertical dashedline) and radiosonde trajectories (four solid black curves) are overlaid.


4 of 18

D09S14

scribed in more detail in section 3), the mean differences(biases) and RMS differences for 1 km (temperature) and2 km (water vapor) vertical layering are also shown. Notethat in this process, the individual profiles are first reducedto lower vertical resolution and then the bias and RMSstatistics are computed; this has the dramatic effect ofsuppressing many of the largest errors associated withvertically fine-scale structure present in the high–verticalresolution profiles. Throughout the troposphere for both dayand night cases, the temperature RMS differences rangefrom 0.5 to 1.0 K, and the water vapor RMS percentdifferences range from 10 to 25%, as shown in Figure 3.Variability at the TWP site is less, with temperature RMSdifferences ranging from 0.3 to 0.7 K, and water vapor RMSpercent differences ranging from 7 to 15%. Considering theAIRS retrieval goals, this temporal variability is significant.This emphasizes the importance of minimizing the temporal

mismatches between the ARM and AIRSmeasurement timeswhen attempting to accurately characterize the AIRS retriev-als, and the need for the temporal interpolation approachdescribed previously. Note that the mean temperature differ-ence for the lowest altitude bin for daytime overpasses isnonzero, which can be attributed to the warming typicallyexperienced at the SGP site in the early afternoon.[12] The AIRS retrievals are performed using a combi-

nation of a 3 by 3 array of AIRS footprints and one AMSUfootprint and have footprint diameters ranging from 50 kmat nadir to �150 km at the edge of scan. Spatial gradientscan be significant on this scale, and the local ARM siteprofiles are therefore adjusted to account for large-scalespatial gradients and the location of the ARM site within theAMSU footprint. At the SGP site, this is accomplishedusing 4 km spatial resolution Geostationary EnvironmentalOperation Satellite (GOES) retrievals [Ma et al., 1999]. The

Figure 3. Short-term temporal variability at the SGP site. Differences between pairs of dedicatedradiosonde profiles are shown, one launched �60 min before the Aqua overpass and one launched at theoverpass time. (top) Daytime overpasses and (bottom) nighttime overpasses; (left) temperaturedifferences and (right) percent difference in water vapor amounts. In each panel the light shaded curvesare differences for individual profiles, and the dashed and solid curves are the mean and RMS differences,respectively, for 1 km (temperature) and 2 km (water vapor) thick layers. MWR total column water vaporscaling has been applied to each radiosonde water vapor profile as discussed in the text.


5 of 18

D09S14

hourly GOES fields are linearly interpolated to the overpasstime. As with the temporal interpolation, the spatial adjust-ment is performed in a relative sense, where relative differ-ences in the GOES fields are computed between the GOESfootprint closest to the ARM site and the mean of all GOESfootprints within the AMSU footprint. The profile adjust-

ments are performed using temperature differences andwater vapor amount ratios and are done independently foreach altitude bin of the GOES retrievals. Sample spatialgradients for the 25 July 2002 SGP overpass are shown inFigure 4. Analogous to Figure 3 in regard to short-termtemporal variability, Figure 5 displays the differences be-

Figure 4. Spatial maps of 4-km GOES 8 products for the SGP Aqua overpass on 25 July 2002 at08:35 UTC: (a) band 8 (�11.0 mm) brightness temperatures, (b) near-surface pressure, (c) near-surface airtemperature, and (d) near-surface water vapor. In each panel the 3 by 3 array of AIRS footprints, theAMSU footprint, and the ARM site location (star) are shown.

Figure 5. Spatial variability at the SGP site. Differences between local ARM site profiles and thosewhich have been adjusted to account for large-scale spatial gradients and the location of the ARM sitewithin the AMSU footprints. (left) Temperature differences. (right) Percent difference in water vaporamounts. In each panel the light shaded curves are differences for individual cases, and the dashed andsolid curves are the mean and RMS differences, respectively, for 1 km (temperature) and 2 km (watervapor) thick layers.


6 of 18

D09S14

tween local SGP site profiles and those which have beenadjusted to account for large-scale spatial gradients withinthe AMSU footprints as discussed above. As with thetemporal differences, spatial gradients at the SGP site aresignificant in terms of the ability to validate the AIRSretrievals. Spatial variability for the tropical ocean is farless than for land sites in the middle and high latitudes.Attempts have been made to use data from the Geostation-ary Meteorological Satellite (GMS-5, prior to May 2003)and GOES 9 (after May 2003) to account for the spatialgradients at TWP. Various issues have been encountered inobtaining and using high-spatial and high–temporal reso-lution atmospheric profiles from these data, however, andspatial adjustments are not performed for the TWP siteprofiles. Future work will involve the use of atmosphericprofile retrievals from the Moderate resolution ImagingSpectroradiometer (MODIS) [Seemann et al., 2003] onAqua to account for the spatial gradients within the AMSUfootprints at all three sites.[13] The high–vertical resolution pressure, temperature,

and water vapor profiles are interpolated to a standard

altitude grid and written to an output file for each Aquaoverpass of each site. The profiles extend to the maximumheight of the shortest radiosonde used in the processing,with a minimum height of 200 mbar. Three versions of theprofiles are provided in the files, including the unmodifiedprofiles (before time and space interpolation) for the tworadiosondes with launch times bounding the overpass time,the weighted mean profile with only the time interpolationsperformed, and the weighted mean profile with both timeinterpolation and spatial adjustment performed. Variousadditional data included in the output files include cloudbase height data from a Vaisala ceiliometer, total columnliquid water vapor from the MWR, estimates of the skintemperature and surface emissivity, and metadata regardingthe logistics of the case.[14] Sample ARM best estimate profiles and sample

AIRS spectra for the TWP and SGP sites are shown inFigures 6 and 7. In Figure 6, note the relatively lowtemporal variability of the TWP site profiles and the largevariability of the SGP site profiles. Also note the relativelack of high–vertical resolution structure in the TWP

Figure 6. Sample temperature and water vapor profiles at the (top) TWP and (bottom) SGP sites.


7 of 18

D09S14

profiles, as opposed to the SGP profiles. Figure 7 alsodemonstrates the range of thermodynamic and cloud con-ditions that are observed at each site.[15] In the following discussion regarding the accuracy of

the validation profiles and in the AIRS retrieval validationanalysis presented in section 3, only the TWP and SGP sitesare included; characterization and validation of the moredifficult polar atmospheres at the NSA site is deferred to asubsequent paper. Numerous analyses have been performedto characterize the accuracy of the ARM site temperatureand water vapor measurements [e.g., Clough et al., 1994;Lesht and Liljegren, 1995; Tobin et al., 2002; Revercomb etal., 2003; Turner et al., 2003, 2004; Ferrare et al., 2004;Miloshevich et al., 2006; Whiteman et al., 2006]. Consid-ering these various investigations in the context of AIRSretrieval validation, the uncertainty of the ARM best esti-mate temperature profiles is estimated to be �0.2 K below100 mbar and for water vapor the uncertainty is estimated tobe �3% below 500 mbar and �10% from 500 to 100 mbar.These are estimates of the absolute uncertainties in the meanbiases for the TWP and SGP ensembles, due primarily tosensor calibration uncertainties, and should be consideredwhen interpreting the mean biases of the AIRS retrievalsinvestigated in section 3.[16] Uncertainties that contribute to individual profiles

and overpasses are more difficult to quantify and dependdirectly on the temporal and spatial variability of the case,the temporal and spatial collocation of the AIRS and ARMmeasurements, and the uncertainty of the ARM measure-ments for those conditions. When reduced to 1 km and 2 kmvertical layering, the vertical resolution of the best estimateprofiles contributes a negligible uncertainty. When per-forming the microwave scaling, the MWR data are averagedover a 20 minute period coincident with the radiosondeascent through the lower troposphere, and short-term vari-ability in the MWR observations is also negligible. How-ever, despite the previously described approaches to accountfor temporal and spatial variability, we should expect some

real variability which is not accounted for in the bestestimate profiles. For example, even with the MWR scalingapproach, residual radiosonde calibration differences for theradiosonde pairs can produce differences that are interpretedas real atmospheric variability. Also, because of radiosondedrift, the MWR, AERI, and radiosondes do not sample theexact same air mass. Since these profiles are used as ‘‘truth’’in the AIRS retrieval validations, the computed RMSretrieval errors should therefore be considered as upperbounds of the true errors. This is particularly true for theSGP site which experiences large variability. We expectnegligible impact of this effect on the computed meanbiases since the residual variability is random from caseto case. Very rough estimates of the residual RMS contri-butions due to the residual variability at the SGP site,computed as 20% of the observed short-term temporaland spatial variability (Figures 3 and 5), are 0.2 K fortemperature and 7% for water vapor. Further discussion ofthis issue is included in section 3.[17] While significant improvements in radiosonde per-

formance and characterization have been achieved in recentyears, measurements of low water vapor amounts in theupper troposphere remain challenging. In this regard itshould be noted that candidate corrections to the VaisalaRS-90 upper level water vapor profiles [Miloshevich et al.,2006] are not included in this analyses, but are underconsideration for future analyses. The component of thesecorrections which have the largest potential impact on theAIRS retrieval assessments is based on comparisons ofeight nighttime radiosonde profiles with colocated measure-ments by a reference frost point hygrometer. For the TWPphase 1 profiles, for example, these corrections produce anoverall moistening of the radiosonde profiles by approxi-mately 15% from roughly 200 mbar to the tropopause, withsignificant variations in the corrections from profile toprofile. Additionally, the Vaisala radiosonde water vaporprofiles are known to have a diurnal bias, with daytime totalcolumn water vapor amounts 3 to 8% drier than nighttime

Figure 7. Sample AIRS spectra at the (top) TWP and (bottom) SGP sites.


8 of 18

D09S14

total column amounts, as demonstrated with coincidentMWR and downwelling infrared radiance observations[e.g., Turner et al., 2003; Miloshevich et al., 2006]. Interms of total column water vapor, the microwave scalingapproach accounts for this diurnal bias. Analyses of clearsky observed and calculated AIRS radiances for upper levelwater vapor spectral channels using day and nighttimeMWR scaled RS-90 profiles, however, demonstrate thatthe diurnal bias is height-dependent, with the daytime RS-90s drier by 5 to 8% in the upper troposphere [Strow et al.,2006]. Impact of this bias on the AIRS retrieval validation isinvestigated in section 3.[18] As mentioned previously, estimates of the surface

emissivity and of the local surface skin temperature areincluded in the best estimate products. Considering the sizeof the AMSU footprints and the spatial nonhomogeneity ofthe SGP and NSA sites, it is a difficult task to estimate theeffective skin temperature and emissivity without retrievingthe quantities from the AIRS data themselves. The skintemperature and emissivity are included, however, alongwith the best estimate profiles to provide initial estimates asinput for top-of-atmosphere radiance calculations. The localskin temperature estimates are measured with downlookingbroadband infrared radiometers at the ARM sites. Theseestimates are not expected to be representative of the largerAMSU footprints. For observations over ocean, the surfaceemissivity is homogeneous and well known [Masuda et al.,1988; Wu and Smith, 1997]. Over land, however, thespectral surface emissivity can vary largely in both timeand space. Likewise, the remote sensing of land surfacetemperature from satellite requires a detailed knowledge ofinfrared land surface emissivity. Roughly speaking, a 2%error in the knowledge of the land surface emissivity near10 microns leads to an error in the derived surface temper-ature of about 1 K. Since the emissivity of bare soil can varyacross the infrared spectrum by 10% or more, errors in theremote sensing of surface temperature from satellites can besubstantial. To address this issue for the ARM SGP domain,investigations utilizing land type surveys, unique AERIobservations of the distinct surface types, and aircraft-basedsounding and imager data have been used to develop aparameterization of the surface emissivity [Knuteson et al.,2003]. In this parameterization, shown in Figure 8, thesurface emissivity representative of an AIRS footprint iscomputed as the linear combination of pure bare soil

emissivity and pure vegetated surface emissivity, with theweighting determined from vegetation fractions determinedfrom land type surveys. Improvement of this empiricalmodel for the SGP site, and the development of a similarmodel for the NSA site, is the subject of ongoing research.[19] In summary, the ARM site best estimate data set is a

collection of ‘‘microwave scaled’’ Vaisala RS-90 radiosondeprofiles, launched at the approximate Aqua overpass times,interpolated to the exact overpass time using time continuousground-based profiles, adjusted to account for spatial gra-dients within the AMSU footprints, and supplemented withsurface temperature, surface emissivity, and some limitedcloud observations. Awide range of atmospheric, cloud, andsurface conditions (hot/cold, wet/dry, day/night, clear/cloudy,ocean/land/ice) are sampled at the three ARM sites. Relyingon the developed ARM infrastructure and previous andongoing characterization studies of the ARM measurements,the data set provides a good combination of statistics andaccuracy which is essential for assessment of the advancedsounder temperature and water vapor retrievals.

3. Validation of AIRS Temperature and WaterVapor Retrievals

[20] Various validation activities to assess the accuracy ofthe AIRS radiances, forward model, cloud cleared radiances,and retrievals have been conducted to date. Summaries ofmany of these activities and additional references areprovided by Fetzer et al. [2003] and Fetzer [2006]. Inparticular, the absolute accuracy of the AIRS spectralradiances have been validated to �0.2 K in selected windowchannels using sea surface temperatures [Aumann andGregorich, 2006] and for most channels with weightingfunctions peaking below 50 mbar using aircraft-basedspectral radiance observations [Tobin et al., 2006], and theAIRS clear sky radiative transfer model has been validatedto 0.2 K for the majority of spectral channels used fortropospheric temperature and water vapor sounding [Strowet al., 2006]. With validated radiance observations and awell-characterized forward model, validation of the AIRScloud cleared radiance and retrieval products is beingpursued. The AIRS team has taken a tiered approach toretrieval algorithm development and validation, startingwith clear sky conditions over nonpolar ocean, followedby clear and cloudy conditions over nonpolar ocean, then

Figure 8. (left) AERI measurements of the two dominant emissivity types used in the SGP emissivityparameterization: bare soil (shaded) and vegetated (solid). (right) A coarse model of fractional vegetationcover (bare soil (shaded) and vegetated (solid)) as a function of season used in the parameterization. Thetime variation is based on the agricultural growth cycle typical of winter wheat production in north centralOklahoma.


9 of 18

D09S14

nonpolar land conditions, and finally polar conditions.Various validation data have been used to assess theaccuracy of the retrievals, including numerical forecastmodel analysis fields, global radiosondes, dedicated radio-sondes, lidar, and retrievals from high-altitude aircraft.Using the ARM site best estimate profiles, validation ofAIRS Team version 4 temperature and water vapor retriev-als for the tropical ocean TWP and midlatitude land SGPconditions is presented here. Susskind et al. [2003] andJ. Susskind and M. Chahine (Accuracy of geophysicalparameters derived from AIRS/AMSU as a function offractional cloud cover, submitted to Journal of GeophysicalResearch, 2005, hereinafter referred to as Susskind andChahine, submitted manuscript, 2005) provide detaileddescriptions of the AIRS Team retrieval algorithm andproducts. Version 4 is the latest data processing packagedeveloped by the AIRS science team and was delivered tothe NASA Goddard Distributed Active Archive Center inApril 2005 for reprocessing of past and future AIRS data.Relevant to this study, it should be noted that elements of theAIRS fast radiative transfer algorithm used in the version 4retrievals have been adjusted on the basis of the TWP phase 1best estimate data set, as described by Strow et al. [2006].[21] The retrieval validation analysis presented here

includes only satellite overpasses for which the ARM sitebest estimate products are successfully produced, the match-ing AIRS retrievals are available, and which involved tworadiosondes that were each launched within two hours ofthe satellite overpass time and reached at least 200 mbar. Allretrievals (irrespective of their quality) within 120 km of theARM site have been considered for the three phases ofspecial ARM site radiosonde launches, resulting in up to�10 retrieval comparisons per overpass. The resulting dataset includes a total of 79 overpasses and 722 retrievalcomparisons for the TWP site and 151 overpasses and1594 retrieval comparisons for the SGP site, with nearlyequal sampling of day and nighttime cases. To account for arecent finding that routine processing of MWR data produ-ces total column pwv values which are too wet by 3%[Liljegren et al., 2005], all of the best estimate water vaporamounts are reduced by 3% before the retrieval statistics aregenerated. This bias is due to a processing issue and affectsthe routine processing of ARM microwave radiometer datafrom April 2002 through the present study. Additionally,since full atmospheric profiles are required for top-of-atmosphere radiance calculations and analyses, the bestestimate profiles are supplemented with coincident analysisfields from the European Centre for Medium-Range WeatherForecasts (ECMWF). ECMWF temperature fields areinserted above 60 mbar (or lower for radiosondes whichdo not reach 60 mbar) and water vapor fields are insertedabove 200 mbar. Note that the 200 mbar cutoff is below thetropopause height for TWP profiles while the tropopauseheight can vary at SGP, as shown in Figure 6.[22] The process of computing the AIRS retrieval perfor-

mance statistics is consistent with the approach used bySusskind et al. [2003] in assessing the retrieval performanceusing simulated data. For each profile, the ARM bestestimate profile is converted to layer quantities (layer meantemperatures and water vapor layer amounts in units ofmolecules/cm2) consistent with the AIRS fixed 101 pressurelevel grid, and the lowest valid layer values are adjusted to

account for the fractional layer above the true surface. TheAIRS 100 ‘‘pseudo-level’’ retrieval profiles provided in theLevel 2 ‘‘support’’ product files are also converted to layervalues (for temperature) and the bottom fractional layervalues above the surface are similarly adjusted. For thetemperature profile comparisons, the ARM and AIRSprofiles are degraded to �1 km vertical layer values, andmean differences (AIRS minus ARM) and RMS differencesare then computed for each layer. For water vapor compar-isons, the profiles are degraded to �2 km vertical layervalues, and mean percent differences (100 (AIRS-ARM)/ARM) and RMS percent differences are computed for eachlayer.[23] For water vapor, it should be noted that the mean

percent differences (i.e., biases) and RMS percent differ-ences are computed using the convention used in reportingAIRS retrieval statistics [e.g., Susskind et al., 2003], wherewater vapor layer amounts are used to weight the observedpercent differences. The weighting is done independentlyfor each �2-km thick layer. For ensembles with higherwater vapor variability, this process has the effect of down-weighting percent errors for cases with lower water vaporamounts. The weights are normalized for the ensemble, andthe weighting therefore has lower impact on the results forensembles with lower water vapor variability (e.g., in thetropics and in the upper troposphere).[24] As mentioned in section 2, the process of degrading

the profiles to lower vertical resolution (e.g., 1 km fortemperature) before computing the mean and RMS differ-ences has the dramatic effect of suppressing large errorsassociated with fine-scale vertical structure in the ARMprofiles. The satellite soundings have limited vertical reso-lution and cannot resolve this fine structure. The compar-isons presented here therefore suppress most of this ‘‘null-space’’ error [Rodgers, 1990] associated with the retrievals.[25] Temperature and water vapor profile validation

statistics are shown for four ensembles of profiles, cor-responding to four different selections of retrievals accord-ing to the version 4 retrieval quality flags. The reader isreferred to Susskind and Chahine (submitted manuscript,2005) for a detailed description of the quality flags and howtheir values are determined. Quality flags are provided forvarious retrieved quantities including the surface fields(Qual_Surf), water vapor profile (Qual_H2O), and temper-ature profile for three vertical regions (Qual_Temp_Profile_Top, above 200 mbar; Qual_Temp_Profile_Mid, 3 km to200 mbar; Qual_Temp_Profile_Bot, below 3 km). Each flagcan have values of 0, 1, or 2 corresponding to retrievals ofhighest, good, and poor (i.e., not accepted) quality, respec-tively. Sea surface temperatures are used in evaluatingQual_Surf and this flag is applicable only to nonfrozenocean profiles (and is therefore not used for the SGP siteselections). The criteria for the four selections of qualityflags are given in Table 2. A range of selections is includedto demonstrate how the accuracy and yield of the retrievalsvary as a function of the assigned quality flags. The firstselection (denoted QC1) includes retrievals for which all ofthe products (including the surface fields, water vaporprofile, and temperature profile at all levels) of a givenprofile are flagged as highest quality. The last selection(QC4) consists of retrievals for which the selected products(temperature at various levels, water vapor) are accepted.


10 of 18

D09S14

That is, for example, if Qual_H2O = 0 or 1 then the watervapor profile is selected irrespective of the temperature andsurface field quality flags. The other two somewhat arbi-trary selections represent data sets of intermediate quality.The QC2 ensemble are retrievals for which all of theproducts for a given profile are accepted (all qualityflags = 0 or 1) and QC3 are retrievals for which both thewater vapor and temperature profiles at a given level are ofhighest quality. Selections QC1, QC2, and QC4 (for watervapor) have yields (percent of selected cases out of allcases) that are independent of altitude while selections QC3and QC4 (for temperature) have yields that are altitude-dependent.[26] Assessment of the AIRS temperature and water

vapor retrievals for the TWP site is shown in Figure 9.The highest-quality retrievals (QC1) demonstrate verygood performance for both temperature and water vapor.The yield of these retrievals is 10%, and inspection of thecoincident AIRS spectra and retrieved cloud fractionsshow that these are primarily clear sky cases. The 1 kmlayer temperature biases are less than 1 K at all levels,with small but clearly defined oscillating biases as afunction of altitude (AIRS warmer than ARM at �500and �150 mbar, AIRS colder than ARM at �350 mbar,).The temperature RMS errors are �0.6 K or less for nearlyall layers below 200 mbar. The 2 km layer water vaporbiases are less than 5% below 400 mbar and approximatelyminus 10% (AIRS drier than ARM) from 300 to 100 mbar.Considering the estimated accuracy of the best estimateupper level water vapor amounts (�10%) the significanceof the bias reported for the upper troposphere is question-able. Application of the suggested radiosonde corrections[Miloshevich et al., 2006] would produce a negligibleimpact on these comparisons below �200 mbar but wouldmoisten the TWP profiles by �15% above �200 mbar.Note that above 200 mbar the water vapor comparisons arewith ECMWF, because of the radiosonde cutoff height.The water vapor RMS errors are very good, with values of�10% below 400 mbar and increasing to �25% at 150mbar. As discussed in section 2, the reported retrievalstatistics should be interpreted as upper bounds of theAIRS retrieval performance since the ARM validationprofiles contain uncertainties that also contribute to thecomputed statistics. The observed RMS performance atTWP, however, is in fact very similar to the retrievalperformance predicted for global clear sky cases using

simulated AIRS data. See, for example, Susskind et al.[2003, Figures 5 and 7]. This confirms that the AIRSretrieval algorithm is behaving as expected for clear skytropical atmospheres over a well-known (ocean) surface. Interms of the best estimate profile accuracy, this result isalso significant because the simulation studies include noerrors in the ‘‘truth’’ profiles or in the radiative transferalgorithms, which suggests that the uncertainty in theARM profiles do not contribute significantly to thereported RMS differences. For the other quality flagselections, the temperature and water vapor mean biasesdo not change significantly with respect to the QC1results. Going from QC1 to QC4, the yields increase andthe temperature and water vapor RMS errors degradegradually, as shown in Figure 9. Considering all acceptedtemperature and water vapor products (i.e., QC4), thetemperature yield is 55% (89%) below (above) 200 mbarand the water vapor yield is 88%. For these retrievals, thetemperature RMS errors are �1 K or less below 200 mbarand the water vapor RMS errors are 20% or less below400 mbar. Retrieval statistics for the QC1 and QC4ensembles are listed in Tables 3 and 4.[27] As mentioned previously, analyses of clear sky

observed and calculated AIRS radiances suggest that theARM best estimate water vapor profiles possess a height-dependent diurnal bias, with daytime profiles 5 to 8% drierin the upper troposphere. The bias is suspected to be dueto effects of solar radiation on the Vaisala radiosondewater vapor sensor. To assess the impact of the bias onthe AIRS retrieval profile validation, the TWP site profilestatistics were generated separately for the day and nightcases using the highest-quality (QC1) retrievals. With thisapproach, the AIRS retrievals are assumed to not have adiurnal bias and in this sense are used as a relativereference to assess the ARM profiles. As shown inFigure 10, the presence of coherent differences from day tonight are not obvious in the retrieval comparisons. Takenliterally, the comparisons in the 200 to 300 mbar regionsuggest that the daytime profiles have higher variability(�5% RMS) and that the daytime ARM profiles are wetterthan the nighttime profiles by 5 to 10%. Further investigationis required in order to resolve this finding with respect to theobserved minus calculated spectral analyses.[28] Analogous to Figure 9 for the TWP site, assess-

ment of the SGP site retrievals is shown in Figure 11. Asopposed to the TWP site results, the SGP site retrieval

Table 2. Retrieval Quality Flag Selections

Ensemble Quality Flag Criteria

QC1 Qual_Surf = 0a AND Qual_H2O = 0 AND Qual_Temp_Profile_Top = 0 ANDQual_Temp_Profile_Mid = 0 AND Qual_Temp_Profile_Bot = 0

QC2 Qual_Surf 6¼ 2a AND Qual_H2O 6¼ 2 AND Qual_Temp_Profile_Top 6¼ 2 ANDQual_Temp_Profile_Mid 6¼ 2 AND Qual_Temp_Profile_Bot 6¼ 2

QC3 Qual_Temp_Profile_Top = 0 AND Qual_H2O = 0, above 200 mbarQual_Temp_Profile_Mid = 0 AND Qual_H2O = 0, 3 km to 200 mbarQual_Temp_Profile_Bot = 0 AND Qual_H2O = 0, below 3 km

QC4 Qual_Temp_Profile_Top 6¼ 2, temperature above 200 mbarQual_Temp_Profile_Mid 6¼ 2, temperature 3 km to 200 mbarQual_Temp_Profile_Bot 6¼ 2, temperature below 3 kmQual_H2O 6¼ 2, water vapor

aQual_Surf criteria is used for the TWP (ocean) site but not the SGP (land) site.


11 of 18

D09S14

statistics do not vary significantly between the fourdifferent quality flag selections. The 1 km layer temper-ature biases are 1 K or less and display vertical oscil-lations similar in character and magnitude to the TWPtemperature biases above 700 mbar. Temperature RMSerrors are 2 K near the surface, decrease to 1 K from 500to 400 mbar, and increase to 2 K at 200 mbar. Divakarlaet al. [2006] report similar temperature biases as afunction of altitude and also discuss potential retrievalissues responsible for the biases. The 2 km layer watervapor biases are 5% or less below 400 mbar and increaseto approximately minus 10% at 200 mbar, similar to theTWP result. The water vapor RMS errors are larger than

those for TWP, with values of 25% below 500 mbar andincreasing to 35% at 200 mbar. There are no significantchanges in the results shown in Figure 11 if the timewindow (radiosonde launch minus overpass time) isreduced from 2 hours to 1 hour and the spatial window(distance from center of AMSU footprint to the ARMsite) is reduced from 120 km to 50 km. As discussedpreviously, AIRS retrievals at the SGP site are moredifficult (compared to the TWP site) because of severalissues, including the need to account for land surfaceemissivity effects, much larger atmospheric state variabil-ity, and the interpretation of microwave observations ofland surfaces for infrared cloud clearing.

Figure 9. Validation of AIRS version 4 atmospheric profile retrievals using ARM TWP site profiles.(top left) The 1 km layer temperature differences (AIRS-ARM), (top right) temperature retrieval yield,(bottom left) percent difference in 2 km layer water vapor amounts (100(AIRS-ARM)/ARM), and(bottom right) water vapor retrieval yield. Each panel includes the bias (dashed curves) and RMSE (solidcurves) for four selections of quality control flags (QC1, black; QC2, green; QC3, purple; QC4, blue) asdiscussed in the text. Note that when computing the mean and RMS water vapor profiles, the percentdifferences are weighted by the ARM water vapor amounts, as discussed in the text.


12 of 18

D09S14

[29] The SGP ensemble includes a wide range of hot/wet and cold/dry profiles, and it should be noted that thewater vapor weighting technique used in computing thewater vapor statistics has a large effect on the SGPstatistics reported here. Computing the statistics withoutthe water vapor weighting, the mean bias and RMSpercent errors are approximately 20 and 55%, respectively,in the lower troposphere (as opposed to values of �5 and25%, shown in Figure 11). This is due to fractional watervapor errors that have a strong dependence on the totalcolumn pwv, as shown in Figure 12. For low-pwv casesencountered at the SGP site, the fractional pwv retrievalerrors have larger variability and larger mean values, withthe mean pwv error approaching 25% (AIRS wetter thanARM) at 1 cm pwv. The ARM best estimate pwv valuesare determined by the MWR retrievals, which have anestimated absolute accuracy of 3% for all but very dry(<0.2 cm pwv) conditions. For the TWP site, the watervapor variability is much less and the water vaporweighting therefore has little impact on the statisticscomputed for the TWP ensemble. Similarly, water vaporvariability in the upper troposphere is relatively low at

both SGP and TWP sites, and the water vapor weightinghas little impact on the statistics computed for the uppertroposphere.[30] Last, the AIRS retrievals are examined as a function

of cloudiness. The 1 km temperature and 2 km water vaporstatistics are computed for the accepted (i.e., QC4) productsfor a range of cloud fraction bins and displayed as afunction of the AIRS retrieved cloud fraction in Figures 13and 14 for the TWP and SGP sites, respectively. For bothsites, the small-scale vertical oscillations in the temperaturebiases are largely independent of cloud fraction but with thepositive bias at �500 mbar increasing slightly with cloudfraction, particularly for the TWP site. The upper levelwater vapor bias increases slightly with cloud fraction at theTWP site. For both the SGP and TWP sites, the temperatureand water vapor RMS differences increase slightly withincreasing cloud fraction.

4. Conclusions and Summary

[31] Validation of products from the new generation ofadvanced satellite sounders requires correlative data sets

Table 4. Summary of 2 km Layer Water Vapor Percent Differences (100(AIRS-ARM)/ARM) for the QC1 and QC4 Quality Flag

Selectionsa

PressureRange, mb

MeanAltitude,b

km

TWP Site SGP Site

QC1 QC4 QC1 QC4

Yield, % Bias, % RMS, % Yield, % Bias, % RMS, % Yield, % Bias, % RMS, % Yield, % Bias, % RMS, %

99.5–138.1 15.7 10 �9.0 24.7 88 �4.5 30.9 21 1.4 28.6 87 �4.8 33.1138.1–185.1 13.9 10 �14.2 26.1 88 �14.3 34.6 21 5.2 29.6 87 �0.7 36.5185.1–266.4 11.8 10 �3.4 20.1 88 �15.9 31.3 21 �13.5 34.0 87 �11.1 34.3266.4–336.1 9.7 10 �12.0 19.6 88 �5.4 27.3 21 �10.4 29.4 87 �5.2 33.4336.1–468.8 7.6 10 �1.8 9.2 88 1.8 18.5 21 �1.3 28.7 87 0.4 29.1468.8–606.8 5.3 10 5.8 12.7 88 4.9 20.2 21 3.8 23.3 87 �1.7 25.9606.8–765.6 3.3 10 �3.1 8.9 88 �0.3 13.9 21 1.9 22.2 87 4.4 26.0765.6–995.9 1.2 10 5.4 9.5 88 3.4 8.2 21 1.8 25.8 87 2.6 25.9

aData are plotted in Figures 9 and 11.bApproximate.

Table 3. Summary of �1 km Layer Temperature Differences (AIRS-ARM) for the QC1 and QC4 Quality Flag Selectionsa

PressureRange, mb

MeanAltitude,b

km

TWP Site SGP Site

QC1 QC4 QC1 QC4

Yield, % Bias, K RMS, K Yield, % Bias, K RMS, K Yield, % Bias, K RMS, K Yield, % Bias, K RMS, K

99.5–121.7 16.1 10 �0.8 1.7 89 �0.3 1.7 21 �0.6 1.8 89 �0.5 1.8121.7–138.1 15.2 10 �0.3 1.5 89 0.4 2.0 21 �0.3 1.8 89 �0.3 1.9138.1–155.8 14.4 10 0.1 1.2 89 0.7 1.9 21 0.4 1.9 89 0.3 1.9155.8–185.1 13.5 10 0.1 0.6 89 0.3 1.3 21 1.1 2.1 89 0.8 2.1185.1–217.7 12.5 10 0.1 0.7 55 �0.1 0.9 21 0.8 1.7 71 0.6 1.7217.7–266.4 11.3 10 0.1 0.6 55 �0.3 1.1 21 �0.3 1.5 71 �0.2 1.6266.4–307.0 10.1 10 �0.3 0.6 55 �0.5 1.0 21 �0.7 1.5 71 �0.6 1.4307.0–336.1 9.2 10 �0.5 0.9 55 �0.5 1.0 21 �0.6 1.1 71 �0.6 1.2336.1–399.1 8.3 10 �0.7 0.9 55 �0.3 0.9 21 �0.4 1.0 71 �0.3 1.0399.1–468.8 7.0 10 �0.0 0.5 55 0.2 0.8 21 �0.1 1.0 71 �0.1 1.0468.8–525.4 �5.9 10 0.4 0.7 55 0.7 1.1 21 0.1 1.0 71 0.0 1.0525.4–606.8 4.9 10 0.0 0.5 55 0.4 0.9 21 0.5 1.0 71 0.3 1.1606.8–672.4 3.9 10 �0.1 0.6 55 0.2 0.9 21 0.4 1.0 71 0.3 1.1672.4–765.6 2.9 10 0.3 0.7 55 0.4 0.8 21 �0.1 1.1 71 �0.0 1.2765.6–865.6 1.8 10 �0.0 0.5 55 �0.0 0.8 21 �0.5 1.4 71 �0.3 1.4865.6–995.9 0.7 10 �0.1 0.6 55 �0.5 1.2 21 0.5 2.0 71 0.5 2.1

aData are plotted in Figures 9 and 11.bApproximate.


13 of 18

D09S14

which are accurate, well characterized and statisticallysignificant. Using various ARM data, including radiosondeslaunched at the Aqua overpass times, ensembles of ‘‘bestestimate’’ temperature and water vapor profiles for Aquaoverpasses of the ARM sites have been constructed, asdescribed in section 2. This data set has been used to assessthe accuracy of the AIRS radiative transfer algorithm and, insection 3 of this paper, assess the accuracy of the AIRSTeam version 4 temperature and water vapor retrievals at theTWP and SGP ARM sites. Specific findings of the retrievalvalidation study include:[32] 1. Short-term and small-scale variability of the

atmosphere is significant and should be taken into accountwhen validating large area footprint retrievals from polarorbiting satellites. This is particularly important for assess-ing RMS errors of the retrievals, opposed to the meandifferences (biases) for which the random variability cancelsfor a large ensemble of cases. RMS errors computed usingvalidation profiles that do not fully account for the vari-ability should be interpreted as upper bounds of the trueerrors.[33] 2. The conventional approach used to report AIRS

water vapor retrieval statistics (and used in this paper) ofweighting the observed percent errors by water vaporconcentration can produce significant differences from thetraditional, unweighted, calculations. This is particularlytrue for ensembles with high water vapor variability, suchas the SGP site lower troposphere.[34] 3. The yields of AIRS retrievals with temperature,

water vapor and surface products flagged with highestquality (i.e., the QC1 ensembles) are 10 and 21% for theTWP and SGP sites, respectively. Considering all accept-ed products (i.e., the QC4 ensemble) at the TWP site, themiddle and bottom level (below 200 mbar) temperatureyield is 55%, the top level (above 200 mbar) temperature

yield is 89%, and the water vapor yield is 88%. Analo-gous yields for the SGP site are 71, 89, and 87%,respectively.[35] 4. AIRS retrievals for the tropical ocean TWP site

have very good performance, with RMS errors approachingthe theoretical limit predicted by retrieval simulation studiesperformed with no errors in the truth data or radiativetransfer algorithms. Retrievals for which the temperature,water vapor, and surface products are flagged as highestquality (QC1) have the best performance, and the perfor-mance degrades gradually as retrievals flagged with lowerquality (and more cloudy scenes) are included. For allaccepted temperature and water vapor products (QC4),1 km layer temperature RMS errors are �1 K or less below200 mbar and 2 km layer water vapor RMS errors are 20%or less below 400 mbar.[36] 5. AIRS retrievals for the midlatitude land SGP site

have poorer RMS performance with respect to the TWP siteresults for both temperature and water vapor. 1 km layertemperature RMS errors range from 1 to 2 K and 2 km layerwater vapor RMS errors range from 25 to 35%. Thisperformance is largely independent of the retrieval qualityflags, yield, and cloud fraction.[37] 6. AIRS total column precipitable water vapor (pwv)

fractional errors are higher for lower-pwv conditions en-countered at the SGP site, with mean fractional errors of�25% (AIRS wetter than ARM) at 1 cm pwv. These largerpercent errors observed for lower water vapor amounts aresuppressed when the water vapor weighting approach isused to compute the AIRS water vapor mean and RMSdifferences.[38] 7. For both the TWP and SGP ensembles, small-scale

(�0.5 K) vertical oscillations are present in the AIRStemperature retrievals (AIRS retrievals are too warm at�600 mbar, too cold at �300 mbar, too warm at

Figure 10. Bias (dashed) and RMSE (solid) differences for daytime (shaded) and nighttime (solid) fullyaccepted retrievals (QC1) at TWP. (left) The 1 km layer temperature differences, (middle) percentdifference in 2 km layer water vapor amounts, and (right) yield.


14 of 18

D09S14

�150 mbar). The magnitude of the oscillations are largelyindependent of retrieved cloud fraction.[39] 8. For both the TWP and SGP sites, water vapor

biases are �5% or less below 400 mbar and increase tominus �10% (AIRS drier than ARM) at 200 mbar. Thebiases are largely independent of retrieved cloud fraction.The significance of the reported upper troposphere bias isquestionable given the estimated absolute accuracy of theARM profiles (�10%) at these levels.[40] 9. Evidence of diurnal biases in the upper level water

profiles is not evident in the ARM/AIRS comparisonsshown here.[41] The AIRS Team version 4 retrieval results reported

here for the ARM TWP site demonstrate the potential ofthe new generation of advanced high–spectral resolutionsatellite sounders. The TWP site retrievals meet orsurpass the 1 K/1-km and 20%/2-km RMS performancerequirements throughout the middle and lower tropo-sphere for clear and partly cloudy conditions. The high

accuracy is due in part, however, to the well-known(ocean) surface and to the tropical profiles which havelittle vertical structure and low spatial and temporalvariability. The results for the SGP midlatitude land siteare more indicative of the AIRS version 4 soundingaccuracy under more typical meteorological condition.For the SGP site, the 1-km temperature and 2-km watervapor RMS performance requirements are not met, moti-vating further improvement of the AIRS science teamretrieval algorithm. Retrievals at the SGP site are difficultbecause of several potential issues, including the need toaccount for land surface emissivity effects, much largeratmospheric state variability, and the interpretation ofmicrowave observations of land surfaces for infraredcloud clearing. The AIRS Team is currently consideringseveral different algorithm changes to address theseissues. An improved method to retrieve and account forthe spectral surface emissivity, for example, is currentlybeing evaluated. Improved temperature and water vapor

Figure 11. Same as Figure 9 but for the SGP site.


15 of 18

D09S14

retrieval performance over land is expected to be provided in asubsequent AIRS Team algorithm release.[42] The ARM site best estimate products described in

this paper will continue to be used to provide systematic andobjective evaluations of future AIRS Team algorithmchanges. Further applications include assessment of retriev-als performed using combined AIRS and MODIS observa-

tions [Li et al., 2005], assessment of the AIRS Team version4 retrievals for the polar NSA site, and assessments ofNOAA 16 ATOVS and Aqua MODIS temperature andwater vapor retrievals. Case studies and statistical correla-tions of the retrieval performance versus retrieval andvalidation parameters are also being used to diagnosespecific issues within the retrieval algorithms. Further

Figure 12. Total column precipitable water vapor (pwv) ratios (pwvAIRS/pwvARM) as a function ofpwvARM for the SGP (shaded) and TWP (solid) site profiles for QC4 selected profiles (all cases withQual_H2O = 0 or 1). Note that pwvARM is determined by the MWR observations.

Figure 13. TWP site retrieval bias and RMS as a function of AIRS retrieved cloud fraction, using QC4selected profiles: (a) temperature mean difference, (b) temperature RMS differences, (c) water vapormean percent difference, and (d) water vapor RMS percent difference.


16 of 18

D09S14

development of the ARM best estimates products will alsobe pursued; specific activities include continued studies tocharacterize the accuracy of the ARM upper level watervapor measurements, improved representation of the SGPand NSA site surface emissivities, and the use of AquaMODIS products to account for spatial gradients within theAMSU footprints. The data set will also be extended toinclude additional radiosonde launch campaigns.

[43] Acknowledgments. This research was supported by the EOSScience Project Office under NASA grant NNG04GG22G. We gratefullyacknowledge numerous ARM infrastructure personnel, David Starr of theEOS Project, and the AIRS project at JPL for facilitating this work. Thanksto Chris Barnet, Bill Smith, and Dave Turner for discussions regardingvarious aspects of the analysis and comments provided by three anonymousreviewers. ARM data were obtained from the Atmospheric RadiationMeasurement (ARM) Program sponsored by the U.S. Department ofEnergy, Office of Science, Office of Biological and Environmental Re-search, Climate Change Research Division.

ReferencesAckerman, T., and G. Stokes (2003), The Atmospheric Radiation Measure-ment program, Phys. Today, 56, 38–45.

Aumann, H. H., and D. Gregorich (2006), Validation of the AIRS radio-metric calibration for climate research using sea surface temperaturemeasurements, J. Geophys. Res., doi:10.1029/2005JD006085, in press.

Aumann, H. H., et al. (2003), AIRS/AMSU/HSB on the Aqua mission:Design, science objectives, data products, and processing systems, IEEETrans. Geosci. Remote Sens., 41, 253–264.

Benjamin, S. G., G. A. Grell, J. M. Brown, T. G. Smirnova, and R. Bleck(2004), Mesoscale weather prediction with the RUC hybrid isentropic–terrain-following coordinate model, Mon. Weather Rev., 132, 473–494.

Clough, S. A., P. D. Brown, N. E. Miller, J. C. Liljegren, and T. R. Shippert(1994), Residual analysis of surface spectral radiances between instru-ment observations and line-by-line calculations, paper presented atFourth Atmospheric Radiation Measurement (ARM) Science Team Meet-ing, U. S. Dep. of Energy, Washington, D. C.

Divakarla, M. G., C. D. Barnet, M. D. Goldberg, L. M. McMillin,E. Maddy, W. Wolf, and L. Zhou (2006), Validation of AIRS temperatureand water vapor retrievals with matched radiosonde measurements andforecasts, J. Geophys. Res., doi:10.1029/2005JD006116, in press.

Feltz, W. F., H. B. Howell, R. O. Knuteson, H. M. Woolf, and H. E.Revercomb (2003), Near continuous profiling of temperature, moisture,

and atmospheric stability using the Atmospheric Emitted Radiance Inter-ferometer (AERI), J. Appl. Meteorol., 42, 584–597.

Ferrare, R. A., et al. (2004), Characterization of upper troposphere watervapor measurements during AFWEX using LASE, J. Atmos. OceanicTechnol., 21, 1790–1808.

Fetzer, E. J. (2006), Preface to topical collection on AIRS, J. Geophys. Res.,doi:10.1029/2005JD007020, in press.

Fetzer, E. J., et al. (2003), AIRS/AMSU/HSB validation, IEEE Trans.Geosci. Remote Sens., 41, 418–431.

Knuteson, R., H. Revercomb, D. Tobin, and B. Osborne (2003), Landsurface product validation using the DOE ARM Southern Great PlainsSite, Eos Trans. AGU, 84(46), Fall Meet. Suppl., Abstract B21E-0751.

Lesht, B. M., and J. C. Liljegren (1995), Comparison of precipitable watervapor measurements obtained by microwave radiometry and radiosondesat the Southern Great Plains CART site, paper presented at 5th ARMScience Team Meeting, U. S. Dep. of Energy, Washington, D. C.

Li, J., C.-Y. Liu, H.-L. Huang, T. J. Schmit, X. Wu, W. P. Menzel, and J. J.Gurka (2005), Optimal cloud-clearing for AIRS radiances using MODIS,IEEE Trans. Geosci. Remote Sens., 43(6), 1266–1278.

Liljegren, J. C., S.-A. Boukabara, K. Cady-Pereira, and S. A. Clough(2005), The effect of the half-width of the 22-GHz water vapor line onretrievals of temperature and water vapor profiles with a 12-channelmicrowave radiometer, IEEE Trans. Geosci. Remote Sens., 43(6),1102–1108.

Ma, X. L., T. J. Schmit, and W. L. Smith (1999), A nonlinear physicalretrieval algorithm—Its application to the GOES-8/9 sounder, J. Appl.Meteorol., 38, 501–513.

Masuda, K., T. Takashima, and Y. Takayama (1988), Emissivity of pure andsea waters for the model sea surface in the infrared window regions,Remote Sens. Environ., 24, 313–329.

Miloshevich, L. M., H. Vomel, D. N. Whiteman, B. M. Lesht, and F. J.Schmidlin (2006), Absolute accuracy of water vapor measurements fromsix operational radiosonde types launched during AWEX-G, and implica-tions for AIRS validation, J. Geophys. Res., doi:10.1029/2005JD006083,in press.

Paukkunen, A., V. Antikainen, and H. Jauhiainen (2001), The accuracy andperformance of the new Vaisala RS90 radiosonde in operational use, paperpresented at 11th Symposium on Meteorological Observations and Instru-mentation, Am. Meteorol. Soc., Albuquerque, N. M., 14–19 Jan.

Phillips, N. (1980), Two examples of satellite temperature retrievals in theNorth Pacific, Bull. Am. Meteorol. Soc., 61, 712–717.

Revercomb, H. E., et al. (2003), The ARM program’s water vapor intensiveobservation periods: Overview, initial accomplishments, and future chal-lenges, Bull. Am. Meteorol. Soc., 84, 217–236.

Rodgers, C. D. (1990), Characterization and error analysis of profiles re-trieved from remote sounding measurements, J. Geophys. Res., 95,5587–5595.

Figure 14. Same as Figure 13 but for the SGP site.


17 of 18

D09S14

Seemann, S. W., J. Li, W. P. Menzel, and L. E. Gumley (2003), Operationalretrieval of atmospheric temperature, moisture, and ozone from MODISinfrared radiances, J. Appl. Meteorol., 42, 1072–1091.

Smith, W. L. (1991), Atmospheric soundings from satellites—False expec-tation or the key to improved weather prediction, Q. J. R. Meteorol. Soc.,117, 267–297.

Smithsonian Institution (1984), Smithsonian Meteorological Tables, 5th ed.,350 pp., Smithsonian Inst. Press, Washington, D.C.

Stokes, G. M., and S. E. Schwartz (1994), The Atmospheric RadiationMeasurement (ARM) program: Programmatic background and designof the cloud and radiation test bed, Bull. Am. Meteorol. Soc., 75,1201–1221.

Strow, L. L., S. E. Hannon, S. De Souza-Machado, and D. C. Tobin (2006),Validation of version 4 AIRS Radiative Transfer Algorithm, J. Geophys.Res., doi:10.1029/2005JD006146, in press.

Susskind, J., C. D. Barnet, and J. M. Blaisdell (2003), Retrieval ofatmospheric and surface parameters from AIRS/AMSU/HSB data inthe presence of clouds, IEEE Trans. Geosci. Remote Sens., 41, 390–409.

Tobin, D. C., H. E. Revercomb, D. D. Turner (2002), Overview of theARM/FIRE Water Vapor Experiment (AFWEX), paper presented atTwelfth ARM Science Team Meeting, U. S. Dep. of Energy, St. Peters-burg, Fla. (Available at http://www.arm.gov/publications/proceedings/conf12/extended_abs/tobin2-dc.pdf)

Tobin, D. C., et al. (2006), Radiometric and spectral validation of Atmo-spheric Infrared Sounder observations with the aircraft-based scanninghigh resolution interferometer sounder, J. Geophys. Res., doi:10.1029/2005JD006094, in press.

Turner, D. D., and J. E. M. Goldsmith (2005), The refurbishment andupgrade of the ARM Raman Lidar, paper presented at Fifteenth ARMScience Team Meeting Proceedings, U. S. Dep. of Energy, DaytonaBeach, Fla., 14–18 March.

Turner, D. D., B. M. Lesht, S. A. Clough, J. C. Liljegren, H. E. Revercomb,and D. C. Tobin (2003), Dry bias and variability in Vaisala RS80-Hradiosondes: The ARM experience, J. Atmos. Oceanic Technol., 20,117–132.

Turner, D. D., D. C. Tobin, S. A. Clough, P. D. Brown, R. G. Ellingson, E. J.Mlawer, R. O. Knuteson, H. E. Revercomb, T. J. Shippert, and W. L.Smith (2004), The AERI LBLRTM QME: A closure experiment fordownwelling high spectral resolution infrared radiance, J. Atmos. Sci.,61, 2657–2675.

Whiteman, D. N., B. Demoz, F. Schmidlin, L. Miloshevich, B. Lesht,H. Vomel, Z. Wang, I. Veselovskii, F. Russo, and S. Hannon (2006),Analysis of Raman lidar and radiosonde measurements from theAWEX-G field campaign and its relation to Aqua validation, J. Geophys.Res., doi:10.1029/2005JD006429, in press.

Wick, G. L. (1971), Nimbus weather satellites: Remote sounding of theatmosphere, Science, 172, 1222–1223.

Wu, X., and W. L. Smith (1997), Emissivity of rough sea surface for 8–13mm: Modeling and verification, Appl. Opt., 36, 2609–2619.

��T. S. Cress, Pacific Northwest National Laboratory, Richland, WA 99352,

USA.W. F. Feltz, R. O. Knuteson, L. A. Moy, H. E. Revercomb, and D. C.

Tobin, Cooperative Institute for Meteorological Satellite Studies, SpaceScience and Engineering Center, University of Wisconsin, Madison, WI53706, USA. ([email protected])E. J. Fetzer and S. E. Hannon, NASA Jet Propulsion Laboratory,

Pasadena, CA 91109, USA.B. M. Lesht, Argonne National Laboratory, Argonne, IL 60439, USA.L. L. Strow, Department of Physics, University of Maryland Baltimore

County, Baltimore, MD 21250, USA.


18 of 18

D09S14

atmospheric radiation measurement site atmospheric state best

Documents