asia-pacific remote sensing symposium

Estimation of Upper Tropospheric Humidity from water vapour channel of Very High Resolution Radiometer onboard INSAT-3A and

Kalpana Satellites

P K Thapliyal, M Vinayak, K S Ajil, S Shah, P K Pal and P C Joshi

Atmospheric Sciences Division Meteorology & Oceanography Group

Space Applications Centre, ISRO Ahmedabad-380015, I NDIA

ABSTRACT Present study describes a methodology to establish an empirical expression to estimate the upper tropospheric humidity (UTH) from brightness temperature observations in water vapour channel of Very High Resolution Radiometer (VHRR) onboard Indian geostationary satellites INSAT-3A and Kalpana. Radiative transfer simulations for VHRR water vapour channel were made using SBDART model for tropical atmosphere with different upper tropospheric relative humidity values and varying zenith angles. INSAT-3A and Kalpana VHRR sensor response functions (SRF) for water vapour channel were used to simulate the convolved radiances. Estimated UTH values have been compared with corresponding Meteosat-5 observations. Comparison of retrieved UTH is also made with radiosonde observations of relative humidity weighted by water vapour channel weighting function. Keywords: UTH, weigthing function, INSAT-3A, Kalpna, Meteosat, VHRR, radiosonde

1. INTRODUCTION Water vapour plays an important role in the weather and climate system. It affects the radiation budget of the earth-atmosphere system by absorbing outgoing infrared radiation and contributing to the greenhouse effect. Various studies have demonstrated the importance of water vapour in the earth-atmosphere system (Raval and Ramanathan, 1989; Stephans, 1990). Upper tropospheric humidity also indicates the deep convection by pumping the moisture into the upper atmosphere. Therefore, estimates of upper tropospheric humidity could help in the diagnosis of convective processes in the atmospheric model simulations. In view of the importance of water vapour it is essential to have a frequent and reliable observing system. Surface based observations includes radiosonde data which is sparse especially over vast oceanic region. Moreover, in the upper troposphere the conventional radiosonde observations may be unreliable (Elliot and Gaffen, 1991). With the advancement in the satellite based observation system reliable water vapour information is easily available globally. Water vapour channel of geostationary satellites, e.g. GOES, Meteosat, INSAT, are located near the center of 6.7 µm water vapour absorption band and are sensitive to the relative humidity averaged over broad upper tropospheric layer. Upper tropospheric humidity (UTH) is routinely estimated from these satellite observations. UTH is an estimate of the weighting function (of water vapour channel) weighted mean relative humidity of the atmosphere between approximately 500 hPa and 200 hPa. This implies that UTH is more likely a representative of the relative humidity around the atmospheric layer where weighting function of water vapour channel peaks. Several methods have been developed for retrieving UTH from water vapour channel measurements (Schmetz and Turpeinen, 1988; Soden and Bretherton, 1993, 1996; Stephens et al, 1996). Here we present a simple methodology to estimate the UTH from INSAT-3A and Kalpana water vapour channels using radiative transfer simulations with SBDART model.

Remote Sensing of the Atmosphere and Clouds, edited by Si-Chee Tsay, Teruyuki Nakajima, Ramesh P. Singh, R. Sridharan,

Proc. of SPIE Vol. 6408, 640807, (2006) · 0277-786X/06/$15 · doi: 10.1117/12.693982

Proc. of SPIE Vol. 6408 640807-1

Downloaded from SPIE Digital Library on 15 Sep 2011 to 210.212.130.71. Terms of Use: http://spiedl.org/terms

2. DESCRIPTION OF THE RADIATIVE TRANSFER MODEL – SBDART

SBDART (Santa Barbara DISORT Atmospheric Radiative Transfer) model developed by Ricchiazzi et al (1998) at University of California, has been used for INSAT-3A/Kalpana VHRR radiance simulations. SBDART has been designed to simulate radiances received by satellite sensor using atmospheric profile of temperature, humidity and ozone mixing ratio along with the sensor response function (SRF). Development of SBDART is primarily based on highly developed and reliable physical models. SBDART uses low resolution band models developed for the LOWTRAN 7 atmospheric transmission code (Pierluissi and Marogoudakis, 1986) which provide the clear sky atmospheric transmisson from 0 to 50000 cm-1 including the effects of all radiatively active molecular species the atmosphere. Spectral resolution of SBDART is 20 cm-1 (translating to a wavelength resolution of ~5 nm in visible and ~200 nm in the thermal infrared). SBDART has six standard atmospheric profiles representing the following typical climatic conditions: tropical, midlatitude summer, midlatitude winter, subarctic summer, subarctic winter and US62. These model atmospheres (McClatchey et al, 1972) have been extensively used to provide standard vertical profiles of pressure, temperature, water vapor and ozone density in radiative transfer models as input. Besides these user can specify separate model atmosphere, e.g, a radiosonde profile. SBDART can compute the radiative effects of several aerosol types in the atmosphere. DISORT (DIScreet Ordinate Radiative Transfer, Stamnes et al, 1988) is used to numerically integrate the radiative transfer equation. DISORT provides a numerically stable algorithm to solve the equations of plane-parallel radiative transfer in a vertically inhomogeneous atmosphere. SBDART is configured to allow up to 65 atmospheric layers and 40 radiation streams (40 zenith angles and 40 azimuthal modes). SBDART uses six basic surface types - ocean water, lake water, vegetation, snow and sand to parameterize the spectral reflectivity of the surface.

3. ALGORITHM AND METHODOLOGY The UTH estimation is in principle the computation of weighted mean column values of the upper tropospheric relative humidity. It involves quantitative description of the transfer of radiation by radiative transfer model in the water vapour channel from the surface to the satellite sensor through atmosphere. The transfer calculations are performed for a set of different constant humidity values for the upper tropospheric atmosphere for standard atmosphere. Advantage of using constant relative humidity through out middle and upper atmosphere is that the computation of UTH become very simple (equal to the constant relative humidity value) and does not require to be weighted by the water vapour channel weighting function. Since geostationary satellites are located over equator, we have used standard tropical atmosphere to represent the vertical temperature structure. VHRR water vapour radiances are simulated using SBDART model. Filter response function for the INSAT-3A and Kalpana-1 WV channel were provided as input to the SABDART in order to convolve the spectral radiances. Standard tropical profile was modified by replacing specific humidity at levels above 800 hPa to represent constant relative humidity ranging from 2 to 100% with the interval gradually increasing from 1% for lower values to 5% for the higher values. Satellite zenith angle (β) was varied from 0° to 70° at an interval of 5° for each constant relative humidity profile (or UTH values). Radiances were simulated for each set of profiles and zenith angles at the top of atmosphere (100 km) from SBDART for WV channel. Brightness temperature of water vapour channel is computed from these simulated radiances. Random gaussian noise with standard deviation equal to water vapour channel NeDT (=0.06K at 300K) is added to the simulated brightness temperatures. For regression analysis we use the analytical expression derived by Soden and Bretherton (1993), which is of the following form:

bTbaUTHp .)cos(

).(ln 0 +=⎟⎟⎠

⎞⎜⎜⎝

⎛β

------------------- (1)

where, β is zenith angle and Tb is the water vapour channel brightness temperature. Here p0 is reference pressure, which is equal to the pressure of the 240 K isotherm divided by 300 hPa. For most of the tropical region p0 may be considered constant and equal to 1.0 (Soden and Bretherton, 1993). Therefore, above expression simplifies to:



bTbaUTH .)cos(

ln +=⎟⎟⎠

⎞⎜⎜⎝

⎛β

--------------------- (2)

A bilinear regression relationship is established between ln{uth/cos(β) and water vapour channel brightness temperature Tb to obtain the ‘a’ and ‘b’ coefficients (for Tb < 245 and Tb ≥ 245). Fig.1-2 shows the regression analysis between water vapour Tb and UTH alongwith the corresponding linear regression equation for INSAT-3A and Kalpana satellites respectively. Coefficient of correlation is greater then 0.99 in all the cases. Regression coefficients ‘a’ and ‘b’ for both satellites are summarized in the following table-1:

Table-1: Regression coefficients ‘a’ and ‘b’ for ln{uth/cos(β) and Tb Tb < 245K a = 36.66 b = -0.1236 INSAT-3A Tb ≥ 245K a = 33.64 b = -0.1359 Tb < 245K a = 36.91 b = -0.1366 Kalpana Tb ≥ 245K a = 34.01 b = -0.1247

Finally UTH may be retrieved from water vapour Tb using following relationship:

UTH = cos(β).exp(a + b.Tb) ---------------- (3)

The algorithm uses water vapour channel of the INSAT-3A/Kalpana-1 VHRR (5.7-7.1µ) to estimate the upper tropospheric humidity in absence of middle and upper level clouds. The generation of the UTH product is directly based on the WV brightness temperatures. Each water vapour image is converted into different segments of 5 x 5 pixels (approximately 50 km x 50 km) as the averaging would reduce the error in measurements and also helps removes the pixels with partial cloud cover. The WV image pixels of each segment not belonging to contamination with medium and high cloud clusters are averaged (if there are any, otherwise no UTH value is derived for such a segment). Presently, pixels with middle and upper clouds are removed using a threshold value of IR brightness temperature of 270 K. However, this approach needs further modification. Out of cloud free pixels 5 pixel with highest water vapour brightness temperature are averaged to further reduce the possibility of partial cloud cover. These segment WV brightness temperatures are used to derive an upper tropospheric relative humidity by using the empirical relation (2). Brightness temperatures depend strongly upon the satellite zenith angle since it increases the path length for the radiances reaching the satellite. Therefore, zenith angle is computed for each segment to be used with the regression equation. Satellite zenith angles are computed for every pixel using satellite-observer geometry using the following formula:

β = (π/2) – arc cos[sin(η)/sin(ρ)] ----------------------- (4) where, ρ=arc sin[Re/(Re+H)]

η=arc tan[sin(ρ)*sin(λ)/(1-sin(ρ)*cos(λ)] λ=arc cos[sin(δS)*sin(δT)+cos(δS)*cos(δT)*cos(∆L)]

Re = radius of Earth = 6378.16 km H = Altitude of satellite = 36000 km

∆L = | LS – LT | δT,LT) is latitude and longitude of the target δS,LS) is latitude and longitude of the sub-satellite point

Only the pixels of surface and low cloud clusters are used, excluding the pixels of contamination with medium and high cloud clusters from the processing.



Fig.2: Regression analysis between Kalpana WV brightness temperature (Tb) and Upper Tropospheric Relative Humidity (UTH)

y = -0.1366x + 36.91R2 = 0.993 (Tb < 245)

y = -0.1247x + 34.01R2 = 0.996 (Tb > 245)

0

1

2

3

4

5

6

225 230 235 240 245 250 255 260 265 270Tb

ln{U

TH/c

os(z

)}

Fig.1: Regression analysis between INSAT-3A WV brightness temperature (Tb) and Upper Tropospheric Relative Humidity (UTH)

y = -0.1359x + 36.66R2 = 0.993 (Tb < 245)

y = -0.1236x + 33.64R2 = 0.995 (Tb > 245)

0

1

2

3

4

5

6

225 230 235 240 245 250 255 260 265 270Tb

Ln{U

TH/c

os(z

en)}



4. COMPARISON AND VALIDATION

We have validated Kalpana derived UTH with the collocated radiosonde observations (within 1 hour in temporal and 50 km in spatial domain) during July 2006. These collocated observations were examined for presence of cloud cover, and only those observations were retained which are clear or have low level clouds (below 700 hPa). Since UTH by definition is weighting function weighted average of the relative humidity, we have computed weighting function (sensitivity of relative humidity at different atmospheric level to the brightness temperature observed by satellite) using perturbation method. Fig.3 shows the weighting function of Kalpana water vapour channel, i.e. relative sensitivity of Kalpana WV-channel brightness temperature for local perturbations of thin layer relative humidity at different atmospheric levels. Kalpana WV-channel Tb shows sensitivity to the relative humidity changes in broad layer between 500-200 hPa with peak sensitivity at ~350 hPa. These weights were used to compute UTH values corresponding to different radiosonde observation (referred to radiosonde observed UTH hereafter). Our preliminary study using collocated radiosonde and satellite observations shows that there is a small bias in Kalpana observed Tb and SBDART simulated Tb for water vapour channel. Kalpana WV-Channel Tb were, therefore, corrected for this bias before deriving UTH using Eq.3. Fig.4 shows scatter plot and best linear fit for radiosonde observed UTH with corresponding Kalpana derived UTH. This shows rms error of 1.9% in satellite observation of UTH and coefficient of correlation 0.97. This shows that Kalpana WV-channel observation can provide highly accurate estimates of UTH in cloud free conditions. Further, we have compared Kalpana derived UTH products with UTH products of Meteosat-5 satellite, which is located at 63°E over Indian ocean. Fig.5 shows the sample UTH maps over Indian region derived from Kalpana and corresponding product from Meteosat-5 for 04 November 2004 (12:00Z). Overall patterns are well matched with Meteosat-5 except for the fact that Meteosat-5 has more stringent condition for cloud masking. Exact comparison is not possible because of slightly different response functions of their water vapour channels, thus having small difference in weighting function. However, the dry/moist areas in the Kalpana and INSAT-3A UTH map are clearly seen matching well with the meteosat-5 UTH.

Fig.3: Relative sensitivity (Weighting function) of Kalpana WV-channel brightness temperature for local perturbations in thin layer relative humidity at different atmospheric levels

0

100

200

300

400

500

600

700

800

900

10000 0.2 0.4 0.6 0.8 1

Relative sensitivity

Pre

ssur

e (h

Pa)

Fig.4: Comparison of Kalpana WV-channel derived UTH with the collocated radiosonde observations

R2 = 0.95RMS error in retrieved UTH=1.9%

0

10

20

30

40

50

0 10 20 30 40 50

UTH derived from Kalpana WV-channel observations

UTH

der

ived

from

radi

oson

de o

bser

vatio

ns



Fig.6 shows the scatter plot of Kalpana derived UTH with corresponding UTH derived from Meteosat-5. This figure shows very good match between Kalpana derived UTH with Meteosat-5 UTH product. Coefficients of correlation is 0.84 between Kalpana and Meteosat-5 UTH for this specific case. RMS difference between Kalpana and Meteosat-5 derived UTH is 9.3%. This difference may be primarily due to slightly different sensitivity of brightness temperatures to the relative humidity in difference layers.

5. CONCLUSIONS AND FUTURE SCOPE

We have derived simple empirically based expression to derive upper tropospheric humidity from Indian geostationary satellites INSAT-3A and Kalpana water vapour channel brightness temperatures. For this purpose we have used standard tropical atmospheric profile as input in a radiative transfer model SBDART, as geostationary satellite would give continuous coverage over tropical region. Humidity in this profile has been perturbed to represent different UTH values and different zenith angles. INSAT-3A and Kalpana water vapour channel sensor response functions have been used to generate training dataset for these satellites separately. The humidity derived by using the method described here applies to a deep layer of the atmosphere reaching from 500 to 200 hPa. Kalpana derived UTH have been compared qualitatively as well as quantitatively with the collocated radiosonde observations and Meteosat-5 satellite observations. Validation with limited dataset of collocated radiosonde profiles show RMS error of 1.9% with coefficient of correlation 0.97. Comparison with corresponding meteosat-5 UTH product for a case study shows RMS error of 9.3% and coefficient of correlation 0.84. These comparisons and validations show that UTH can be derived with high accuracy from Kalpana water vapour channel.

Fig.5: Comparison of Kalpana derived UTH with Meteosat-5 derived UTH for 04 November 2004 (12:00Z)

Fig.6: Comparison of Kalpana derived UTH with Meteosat-5 derived UTH product for 04 Nov 2004 (12:00Z)

y = 0.89x + 2.6R2 = 0.71

RMSE = 9.3%

0

10

20

30

40

50

60

70

80

90

100

0 20 40 60 80 100

Meteosat-5 derived UTH

Kal

pana

der

ive

UTH



In order to improve the UTH product, the seasonal and geographical dependent factors needs to be included in the empirical formulation. We are trying to include latitudinal dependence for different seasons from the NCEP analysis through the parameter p0 in Eq.1. For this global diverse radiosonde profiles will be used in radiative transfer simulations. We are also trying to generate a large database of collocated radiosonde profiles and INSAT-3A/Kalpana observations for cloud free conditions. Acknowledgments: Authors would like to thank Director, SAC, Dep. Director, RESIPA and Group Director, MOG for their constant encouragement. Meteosat-5 products from EUTMETSAT is thankfully acknowledged.

REFERENCES 1. Elliott, W. P. and D. J. Gaffen (1991): On the utility of radiosonde humidity archives for climate studies, Bull.

Am. Meteorol. Soc., 72, 1507-1520 2. McClatchey, R.A., R.W. Fenn, J.E.A. Selby, F.E. Volz, J.S. Garing, (1972): Optical properties of the

atmosphere, (third edition), Air Force Cambridge Research Laboratories, Report AFCRL-72-0497. 3. Pierluissi, J.H., and Maragoudakis, C.E. (1986): "Molecular Transmission Band Models for LOWTRAN",

AFGL-TR-86-0272, AD A180655. 4. Raval, A., and V. Ramanathan, (1989): Observational determination of the greenhouse effect, Nature, 342,

758-762. 5. Ricchiazzi, P., Yang, S., Gautier, C., and Sowle, D. (1998): SBDART: A Research and Teaching Software Tool

for Plane-Parallel Radiative Transfer in the Earth’s Atmosphere, Bull. Amer. Met. Soc., 79 (10), 2101-2114. 6. Schmetz, J., and O.M. Turpeinen, 1988: Estimation of the Upper Tropospheric Relative Humidity Field from

METEOSAT Water Vapour Image Data. J. Appl. Meteor., 27, No. 8, 889-899. 7. Shettle, E. P., and R. W. Fenn, 1975: "Models of the atmospheric aerosols and their optical properties."

AGARD conference proceedings no. 183, Optical Propagation in the Atmosphere, 700 pages, presented at the Electromagnetic Wave Propagation Panel Symposium, Lyngby, Denmark 27-31 October 1975, sponsored by North Atlantic Treaty Organization, Advisory Group for Aerospace Research.

8. Soden B. J. and F. P. Brefherton (1996): Interpretation of TOVS water vapour radiances in terms of layer-

average relative humidities: method and climatology for the upper, middle, and lower troposphere, J. geophy. Res., 101, 9333-9343.

9. Soden B.J. and F. P. Brefherton (1993): Upper tropospheric relative humidity from GOES 6.7 �m channel:

method and climatology for July 1987, J. geophy. Res., 98, 16669-16688. 10. Stamnes, K., S. Tsay, W. Wiscombe and K. Jayaweera, 1988: "Numerically stable algorithm for discrete-

ordinate-method radiative transfer in multiple scattering and emitting layered media." Appl. Opt., 27, 2502-2509. 11. Stephens, G. L. (1990): On the relationship between water vapour on the oceans and sea surface temperature, J.

Climate, 3, p634 12. Stephens G., D. Jackson and I. Wittmeyer (1996): Global observations of upper tropospheric water vapor

derived from TOVS radiance data. J. Climate, 9, 305-326. 13. Tanre D., C. Deroo, P. Duhaut, M. Herman, J. Morerette, J, Peros and P. Y. Deschamps (1990) : “ Description

of a computer code to simulate the satellite signal in the solar spectrum : the 5S code”, Int, J, Rem. Sens., Technical note, 11, 659-668.



asia-pacific remote sensing symposium

Documents