IAPRS & SIS, Vol.34, Part 7, "Resource and Environmental Monitoring", Hyderabad, India, 2002 
  
The sampling sites for calibrating the LSTs derived from 
AVHRR data (with ground data measuring 1.25m height) are 
different from Gujarat Agricultural University (GAU) agro- 
meteorological stations in Gujarat. 
2.2 Data Sets 
A 1 km LAC (Local Area Coverage) data of afternoon over 
pass (2.00 to 2.30 A.M) were taken from the AVHRR on the 
NOAA-14 operational sun-synchronous polar orbiting satellite. 
Our study was carried out particularly in the rabi season 
covering five dates in December 98 to April 99. AVHRR-LAC 
images were acquired approximately during mid of every 
month (i.e 15 December 1998, 15 January 1999, 18 February 
1999, 17 March 1999 and 14 April 1999) while taking into 
account least count contamination. 
Surface weather data (soil temperature at 5, 10 and 20 cm 
depths and ambient air temperature) measured around 14.30h 
IST at different agro-meteorological station of GAU were used 
for validation in corresponding to afternoon satellite overpass 
on each dates. 
3. METHODOLOGY 
3.1 Preprocessing of Datasets 
The AVHRR images at level 1B were pre-processed for data 
conversion and geometric correction. The brightness 
temperatures of AVHRR channels 4 and 5 (T4, T5) were 
derived using the on board coefficients. For the determination 
of reflectance in channels 1 (pl) and (p2) the post launch 
coefficients (Rao and Chen, 1996) were used. Cloud screening 
was performed using channel 1 reflectance (pl) less than 0.15 
and channel 4 greater than 273 K as an indicator of the absence 
of clouds (Czajkowshi er al. 1997). 
3.2 Processing of satellite ground combined data sets 
After the cloud screening NDVI images of each dates were 
generated using the following typical equation, 
NDVI = (p2-p1) / (p2+p1)......(1) 
Where pl and p2 are visible and near infra-red reflectance in 
channel 1 and channel 2, respectively. 
Land surface temperatures (LST) was derived from split 
window equation including the emissivity (E) correction based 
on NDVI. 
LST = T4 + 2.63* (T4 - T5) + 1.274 + (T4 + T5) / 2 * 
(0.156+3.98* (T4 - TS) / (T4 + T5))*(1-e)/ e…(2) 
Where surface emissivity (E) is based on the following 
equation (Van de Griend and Owe.1993). 
Emissivity (E) = 1.0094 + 0.043 In(NDVI + 0.2) ...(3) 
594 
In the above equation 0.2 is added to NDVI to account roughly 
for the atmospheric effect. 
Least square regression analysis were attempted involving LST 
as dependent and NDVI as independent variable for old 
eighteen districts of Gujarat. A regression analysis was also 
done to relate average LST of 9 x 9 pixel window surrounding 
selected agro-meteorological stations with ground observed 
temperatures. 
4. RESULTS AND DISCUSSION 
4.1 LST-NDVI Relationship 
.It has been established fact that regional differences in 
vegetation activity and soil moisture often leads to vary the 
relationship between LST and NDVI (i.e the slope of LST 
versus NDVI, now referred to as TVX after Prihodiko and 
Goward, 1997). As a result, it was thought worthwhile to 
evaluate trends in LST versus NDVI relationships with respect 
to changes through the entire rabi growing season (December 
through April) for Gujarat state having large agro-climatic 
variability. The results from the regression of LST on NDVI at 
district level, for each separate day of satellite over pass are 
summarized in table 1. The R? values were also significantly 
different from zero at the 95 per cent confidence level and 
ranged from 0.62 to 0.99, except for Valsad and Bharuch 
district on 18" February due to high cloud coverage which was 
not removed with particular cloud screening threshold 
Table 1 also revealed that a strong negative relation exists 
between mean NDVI and land surface temperature, for all 
districts on each day of satellite over pass. The slope of 
regressions were variable among districts and over different 
dates of satellite over pass ranging from -20.5 to —64.5. 
Districts with a low irrigation infrastructure, such as Amreli, 
Rajkot and Bhavnagar and Ahmedabad has higher negative 
slopes in comparison with Kheda, Bhaunch and Vadodara 
districts with better irrigation facilities. Districts with similar 
regression slopes generally were not spatially clustered, 
although there was a tendency for districts in central and South- 
West of Gujarat to have similar slopes. However, all districts 
showed increasing trend in slope of regression towards end of 
maturity compared to early growth stages (i.e December and 
January). The results presented also have implications for using 
LST-NDVI relationship to evaluate place to place differences in 
negative stress or variation in evaporation fluxes. 
4.2 Steepness and Scatterogram of LST-NDVI relationship 
Typical examples of LST-NDVI scatter plots are given in 
figure 1 and figure 2, for better examination of LST-NDVI 
scatterogram with respect to site differences and progress of 
phenology during rabi crop growing season. These scatterplots 
revealed that all the relationships are linear which was 
consistent with the results obtained for other locations by 
Goward et al, 1985; Nemani and Running, 1989; Gupta et al, 
1997). The scatterogram for typical districts (figure 1) revealed