Spectro-meteorological modelling of sorghum yield using single date IRS LISS-I and rainfall distribution data

The yield of grain Sorghum cultivated in dry-land regions in India fluctuates widely in relation to its critical growth phases depending on the weather conditions. Vegetation indices derived form remote sensing data acquired at the time of maximum vegetative growth are indicative of crop growth and vigour and consequent potential grain yields. In this paper we investigate rabi (winter) sorghum yields using Indian Remote Sensing Satellite's Linear Imaging and Self Scanning-I (IRS LISS-I) sensor data and monthly rainfall distribution data of the recent four seasons for the 37 tehsils (sub-units of districts) that constitute the three principal sorghum producing districts of the central Maharashtra state (India). The multiple linear regression yield models with both the spectral and spectro-meteorological parameters have been developed for tehsil, as well as the district yields, by identifying critical parameters with model estimation errors of about 22 per cent on tehsil yields and about 5 per cent on district yields. The yields are found to be correlated significantly with monsoon rainfall about 1 to 2 months before sowing. This study brings out the problems of yield modelling of the semi-arid tropical crop in a small region using remote sensing data only, and shows that the vegetation indices deduced from remote sensing data are found to be good indicators of the yield on large spatial scales, as the effects of varying rainfall on yields largely cancel out.