Large area operational wheat yield model development and validation based on spectral and meteorological data

Remote sensing (RS) data acquired by satellite have wide scope for agricultural applications owing to their synoptic and repetitive coverage. Especially the spectral indices deduced from visible and near-infrared RS data have been extensively used for crop characterization, biomass estimation, crop condition monitoring, crop yield condition monitoring and crop yield monitoring and forecasting. This paper reports the development of operational spectrometeorological yield models of wheat crop using spectral index viz., normalized difference vegetation index ( NDVI), derived from NOAA-AVHRR data and monthly rainfall data. The AVHRR data spanning seven crop growing seasons, rainfall data from the rain gauge stations and crop yield data from crop cutting experiments (CCE) conducted by the state Directorate of Economics and Statistics ( DES) are the basic input parameters. The statistical multiple linear regression yield models have been developed for fifteen geographically large wheat growing districts of Rajasthan state in India. The spectrometeorological yield models were validated by comparing the predicted district-level yields with those estimated from the crop cutting experiments. The yield models based on NDVI (spectral), rainfall (meteorological) and both NDVI and rainfall (spectrometeorological) have been tried. The models have been developed for districts, groups of districts comprising agroclimatic zones, and groups of agroclimatic zones. The incorporation of monthly rainfall in the regression yield models in addition to NDVI improves the model performance significantly. Amongst the three categories of models attempted, the spectrometeorological yield models have highest predictive capability as shown by the validation results. The district-level models show highest correlation with yield, followed by agroclimatic zonal and group-of-zones level models.