Development of Agromet-Spectral Cotton yield models using IRS data

This study was conducted in the five major cotton producing districts which account for 95.5 per cent of cotton production in Punjab State. The spectral growth profile parameters computed are: Peak value of profile (Gmax) at the time of peak occurrence (tmax),Total area under the curve (AREASUM), and Area under different growth phases (Phase-I to Phase-IV). The best subset of variables for regression analysis with yield was selected using the backward elimination procedure and detailed multiple regression analysis was carried out. The correlation matrices of NDVI and TSAVI derived from growth profile parameters indicate that growth profile area under Phase-II, Phase-III and peak value of profile (Gm) are highly correlated with district-level cotton yields. The performance of regional Spectral-Yield Models developed using both NDVI and TSAVI in terms of their yield prediction capability was evaluated by computing Mean Absolute Per cent Error (MAPE). It was observed that the predicted yields of five districts during five crop seasons (1999 to 2003) using NDVI and TSAVI based growth profile parameters, in general, the MAPE of the predicted yields were within 15 per cent.