A Crop-Weather Model for Prediction of Rice (Oryza sativa L.) Yield Using an Empirical-Statistical Technique

Abstract

Rice is the staple food in many countries and is grown in varied climates from per-humid to semiarid areas. Crop–weather models were used to predict rice yield in India. However, in spite of a significant influence of solar radiation on rice yield, none of these models used solar radiation as one of the predictors. In this paper, an attempt was made to predict the first season (June–September) rice yield at Coimbatore, Tamil Nadu, India by including solar radiation as one of the predictors. Ten years (1987/88–1996/97) data were used for the study. Seven predictors viz., percentage of rice area during first season (X1), number of days with minimum temperature below 22 °C in August and September (X2), average daily maximum temperature for three months (July, August and September; X3), average daily minimum temperature for three months (July, August and September; X4), total of average sunshine hours in August and September (X5), and total rainfall of July, August and September (X6) total average solar radiation of August and September (X7) were selected based on earlier report. Full model and stepwise regression analysis were performed using MSTAT computer package. The full model regression without solar radiation as predictor (Model I) recorded comparatively less R2 (0.6292). Inclusion of solar radiation (Model II) enhanced the R2 value considerably (R2=0.9464). Seven variables were further subjected to stepwise regression analysis and only four predictors were retained in the final model (Model III) with an R2 value of 0.9234. The model III with minimum parameters Y=22119.5758 + 19.6898, X1 − 150.9261, X2 − 1126.7501, X4 + 0.7179 X7 can be used to predict the first season rice yield (Y) at Coimbatore, India.