Different statistical models based on weather parameters in Navsari district of Gujarat

Abstract

Agriculture plays very important role in development of country. Rice is a staple food for more than half of world’s population. Timely and reliable forecasting provides vital and appropriate input, foresight and informed planning. The present investigation was carried out to forecast Kharif rice yield using two different statistical techniques, viz., discriminant function analysis and logistic regression analysis. The statistical models were developed using data from 1990 to 2012 and validation of developed models was done by using remaining data, i.e., 2013 to 2016. It was observed that value of adjusted R2 varied from 73.00 per cent to 93.30 per cent in different models. The best forecast model was selected based on high value of adjusted R2, Forecast error and RMSE. Based on obtained results in Navsari district, the discriminant function analysis technique (Model-5) was found better than logistic regression analysis (Model-12) for pre-harvest forecasting of rice crop yield. The results revealed that Model-5 showed comparatively low forecast error (%) along with highest value of Adj. R2 (93.30) and lowest value of RMSE (120.07). Also Model-5 is able to generate yield forecast a week earlier (39thSMW) than Model-12 (40thSMW).