Crop yield forecasting of paddy, sugarcane and wheat through linear regression technique for south Gujarat

Abstract

Multiple regression models were developed for yield forecasting of paddy, sugarcane and wheat for two districts of Gujarat (Navsari and Bharuch). The historical weather and crop yield data of 31 years of Navsari (1980-2010) and 27 years of Bharuch (1984-2010) were used. The data of de-trend yield and generated weather variables for 27 years of Navsari (1980-2006) and 23 years of Bharuch (1984-2006) were used for generation of the model for both districts. Significant weather variables are obtained on the basis of highest R2 and significant P-value. The multiple regression analysis was executed by trial and error method. The models were validated with 4 years independent data set (2007 to 2010) of these two districts. During the validation period, Navsari district model deviations for paddy, sugarcane and wheat were between -7.30 to 3.41%, 1.68 to 2.05% and -8.27 to 11.51% respectively. Similarly Bharuch model deviations for paddy, sugarcane and wheat were between 5.35 to 11.76%, -12.65 to 7.18% and -12.07 to 6.86% respectively.