A multiple linear regression model for precipitation forecasting over Cuttack district, Odisha, India

Abstract

Estimation of precipitation is necessary for optimum utilization of water resources and their appropriate management. The economy of India being heavily dependent on agriculture becomes vulnerable due to lack of adequate irrigation facilities. In this paper, a multiple linear regression model has been developed to reckon annual precipitation over Cuttack district, Odisha, India. The model forecasts precipitation for a year considering annual precipitation data of its three preceding years. The model testing was performed over a century-long dataset of annual precipitation i.e. for 1904–2002. Assuming the intercept or constant of the multiple linear regression model as zero, the equation developed thereby displayed a superb result. The model predictions showed an excellent association with the observed data i.e. the coefficient of determination (R2) and adjusted R2 value was obtained to be 0.974 and 0.963 respectively. This reconciliation justifies the application of the developed model over the study area to forecast rainfall, thereby aiding in proper planning and management.