Forecasting crop yield through weather indices through LASSO

Reliable forecast of crop production before the harvest is important for advance planning, formulation and implementation policies dealing with food procurement, its distribution, pricing structure, import and export decisions, and storage and marketing of the agricultural commodities. Weather plays a very important role in crop growth and development. Therefore, model based on weather variables can provide reliable forecast. Weather variables used can be employed for crop production forecast by making appropriate models. In this study, a statistical model is used for crop yield forecast at different growth stages of wheat crop. This model uses maximum and minimum temperature, rainfall, morning and evening relative humidity during crop growing period. The forecast model was developed using generated weather indices as regressors in model. In order to select significant weather variables affecting the yield of crop least absolute shrinkage and selection operator (LASSO) as well as stepwise regression methodology is applied. The result of lasso gives a better result as compared to stepwise regression. The R-2 of lasso and stepwise regression are 0.84 and 0.85, respectively. The mean square error (MSE) and root mean square error (RMSE) of Lasso regression were better than stepwise regression, which leads to improvement of crop yield forecasting. It can be inferred that for the data under consideration, lasso works better than stepwise regression for variable selection.