Pantnagar Journal of Research

Abstract

The importance of precise crop yield forecasts cannot be overstated, as they serve as a critical input for policy formulation and implementation related to crop procurement, distribution, price structure, and import/export decisions. Our objective is to develop a reliable forecasting model that can inform evidence-based decision-making for policymakers and stakeholders. A 21-year dataset was split into training (17 years) and testing (4 years) subsets. Weather indices were developed using weekly data, in accordance with the method outlined by Agrawal et al. (1983). This study explored three multivariate models for predicting maize yields based on weather variables: SMLR (Stepwise Multiple Linear Regression), PCA (Principal Component Analysis) and DFA (Discriminant Function Analysis). The performance of the model was assessed using two primary metrics: Adjusted R-squared (adj. R²), which indicates the proportion of variance explained by the model, and Root Mean Square Error (RMSE), which measures the average magnitude of prediction errors. On the basis of adj R2 (89.98 %) and RMSE (218.54 Kg/ha), the DFA-SMLR model performed best for maize yield prediction compared to SMLR and PCA-SMLR models in the studied region.