Abhishek Yadav, K.K. Mourya, K.K. Pandey, Utkarsh Pandey and Amit Kumar Yadav
Abstract
Predicting crop yield before the harvest for sound policy making at various levels. Rapeseed-mustard is very much sensitive to local weather during rabi season. In this study the focus on the utility of weather-based statistical models for forecasting rapeseed-mustard yield in Ayodhya district, Uttar Pradesh, India, using 32 years span of data from 1991-92 to 2022-23. Development of crop yield forecasting models based on seven meteorological parameters viz. maximum temperature, minimum temperature, rainfall, relative humidity, sunshine hours, wind velocity, and evaporation observed over 21 Standard Meteorological Weeks (42th SMW to 52nd SMW for the first year and 1st SMW to 10th SMW for the next year). In this context unweighted and weighted weather indices has been generated for individual variables and their interactions to develop pre- harvest yield forecast model by using Principal Component Analysis (PCA) and Multiple Regression Latent Variables (MRL) techniques by new generated weather variable Latent variables and to examine the real impact of weather parameters on crop yield. The efficiency of the model has been based on the coefficient of determination (R2), Root Mean Square Error (RMSE), RMSEn values. The present study can be used to develop model for prediction of crop yield close to observed yields, with Adjusted R2 (0.62), RMSE (1.03) and RMSEn (12.01). The findings demonstrate that weather indices combined with PCA-MRL modelling offer a robust, operationally feasible approach to pre-harvest crop yield forecasting.