Statistical modelling for forecasting of wheat yield based on weather variables

Abstract

Forecasting of crop yield based on historical data and pertinent external climatic information is considered. To this
end, Autoregressive Integrated Moving Average with Exogenous variables (ARIMAX) time-series model along with its estimation procedure is studied. In the present investigation, five models at five important stages of wheat growth are developed by including the most important weather variables. The weekly maximum temperature at crown root initiation (CRI) stage, tillering stage, anthesis stage, milk stage and dough stage and evapotranspiration at CRI stage are used for model development. As an illustration, ARIMAX models are employed for forecasting of wheat yield in Kanpur district of
Uttar Pradesh. Comparative study of the fitted models is carried out from the viewpoint of Relative mean absolute prediction error (RMAPE). It is demonstrated that the ARIMAX methodology is able to provide pre-harvest forecasts based on weather variables at various stages of wheat crop growth, starting from CRI stage (21 days after sowing) to dough stage (126 days after sowing). It is observed that, as wheat crop grows towards maturity, pre-harvest forecasts get closer to actual values.