Abstract

Changing climate has increased the vulnerability of agricultural systems. Due to the changing climate, Pakistan has experienced frequent heat waves and variability in monsoon rainfall which have affected the cropping patterns, soil health and water resources. For the last two decades, no significant improvement has been seen in the wheat yield. Many studies reported that climate change is negatively impacting the wheat yield in Pakistan. The joint effect of all these uncertainties leads to the problem of severe food security, poverty, and water crisis. Due to this, we face difficulty in predicting the yield and formulating our policies accordingly. Based on this narrative, we have developed a forecasting model for the wheat yield in Pakistan that considers scenarios of varying weather conditions and makes projections. Here in our analysis, we applied and compared three approaches: one is the conventional ARMA technique, the second one is the ARMA I-S, and the third and last one is the ARMA-X technique. Faisalabad district is selected for our analysis. We used times series data of 31 years from 1990 to 2021. The key variables of the study are yield, maximum and minimum temperature, average rainfall, relative humidity, sunshine hours and wind speed. We have divided the period into 3 different stages naming sowing, growing and harvesting for the analysis. In all three phases of the wheat crop, we found the wheat yield is influenced by the past year’s yield and past shocks with strong oscillations that persist over the years. In the sowing phase, ARMA-X model shows that among the weather variables, only minimum temperature and average rainfall are positively affecting the wheat yield. In the growing phase, we found that minimum temperature, average rainfall, relative humidity, and sunshine hours have a positive and significant impact on the yield. In the harvesting phase, minimum temperature and sunshine hours have a positive while maximum temperature and average rainfall have a significant negative impact on the wheat yield. After a detailed analysis of wheat yield using different forecasting techniques like ARMA, ARMA I-S and hybrid approach. We found that hybrid framework performed well in all three phases based on higher Log-likelihood, lower AIC and low variance of the residuals. It not only projects the yield based on the projected values of the weather variables which serve the purpose of scenario building but also shows that how the weather variables are affecting the wheat yield.