Estimating non-additive within-season temperature effects on maize yields using Bayesian approaches.

Detrimental impacts of extreme heats on the U.S. crop yields have been well-documented by a number of empirical studies. However, less have focused on within-growing season weather variation and the interaction between temperature and precipitation. The objective of this study is to emphasize the importance of disaggregating temperature exposures within growing season. To achieve our objective, we estimate the impact of within-season monthly temperature and precipitation variations on maize yields in the U.S. corn belt region. We provide a discussion on variable selection methods in the context of estimating crop yield responses to climate variables. We find that the models that utilize within-growing season monthly variations performs better compared to the models with growing season aggregated weather variables and show the strength of Bayesian estimations. We also find that the warming impacts predicted by the models that utilize within-growing season variations are smaller than the predicted impacts of the models with aggregated weather variables. The findings indicate that the temperature effects are not additive across months within growing season.