The contribution of weather to recent maize and wheat yield trends in the US: a comparison of two approaches

Evaluating the contribution of weather and its individual components (rain, temperature, radiation) to recent yield trends can be useful in predicting the response of crop production to future climate change. In this study, the Decision Support System for Agrotechnology Transfer (DSSAT) model was used to simulate maize and wheat yields in sites located in six of the top ten US producing states from 1984 to 2008. By varying weather variables individually, we evaluated the contribution of changes in temperature, radiation, and rain to simulated yield trends. A parallel analysis was conducted by regressing observed county-level yields against weather data and multiplying regression coefficients by observed weather trends to estimate weather contributions to empirical yield trends. When using DSSAT, weather had moderate and high contributions to simulated yield trends for maize and wheat, respectively, with rain having the largest effect. In contrast, when using empirical regressions, weather had a lower overall contribution, and rain was relatively less important to yield trends. Along with evidence from previous studies, this suggests that crop models such as DSSAT may be too sensitive to water, which could inflate the importance of rain as a driver of yield trends in simulation studies. The time period used to compute yield trends also had a large effect on the importance of weather and its individual components. These results highlight the importance of choosing a computational approach and a time period when estimating climate-related yield trends, and thus the need to consider these two sources of uncertainty.