Quantifying the Impacts of Compound Extremes on Agriculture
Statistical crop models have been used to both elucidate drivers of crop yield trends and variability, and to project the variation of crop production in the future. However, these models typically use seasonally averaged water availability metrics (e.g., total growing season precipitation), and utilize precipitation more often than soil moisture. Generally, if there is an expected change in the within-season distribution of soil moisture, using the mean variable will create biased yield projections. Because climate models project significant changes in the frequency and intensity of both extreme precipitation and temperature, the mean metrics of water availability – especially mean precipitation - are not sufficient to capture the impacts on yields. We employ a fine-scale, high temporal resolution dataset to investigate the conditional marginal value of soil moisture and heat in US corn yields for the 1981-2015 period employing a statistical framework.
This study serves to bridge the gap between statistical studies of the impacts of hydroclimatic extremes on crops and their biophysical counterparts by recognizing the central role of soil moisture in understanding crop yields. It identifies new water availability metrics that improve the predictive power of statistical corn yield models. The first key finding is that seasonal mean soil moisture performs better than average precipitation in statistically predicting corn yield. The major contribution of this study is showing that the estimated coefficient on yield response to extreme heat is significantly different while considering daily interactions with soil moisture, emphasizing the importance of compound hydroclimatic conditions. While predictive power is an important outcome of this analysis, the insights gained from incrementally adding higher temporal-resolution metrics of water extremes to the models are also valuable for understanding the drivers of corn yield variability, and for revealing the resolution of water availability data required to capture future extremes.
Agricultural production and food prices are affected by hydroclimatic extremes. There has been a large literature measuring the impacts of individual extreme events (heat stress or water stress) on agricultural and human systems. Yet, we lack a comprehensive understanding of the significance and the magnitude of the impacts of compound extremes. This study combines a fine-scale weather product with outputs of a hydrological model to construct functional metrics of individual and compound hydroclimatic extremes for agriculture. Then, a yield response function is estimated with individual and compound metrics focusing on corn in the United States during the 1981-2015 period. Supported by statistical evidence, the findings suggest that metrics of compound hydroclimatic extremes are better predictors of corn yield variations than metrics of individual extremes. The results also confirm that wet heat is more damaging than dry heat for corn. This study shows the average yield damage from heat stress has been up to four times more severe when combined with water stress.