Rainfed agriculture in the Southeastern US is vulnerable to hydrological extremes, especially droughts. The nature of a drought event (duration, severity, intensity, and recurrence interval) has varying impact on crop yields. Drought characteristics and associated impacts are expected to change as the climate changes, thus threatening crop productivity and in turn local and state economies. Understanding the risks that drought of poses to agriculture is a critical step in mitigating the effect of climate change. In this study, we examine and compare the univariate and multivariate relationships between the recurrence period of different drought characteristics on the yield of four major crops, namely corn, cotton, peanuts, and soybeans. Using the theory of runs, we extract drought duration and severity, at the county level for North Carolina, South Carolina, and Georgia, from the Standardized Precipitation and Evapotranspiration Index (SPEI). We first investigate the linear relationship between these characteristics and crop yield variability. We then derive the multivariate return period of these drought events, using copula, and develop a model between the later and crop yield. Copulas are an effective way to model the dependence between the variables because they are distribution independent. We hypothesize that multivariate models can better explain crop yield losses compared to univariate models.
