Improving the validation of model-simulated crop yield response to climate change: An application to the EPIC model

Crop models have been used extensively to simulate yield response to various scenarios of climate change. Such simulations have been inadequately validated, limiting their utility in policy analysis. In this research, it is argued that the performance of crop models during recent years of extreme weather conditions relative to current normals may give a better indication of the validity of model simulations of crop yields in response to climate change than performance during the full range of climate conditions (as is done now). Twenty years of the climate record (1971-1990) are separated into different growing season temperature and precipitation classes (normal years, hot/cold extremes, wet/dry extremes) for 7 weather stations in eastern Nebraska, USA. The Erosion Productivity Impact Calculator (EPIC), a crop growth model, is used to simulate crop yields with each of the above weather classes. Statistical comparisons are made between simulated yields, observed yields and observed yields detrended of technology influences. Based on these comparisons, we conclude that EPIC reliably simulates drop yields under temperature extremes, some which approach the types of climate conditions that may become more frequent with climate change. Simulations with precipitation extremes are less reliable than with the temperature extremes but are argued still to be credible. Confidence in crop simulations during years mimicking climate warming scenarios appears warranted.