Reliability and input-data induced uncertainty of the EPIC model to estimate climate change impact on sorghum yields in the U.S. Great Plains

Crop simulation models are frequently used to estimate the impact of climate change on crop production. However, few studies have evaluated the model performance in ways that most researchers practiced in climate impact studies. In this article, we examined the reliability of the EPIC model in simulating grain sorghum (Sorghum bicolor (L.) Moench) yields in the U.S. Great Plains under different climate scenarios, namely in years with normal or extreme temperature and precipitation. We also investigated model uncertainties introduced by input data that are not site-specific but commonly used or available for climate change studies. Historical field trial data of sorghum at the Mead Experimental Center, NE, were used for model evaluations. The results showed that overall model reliability was about 56%. The mean absolute relative error (absRE) was about 29%. The degree of accuracy and reliability varied with climate-classes and nitrogen (N)-treatments. The largest bias occurred in drought years (RE = -25%) and the most unreliable results were found in N-0 treatment (reliability = 32%). There was more than 69% probability that input-data-induced uncertainties were limited to less than 20% of absRE. Our results support the application of the EPIC model to climate change impact studies in the U.S. Great Plains. However, efforts are needed to improve the accuracy in simulating crop responses to extreme water- and nitrogen-stressed conditions.