Climate variability and changes affect crop yields by causing climatic stresses during various stages of the plant life cycle. A crop growth model must be able to capture the observed relationships between crop yields and climate stresses before its credible use as a prediction tool. This study evaluated the ability of the geographically distributed cotton growth model redeveloped from GOSSYM in simulating U.S. cotton ( Gossypium hirsutum L.) yields and their responses to climate stresses during 1979 to 2005. Driven by realistic climate conditions, the model reproduced long-term mean cotton yields within 10% of observations at the 30-km model resolution across virtually the entire U.S. Cotton Belt and correctly captured the critical dependence of their geographic distributions on regional climate characteristics. Significant correlations between simulated and observed interannual variations were found across 87% of the total harvest grids. The model also faithfully represented the predictive role of July to August air temperature and August to September soil temperature anomalies on interannual cotton yield changes on unirrigated lands, with a similar but weaker predictive signal for irrigated lands as observed. The modeled cotton yields exhibited large, positive correlations with July to August leaf area index. These results indicate the model's ability to depict the regional impact of climate stresses on cotton yields and suggest the potential predictive value of satellite retrievals. They also provide a baseline reference for further model improvements and applications in the future study of climate-cotton interactions.
