Non-stationary thermal time accumulation reduces the predictability of climate change effects on agriculture

Current modeling studies on the impacts of climate change on agriculture widely assume that thermal time accumulation of crops during the growing season remains constant under various climate conditions. However, in this study, a 20-year single rice variety, experimental dataset indicates that the thermal time accumulation for the entire growing season is not constant. As a result, a crop model based on constant thermal time accumulation significantly underestimates the observed phenological trend exhibited over the two decades of research-despite comparably accurate simulations of short periods. This deviation can result in misleading yield simulations, whereas the model simulations, using observed phenology data, show a similar yield trend as the observation. This study casts serious doubt on the assumptions of constant thermal time accumulation made in previous modeling studies, and, moreover, it highlights the critical requirements needed to improve phenology simulations on a larger scale so that predictions of the eventual yield trends due to climate change can more accurately reflect the results of yield trends in reality. (C) 2008 Elsevier B.V. All rights reserved.