Extreme heat has occurred more frequently in recent years and will intensify in the future, and this change has serious impacts on rice (Oryza sativa L.) yields. Thus, it was crucial to evaluate its influence on rice yield reductions. Recent papers have shown that a lack of experimental data makes it difficult for most crop models to capture the impacts of heat stress. Therefore, this paper explored how to improve the performance of crop models under extreme heat stress based on the Decision Support System for Agrotechnology Transfer (DSSAT) CERES-Rice model. This study primarily focused on (i) quantifying spikelet fertility based on daily temperature and durations derived from controlled experiments, (ii) improving the performance of the CERES-Rice model under extreme heat stress, and (iii) simulating historical and future rice yields using the improved model. Specifically, a meta-analysis method was utilized to build a new heat stress function between spikelet fertility and temperature and heat day duration with high realization. Subsequently, independent artificial controlled experiments at two sites were proposed to calibrate and validate the CERES-Rice model. The results showed a higher R² (> 0.739) and a lower RMSE that was reduced by 38~68% after incorporating the new heat stress function in the CERES-Rice model compared with that of the original model. Furthermore, a historical simulation (1980–2010) demonstrated that an improved CERES-Rice model could better capture rice yield in response to extreme heat. Using an ensemble of five climate model datasets and four Representative Concentration Pathways (RCPs), the analysis of the projected future (2020–2099) rice yield showed that the rice yield reduction caused by high temperature was considerable; however, the rice yields were overestimated by 34% and 18%, respectively, at the two sites. Some regions rarely affected by heat are likely to experience yield reductions in the future due to climate change.

近年来，极端高温的发生频率越来越高，将来还会加剧，这种变化对稻米（Oryza sativa L.）的收益。因此，评估其对水稻减产的影响至关重要。最近的论文表明，缺乏实验数据使得大多数农作物模型难以捕捉到热胁迫的影响。因此，本文基于农业技术转移决策支持系统（CESAT-Rice）模型，探索了如何改善极端高温胁迫下作物模型的性能。这项研究主要集中在（i）根据每日温度和受控实验得出的持续时间来量化小穗的繁殖力，（ii）在极端高温胁迫下改善CERES-Rice模型的性能，以及（iii）使用以下方法模拟历史和未来水稻产量改进的模型。特别，运用荟萃分析的方法在小穗受精率与温度和热天持续时间之间建立了新的热应激函数，具有很高的认识。随后，提出了两个站点的独立人工控制实验，以校准和验证CERES-Rice模型。结果显示较高与原始模型相比，在CERES-Rice模型中加入了新的热应力函数后，R ²（> 0.739）和更低的RMSE降低了38〜68％。此外，一项历史模拟（1980-2010年）表明，改进的CERES-Rice模型可以更好地捕获极热条件下的水稻产量。使用五个气候模型数据集和四个代表性浓度路径（RCP）的集合，对预计未来（2020-2099年）水稻产量的分析表明，高温造成的水稻产量下降是可观的；然而，两个地点的稻米产量分别高估了34％和18％。一些极少受热影响的地区，由于气候变化，将来可能会降低产量。