In extreme event attribution, which aims to answer whether and to what extent a particular extreme weather event can be attributed to global warming, the probability of an event is generally estimated through large ensemble simulations, using an atmospheric general circulation model (AGCM). In islands, such as Japan, it has been considered that surface air temperature (SAT) can be significantly affected by the surrounding sea surface temperature (SST), which mostly is affected by atmospheric circulation at mid- and high-latitudes. Therefore, the absence of SST responses to atmospheric variability in AGCMs impacts the estimation of the occurrence of extreme events, such as heat waves in Japan. In this study, we examined the impact of air–sea coupling on the probability of occurrence of severe heat waves that occurred in Japan in the summer of 2010 by analyzing the probability differences obtained from AGCM and coupled general circulation model (CGCM) large-ensemble experiments. The observed ocean temperature, salinity, and sea ice were assimilated in the 100-member CGCM experiments, as they were assigned as boundary conditions in the 100-member AGCM experiments. The SAT around Japan in the northern summer is largely related to the Bonin high, whose interannual variability is largely affected by the Silk Road and Pacific-Japan (PJ) pattern teleconnections in the real world. The SAT anomaly over Japan was related to the pressure variability due to the Silk Road and PJ patterns in the CGCM experiment. By contrast, the SAT over Japan simulated by AGCM was less sensitive to such pressure variability, and the SAT ensemble spread became narrower in AGCM. The results suggest that the probability of occurrence of the 2010 heat wave in Japan would tend to be underestimated by the AGCM ensemble compared to the CGCM ensemble, provided that the ensemble averages of the SAT anomalies were equal between CGCM and AGCM experiments. This study raised the issue of the absence of SST response to atmospheric variability in AGCMs, which can critically impact the estimation of extreme event probability, particularly in mid-latitude islands, such as Japan.

在极端事件归因（旨在回答特定的极端天气事件是否可以归因于全球变暖的程度以及在何种程度上归因于气候变暖）中，通常使用大气总循环模型（AGCM）通过大型整体模拟估算事件的概率。在日本这样的岛屿上，人们认为地表气温（SAT）可能会受到周围海面温度（SST）的显着影响，而海面温度（SST）主要受中高纬度大气环流的影响。因此，AGCM中没有SST对大气变化的响应，影响了对极端事件（例如日本的热浪）的发生的估计。在这个研究中，我们通过分析从AGCM和耦合总体循环模型（CGCM）大集合实验获得的概率差异，研究了海海耦合对2010年夏季日本发生严重热浪的可能性的影响。在100成员CGCM实验中，将观测到的海洋温度，盐度和海冰同化，因为在100成员AGCM实验中将它们指定为边界条件。北部夏季在日本周围的SAT很大程度上与Bonin高有关，其年际变化很大程度上受现实世界中的丝绸之路和日本太平洋（PJ）模式遥相关。日本的SAT异常与CGCM实验中的丝绸之路和PJ模式导致的压力变化有关。相比之下，AGCM模拟的日本SAT对这样的压力变化不太敏感，并且AGCM中的SAT集合传播变窄。结果表明，如果CGCM和AGCM实验之间SAT异常的平均平均值相等，则AGCM集合与CGCM集合相比，日本2010年热浪发生的可能性会被低估。这项研究提出了AGCM不存在SST对大气变化的响应的问题，这可能严重影响极端事件概率的估计，尤其是在日本等中纬度岛屿上。结果表明，如果CGCM和AGCM实验之间SAT异常的平均平均值相等，则AGCM集合与CGCM集合相比，日本2010年热浪发生的可能性会被低估。这项研究提出了AGCM不存在SST对大气变化的响应的问题，这可能严重影响极端事件概率的估计，尤其是在日本等中纬度岛屿上。结果表明，如果CGCM和AGCM实验的SAT异常的平均平均值相等，则AGCM集合与CGCM集合相比，日本2010年热浪发生的可能性会被低估。这项研究提出了AGCM中没有SST对大气变化的响应的问题，这可能严重影响极端事件概率的估计，尤其是在中纬度岛屿（例如日本）。