Attribution of historical extreme heat events in different climate zones of China

Abstract China has a vast territory with diverse climates, including the arid, semi-arid, semi-humid and humid regions. Previous studies on extreme heat event attribution mainly focus on individual events in a specific region, with less attention paid to comparisons between historical events in different climate zones. Here, we use the number, seasonal length and intensity of hot days with daily maximum temperature exceeding 35 °C, to investigate human influence on extreme heat events in early and recent periods. It is clear that all three heat indicators have shown obvious increase across China since the early 1960s, with a rapid rise in recent years and the hottest event occurring in 2022. The Coupled Model Intercomparison Project Phase 6 models generally capture the temporal evolution of these indicators, but some biases exist. We utilize an annual cycle-based method to correct the model biases in climate state and then use the adjusted model results to conduct event attribution in different historical periods. We find that human influence has greatly increased the probability of recent events across all regions, while having no impact on early historical events. For the hottest 2022 event, the risk ratios for seasonal length of hot days in a few regions could not be estimated due to zero probability in the natural world, indicating that such events would not happen without human influence. In different climate zones, the risk ratios for all indicators in arid northwestern China exceed those in other regions when using consistent observational thresholds, indicating a greater response of extreme heat to anthropogenic forcing in this area. For the same event, attribution results of different indicators yield varying risk ratios, highlighting the importance of considering multiple indicators in event attribution. Additionally, model performance notably affects attribution results; without bias correction, human influence may be incorrectly estimated.