The occurrence of very high temperatures (hot extremes) is often linked with negative impacts in human health, natural ecosystems and the economy (e.g., energy, water supply and agriculture). Studies have invariably shown that the intensity and frequency of hot extremes will increase in the future thus increasing their associated risks. While much progress has been made in quantifying and understanding hot temperature extremes and their future changes, there are still open questions. This paper focusses on the sources of hot extremes and their changes by applying a simple and unambiguous methodology that describes daily hot extremes as the superposition of four well known physical terms that include information on the annual mean temperature, the amplitude of the annual cycle, the diurnal temperature range and the local temperature anomaly on the day of the extreme. The methodology was applied to 30-year daily temperature records from 6 observation-based datasets and 31 atmosphere-ocean global climate models from the Coupled Model Intercomparison Project Phase 5 (CMIP5). The comparison between observed and simulated hot extremes shows a remarkably consistent picture where most CMIP5 models overestimate the term describing the local temperature extreme anomaly over most regions of the globe regardless of the observed dataset considered. Simultaneously, CMIP5 models show a systematic cold bias in the annual mean temperature and in the diurnal temperature range terms leading to substantial error compensation over some regions. This prompted us to define a new error estimator as the sum of errors in individual terms that appears to be much more effective at characterising model's performance compared to the traditional bias estimator. The assessment of future changes in hot extremes shows that changes are dominated by changes in annual mean temperatures with varying contributions from the other terms that strongly depend on the specific region being considered. Western Europe appears as a hot spot for extreme temperature changes (increases of $~$8 ^∘C by the end of the century) due to significant contributions from all decomposition terms including the summer mean anomaly, the diurnal temperature range and the daily extreme anomaly. Tropical South America also appears as a hot spot for extreme temperature changes (increases of $~$7 ^∘C) largely due to an increase in the daily extreme anomaly term (explaining about 30% of the full change) making this region one of the most sensitive regions in the world in terms of hot extremes. The analysis reveals that the separation of future changes according to terms describing mean, variability and tails is very sensitive to the specific way the mean component is defined including assumptions about stationarity.

高温（极端高温）的发生通常与对人类健康，自然生态系统和经济（例如，能源，供水和农业）的负面影响有关。研究始终表明，未来极端高温的强度和频率将增加，从而增加其相关风险。尽管在量化和了解高温极端及其未来变化方面已经取得了很大进展，但仍然存在未解决的问题。本文通过采用一种简单明确的方法，将极端高温的源头及其变化集中起来，将每日高温极端描述为四个众所周知的物理术语的叠加，其中包括年平均温度，年周期振幅，极端一天的昼夜温度范围和局部温度异常。将该方法应用于来自6个基于观测的数据集和来自耦合模型比较项目第5阶段（CMIP5）的31个大气海洋全球气候模型的30年每日温度记录。观测到的和模拟的极端高温之间的比较显示出非常一致的图景，其中大多数CMIP5模型都高估了描述地球上大多数地区的局部极端温度异常的术语，而不考虑所考虑的观测数据集。同时，CMIP5模型在年平均温度和昼夜温度范围方面显示出系统性的冷偏差，从而导致某些区域的误差补偿。这促使我们定义一个新的误差估计器，以单个术语表示的误差之和，与传统的偏差估计器相比，在表征模型性能方面似乎更为有效。对未来极端天气变化的评估表明，变化主要由年平均温度变化决定，而其他术语的贡献则在很大程度上取决于所考虑的特定区域。西欧似乎是极端温度变化的热点（对未来极端天气变化的评估表明，变化主要由年平均温度变化决定，而其他术语的贡献则在很大程度上取决于所考虑的特定区域。西欧似乎是极端温度变化的热点（对未来极端天气变化的评估表明，变化主要由年平均温度变化决定，而其他术语的贡献则在很大程度上取决于所考虑的特定区域。西欧似乎是极端温度变化的热点（$〜$8 ^∘下用本世纪由于来自所有分解术语包括夏季显著贡献的端部）的意思是异常，昼夜温度范围和每日极端异常。南美热带地区也出现了极端温度变化的热点（$〜$7 ^∘ C）主要是由于增加了每日极端异常术语（解释关于充满变化的30％），使得在世界上最敏感的区域的在热极端而言，这区域中的一个。分析表明，根据描述均值，变异性和尾部的术语来区分未来变化对均值成分的特定定义（包括平稳性假设）非常敏感。