Global warming, significantly accelerating since the Industrial Era, poses escalating risks to both human society and the environment. The impacts of climate change vary considerably across regions due to the complex interplay of physical processes triggered by increased greenhouse gas forcing and inherent regional vulnerabilities. Many areas have already experienced temperature increases exceeding the Paris Agreement's most conservative limits, and extreme weather events, particularly heat extremes, have dramatically altered. Climate models project a further acceleration of these changes, emphasizing the urgent need for mitigation. Despite uncertainties in scenario projections, the robust and heterogeneous changes in extreme events necessitate a deeper understanding of their amplification mechanisms. This study focuses on the statistical properties of atmospheric temperatures to quantify how and to what extent these properties influence the future trajectories of heat extremes. The research hypothesis is that daily temperature variability is the primary driver of changes in the frequency and intensity of heat extremes, potentially exceeding the impact of background warming in many regions.

Existing research highlights the complex interaction between changes in higher-order moments of thermal distributions (e.g., variability, skewness, kurtosis) and the response of heat extremes to greenhouse gas forcing. Studies suggest that changes in higher-order moments can either amplify or dampen the effects of background warming, but the overall impact remains unclear. A decrease in mid-latitude temperature variability is projected for the coming decades, potentially contributing to a rapid weakening of cold-season cold extremes. Conversely, regional increases in variability are predicted to exacerbate hot extremes, particularly over tropical lands. However, limitations in historical data coverage and observational evidence of changes in higher-order moments have led to ongoing controversies in this area. Furthermore, the inherent non-Gaussian properties of atmospheric temperatures have shown significant regional influence on extreme event probabilities.

The study uses daily maximum (TX) and minimum (TN) temperature data from a multi-model ensemble of CMIP6 simulations under various Shared Socioeconomic Pathways (SSPs). The analysis focuses on changes in the magnitude of extreme temperature anomalies (hottest day TXx and coldest night TNn) and the changes in the exceedance probabilities of high-magnitude temperature anomalies. The authors investigate the relationship between changes in the extremes and changes in the leading moments of the temperature distributions, specifically the mean and standard deviation. They employ regression analysis to quantify the contribution of changes in the mean and standard deviation to changes in the extremes. A Gaussian approximation is used to analyze the scaling behavior of exceedance probabilities, considering the effects of both mean and standard deviation. The analysis considers seasonal variations and compares results for different regions and warming levels. Data from the Berkeley Earth dataset are used for a preliminary comparison with observational data.

The study reveals that heat extremes exhibit near-linear increases with global-mean surface air temperature (GSAT) across various SSPs. Changes over land are more pronounced than global averages. Regional and seasonal changes in extremes show significant heterogeneity, differing considerably from changes in mean conditions. The analysis finds that the changes in standard deviation (variability) of daily temperatures are strongly linked to the regional differences in warming rates of both hot and cold extremes. In mid-to-high latitudes, decreasing variability accelerates the warming of coldest nights while suppressing the warming of hottest days. Conversely, regional increases in variability over lower latitudes amplify the warming of hottest days. Changes in variability explain a substantial portion (over 90% in many cases) of the spatial variation in extreme warming rates. The analysis shows that variability plays a dominant role in determining future changes in the frequency of hot events, often surpassing the effects of background warming. Specifically, regions with lower inherent variability, such as the tropics, experience a disproportionate increase in hot days. The study demonstrates a non-linear relationship between GSAT and exceedance probabilities of hot events, with probabilities increasing more rapidly as GSAT rises, particularly in regions with low initial variability. A Gaussian approximation provides a good fit to the observed patterns of change in exceedance probabilities, highlighting the importance of native variability in shaping future frequencies of hot events. A preliminary comparison with historical observations reveals some discrepancies in model simulations of historical variability, suggesting potential underestimation of the increase in future hot event frequencies in certain regions.

The findings highlight the importance of considering temperature variability alongside background warming when projecting future changes in heat extremes. The dominant role of native variability in shaping the frequency of hot events underscores the need for accurate representation of historical variability in climate models. The non-linear relationship between GSAT and exceedance probabilities underscores the potential for a disproportionate increase in extreme heat events in regions with initially low variability, especially the tropics. The substantial contribution of changes in variability to the warming rates of extremes has important implications for regional climate risk assessment. The study also points to the importance of physical processes affecting variability, such as the weakening of the meridional gradient at midlatitudes due to Arctic amplification.

This research demonstrates the crucial role of daily temperature variability in determining the future intensification of heat extremes. Native variability, often overshadowing background warming, significantly impacts the frequency of hot events, especially in the tropics. Improving climate models' representation of historical variability is critical for more accurate predictions of future heat extremes. Future research should focus on refining model simulations of variability, exploring the contributions of higher-order moments at regional scales, and investigating the impacts of changing variability on various sectors and ecosystems.

The study relies on climate model outputs, which inherently contain uncertainties. The preliminary comparison with observational data is limited in scope, requiring more comprehensive analysis to fully validate the model projections. The Gaussian approximation used for exceedance probabilities simplifies the complex dynamics of temperature distributions, potentially overlooking subtle impacts of non-Gaussian behavior. The regional analysis relies on specific definitions of hotspot regions, and alternative delineations could yield slightly different results.