| 04 Mar 2026
Increasing heat stress across Southeast Asia driven by compound warming and moistening
Abstract. Heat stress is intensifying across Southeast Asia under global warming, yet the relative influences of atmospheric warming and moistening across timescales remain insufficiently quantified. This study investigates the thermodynamic drivers of warm-season (April–October) heat-stress intensity, measured by the daily maximum wet-bulb globe temperature (WBGTmax), and frequency, defined as the annual number of extreme heat-stress days (Nxday), across Southeast Asia and its 20 climatic sub-regions. Using observations together with dynamically downscaled CORDEX–SEA simulations, we apply a unified attribution framework to separate the effects of air temperature, specific humidity, and residual nonlinear processes on historical (1985–2014) trends, interannual variability, and projected late-century changes (2071–2100 relative to 1985–2014) under SSP5–8.5. Historical increases in WBGTmax and Nxday are dominated by temperature across much of the region. However, humidity already provides important amplification in monsoon-influenced lowlands, including Indochina, the Philippines, and parts of the Malay Peninsula. Future projections indicate a coherent basin-wide shift toward compound warm–humid conditions. In many monsoon regions, rising moisture contributes roughly 30–50 % of the increase, consistent with enhanced atmospheric water-holding capacity in a warmer climate. In contrast, interannual variability, particularly over the Maritime Continent, is strongly governed by nonlinear temperature–humidity interactions, which generate substantial unexplained components. Because Nxday depends on threshold exceedance, it shows a stronger amplification of future change than WBGTmax, even though its long-term evolution remains primarily temperature controlled. Overall, the results demonstrate that heat-stress escalation in Southeast Asia increasingly reflects rising atmospheric moist enthalpy rather than dry-bulb warming alone, underscoring the need for adaptation strategies that explicitly consider both temperature and humidity.
The author used three different types of attribution methods, namely trend decomposition, variability attribution, and mean-state change attribution, to quantify the relative contributions of air temperature and specific humidity to historical and projected heat stress across Southeast Asia. An interesting study, but there are a few substantial concerns that need to be addressed before acceptance.
1. The manuscript's entire attribution framework relies on an assumption that the contributions of air temperature (T) and specific humidity (Q) to wet-bulb globe temperature (WBGT) can be treated as linearly separable, additive components. This assumption fundamentally misunderstands the thermodynamic relationship between the two variables. In reality, the relationship between T, Q and moist thermodynamic quantities (WBGT in this context) is inherently non-additive. The physical relationship between WBGT and its drivers is nonlinear due to the saturation of evaporative cooling at high humidity and the nonlinear dependence of wet-bulb temperature on T and Q.
T and Q to WBGT is not as analogous to adiabatic and diabatic warming to the total warming, because adiabatic and diabatic contributions are genuinely additive science they both operate on the same variable (T) through physically distinct. See https://doi.org/10.1038/s41561-023-01126-1 and https://doi.org/10.1038/s41612-024-00797-w for details.
Therefore, a substantial portion of interannual variability is attributed to an "unexplained (residual)" term. The manuscript attributes this residual to "nonlinear temperature-humidity interactions" and "compound anomalies" without further analysis. Actually, the high residual fractions in interannual variability indicate that linear models are inadequate for capturing the true dynamics.
2. The author’s attempt to express contributions as percentages gives a false sense of precision in Figures 4, 7 and 10. If the author insists on empirical attribution, uncertainty levels and confidence intervals should be provided. The contributions depend on the specific model specifications and could change with alternative methodological choices (e.g., different regression models, inclusion of interaction terms, or alternative bias-correction methods).
3. The study explicitly states that wind speed and radiation are included in the WBGT calculation but are not attributed.