El Niño Enhances Exposure to Humid Heat Extremes With Regionally Varying Impacts During Eastern Versus Central Pacific Events Abstract Humid heat extremes, characterized by high wet bulb temperature (Tw), pose significant health risks. While strong El Niño events are known to affect the frequency of extreme Tw days, the distinct impacts of Central Pacific (CP) and Eastern Pacific (EP) El Niño events remain understudied. Using a 12‐member CMIP6 ensemble at discrete global warming targets (+1.5, 2, 3, 4°C), this study shows progressively enhanced humid heat extent during EP events primarily in Mainland Southeast Asia, while South Asia experiences regionally opposing effects from EP and CP events. EP and CP events drive distinctly different, regionally varying patterns of dangerous Tw, yet both significantly increase the affected population and area impacted by humid heat extremes at all global warming levels. This amplification surpasses the impact of an additional degree of global warming, highlighting El Niño's compounding effect on heat stress threats across warmer climates. Plain Language Summary Humid heat extremes, characterized by high wet bulb temperature (Tw), pose health risks even to young, healthy individuals. While strong El Niño events are known to affect extreme Tw days, the impact of different El Niño types (Central Pacific and Eastern Pacific) has not been well studied. Using historical data and future climate projections, we examined how these El Niño types affect the frequency and spatial extent of dangerous Tw. Our analysis shows that under future warming, Eastern Pacific and Central Pacific El Niño events drive distinctly different, regionally varying patterns of dangerous Tw, yet both significantly increase the affected population and area impacted by humid heat extremes at all global warming levels. Even at low global warming levels, during El Niño events, the population exposed to dangerous Tw is expected to be equal to that exposed regularly when the mean warming is more than four times higher. This highlights the need to consider El Niño diversity in assessing the additional heat stress in heavily populated regions as the planet warms and approaches the critical threshold of heat stress. 1. Introduction Heat‐related fatalities are a leading cause of death during extreme weather events worldwide (Barriopedro et al., 2011; Buzan & Huber, 2020; Buzan et al., 2015). Elevated environmental humidity further exacerbates heat stress by diminishing the effectiveness of evaporative cooling, the primary mechanism through which humans regulate their internal body temperature (Buzan & Huber, 2020; Coffel et al., 2018; Foster et al., 2021; Im et al., 2017; Raymond et al., 2020; Schär, 2016; Sherwood & Huber, 2010). Humid heat extremes are charac- terized by high wet bulb temperature (Tw), a measure of the moist enthalpy of the atmosphere (Song et al., 2022). Until recently, most studies used a critical Tw threshold of 35°C which was proposed as the theoretical upper bound of human thermal tolerance, above which the human body cannot dissipate metabolic heat through sensible or latent cooling (Sherwood & Huber, 2010). However, recent findings indicate a lower threshold, dependent on local temperature and humidity conditions, as being more accurate (Vanos et al., 2023; Vecellio et al., 2022, 2024). In high‐humidity environments, such as the tropics, exposure to a Tw of 30.6°C for 6 hr poses risks of hyperthermia, heat stroke, or death, even for young, healthy individuals performing minimal metabolic workloads under ideal conditions, that is, full shade, drinking water, minimal exertion (Vecellio et al., 2022). Both the mean moist enthalpy and humid heat extremes are projected to increase in the tropics, spatially coinciding with the fastest projected population growth, underscoring the importance of understanding exposure to Tw (Matthews, 2018). Superposed on the warming trends, natural climate variability in the tropics, dominated by the El Niño Southern Oscillation (ENSO) phenomenon, modulates Tw extremes. Previous studies focused on ENSO impacts on Tw under current climate conditions and found that extreme Tw days are twice as likely during very strong El Niño years compared to neutral years and thrice as likely compared to La Niña years worldwide (Speizer et al., 2022) with the most intense El Niño events resulting in the largest annual average land area affected by extreme Tw days (Rogers et al., 2021). However, the so‐called ENSO diversity (Capotondi et al., 2020), that is, the different patterns of SST anomalies during El Niño events whereby the peak SST anomalies are either concentrated in the Eastern Pacific (EP) or Central Pacific (CP) have distinct impacts on extreme events (McKenna & Karamper- idou, 2023) and the climate worldwide (Ashok et al., 2007; Capotondi et al., 2020). In this study, we evaluate the escalating impact of the additional heat load imposed by El Niño flavors (EP vs. CP events) in the tropics as the planet approaches the critical threshold of heat stress under discrete global warming targets (GWT) (+1.5, 2, 3, 4°C) using the CMIP6 Tw data set published by Vecellio et al. (2023). We thus address two significant gaps in the literature of Tw extremes in a changing climate: we quantify the superposed impact of El Niño variability on the global warming trend, and we distinguish between the two El Niño flavors, showing the significance of considering their distinct impacts for regional Tw extremes. 2. Data and Methods 2.1. Data To compute the impact of El Niño flavors on Tw extremes under future climate change scenarios, we use the Tw data set published in Vecellio et al. (2023), who used 2 m temperature, specific humidity, and surface pressure from 12 CMIP6 coupled general circulation models (GCMs) at 3‐hourly resolution to calculate Tw under future discrete GWT using the Davies‐Jones adiabat Tw approach (Davies‐Jones, 2008). The calculated Tw values were further bias‐corrected following a “pseudo‐global warming” approach. First, Tw was calculated for both the ERA5 reanalysis (Hersbach et al., 2023) and each CMIP6 model in the baseline period (1950–1976) that precedes the acceleration of global warming. CMIP6 Tw was also calculated for target years representing 1.5, 2, 3, and 4°C of global warming under the Shared Socioeconomic Pathway (SSP)585 scenario. The SSP585 scenario was chosen to maximize available warming targets, and the warming level‐oriented approach makes the results robust to the choice of emission scenarios (Seneviratne & Hauser, 2020; Senevir- atne et al., 2016). Then, the differences in 3‐hourly Tw between the average year of the baseline period and that of years representing a certain warming target were computed for each CMIP6 model and bilinearly interpolated to the ERA5 horizontal grid. The CMIP6‐simulated temporal changes in Tw were finally added to the ERA5 Tw for each year within the baseline, providing bias‐corrected estimates of Tw under each GWT. This approach captures seasonal and diurnal variations in Tw changes with global warming, considering potential changes in background circulation attributed to anthropogenic warming. It preserves climate change signals from CMIP6 models while correcting biases in baseline climate state and provides adequate resolution for capturing geographic features and physical processes important for accurate Tw calculations (Freychet et al., 2022; Im et al., 2017; Vecellio et al., 2023). However, the pseudo‐global warming approach imposes background warming on historical extreme Tw patterns during El Niño years (Figure S1 in Supporting Information S1), and thus it does not account for changes in El Niño teleconnections, which would alter the response of Tw extremes to El Niño events, or changes in El Niño diversity itself due to global warming. The implications of this methodological approach for interpreting the results are further addressed in Section 4. 2.2. El Niño Diversity Indices To characterize EP and CP El Niño years, we use the E and C indices proposed by Takahashi et al. (2011). An El Niño year spans from June (Year 0) to May of the following year (Year +1), and EP and CP events are defined when the DJF E and C indices exceed one standard deviation, respectively (Figure S2 in Supporting Informa- tion S1). During the period 1950–1976 used in Section 3, we consider 4 EP events (1951, 1957, 1965, 1972) and 3 CP events (1958, 1963, 1968) which agrees with the consensus in prior studies (e.g., Capotondi et al., 2020; McKenna & Karamperidou, 2023). During the selected EP and CP events, the average June–May global mean temperature anomaly is 0.07 and 0.09°C, respectively. Therefore, the events do not contribute to an average global warming that exceeds any of the GWTs considered for our future scenario analysis. 2.3. Humid Heat Metric Calculations For the projected impacts of El Niño diversity (Section 3), we calculate the number of dangerous Tw (dTw) days as the number of days at each grid cell in which the daily maximum Tw exceeds a threshold of 31°C. The de- parture from the number of days above the threshold under climatological conditions is used to quantify the projected El Niño impact and is hereafter referred to as anomalous dTw days. On a day when the temperature reaches this threshold, an individual would face prolonged exposure to hazardous Tw values, posing a severe threat to their health and contributing to the loss of a working day (Buzan & Huber, 2020), impacting the region socially and economically. Significance testing was conducted using a “bootstrapped” 1,000‐member ensemble of EP and CP El Niño composites comprised of random samples of equal size to the ones tested (4 and 3 years, respectively), selected with replacement from every GCM ensemble member under each GWT. Differences in dTw compared to climatology were considered significant if they fell above (below) the 90th (10th) percentile of the difference values calculated from the 1,000‐member ensemble for more than half of the models (Figure S3 in Supporting Information S1). To assess whether the duration of dangerous Tw conditions during EP and CP years changes in the future, we compute the maximum number of consecutive (MC) days exceeding the 31°C threshold. To estimate changes in the spatial extent of the impacts, we calculate the total land area in square kilometers in five regions (West Africa, the Middle East, South Asia, East Asia, and Mainland Southeast Asia) and, for each region, the total land area experiencing dTw for at least 1 day during each El Niño year (Section 3.2). We conduct an independent t‐test across all models and El Niño events for each region to assess the statistical significance of differences in the total land area affected. To determine the population affected by dTw for at least 1 day during every El Niño year for each GWT, we use the population projections from the SSP2 scenario for the same five regions (Jones & O'Neill, 2016). SSP2 is a “middle‐of‐the‐road” scenario and is considered the most‐likely pathway. It assumes medium mortality, fertility, and migration at the country level and is compatible with all GWTs considered. Since the global population under SSP2 is projected to stabilize around 2040, we use the 2050 SSP2 population for all GWTs, consistent with Vecellio et al. (2023). This approach avoids additional complexities from population changes when quantifying the relative impact of El Niño events and additional global warming, which is the focus of this study, and is robust to the choice of population levels (cf. Figure 3; Figure S8 in Supporting Information S1). 3. Impacts of El Niño Diversity on Tw in Warmer Climates 3.1. Dangerous Tw Days Under future GWT conditions, we find that EP and CP El Niño events impact the frequency, duration, and spatial extent of dTw days in five regions: West Africa, the Middle East, South Asia, East Asia, and Mainland Southeast Asia (Figure 1; see Figure S4 in Supporting Information S1 for a global map and S5 for the relative change in the number of dTw days from one GWT to the next). In all regions, few to no days exceed the 31°C threshold under climatological or El Niño conditions for the first two GWTs (1.5, 2°C). However, consistent with findings by Vecellio et al. (2023), the number of dTw days increases substantially under climatological conditions for the latter two GWTs (3, 4°C), while our analysis additionally reveals that the role of El Niño in modulating the heat stress also becomes significant under these high GWTs. In the Middle East, Mainland Southeast Asia, and to a lesser extent East Asia, the climatological conditions result in few dTw days, even at a 4°C GWT. However, at the lower GWT of 3°C, El Niño events lead to a substantial increase in dTw days, underscoring El Niño's significant role in exceeding the critical physiological threshold. A similar yet weaker signal is found in the MC days exceeding the threshold, detailed in Supporting Information S1 (Figures S6 and S7). El Niño has the weakest impact in West Africa (Figure 1a), where a marginal increase is found for both flavors, adding approximately 8–12 dTw days to the weak climatological signal under a 4°C GWT. The change in MC days above the threshold follows similar patterns and increases by about 2–4 days (Figure S7a in Supporting Information S1). CP events primarily impact the northern portions of central Africa, increasing the MC days by up to a week. The primary impact of El Niño on dTw occurs in South Asia (Figure 1b). Under a GWT of 4°C, the entire Ganges River and Indus Valley regions experience an average of around 40 dTw days, with a minimum of nearly 30 days in the former and reaching a maximum of approximately 65 days in the latter. During EP years, the largest in- creases of dTw days (20–30 days) are found in the Ganges River delta, doubling the climatological signal. In contrast, in the Indus Valley, where the climatology reaches its peak impact, there is a decrease of 20–30 dTw days and nearly a 6–30 days shorter MC streak (Figure S7b in Supporting Information S1). During CP years, much of the Indus Valley and Ganges River regions experience an increase of 12–20 dTw days, while the Ganges River delta and the adjacent region of India have an increase of approximately 25–35 days. This dipole in the impact of the two El Niño flavors can be explained by the strong mitigating effect that EP events have on daily maximum Tw in the Indus Valley region, which does not allow the climatological increase of Tw under the high GWTs to result in exceedance of the 31°C threshold (Figure S1 in Supporting Information S1).

Wet-bulb globe temperature (WBGT)–a standard measure for workplace heat stress regulation–incorporates the complex, nonlinear interaction among temperature, humidity, wind and radiation. This complexity requires WBGT to be calculated iteratively following the recommended approach developed by Liljegren and colleagues. The need for iteration has limited the wide application of Liljegren’s approach, and stimulated various simplified WBGT approximations that do not require iteration but are potentially seriously biased. By carefully examining the self-nonlinearities in Liljegren’s model, we develop a zero-iteration analytic approximation of WBGT while maintaining sufficient accuracy and the physical basis of the original model. The new approximation slightly deviates from Liljegren’s full model—by less than 1oC in 99\% cases over 93\% of global land area. The annual mean and 75-99\% percentiles of WBGT are also well represented with biases within ±0.5oC globally. This approximation is clearly more accurate than other commonly used WBGT approximations. Physical intuition can be developed on the processes controlling WBGT variations from an energy balance perspective. This may provide a basis for applying WBGT to understanding the physical control of heat stress.

A robust estimation of how heat stress is changing worldwide is complicated by the variety of heat stress metrics in use. This study compares heat stress changes between 1979–2000 and 2001–2023 calculated from five commonly used metrics based on ERA5 reanalysis. We identify regions where all metrics indicate significant increases in heat stress, underscoring the need for urgent adaptation efforts. Conversely, we also find regions where metrics disagree, even on the direction of change. Substantial inter-metric spread in population heat exposure exceeds the spread from reanalysis choices and is comparable to the spread across an ensemble of CMIP6 climate models. Detailed local examinations suggest these inter-metric discrepancies arise from differing temperature-humidity relative weighting across metrics. Our findings highlight metric choice as a significant source of uncertainty in heat stress projections and emphasize the need for a better understanding of the suitability of different metrics for specific climate regimes and impacts.

Wet bulb globe temperature (WBGT) is a widely applied heat stress index. However, most applications of WBGT within the heat stress impacts literature do not use WBGT at all, but one of the ad hoc approximations, typically the simplified WBGT (sWBGT) or the environmental stress index (ESI). Surprisingly little is known about how well these approximations work for the global climate and climate change settings that they are being applied to. Here we assess the bias distribution as a function of temperature, humidity, wind speed and radiative conditions of both sWBGT and ESI relative to a well-validated, explicit physical model for WBGT developed by Liljegren, within an idealized context and the more realistic setting of ERA5 reanalysis data. sWBGT greatly overestimates heat stress in hot-humid areas. ESI has much smaller biases in the range of standard climatological conditions. However, both metrics may substantially underestimate extreme heat especially over subtropical dry regions. These systematic biases demonstrate that sWBGT-derived estimates of heat stress and its health and labor consequences are significantly overestimated over much of the world today. We recommend discontinuing the use of sWBGT. ESI may be acceptable for assessing average heat stress or integrated impact over a long period like a year, but less suitable for health applications, extreme heat stress analysis, or as an operational index for heat warning, heatwave forecasting or guiding activity modification at workplace. Nevertheless, Liljegren’s approach should be preferred over these ad hoc approximations and we provide a Python implementation to encourage its widespread use.