Large humidity effects on urban heat exposure and cooling challenges under climate change

Many urban climates are characterized by increased temperature and decreased relative humidity, under climate change and compared to surrounding rural landscapes. The two trends have contrasting effects on human-perceived heat stress. However, their combined impact on urban humid heat and adaptation has remained largely unclear. Here, we use simulations from an earth system model to investigate how urbanization coupled with climate change affects urban humid heat stress, exposure, and adaptation. Our results show that urban humid heat will increase substantially across the globe by 3.1 °C by the end of the century under a high emission scenario. This projected trend is largely attributed to climate change-driven increases in specific humidity (1.8 °C), followed by air temperature (1.4 °C)—with urbanization impacts varying by location and of a smaller magnitude. Urban humid heat stress is projected to be concentrated in coastal, equatorial areas. At least 44% of the projected urban population in 2100, the equivalent of over 3 billion people worldwide, is projected to be living in an urban area with high humid heat stress. We show a critical, climate-driven dilemma between cooling efficacy and water limitation of urban greenery-based heat adaptation. Insights from our study emphasize the importance of using urban-explicit humid heat measures for more accurate assessments of urban heat exposure and invite careful evaluation of the feasibility of green infrastructure as a long-term cooling strategy.


Introduction
Heat stress has severe socioeconomic implications (Patz et al 2005, Anderson and Bell 2011, Hsiang et al 2017, Anderson et al 2018, especially in urban areas-where 85% of the global population is projected to be living by 2100 (OECD 2015). Urban surfaces generate local warming by modifying local biophysical processes (Oke 1973, Arnfield 2003, Stewart and Oke 2012, Zhao et al 2014. This urban-induced warmth adds to non-localized, greenhouse gas (GHG)-induced warming, is comparable in magnitude to background GHG warming (Grimmond 2007, Georgescu et al 2013, Zhao et al 2017, Krayenhoff et al 2018, and exacerbates extreme heat events (Li and Bou-Zeid 2013, Zhao et al 2018. Despite the critical societal impacts of urban heat stress, there is a considerable lack of urban-specific and globally-consistent analysis on heat stress in both observational- (Hirschi et al 2011, Mora et al 2017, Tuholske et al 2021 and modeling-based (Hayhoe et al 2004, Meehl and Tebaldi 2004, Beniston et al 2007 research. This is mainly because of two reasons. First, globally and compared to other landscapes, urban areas are substantially underrepresented in ground-based observations (Ren et al 2014, Herrera et al 2017. 'Urban' weather stations, which are already sparse, are not representative of the urban 3D characteristics and built environments. Most of these stations are located at airports or in suburban areas; others, even if located around urbanized land (Sun et al 2016, Jiang et al 2020, are biased against built environments due to the World Meteorological Organization's standards for weather station placement (Hamdi et al 2020, WMO 2021. The standards require weather stations to be situated in areas of open, clear-cut land covered by grass and distant from built structures. The resulting data gap makes nearly all station-based datasets highly biased toward rural landscapes (Muller et al 2013, Azevedo et al 2016, Droste et al 2017. Second, current Global Circulation Models and Earth System Models (ESMs) almost universally lack physical representation of urban areas (Best 2006, Oleson et al 2008. These models essentially simulate 'non-urban' surface climates, regardless of whether the locations (model grid cells) have an urban landscape or not in reality. Therefore, urban heat stress, exposure, and impacts derived from these conventional datasets, observational or modeled, are highly biased .
Previous studies primarily used air temperature to measure urban heat, or namely, urban dry heat. However, humid heat (i.e. temperature with humidity factored in) has increasingly been recognized as being more indicative of the human experience of heat (Sherwood and Huber 2010, Oleson et al 2015, Kjellstrom et al 2016, Sherwood 2018) than air temperature alone. This recognition has motivated growing research into human-perceived heat stress using humidity-temperature combinations (Lobell et al 2008, Fischer et al 2012, Dunne et al 2013, Oleson et al 2015. Cities are often hotter (Oke 1982, Zhou et al 2004, Imhoff et al 2010, Peng et al 2012 and drier (Lokoshchenko 2017, Morris et al 2017, Mahmoud and Gan 2018) than their surrounding environments. In the future, urbanization coupled with climate change is projected to lead to warmer and drier urban climates (Huang et al 2019. However, the resulting impacts on urban humid heat are unclear, given the nonlinear interactions of urbanization and climate change (McCarthy et al 2010) and the contrasting effects of urban warming and drying on humid heat. These trends and interactions have raised an important yet unresolved question: would urban drying counter the combined heating from climate change and urban heat island, thus potentially avoiding dangerous levels of urban humid heat under climate change?
Here we use a global ESM and an urban population projection to assess spatiotemporal variation of, projected exposure to, and adaptation to future urban humid heat stress under one of the most plausible Shared Socioeconomic Pathway-Representative Concentration Pathway scenarios. Urban parameterization within global ESMs is a new approach that allows for dynamical interaction between small-scale urban and large-scale climate processes through a globallyconsistent methodology. We further propose a simple proxy of urban green infrastructure (UGI) cooling efficacy based on urban 2 m wet-bulb temperature (T W ) and dry-bulb 2 m air temperature (T A ) to evaluate the cooling potential and feasibility of urban greenery-based heat adaptation. The specific objectives of this study are: (i) to assess future changes in, exposure to, and drivers of urban humid heat stress and (ii) to characterize the climate-driven cooling challenges of urban greenery as an urban climate adaptation strategy.

Model simulation
We used the Community Earth System Model version 2 (CESM2) (Danabasoglu et al 2020) to simulate future global patterns of urban humid heat stress and adaptation under a changing climate. We conducted two offline (land-only) simulations: one historical run from 1850 to 2014 and one future projection from 2015 to 2100 under SSP3-7 (a high-emission and one of the most plausible climate change scenarios defined by the Intergovernmental Panel on Climate Change, IPCC). The historical simulation was driven by a careful reconstruction of the observed climatology-Global Soil Wetness Project Phase 3 (GSWP3v1) data (Lawrence et al 2019). The future climate simulation utilized an 'Anomaly Forcing' method supported by CESM2, where the monthly anomalies obtained from a three-member ensemble of fully coupled future projection runs are added on top of a repeating 10 year cycle (2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014) of the high-frequency GSWP3 forcing. This configuration essentially serves as a bias-corrected future climatology forcing run. Both simulations were conducted at a spatial resolution of 0.9 • latitude × 1.25 • longitude globally.
We used the simulation results of the first (2000-2009) and last (2091-2100) decades of the 21st century to represent the present-day condition and a projected future climate, respectively. By comparing the first and last decades, we highlight the long-term effects of climate change on urban climate. Because our climate simulations for the entire time range (2000-2100) are documented and publicly accessible (see data availability), urban projections for any timescales of interest (such as mid-century or other time frames) can be obtained from our data. We utilize decadal seasonal means for June-July-August (JJA) as the temporal scale of analysis. The atmospheric variables analyzed are the monthly mean 2 m air, or dry-bulb, temperature (T A ), monthly mean 2 m relative humidity (R h ), monthly mean precipitation, and 2 m wet-bulb temperature (T W ) (Stull 2011) output from the two offline CESM simulations conducted. We leveraged the CESM model structure to extract both urban and rural climatic values for all variables except for precipitation, which only has gridcell mean output. In CESM, 'rural' landscape refers to the area-weighted average of all non-urban land covers in each grid cell. Please see the supplementary information for more details on model simulation.

Spatially explicit population projection
We utilized a global, spatially-explicit urban population projection (Jones and O'Neill 2016) under SSP3-7, the scenario consistent with the CESM simulations analyzed herein, to estimate future urban population exposure to high humid heat. This dataset can be accessed at the following website: www.cgd.ucar. edu/sections/iam/modeling/spatial-population. We paired the urban 1/8-degree projections for the base data for 2000 and for the 2090s with the urban climate projections for the first and last decades, respectively. For both periods, we then aggregated the urban population by country and by bounding boxes we defined. Shapefiles of world cities and countries were obtained from the Environmental Systems Research Institute's (Esri) ArcGIS Hub, with data sources from Arc World/GMI and Esri/Garmin, respectively.

Urban humid heat stress and its drivers 2.3.1. Urban wet-bulb temperature
Combined T A and humidity metrics have been suggested as a better predictor of heat stress impacts and a more accurate indicator of human-perceived heat stress than T A alone (Mora et al 2017, Raymond et al 2017. Among a range of such metrics (Epstein and Moran 2006, Buzan et al 2015, MartilMartilli et al 2020, T W is preferable. It is inherently tied to thermodynamic principles and is thus advantageous as a consistent and universal metric over large scales (Sherwood and Huber 2010, Pal and Eltahir 2016, Coffel et al 2018, Sherwood 2018. T W also provides an upper threshold to physiological human survivability (Sherwood andHuber 2010, Raymond et al 2020). Finally, it has the added advantage of serving as an indicator of the efficacy of evaporative cooling.
There are multiple empirical ways to estimate wet-bulb temperature. In this study, urban-specific T W was derived from relative humidity (R h ) and Critically, ideal behavioral, physiological, and climatic conditions that alleviate heat stress are rarely met. Thus, serious mortality and morbidity impacts are observed at T W values much lower than the 35 • C survivability limit proposed by (Sherwood and Huber 2010). It is commonly assumed that a healthy human could not survive outdoors at this threshold for over 6 h (Kang and Eltahir 2018), and lower thresholds around 31 • C have been proposed for physical labor (Liang et al 2011, Monteiro and Caballero 2019). Further, a recent experimental study found critical, short-term wet-bulb temperatures significantly lower than 35 • C, 25 • C-28 • C in dry-hot environments and 30 • C-31 • C in warm-humid environments (Vecellio et al 2021). Here, for illustrative purposes, we define decadal JJA mean T W > 26 • C as dangerous and reflective of potential chronic humid heat stress. For context, the combinations of air temperature and relative humidity that yield T W > 26 • C would be classified as 'Caution,' 'Extreme Caution,' or 'Dangerous,' using the United States' National Weather Service's Heat Index, another measure of combined heat and humidity (Smith et al 2013, Im et al 2017, Kang and Eltahir 2018.

Decomposition of projected urban humid heat changes
To elucidate the impact of drivers of projected urban humid heat, we decompose projected urban T W into contributions from changes in T A and Q (specific humidity), and further into contributions from local urbanization and GHG-induced global climate change. We estimate contributions from T A and Q by first differentiating equation (1) into its component parts (T A and R h ), then using the relationship between R h , T A , and Q (based on the definition of relative humidity and the Clausius-Clapeyron relation) to further decompose R h into T A and Q contributions.
The decomposed equation is: Then, using equation (2) (see the supplementary information for more details on the derivation) with the simulation results, we are able to attribute the increased urban humid heat into the contributions of changes in T A due to urbanization and climate change, and changes in Q due to urbanization and climate change. Here, we use 'urbanization' to refer to the impact of urban biophysical processes, approximated as the urban-rural difference in climatic variables, rather than growth of urban extent. This is because urban land is fixed in this version of CESM. Thus, our static urban simulations are a lower bound of urban humid heat exposure. In CESM, the 'rural' climate of each grid cell refers to the areaweighted climate of all non-urban land in that grid cell, rather than any specific rural land cover types (Wang et al 2017, Jiang et al 2019. This serves to highlight the overall impact of urbanization, using the area-weighted background climate as a reference.

Urban green cooling efficacy
To assess the impacts of local urban climate on the efficacy of green infrastructure at cooling, we suggest a new metric referred to as Urban Green Cooling Efficacy (UGC), defined as follows: Since T W represents the lowest air temperature that can be attained by a water-sufficient surface evaporating at its potential rate at constant air pressure, the difference between T A and T W reflects the maximum evaporative cooling potential and can be used as a proxy for cooling efficacy of urban greenery. The greater this difference, the greater the efficacy of evaporative cooling. A UGC value of zero indicates that T A is equal to T W , and evaporative cooling will theoretically be unable to achieve any heat mitigation. Higher UGC values reflect low R h and high T A , a combination that increases latent heat flux and cooling efficacy, given sufficient water is available to evaporate. Lower UGC values reflect high R h and low T A , which decrease evaporative cooling efficacy.
A certain UGC value indicates the theoretical maximum temperature reduction achievable purely through evaporative cooling, although it may not be achievable purely through passive evaporation. We note that it is beyond the scope of this paper to suggest a numerical relation between UGC and the quantitative efficiency of cooling achievable. Nonetheless, we propose that UGC is useful as a simple and universally quantifiable metric that reflects the impact of urban climatic conditions on vegetative evaporative cooling efficacy. Specifically, the thermodynamic basis of UGC enables it to provide the upper bound to evaporative cooling potential as well as a general indication of cooling efficacy. Here, UGC is derived from uniquely urban climatic variable outputs from CESM and reflects both urbanization and climate change and their interactions.

Future dangerous urban humid heat stress and exposure
Our results show elevated urban humid heat in nearly all urban areas globally under a high emission climate change scenario by the end of the century (figure 1). The global mean urban T W is projected to increase by 3.1 • C by the end of the century under SSP3-7. Coastal Central America, Southern and Southeast Asia, and Western Africa exhibit the most pronounced high urban humid heat (figure 1(e)). Notably, when compared with urban air temperature (T A ), future hotspots of T W overlap with those of T A but nonetheless reveal a markedly different spatial distribution (figures 1(b) and (e)). T W hotspots are shifted toward regions that are close to the equator, experience high humidity, and are near coastlinescorroborating spatial patterns found by studies on present and future humid heat stress (Li et al 2020, Raymond et al 2020). This indicates that the traditional T A approach may misrepresent the spatial distribution of where heat mitigation is most urgently needed, by underemphasizing regions with high combined heat and humidity. We suggest that humid adaptation, the usage of cooling measures that explicitly focus on combined heat and humidity, will more effectively reduce human-perceived heat stress.
Globally, we project an over fivefold increase in areal exposure to dangerous urban humid heat (decadal mean JJA T W > 26 • C) under SSP3-7, from 4.7% of urban grid cells exposed in the 2000s to 26.7% in the 2090s. Urban areas with the greatest projected humid heat will experience it during almost every month of the JJA season ( figure 1(e)). This is concerning as chronic heat stress is linked with significant negative health, well-being and productivity effects (Kjellstrom et al 2009, Oppermann et al 2021, Ramsay et al 2021. Spatially explicit projections of the intersections of urban populations with dangerous humid heat are crucial for informing effective, locally-tailored heat mitigation strategies. We find that at least 44.6% of the projected urban population will be living in urban areas with dangerous humid heat stress in the 2090s, an over three-fold increase from 13.4% in the 2000s. Our estimates of both current and projected urban humid heat stress exposure are lower and more concentrated in the low latitudes than a study estimating recent exposure increases (Tuholske et al 2021)-a difference that shrinks when we use a lower threshold of exposure. Notably, they focus on daily extremes and use non-urban climate, different humid heat metrics, and finer spatial resolution. We find that dangerously exposed urban areas are concentrated in the coastal tropics-where T W is high and cities are growing rapidly (McGranahan et al 2007, Seto et al 2011, Mora et al 2017. Of the 81 countries with urban citizens exposed to humid heat stress, 31 will have over 95% of their urban citizens exposed, equivalent to 1.1 billion people ( figure 2(b)). These findings suggest a tendency toward dangerous humid heat in areas where urban populations are concentrated.
Increased humid heat and urban population growth are projected to result in a perfect storm of climate hazard and exposure. Globally, the growth   of urban humid heat exposure far outpaces that of the total urban population by the 2090s. We note five 'hotspot' regions with particularly large urban populations susceptible to dangerous humid heat: Central America/Caribbean, Western Africa, Middle East, Southern Asia, and Southeast Asia ( figure 1(e)). These regions are all either projected to experience a severalfold increase in exposure (Middle East, Central America/Caribbean) and/or have over 75% of the urban population exposed by the end of the century (Western Africa, Southern Asia, Southeast Asia). In Western Africa, where total urban population growth is the fastest out of all five regions, urban humid heat exposure increases from 0 to 800 million urban residents (figure 3)-from 0% to 95% exposure. The Central America/Caribbean and Middle East regions similarly begin the century with a low percentage of exposure, and end it with 36% and 28% exposure, respectively. In the Middle East region, total urban population increases by a factor of 7, whereas urban population exposed to dangerous humid heat increases by a factor of 42. This rapid growth in exposure, in the absence of sufficient cooling measures, will severely limit the livability of cities. Similarly rapid growth in exposure is projected for Southern and Southeast Asia, but builds on 30% and 39% presentday exposure, respectively. Across the five hotspots, we find large increases of humid heat exposure in coastal cities. This intersection of urban population growth and humid heat stress is particularly concerning given the unique vulnerabilities of coastal areas to climate change (Neumann et al 2015).

Decomposition into drivers of increased humid heat stress
We find that the near-universal elevated urban humid heat stress is largely driven by climate change and the substantial resulting specific humidity increases (figure 4). Specifically, averaged globally, climate change-driven contributions to overall urban T W are approximately 1.4 • C and 1.8 • C from increases in air temperature and specific humidity, respectively. Urbanization has a negligible effect on global urban T W changes, with increased air temperature contributing 0.2 • C and decreased specific humidity contributing −0.2 • C. Importantly, our results demonstrate a consistent, dominating effect of broad-scale, GHGdriven humidifying and warming on elevated urban T W . This further shows that local humid heat adaptation measures are thus insufficient to counter humanperceived urban heat stress in the absence of simultaneous global climate change mitigation (Zhao et al 2017, Krayenhoff et al 2018, Zhao 2018. Our decomposition underscores the critical importance of mitigating non-local GHG emissions, in addition to urban interventions, for local urban climate benefits. The physical mechanisms driving changes in air temperature and specific humidity vary by both urban characteristics and background climate, which also interact and result in varying magnitudes of urban humid heat increases. Both air temperature and specific humidity increases under climate change are mainly due to GHG emissions, with the specific humidity increases due to the ability of warmer air to hold more moisture. Air temperature increases from urbanization can be attributed to a variety of factors, including reduction of evaporative cooling, anthropogenic heat, building materials, and albedo differences (Zhao et al 2014). Specific humidity decreases from urbanization are largely due to locally reduced surface evapotranspiration that comes from removal of vegetation. The range in relative contributions of air temperature and specific humidity changes from climate change and urbanization is apparent in the five hotspot regions defined, which are all in the tropics and have significant urban settlements along coastlines. The substantial impact of specific humidity increases is especially striking in these regions compared to the global average . In Southern and Southeast Asia, the contributions of climate change-driven specific humidity are over twice as large as those of climate change-driven air temperature. The impacts of urbanization vary-air temperature contributions are consistently around 0.1 • C-0.2 • C, whereas specific humidity contributions range from 0 to −0.7 • C-patterns driven by variation in both urban characteristics and background climates. These results highlight the importance of urban-rural differences-especially for understanding the physical mechanisms affecting urban climate, but not necessarily directly relevant to urban heat stress and mitigation itself (Martilli et al 2020, Krayenhoff et al 2021).

Climate-driven feasibility-efficacy tradeoff of urban greenery
The projected dangerous urban humid heat and exposure illustrated above call for urgent and effective urban humid adaptation. Urban green infrastructure (UGI) has been widely proposed (and implemented in certain regions) as an effective heat adaptation measure across scales (Alexandri and Jones 2008, Bowler et al 2010, Krayenhoff et al 2021. Recent studies assessing the feasibility and benefits of UGI found climate-driven trade-offs between hydrological retention and cooling potential benefits of UGI (Cuthbert et al 2022) and reviewed research on the heat-water trade-off for irrigation of UGI (Wang 2021). These studies suggest that UGI is not a one-size-fits-all solution. Here, we evaluated the cooling efficacy of urban greenery using the proposed indicator-Urban Green Cooling Efficacy (UGC). Previous studies have shown that outdoor water needs (mainly for watering greenery) correlate strongly with precipitation (Grimmond and Oke 1986, Gutzler and Nims 2005, Oke et al 2017; long-term average precipitation could thus be used as a proxy for urban greenery irrigation needs. Therefore, we propose to situate cities (urban grid cells) in a cooling efficacy-water needs space, composed of these two simple indicators, to elucidate the climate-driven feasibility of UGI on the global scale.
Our results reveal a globally consistent pattern of trade-offs between water availability and cooling efficacy of urban greenery (figure 5). Results clearly show an exponential decay of UGC with the increase of long-term mean precipitation. This can be understood from a heat mitigation perspective by broadly categorizing global cities into two regimes: waterlimited and efficacy-limited regimes. In the waterlimited regime, high irrigation needs may limit the cooling supported by high cooling efficacy. In contrast, the efficacy-limited regime characterizes conditions where low cooling efficacy constrains the cooling supported by low irrigation needs. This cautions previous assessments of urban cooling strategies, as most numerical modeling studies of UGI assume a sufficient (or even unlimited) water supply for evapotranspiration (Georgescu et al 2014, Krayenhoff et al 2021. This assumption may be unreasonable in dry climates and/or water-scarce urban areas (Nouri et al 2019). For example, irrigation is necessary all year round in some urban climates. Cooling effects may be overestimated if available water resources are unable to support urban greenery. While the schemes identified are particularly salient for urban grid cells in the top-left and bottom-right corners of the irrigationefficacy space (figure 5), most urban grid cells will tend toward one regime or the other. This result highlights a critical climate-driven dilemma of UGI-based heat intervention in practice: cities with ample atmospheric water supply to sustain UGI can be cooled only marginally through evapotranspiration; whereas cities where UGI could cool efficiently lack the atmospheric water supply to support the cooling.

Implications for long-term heat mitigation planning
Finally, we examine how climate change will affect this climate-driven feasibility-efficacy paradigm of urban greenery. By the 2090s, 22% of global urban grid cells are projected to have statistically significant changes (dependent student's t-test for paired samples, significant if p-value ⩽ 0.05) in both precipitation and UGC under SSP3-7-impacting longterm urban heat mitigation planning (figure 6). Of those urban grid cells, those with the highest future UGC are concentrated in Eastern Europe and scattered throughout areas of the Middle East, South Asia, and Western Africa. This is likely due to relatively low seasonal JJA humidity and high air temperatures, which drive the large difference between T W and T A (supplemental figure 1). A substantial majority (over 93%) of the 22% of urban grid cells will experience an increase in UGC (figure 6). However, of this percentage, over 83% will experience a decrease in precipitation. In cities where both UGC and precipitation increase, greater irrigation may not be required to support the higher cooling efficacy of urban greenery. Such benefits are especially likely for cities with low current UGC, such as Pucallpa (Peru) and Vientiane (Laos), as their percent increase in UGC will be higher. In cities with large increases in UGC and decreases in precipitation, such as Bamako (Mali) and Guadalajara (Mexico), greenery will increasingly effectively help cool urban areas but may require more irrigation. This is particularly concerning for cities in dry climates with low precipitation and high water needs. In such cases, a diverse cooling strategy portfolio (i.e. albedo modification, building material selection) that relies marginally on urban greenery may be more resilient and feasible. Our findings underscore the need to assess the impact of climate change and urbanization not only on urban heat stress, but also on urban heat mitigation strategies themselves. Comprehensive cooling plans that consider future placements of cities in the irrigation-efficacy space can more effectively and resiliently cool in changing urban climates.
Elevated dangerous urban heat stress and exposure call for not only innovative and effective heat mitigation solutions, but also for increased knowledge sharing between cities. The applicability of shared knowledge will be shaped by the selection criteria for peer cities. Our approach of situating cities in a UGI irrigation-efficacy space facilitates sharing urban heat adaptation knowledge across diverse urban contexts. Importantly, criteria such as physical proximity, culture, and country do not necessarily lead to closeness in irrigation needs and cooling efficacy. For example, Riyadh and Jeddah are two water-limited cities in Saudi Arabia. Riyadh is the capital and the most populated city, and Jeddah is the second largest city and the principal gateway to Mecca. From a heat mitigation standpoint, both cities are water-limited, with negligible average JJA precipitation at both the beginning and end of the century. Urban greenery in either city will require irrigation. However, Jeddah has projected humid heat stress that is 9.2 • C higher than Riyadh. Despite its higher humid heat, Jeddah has a lower UGC value of 6.6 • C, less than onethird of that of Riyadh. Although there are similarities in size, cultural and economic importance, and precipitation levels for these two cities, their cooling contexts present very different opportunities and challenges.
Cities in diverse geographic and cultural contexts may have similar placements in the irrigationefficacy space and thus serve as unexpected peer cities for sharing of UGI knowledge and research. For example, Hanoi (Vietnam) and Bucaramanga (Colombia) are two cities located in very different geographic and cultural contexts. Additionally, Hanoi has a monsoon-influenced humid subtropical climate and Bucaramanga has a tropical savanna climate (Beck et al 2018). However, they are in close proximity in the irrigation-efficacy space. Hanoi and Bucaramanga have projected future JJA T W of 30.4 • C and 30.0 • C, monthly precipitation of 278.8 mm and 264.6 mm, and UGC of 2.5 • C and 2.4 • C, respectively. Thus, although the two cities are different in language, culture, location, and size, they may face similar climate-driven challenges and opportunities for humid heat mitigation. The inclusion of urbanspecific peer selection criteria, such as precipitation and UGC, can be a valuable addition to heat mitigation efforts. This is especially important given the urgency of heat stress and the spatial heterogeneity of urban heat research. Through this approach, we can better utilize limited resources for heat mitigation and expand networks for worldwide collaboration and knowledge transfer between diverse cities. Cities' placement in this irrigation-efficacy space can supplement existing selection criteria for peer cities and strengthen the applicability of shared urban cooling knowledge.

Conclusions and implications
Urban areas are unique built environments where human populations interface with local urbanization and global climate change, resulting in climate-driven risks as well as opportunities for change. Our analysis highlights substantially elevated urban humid heat stress despite projected urban drying under climate change, and large intersections of humid heat stress and urban population growth-especially in the coastal tropics. Our decomposition of the drivers of urban humid heat increases reveals a considerably larger effect of humidification (driven by climate change) in coastal, tropical urban areas, compared to the global average. With this in mind, there is an especially urgent need for humid adaptation in tropical, coastal urban areas, which are also vulnerable to climate change risks such as higher sea levels, air temperatures, storm surges, and greater probability of flooding (IPCC 2021). Our results also emphasize the dominating impacts of climate change and highlight that local heat adaptation strategies are important but insufficient in the absence of climate change mitigation.
We propose an irrigation-efficacy space to support cooling assessments and knowledge sharing across diverse urban climates. Such urban assessments are currently limited by challenges such as lack of urban water supply data, global disparities in urban data collection and research (Cohen 2004, Grimmond 2006, Manley and Dennett 2019, and lack of comparable, explicit urban assessments. Precipitation and UGC are easily derived from climate change projections at the spatial coverage needed and can thus serve as the first step to support global, urban comparisons. Using this irrigation-efficacy space, we find globally consistent tradeoffs between water availability and cooling efficacy that characterize a general dilemma for UGI implementation. This space further reveals significant climate-driven changes in cooling context and that cooling solutions are impacted by the same climatic change driving the need for cooling. Thus, long-term cooling strategy development should take both local urbanization and global climate change into account. Our irrigation-efficacy framework can provide insights into opportunities and challenges for cooling and support development of urban cooling portfolios. Urban-explicit projections of future cooling contexts are crucial for improving the resiliency of urban cooling and the livability of cities in a changing climate. We note two main limitations to this study. First, the urban fraction and properties in each grid cell are time-invariant in the current version of the CESM, meaning future urban development and land expansion are not taken into consideration. Our results are projections of exposure over existing urban extent and are therefore likely to represent a lower bound of future exposure. Second, we consider seasonal decadal means rather than the hourly or daily timescales that characterize extreme heat events. Therefore, our results are more indicative of chronic heat stress than short-term heat waves. Next steps for this area of research include explicitly modeling the effects of future urban development and expansion, or assessing smaller (i.e. daily or hourly) time scales that can capture short-term extremes. Specifically, global climate models with explicit urban parameterization can facilitate research that improves our understanding of future urban climate-driven risks, and assessment of potential adaptation strategies. Finally, findings from this area of research have the potential to support urban knowledge transfer and climateconscious decision-making.

Data availability statement
The data that support the findings of this study are openly available at the following URL/DOI: https:// doi.org/10.13012/B2IDB-9627482_V1. The population data used in this study are available via www.cgd. ucar.edu/sections/iam/modeling/spatial-population.