Urban areas vary drastically in terms of climate conditions on a global scale. Some climate features can affect not only the health of urban dwellers but also their well-being and satisfaction levels. Taking these factors into account, it is important to analyze objectively how climate change makes urban areas more or less livable.

The key concept of this study is urban thermal comfort (UTC). The first works dealing with this concept date back to the early 1970s and were mostly written by meteorologists (Clarke & Bach 1971; Mayer & Höppe 1987). These publications did not always use the exact term “urban thermal comfort” but they focused on problems related to the condition of city dwellers in an urban microclimate context. According to research standards of that period, the main method of quantitative analysis was the development of mathematical models. Two popular models of this kind can be mentioned: the Predicted Mean Vote (PMV) model and the Munich Energy-balance Model for Individuals (MEMI) (Höppe 1993). The PMV model is based on the prediction of the percentage of people dissatisfied with the environmental parameters inside a room or office. This model has two main weak points: first, it only works inside the building, and second, it has shown no significant differences in comfort conditions for gender and age (Schaudienst & Vogdt 2017). Despite this, the PMV model is still widely used and has been integrated into the ISO (International Organization for Standardization) system of standards (ISO 7730).

The MEMI model (Höppe 1993) is based on three heat balance equations: the energy balance equation of the total body, the equation for heat flux from the body core to the skin, and the equation for heat flux from the skin through the clothing layer to the outer surface of the clothing. While we do not reproduce the full mathematical derivation here, the model’s key innovation was its focus on the individual as a thermal subject, accounting for metabolic rate, clothing insulation, and environmental parameters (Höppe 1993; Mayer & Höppe 1987). For a detailed exposition of the MEMI equations, readers are referred to Höppe (1993) and Zhao et al. (2021).

Without looking into the mathematical component of this model, the initial focus can be put on a person as a subject of comfort. There are dozens of models using the same approach (e.g., those reviewed by (Zhao et al. 2021)). This approach is correct, but in our case, it needs to be more customized. The personal characteristics of people are too changeable to be incorporated into city management processes. Therefore, we should specifically focus on the environmental properties that, with a certain degree of probability, determine comfortable conditions for the diverse population of large urban areas.

The most conventional and contemporary approach to define UTC is based on the idea that UTC is a “condition of mind that expresses satisfaction with the thermal condition of the surrounding environment” (Chan et al. 2017; Aghamolaei et al. 2023). It means that it is influenced by “psychological and behavioral factors in addition to the physical condition of outdoor spaces” (Aghamolaei et al. 2023). As Lai et al. have shown in their comprehensive review of outdoor thermal comfort studies, this topic can be studied through the analysis of indirect and direct influences (see Fig. 1).

While thermal comfort is conventionally assessed using microclimatic variables such as air temperature, humidity, wind speed, and mean radiant temperature (Mayer & Höppe 1987; Höppe 1993), our study adopts a broader urban-scale perspective. We use land surface temperature and vegetation indices as proxies for thermal exposure across large urban areas, recognizing that these indicators capture the environmental dimension of comfort at the city scale where microclimatic data are unavailable.

However, as it has been mentioned earlier, it is too complicated to apply the human condition to urban policies. In this respect, this study will thoroughly assess the environmental factors and various population characteristics on a city-wide scale. To do so, the issue of urban thermal comfort will be placed in a broader conceptual framework, more related to urban studies in general than to urban climatology. This also explains our reason for using the term “urban thermal comfort” rather than “outdoor thermal comfort”. The intention is to highlight some unique features of a city that would be unusual for other types of outdoor environments (rural or other areas less affected by anthropogenic impact).

To study the processes that shape the effects caused by outdoor environments, we use the concept of urban health. The concept describes how we can look at outdoor environments and the features that most affect public health (Vlahov & Galea 2003; Vlahov & Galea 2005). It deals with the factors of health and diseases in urban areas and with the urban context itself (Vlahov & Galea 2003).

Even though plenty of nuanced studies link UTC parameters with public health issues (Ormandy & Ezratty 2012; Jiang et al. 2020; Ma 2020), our paper is the first attempt to look at UTC as the main problem in the field of urban health studies. By gaining insight into these elements, the main point is to note that not only does the physical environment matter for urban health, but so does the socioeconomic context. There is more than convincing evidence that urban health disparities are strongly connected to socioeconomic inequality (Fiscella & Williams 2004; Borrell et al. 2013; Brzoska & Razum 2015). While the importance of the socioeconomic context cannot be downplayed, our goal is to make policy recommendations for urban planners, so we will focus on physical and basic demographic factors affecting urban health.

Vlahov and Galea (2005) demonstrated that rapid urbanization impacts urban health. This process is overly complex, so probably not every definition possible will be accurate enough. However, drawing on our research goals and inevitable limitations, we can determine the term “rapid urbanization” very roughly: it is a combination of simultaneous rapid growth of urban population with extensive enlargement of built-up area. This somewhat simplified view of urbanization is well applicable to urban health studies. Both population growth and comprehensive development of greenfields cause multiple risks to public health (Vlahov & Galea 2002; Sarkar & Webster 2017), and these also are zones of responsibility for urban planners. So, our recommendations for dealing with challenges caused by rapid urbanization may be implemented in policy and decision-making processes.

On a city-wide scale, the most contemporary framework for dealing with emerging challenges is urban resilience. This topic came into the mainstream research spotlight during the COVID-19 pandemic, which was a shocking and unexpected crisis for most urban areas worldwide (Glaeser 2022). However, the main conceptual contributions were made in the 2010s, when some of the most thorough reviews on urban resilience were published (e.g. Meerow et al. 2016; Gomes Ribeiro & Jardim Gonçalves 2019).

These reviews, as well as Glaeser’s findings (Glaeser 2022), have shown that the way of thinking about urban areas as systems trying to adapt to constantly emerging new risks, challenges, and even disasters is nothing new. Since the dawn of urban civilization, immunity to epidemics, war damage, and changes in trade routes has been tested for urban settlements across the globe. The new idea is to put this immunity at the center of policies. So, as Meerow et al. put it, “Urban resilience refers to the ability of an urban system (and all its constituent socio-ecological and socio-technical networks across temporal and spatial scales) to maintain or rapidly return to desired functions in the face of a disturbance, to adapt to change, and to quickly transform systems that limit current or future adaptive capacity” (Meerow et al. 2016; p. 39). We consider this definition applicable to our paper.

To summarize the above-described framework, the idea is to highlight the main factors affecting urban health during the rapid urbanization process. Drawing from urban resilience studies, climate change can be highlighted as one of the most relevant risks as well as the most persistent one. To assess its influence without getting too deep into meteorological nuances, it can be linked with the process of rapid urbanization in a given physical context. All the chosen urban areas vary greatly in terms of urbanization dynamics and physio-geographical features. However, there are some universal markers for our research: population dynamics, urban heat island, NDVI, and the climate zone in which the city is situated.

We have already mentioned that population growth is a challenge to cope with if we are dealing with urban health issues. The growing population increases the load on the healthcare system and makes the city more diverse, which in turn expands the range of healthcare requirements (Vlahov et al. 2007; Harpham 2009; Giles-Corti et al. 2016). Within the process of global urbanization, an extremely important challenge is the rapid increase in population in the countries of the Global South, which continue to lag behind the most developed countries in terms of equal access to quality medical services (Vlahov & Galea 2005; Mitlin & Satterthwaite 2013; Corburn et al. 2020).

The nature of climate change has been a very controversial topic of academic writing in recent decades (Hulme et al. 1999; Eriksen et al. 2015; Collins et al. 2012). However, it is quite conventional that urban heat island (UHI) is one of the largest sources of negative climate change in urban environments (Stewart & Mills 2021). UHI has been successfully linked to climate change and population growth (Manoli et al. 2019), urban thermal comfort (Yin et al. 2023), the historical context of urban development (Stewart 2019), which is largely affected by the climate zone factors, and built-up features (Szagri et al. 2023). Therefore, based on this corpus of texts, we can confidently assume that climate change is a main persistent risk for urban resilience to cope with, and its crucial factor is the urban heat island. UHI, in turn, is closely connected to rapid urbanization and climate zone context.

The methodology consists of two blocks: population dynamics and climate change. All the data was collected for 80 large urban areas that are present in the C40 organization. C40 is a global network of mayors of large cities that are united in action to confront the climate crisis1. This organization creates a platform for promotion of practices to cut carbon emissions and limit global heating. We chose this kind of sampling because of the breadth of the organization’s coverage of different urban climates and because of the organization’s awareness of the importance of climate change and building resilient cities to adapt to it. So, we can assume that mayors of these cities are at least interested in lowering urban heat and creating a more comfortable climate context.

Urban boundaries were defined using the administrative city limits as delineated by the Global Administrative Areas (GADM) database (version 3.6). This approach was chosen to ensure consistency across all 80 cities, as administrative boundaries are uniformly available and allow direct comparison with population data from the United Nations.

While socioeconomic inequalities are central to urban health disparities (Fiscella & Williams 2004; Borrell et al. 2013), their city-scale quantification across 80 diverse global cities lies beyond the scope of this paper. Our analysis, therefore, focuses on biophysical and basic demographic drivers (LST, NDVI, population counts), acknowledging that future research must integrate socioeconomic data — such as Gini coefficients at the city level or neighborhood-scale deprivation indices — to achieve a truly holistic assessment (Brzoska & Razum 2015; Sarkar & Webster 2017).

The specific objectives of this study are therefore: (1) to quantify LST and NDVI trends across 80 large urban areas using Landsat 8 imagery (2018–2023); (2) to group these trends by Köppen climate zone; (3) to develop a simplified forecast model linking population growth to changes in built-up area, LST, and NDVI to 2035; and (4) to identify which climate zones face the most concerning LST trajectories.

To assess population change, we use the data from World Urbanization Prospects 2018 compiled by the UN Department of Economic and Social Affairs2. Climate zones were determined according to the Köppen climate classification (Beck et al. 2018).

GIS-based and remote sensing methodologies have shown their effectiveness in UHI research (Guha & Mukherjee 2017; Zhang & Sun 2019; Sharma et al. 2021). To assess UHI, the land surface temperature (LST) and Normalized Difference Vegetation Index (NDVI) are measured (Jabbar et al. 2024). Both indicators are assessed based on the data on the retrospective dynamics of built-up areas compiled from Landsat 8 satellite images (2018–2023) using the Google Earth Engine platform. To calculate NDVI for each analyzed city, the climatic summer was determined (based on average monthly temperatures for the last 5 years for each city) and only these months were used in further calculations. To calculate LST, the hottest months were determined for each analyzed city (based on average monthly temperatures for the last 5 years) and only these months were used in further calculations.

The algorithm for our study goes as follows:

All the trends calculated for the period 2018–2023 are integrated into a system of differential equations solved by first-order numerical methods. For this purpose, we conduct the following steps:

The trend of population changes (the rate and direction of change) was calculated. UN data, including historical data for 2000–2022 and forecast data until 2035, was used. The trend of changes in the built-up area has been calculated. ESRI data was used for the period from 2017 to 2022. The correlation between population and built-up area change rates for 2018–2022 was calculated, as was the ratio of their average rates.

Based on the current and retrospective balance of urban territories and the corresponding NDVI and LST data, correlations between changes in these indicators and changes in built-up areas, as well as ratio coefficients of the average rates of change in parameters, were calculated (Eq. 1):

 (1)

where VLST and VL are mean rate of change of LST and built-up area respectively (Eq. 2).

 (2)

where VNDVI и VL are mean rate of change of NDVI and built-up area respectively.

All average rates of change in values for the specified period were calculated using the Eqs. 3–4:

 (3)

 (4)

Vp and VNDVI were calculated as follows: VL was calculated as follows: According to the basic system of differential equations, the values of the territory’s built-up area, NDVI and LST were extrapolated up to 2035 inclusive. The system of differential equations consists of following Eqs. 5–6:

 (5)

(6)

where i is an index for a calculated year i є[ 2018; 2034], t=1 is for time that is 1 year.

All remote sensing processing was performed using Google Earth Engine and ArcGIS Pro 3.1. Landsat 8 Level 2 surface reflectance and surface temperature products were used. These images have a spatial resolution of 30 m for NDVI and 100 m (resampled to 30 m) for LST. Only scenes with less than 10% cloud cover were retained. For each city, we extracted the median composite of the climatic summer months (for NDVI) and the hottest months (for LST) for each year from 2018 to 2023.

Land surface temperature was retrieved from Landsat 8 Collection 2 Level-2 surface temperature products. These products are generated using the single-channel algorithm developed by Jiménez-Muñoz et al. (2014), which incorporates atmospheric correction using the Atmospheric Correction Parameter Calculator (ATMCORR) and land surface emissivity estimated from the NDVI-based method (Sobrino et al. 2004). Surface emissivity was assigned using a classification scheme based on land cover types, with values ranging from 0.97 for vegetated surfaces to 0.92 for barren areas (Sobrino et al. 2001). The thermal infrared band 10 (10.9 µm) was used for LST retrieval, with band 11 excluded due to higher uncertainty in atmospheric correction (USGS 2019). All processing was performed in Google Earth Engine using the built-in Landsat 8 surface temperature product (USGS 2019). No independent ground-based validation was performed, as this was beyond the scope of the study.

The calculation results broken down into Köppen climate zones will be analyzed. A total of 16 Köppen climate zones are present across the chosen urban areas, including all the larger groups except Polar: Tropical, Dry, Temperate, and Continental. The exact list of Köppen climate zones can be seen at Britannica3. The list of C40 cities can be found at the C40 official website4.

It was not possible to collect the necessary data for some urban areas (e.g., Bogotá, Delhi, Ho Chi Minh, Hong Kong, Los Angeles, Medellín, Mexico City, Quezon City, Tshwane, Yokohama, Heidelberg, Cape Town, Ahmedabad, Montreal, Quito, and San Francisco), so these locations are excluded from the analysis. This problem occurred during the analysis of satellite images. Those dated the same years as all the images analyzed were unsuitable for analysis due to high cloud cover. The rest of the chosen urban areas are shown on the map (Fig. 2) the diagrams (Fig. 3), grouped by 4 higher taxons.

The proportion of people living in areas of higher LST (top 10% of values) is considered a proxy for the proportion of people living inside UHIs (urban heat islands). To make recommendations, we must analyze what is linked to this indicator’s high and low values. It may be noticed that there is no strong correlation or covariation between shares of the population living in the greenest (with higher NDVI values) and the hottest areas (with higher LST values). Another pattern that can be slightly unexpected is that some cities with a very dry climate (Dubai, Dakar, Lima) do not show larger shares of people living near or inside the UHIs. We assume that it is related to two features: 1) the thermal context of urban areas, where almost all urbanized areas are situated inside the zones of high LST (so there are no extremal outliers); 2) the historical process of urbanization took into account the physiographic conditions of urban areas, and people sought to populate the least hot places.

Several explanations may account for this absence of correlation between the share of the population living in high-LST areas and those living in high-NDVI areas. First, vegetation in high-density urban areas is often fragmented or consists of small, isolated patches that do not provide sufficient cooling at the neighborhood scale due to limited evapotranspiration and shadow effects (Ziter et al. 2019). Second, the spatial resolution of Landsat 8 (30 m for NDVI) may smooth over fine-grained variations, failing to capture small green spaces that could locally mitigate LST but are aggregated into mixed pixels (Li et al. 2017). Third, in some cities, the hottest areas correspond not to low-NDVI zones but to specific built morphologies — such as wide paved streets, low-albedo building materials, or canyon geometries that trap radiation — which are not directly offset by nearby parks (Stewart & Oke 2012; Oke et al. 2017). Fourth, the timing of satellite overpass (Landsat 8 acquires data around 10:30 AM local time) may miss diurnal patterns where vegetation cooling is more pronounced in the afternoon (Imhoff et al. 2010). This complexity suggests that greening alone is an incomplete strategy for UHI mitigation without concurrent attention to urban form, building materials, and the spatial configuration of green infrastructure.

The top 3 urban areas with the largest shares of people living inside UHI are in China (Fuzhou, Guangzhou, Wuhan), and another four Chinese urban areas (Beijing, Hangzhou, Dalian, Zhenjiang) are in the top 10. Such a disproportionately high presence of urban areas in this country can be related to a very high population density. There are two pieces of evidence for this assumption. First, the city with the largest share of people living inside UHI, which is not situated in China, is Istanbul. Not only is it located in a climate zone that is absent among the above-mentioned Chinese urban areas, but Istanbul also has a very distinct history of urbanization and planning policymaking. However, there is one thing in common between rapidly growing mega-urban areas such as Guangzhou and Istanbul. The largest city in Turkey is known for its discrepancy between official city limits and densely urbanized central parts. Therefore, even though Istanbul does not have an extremely dense population inside city limits, it is actually a very densely populated urban area. The same can be said about San Salvador (the city with the largest share of people living inside UHI in the respective climate zone) and, with some caveats — about Oslo. It is incorrect to compare the population density of San Salvador and Oslo with the urban areas of China and Istanbul, because the principle of defining boundaries for these urban areas is different. However, in general, both Oslo and San Salvador have high densities, comparable, for example, to the population density of Shanghai (3,900, 2,000, and 4,200 people/km², respectively).

Second, there is no correlation between two shares of the population (inside areas of UHI/high LST and high NDVI). For example, the urban areas of Guangzhou and Wuhan show almost the exact same values of people living inside UHI. Still, the shares of the population close to higher NDVI areas vary greatly: 35% in Guangzhou and 98% in Wuhan (these urban areas are in the same climate zone). Therefore, as we have noted earlier, higher population density could be a better predictor of problems with UHI than climatic features or accessibility of green spaces. Stewart and Oke have shown that UHIs result from an individual or the combination of several factors, such as the reduction in previous land cover, the increase in solar absorption and heat storage material, the increase in anthropogenic heat release, and the decrease in urban ventilation (Stewart & Oke 2012). All these factors may play a role in making the urban areas mentioned above hotter. However, it is intriguing that Oslo is among the most UHI-affected urban areas. Following the findings made by Venter et al. (2020), we may assume that this phenomenon is due to the specifics of Oslo’s built-up area dominated by paved or impervious surfaces.

The observation that several less UHI-affected urban areas are in very dry climate zones yet also represent the Global South (e.g., Dubai, Lima, Nairobi, Karachi, Addis Ababa) invites further investigation. Rapid urbanization in post-colonial societies has distinct features: the juxtaposition of traditional, contemporary, and futuristic building morphologies, the lag of infrastructure development behind growing population needs, and general inconsistency in planning decisions (Datta & Shaban 2016). We hypothesize that these factors — particularly the persistence of vernacular architecture with passive cooling strategies (Foudazi & M’Rithaa 2013) and lower anthropogenic heat emissions from differential industrial and vehicular densities — may explain why some Global South cities show lower LST values despite hot climates. Testing these hypotheses would require integrating city-level socioeconomic and infrastructure indicators (e.g., motorization rates, industrial zoning maps, building age distributions) which is a necessary direction for future research that lies beyond the biophysical scope of the current paper.

The combination of pre-modern and modern architecture is characteristic of many urban areas from the list. It would be noteworthy to look at one specific example where the presence of large areas with 19th-century buildings, as well as newer ones using the same techniques, did not result in a substantial decrease in overheating. Istanbul serves as a case study, as illustrated by the maps of its NDVI and LST values in Figs. 4 and 5. Istanbul was selected as a case study because it combines a hot-summer Mediterranean climate (Csa) with extreme intra-urban density variation, making it an ideal test case for examining the relationship between population density and LST at the neighborhood scale. Additionally, Istanbul is representative of rapidly growing megacities in the Global South where infrastructure development lags behind population growth.

Over the past two decades, Istanbul’s administrative boundaries have been expanded to match those of its province. As a result, the city now combines a dense urban core with extensive peri-urban natural areas, including the Belgrad Forest to the north and the Şile coastal zone to the northeast. Istanbul’s city center is located on both banks of the Bosphorus Strait. These areas, as well as the city’s western and eastern surroundings, are the most urbanized and less covered with canopy than the northeastern and northwestern periphery of Istanbul.

Two specific patterns from the maps support the conclusion about the Bosphorus’s limited role in mitigating LST. First, the hottest LST pixels (dark red in Fig. 5) are consistently located in elevated inland neighborhoods such as Kağıthane, Gaziosmanpaşa, and Esenler — areas situated 60–100 m above sea level and at least 3–5 km from the waterfront. Second, high NDVI areas (light green in Figure 4) are concentrated in the northern periphery (e.g., Sarıyer and Beykoz), far from the densely populated central districts of Fatih, Beyoğlu, and Üsküdar. Proximity to the Bosphorus does not translate into measurable cooling where population density exceeds 15,000 people/km² (Turkish Statistical Institute 2022). This finding aligns with previous work showing that sea breezes in Istanbul are often blocked by coastal topography and building morphology (Ünal et al. 2020). Thus, the Istanbul case supports our broader hypothesis (see Section 4.2) that population density and elevation may be stronger predictors of UHI exposure than proximity to water bodies.

Proximity to the open water of the Bosphorus plays a very limited role in the mitigation of heating. The most heated areas are situated not far away from the city center, but above sea level. Therefore, a mix of factors (population density, lack of green spaces, differentiation of height, and proximity to water) describes the structure of UHI in Istanbul. Overall, the situation in Istanbul is of major concern for urban health. With the addition of evidence that high population density caused negative effects during the COVID-19 pandemic (Paköz & Işık 2022), there are some crucial decisions to be made in Istanbul and other densely populated and overheated urban areas.

To evaluate possible changes in the three main indicators we have used in our analysis (population, LST, NDVI), we present the results of the calculation of our forecast model, which is based on the annual dynamics of the average rate of indicators (Fig. 6).

Being grouped according to Köppen climate zones, urban areas show different dynamics of expected changes in their indicators. The most concerning scenario in terms of rising LST is expected to unfold in the following climate zones:

All these zones are dry, and as it was found earlier, not all of them are linked to higher LST values. Cfa and Dwa urban areas are expected to get hotter than they are now, being the hottest among the urban areas from the list. BWh urban areas are less troubled nowadays, but the value of LST will probably increase significantly. It may be associated with two processes, with the first being a cumulative effect of hyper-urbanization (the case of Chinese mega-urban areas), and the second being a disproportionate effect of global warming, where urban areas in arid climate zones will suffer the most.

It is important to note that in the early 2030s a slight decline in LST values for Cfa and Dwa urban areas is expected to be seen, and only the BWh group will continue to show an increase. Paradoxically, the BWh group is the only one that is expected to seemingly increase its NDVI level (Milovanović et al. 2023). Therefore, only a decisive struggle with policy-making tools will help these urban areas to overcome the negative effects of global warming (Kaur 2022).

The projected slight decline in LST for Cfa and Dwa zones in the early 2030s should be interpreted with considerable caution. Based on our model structure (see Eqs. in Section 2), this predicted decline emerges from the interaction between slowing population growth rates (as forecast by UN data after 2030 for several East Asian cities) and the μ₃ coefficient linking built-up area change to NDVI dynamics. However, due to the short baseline period (2018–2023) and the unavailability of city-level calibration data for each coefficient (see limitations in Discussion), we cannot robustly determine whether this specific inflection point is a genuine signal or an artifact of model assumptions. Independent validation using longer time series (e.g., MODIS LST data from 2000–present) would be required to confirm this prediction. Until such validation is performed, readers are advised to focus on the directional trends (increasing LST for BWh, Cfa, Dwa) rather than the precise timing of any decline.

On the contrary, positive dynamics are predicted for some urban areas, that is, a decrease in temperatures. They represent the following climate zones:

Specifically, they are the urban areas of Cape Town, Madrid, Santiago, Boston, Portland, and Seattle. We have not had the opportunity to discuss these urban areas in detail. All of them, except Madrid and Portland, are situated by the sea. But the two exceptions have a common feature with the rest of the urban areas listed: local climates are not characterized by excessive temperature changes, and the hottest months are inferior to those in groups of urban areas where LST growth is expected. We can assume that global warming will make some temperate urban areas (especially by the sea) more comfortable.

The data used in calculations and analytics imposes several restrictions on the interpretation of the results obtained. On the one hand, the results obtained at the city scale contribute to a comprehensive assessment and identification of socio-ecological risks. On the other hand, they do not allow the identification of the root causes of these risks.

The broader findings of this study suggest that high population density may be a stronger predictor of urban heat island exposure than climate zone or vegetation cover alone. Istanbul offers a useful test case for this hypothesis, as it combines a hot-summer Mediterranean climate with extreme intra-urban variation in population density.

District-level analysis of Istanbul reveals that areas with the highest population densities show substantially higher land surface temperatures compared to low-density districts. Importantly, this pattern persists even when vegetation cover is held approximately constant, indicating that density drives temperature increases independently of green space availability. Furthermore, the temperature difference between high-density and low-density districts within Istanbul exceeds the average temperature difference between Istanbul and other cities in the same climate zone.

These findings confirm that high population density is a robust predictor of thermal discomfort across different neighborhoods within a single city. From a policy perspective, this suggests that density management and targeted cooling interventions in densely built areas — such as cool materials, green roofs, and ventilation corridors — may be more effective than relying solely on city-wide greening strategies to mitigate urban heat.

The main limiting factor of this research is the data needed to scale up the study. To detail the results of the analysis and higher predictive accuracy, it is necessary to conduct research, including using AI, and to use more extensive archival data, the historical depth of which should be at least 10 years with at least monthly differentiation. In addition to the volume of information analyzed, the level of detail is extremely important. This approach not only directly affects the correctness of correlations between indicators but also makes it possible to identify more accurately the very root causes of progressive or regressive changes in the surrounding world, including the urban environment.

This study has analyzed only two key indicators (NDVI and LST) out of many others that influence the environmental aspects of sustainable urban development. At the same time, as part of the methodology’s development, it is necessary to consider such indicators as the content of various pollutants in the air and the territory’s hydrological, climatic, and meteorological characteristics.

In addition, understandably, it is impossible to consider the environmental agenda separately from the socio-economic one. Therefore, a detailed assessment and forecasting of anthropogenic impact on the environment is impossible without a detailed analysis of such indicators as motorization, industrialization, the development of the agricultural industry, and others.

Furthermore, our analysis focuses on land surface temperature and vegetation indices as proxies for urban thermal conditions. We acknowledge that thermal comfort is a multidimensional concept that also depends on air temperature, humidity, wind speed, and human physiological responses (Ormandy & Ezratty 2012; Zhao et al. 2021). However, these variables are not available at the global city scale, and our use of the term “urban thermal comfort” reflects the broader urban planning perspective adopted in this study.

Despite these limitations, our study offers several strengths. It provides a globally consistent, cross-climate-zone comparison of LST and NDVI trends using standardized satellite data. The coupling of population forecasts with built-up area change offers a replicable template for scenario modeling in data-scarce regions. The identification of specific climate zones (BWh, Cfa, Dwa) as LST hotspots directs targeted policy attention. Finally, the finding that population density may be a stronger predictor of UHI exposure than climate zone or NDVI alone redirects attention to urban form and density management.

The most promising vector of development could be a series of in-depth studies of urban areas in individual climate zones followed by recommendations based on them. In addition, it is necessary to consider socio-economic inequalities that affect the ability of urban areas and the countries in which they are located to combat negative climate change. Finally, it would be valuable to conduct an analysis of the legislative framework to better understand in what categories and at what level complex issues, such as emission reduction, should be addressed.

Based on our findings, we recommend the following actions for urban areas in high-risk climate zones (BWh, Cfa, Dwa) and for cities with high population density. First, prioritize cool materials such as reflective coatings in building codes for new developments and retrofits. Second, mandate green roofs and vertical greening in zones where ground-level NDVI cannot be increased due to density constraints. Third, preserve and connect existing high-NDVI patches into cooling corridors aligned with prevailing winds. Fourth, establish heat-health early warning systems linked to LST forecasts rather than air temperature alone. Fifth, use population density thresholds as triggers for mandatory UHI mitigation plans regardless of climate zone.

1. C40 Citites. Available online: https://www.c40.org/ (accessed on 23 June 2025).2. World Urbanization Prospects 2018. Available online: https://population.un.org/wup/DataQuery/ (accessed on 23 June 2025).3. Köppen climate classification. Available online: https://www.britannica.com/science/Koppen-climate-classification (accessed on 23 June 2025).4. C40 Citites. Available online: https://www.c40.org/ (accessed on 23 June 2025).