Urbanization is projected to increase local surface temperature

Global rapid urbanization has intensified the urban heat island (UHI) effect, raising worldwide sustainability concerns. By altering urban land surface properties, urbanization amplifies UHI, a trend well-documented in recent studies. This intensification, in turn, threatens human health through heatwaves, increases energy consumption, and disrupts water–atmosphere interactions, thereby challenging future sustainable development. As a key indicator, surface UHI (SUHI) is particularly valuable due to its high spatiotemporal resolution and is widely applied in urban planning, climate research, public health, and environmental protection.

Satellite-derived land surface temperature (LST) is central to assessing SUHI and its dynamics. SUHI is typically quantified as the urban-rural LST difference. Moreover, LST provides consistent, repeatable observations across diverse spatial (local to global) and temporal (diurnal to interannual) scales, supporting analyses of thermal environments and the estimation of air temperature. With sensors such as the Moderate-Resolution Imaging Spectroradiometer (MODIS), the Advanced Spaceborne Thermal Emission and Reflection Radiometer, and the Land Remote-Sensing Satellite series, large-scale SUHI assessments are feasible owing to their complementary spatial and temporal resolutions. Looking ahead, as global warming and urbanization persist, SUHI is expected to intensify.

However, explicit projections of future urban heat patterns remain limited. While some studies employ Earth System Models and General Circulation Models from the Coupled Model Intercomparison Project Phase 6 (CMIP6) to estimate temperature, these models lack detailed urban representation. Models such as the Community Earth System Model incorporate urban processes but remain constrained by coarse resolution and simplified parameterizations. In contrast, Regional Climate Models offer finer-scale detail through downscaling, yet they are computationally intensive and less applicable for global-scale analyses.

To address this gap, we developed a global 1 km LST dataset (2020–2100) that integrates the combined effects of both climate change and urbanization. First, we quantified global warming driven by climate change using a multi-model ensemble of CMIP6 surface temperature projections. Next, we estimated urbanization-induced warming by establishing dynamic regression models between MODIS-derived LST and impervious surface area (ISA). These models were iteratively updated and applied to projected ISA time series to characterize the urbanization-induced warming. Finally, by combining climate and urbanization components, we generated a 1 km LST dataset for 2020–2100, capturing both large-scale climate impacts and localized urban heat amplification to support advanced thermal analyses.

Historical relationship between changes in ISA and LST

The response of LST change (ΔLST) to ISA change (ΔISA) from 2003 to 2020 exhibits pronounced global spatial heterogeneity (Fig. 1). The historical slopes predominantly range from −0.04 to 0.04 °C per %, with 68% of analyzed cities experiencing increased warming alongside ISA expansion. In  contrast, some cities show a negative ΔISA–ΔLST relationship, where  each unit increase in ΔISA corresponds to a decrease in ΔLST, resulting in slightly lower LST values (Fig. 1a). Cities with high slopes are concentrated in rapidly urbanizing regions, including parts of Africa, East and Southeast Asia, and northern South America, reflecting intensified urban heat effects in these areas. Statistically, more than half of the cities exhibit significant correlations (p < 0.1) (Fig. 1b), particularly those undergoing rapid urbanization (Supplementary Fig. 1). Across climate zones, LST consistently increases with ISA growth, though the magnitude of response varies substantially. Tropical and temperate regions show stronger warming, with average slopes of 0.0187 °C per % and 0.0125 °C per %, respectively (Supplementary Fig. 2), likely driven by rapid urbanization and local climatic conditions. For instance, Guangzhou, a representative rapidly urbanizing city, experienced a 100% increase in ISA, accompanied by a strong positive ΔISA–ΔLST  correlation from 2003 to 2020 (Supplementary Fig. 1). By contrast, despite being located in a temperate zone, Washington D.C. experienced less than half the warming observed in Guangzhou, likely due to its more mature stage of urban development (Supplementary Fig. 1). Arid and cold regions exhibit comparatively modest responses, with average slopes of 0.0022 °C per % and 0.0076 °C per % respectively (Supplementary Fig. 2), likely attributable to lower solar radiation, limited anthropogenic heat sources, and the high-albedo effects of vegetation or snow cover. For example, Helsinki shows only about one fifth of Guangzhou’s LST response, despite a 70% increase in ISA (Supplementary Fig. 1). This climatic variation underscores the crucial role of environmental context in modulating the thermal impacts of urbanization, highlighting the sensitivity of the ΔISA–ΔLST relationship across climatic zones.

a Slope derived from historical observations. b Statistical significance of the linearly fitted model. The analysis included 6359 level-2 administrative units from Database of Global Administrative Areas with urban areas larger than 100 km². Base map source: Esri ArcGIS Online (public domain).

Full size image

Spatial consistency of historical and future LST

Our estimated LST under diverse Shared Socioeconomic Pathways–Representative Concentration Pathways (SSP-RCP) scenarios show strong spatial consistency with MODIS-observed LST during the historical period and preserves these spatial patterns in future projections (Fig. 2 and Supplementary Fig. 3). Globally, projected LST by 2100 under the SSP5-RCP8.5 scenario indicates a substantial increase, with high warming zones concentrated in northern and southern Africa, Australia, and the Middle East, where values may exceed 40 °C (Fig. 2a). Climate change-driven warming dominates this pattern, affecting nearly all regions in accordance with the overall LST distribution (Fig. 2b). In urban contexts, rapid urbanization further amplifies warming, particularly in suburban areas of cities such as Beijing, Indianapolis, Bangalore, and Accra, where warming gradients intensify toward the periphery (Fig. 2b). Although smaller than the effects of climate change, urbanization-induced warming remains pronounced in these expanding regions. At the global scale, estimated LSTs correlate well with MODIS-observed LST in 2020, achieving an R² exceeding 0.9 (Supplementary Fig. 4). At the city scale, ~70% of cities exhibit R² values above 0.7, with representative cases such as Chicago (0.89), New Delhi (0.90), and Beijing (0.82) under the SSP1-RCP2.6 scenario (Supplementary Fig. 5). This strong spatial coherence underscores the robustness of our projections for future thermal patterns.

a Spatial patterns of future LST globally under the compounded warming effect of climate change and urbanization in 2100 under the SSP5–RCP8.5 scenario. b Detailed information of four representative cities in Beijing, Indianapolis, Bangalore and Accra. Note: the terms “Compounded”, “Urbanization” and “Climate change” refer to future LST under the compounded warming effect of climate change and urbanization, warming solely from climate change, and warming exclusively from urbanization, respectively. This nomenclature will be maintained throughout. Same base map source as Fig. 1.

Full size image

Read more on Nature text and related references (deleted here for ease of reading).

**