Urban areas experience significantly higher temperatures than surrounding rural areas, a phenomenon known as the urban heat island (UHI) effect. This effect is exacerbated by global climate change and continued urban population growth, leading to increased heat exposure for urban residents. Current assessments of heat-related mortality and productivity loss often rely on temperature data from non-urban weather stations, which may not accurately reflect urban heat conditions. Many studies implicitly assume that urban and rural temperatures increase at the same rate, neglecting the unique warming dynamics within cities. While some research has attempted to isolate the urban warming signal from surface air temperature (SAT) data, these studies often focus on regional warming rather than specific city core warming and suffer from limitations in data quality and spatial resolution. Other studies utilizing satellite-derived land surface temperature (LST) data have not adequately separated the warming signal of the urban core from the effects of urban expansion. This study addresses these shortcomings by analyzing satellite LST data to quantify and attribute the surface warming trends in global city cores, comparing them to rural background trends, and examining the contributions of background climate change, urban expansion, and urban greening to the observed warming trends. This understanding is crucial for accurate assessments of heat-related risks and effective urban heat mitigation strategies.

Existing literature highlights the significant impact of urbanization on local and regional climate. Studies have shown that urbanization-induced warming can account for a substantial portion of overall warming in rapidly urbanizing regions, particularly in China. However, these studies often focus on regional-scale effects, not on the distinct warming trends within the core of global cities. Investigations using in-situ surface air temperature (SAT) data have limitations due to the scarcity of well-located urban stations and potential biases from station changes and urbanization effects. Satellite-based LST studies have offered an alternative approach but have often struggled to separate the warming signal of the city core from the influence of urban expansion into previously rural areas. Existing LST studies show varying results regarding the global mean LST-based UHI trends, indicating discrepancies in methodology and data handling that necessitate a more comprehensive approach that explicitly distinguishes the urban core from the effects of urban expansion.

This study utilizes three MODIS datasets (2002–2021): LST (MYD11A2), EVI (MOD13A2), and yearly land cover type (MCD12Q1). The LST data, with a spatial resolution of 1000 m, contains two overpasses per day. The EVI data, also at 1000 m resolution, provides a measure of vegetation greenness. The land cover data (resampled to 1000 m) allows for the identification of urban areas. City cluster boundary data from the Global Urban Boundary (GUB) dataset (2018) identifies 2080 city clusters larger than 50 km². These clusters are categorized into four size classes: small, medium, large, and megacities, based on area. Population data (2002-2020) from the Oak Ridge National Laboratory are used, resampled to 1 km resolution. Monthly reanalysis SAT data (GLDAS) from the Goddard Earth Sciences Data and Information Services Center (GES DISC) are used to represent background climate change (BCC), resampled to 1 km resolution. A breakpoint detection algorithm (BFAST) is employed to identify abrupt changes in LST and EVI time series, enabling the classification of pixels into three categories: urban core, rural background, and transitional (initially rural, then urban). Linear regression is used to calculate warming trends in each category. A statistical attribution method, employing least squares regression, quantifies the contributions of BCC, urbanization (URB), and landscape greening (LSG) to the observed urban surface warming trends. The BCC component is estimated from rural reanalysis SAT data. The URB component uses the relationship between LST and population density at the pixel level. The LSG component is estimated from the relationship between EVI and LST in rural areas, accounting for possible differences between rural and urban core relationships.

The study reveals significantly faster surface warming trends in urban cores compared to rural background areas. The global mean surface warming trend in urban cores is 0.50 ± 0.20 K per decade, 29% higher than the rural background trend (0.38 ± 0.21 K per decade). This difference is even more pronounced when compared to the daily mean SAT trend from reanalysis data (0.32 ± 0.083 K per decade). The warming trend increases with city size, with megacities showing the most pronounced warming. Asian cities exhibit the highest warming trends, while European cities show the lowest. The transitional land shows the strongest warming, due to the loss of evaporative cooling from land-use change. Analysis of Enhanced Vegetation Index (EVI) shows an increasing trend of vegetation greening in most areas, particularly in Europe, offsetting background warming. Attribution analysis shows that, globally, background climate change (BCC) is the dominant contributor to urban core warming (0.34 ± 0.13 K per decade), followed by urbanization (URB, 0.27 ± 0.13 K per decade). However, in rapidly urbanizing regions of China and India, URB becomes a more substantial contributor. Landscape greening (LSG) shows a net cooling effect globally (-0.10 ± 0.028 K per decade), primarily due to urban greening in Europe. The study found that increasing population density leads to increased LST trends while increasing EVI decreases LST trends. The surface urban heat island (SUHI) intensity shows substantial increasing trends for the majority of city clusters.

The findings demonstrate that urban areas are warming at a faster rate than rural areas, highlighting the importance of considering these different rates in future climate impact assessments and urban planning. The significant contribution of background climate change underscores the need for global efforts to reduce greenhouse gas emissions. The substantial impact of urban expansion in certain regions, particularly in China and India, highlights the need for sustainable urban development strategies. The cooling effect of urban greening, especially prominent in Europe, emphasizes the importance of implementing and maintaining green infrastructure in cities to mitigate urban heat. The study's results can inform the development of more accurate models for predicting future heat exposure in urban areas and guide targeted mitigation efforts.

This study provides compelling evidence that urban areas are warming significantly faster than surrounding rural areas, a phenomenon amplified in megacities. Background climate change is the primary driver, but urban expansion plays a substantial role in certain regions, while urban greening offers a mitigating effect. These findings highlight the need for integrated approaches to urban heat mitigation, combining emissions reductions with sustainable urban development and green infrastructure initiatives. Future research should focus on improving the accuracy of the statistical attribution methods, exploring the detailed mechanisms underlying the observed trends in different regions, and investigating the socio-economic implications of urban heat in different contexts.

The study uses satellite-derived LST data, which may not perfectly capture the same information as in-situ SAT data. The statistical attribution method relies on certain assumptions about the relationships between LST and population density, and LST and EVI. These relationships might vary across different geographic and climatic contexts. The study does not examine all possible contributing factors to urban warming; other factors like building materials and urban morphology could play a role. The temporal resolution of some datasets might limit the precision of certain analyses, such as the identification of abrupt changes and trends.