Urban Stadiums as Multi-Scale Cool-Island Anchors: A Remote Sensing-Based Thermal Regulation Analysis in Shanghai

Abstract

The intensification of urban heat in high-density cities has raised growing concerns for public health, infrastructural resilience, and environmental sustainability. As large-scale, multi-functional open spaces, sports stadiums play an underexplored role in shaping urban thermal patterns. This study investigates the spatial and temporal thermal characteristics of eight representative stadiums in central Shanghai and the Pudong New Area from 2018 to 2023. A dual-framework approach is proposed: the Stadium-based Urban Island Regulation (SUIR) model conceptualizes stadiums as active cooling agents across micro to macro spatial scales, while the Multi-source Thermal Cognition System (MTCS) integrates multi-sensor satellite data—Landsat, MODIS, Sentinel-1/2—with anthropogenic and ecological indicators to diagnose surface temperature dynamics. Remote sensing fusion and machine learning analyses reveal clear intra-stadium thermal heterogeneity: track zones consistently recorded the highest land surface temperatures (up to 37.5 °C), while grass fields exhibited strong cooling effects (as low as 29.8 °C). Buffer analysis shows that cooling effects were most pronounced within 300–500 m, varying with local morphology. A spatial diffusion model further demonstrates that stadiums with large, vegetated buffers or proximity to water bodies exert a broader regional cooling influence. Correlation and Random Forest regression analyses identify the building volume (r = 0.81), NDVI (r = −0.53), nighttime light intensity, and traffic density as key thermal drivers. These findings offer new insight into the role of stadiums in urban heat mitigation and provide practical implications for scale-sensitive, climate-adaptive urban planning strategies.

1. Introduction

Urban overheating has emerged as a defining environmental risk in rapidly urbanizing cities, particularly in high-density metropolitan areas where ecological disruption and construction intensity converge. In cities such as Shanghai, prolonged and increasingly frequent heatwaves are intensifying public health pressures, threatening infrastructural performance, and compromising overall urban livability [1,2]. The primary contributors to this thermal stress—extensive impervious surfaces, fragmented ventilation corridors, and the declining availability of green and blue infrastructure—have significantly weakened the natural regulation capacity of urban thermal systems [3,4]. As one of the world’s most densely populated cities, Shanghai exemplifies the complex interplay between urban morphology and extreme heat exposure, with record-breaking summer temperatures and expanding zones of thermal risk in recent years [5].

In response, numerous studies have emphasized the role of urban green spaces, plazas, and water bodies in mitigating the urban heat island (UHI) effect. These elements remain critical in buffering temperature extremes, yet their spatial distribution and availability are often constrained in high-density urban cores. This growing limitation calls for a broader re-examination of underutilized urban spaces with thermal regulatory potential. In this context, sports stadiums—characterized by expansive footprints, multi-functional uses, and structurally heterogeneous surfaces—are emerging as alternative climate-responsive spatial units. With surface elements such as grass fields, synthetic tracks, and grandstands exhibiting distinct thermal behaviors, stadiums create pronounced intra-site LST gradients [6,7,8]. Moreover, they are frequently embedded in complex urban zones that include green corridors, transportation hubs, and mixed-use developments, forming micro–meso thermal interaction systems. Despite these attributes, stadiums have received little attention in the literature as active agents of urban thermal mitigation. Most studies either overlook them entirely or simplify them into static land use categories. Recognizing stadiums as climate-functional infrastructure offers a promising conceptual shift—one that expands UHI mitigation beyond natural landscapes and opens up new pathways for integrated spatial design. Unlike previous research that focuses on single-variable cooling effects or isolated land uses, this study introduces a novel multi-dimensional approach that integrates physical, ecological, and anthropogenic factors into a unified thermal regulation framework. By conceptualizing stadiums not merely as passive land parcels but as thermally functional, spatially scalable, and structurally networked urban elements, our work seeks to reframe their role from land use types to dynamic thermal infrastructures. This reconceptualization provides a new theoretical lens for urban climatology and opens up methodological space for modeling thermal diffusion and connectivity beyond conventional site-based analyses.

Although a growing body of research has confirmed the cooling effects of urban parks and waterfronts [9,10,11], systematic investigations into stadium-based thermal regulation remain scarce. Most existing studies rely on fixed-radius buffer analyses or point-based field measurements, offering limited insight into the scale sensitivity and spatial heterogeneity of cooling performance. In high-density urban environments—where thermal gradients fluctuate rapidly over space and time—this single-scale approach fails to capture the complex interactions between stadium morphology, land cover composition, and surrounding urban structures [12]. Moreover, remote sensing studies on the urban thermal environment continue to depend heavily on single-source datasets such as MODIS or Landsat, limiting either spatial granularity or temporal frequency. Despite the availability of complementary satellite systems such as Sentinel-1/2, few studies have integrated these sources to construct continuous, high-resolution temperature time series across urban subzones [13]. In addition, thermal mechanism analyses often rely on simple linear regressions with environmental indicators, without disentangling the nonlinear or interactive effects of physical, ecological, and anthropogenic drivers [14]. These methodological constraints not only narrow the scope of stadium-related thermal research but also hinder the development of generalized frameworks for climate-adaptive spatial planning.

To address these critical research gaps, this study focuses on eight representative stadiums located in the central urban districts and the Pudong New Area of Shanghai—regions that exemplify high-density urban morphology and complex thermal dynamics. Using multi-source remote sensing datasets from 2018 to 2023, this study proposes a two-tiered analytical framework that integrates spatial scale modeling, thermal driver diagnosis, and cold-island network construction. The first framework, the Stadium-based Urban Island Regulation (SUIR) model, treats stadiums as thermal regulatory nodes across micro (intra-site), meso (surrounding 500–1000 m), and macro (city-scale diffusion) scales. The second framework, the Multi-source Thermal Cognition System (MTCS), fuses MODIS, Landsat, and Sentinel-1/2 data to construct high-resolution LST time series, allowing for robust modeling of thermal mechanisms. Together, these frameworks enable a transition from descriptive observation to predictive, scale-sensitive, and spatially integrative thermal modeling. By combining data fusion, spatial statistical modeling, and network analysis, this study offers a novel pathway to interpret, quantify, and simulate the cooling performance of large stadiums in dense urban contexts. The findings aim to contribute both theoretically—by expanding the typology of climate-regulatory urban units—and practically—by informing climate-adaptive planning strategies for megacities facing intensifying thermal risk.

2. Study Area and Data

2.1. Study Area

This study focuses on the central urban districts and the Pudong New Area of Shanghai, located on the eastern coast of China within the Yangtze River Delta. Geographically, the study area extends from 120°52′E to 122°12′E and 30°40′N to 31°53′N, encompassing a low-lying alluvial plain with elevations generally below 10 m. It borders the East China Sea and is interwoven with a dense hydrographic network, forming a complex terrain characterized by flatness, proximity to water, and intensive urban development. The region includes eight administrative districts—Huangpu, Xuhui, Changning, Jing’an, Putuo, Hongkou, Yangpu, and Pudong New Area—that jointly form the urban core of Shanghai and serve as key nodes in the spatial structure of the city [15].

Shanghai experiences a subtropical monsoon climate, with an annual average temperature of approximately 17.6 °C and total yearly precipitation ranging from 1100 to 1200 mm. The hottest period typically occurs from July to August, when average daytime temperatures frequently exceed 35 °C. During this period, the combination of high humidity and anthropogenic heat emissions leads to intensified UHI effects, particularly in densely built-up zones [16,17]. The continuous expansion of impervious surfaces and the decline of green and open spaces have further exacerbated thermal stress, contributing to local microclimatic deterioration in central districts [18].

The selected study area is particularly suitable for investigating stadium-based thermal regulation due to its high concentration and diversity of large-scale sports facilities. These facilities exhibit a “polycentric distribution, complementary functions, and embedded integration” within the urban fabric [19]. They range from national-level venues, such as the Shanghai Oriental Sports Center, to district-level sports complexes (e.g., Putuo, Changning) and community-oriented recreational parks (e.g., Xuhui Sports Park). Beyond their role in hosting athletic events and public fitness activities, these stadiums contribute to ecological ventilation corridors, serve as potential heat-mitigating spatial nodes, and are incorporated into urban emergency planning strategies [20,21].

Based on remote sensing interpretation and spatial planning datasets, eight representative stadiums were identified and spatially coded as primary analysis units (see Figure 1 and

Table 1. Geographic distribution of major stadiums in central Shanghai and Pudong New Area.

No.	District	Stadium Name	Description
1	Huangpu District	Shanghai Stadium (Huangpu Venue)	Classic comprehensive sports complex
2	Yangpu District	Yangpu Sports Center	Newly built large-scale sports complex
3	Xuhui District	Xuhui Sports Park	Community-based recreational sports venue
4	Hongkou District	Hongkou Football Stadium	Home stadium of the Shanghai football team
5	Jing’an District	Jing’an Sports Center	Major venue for sports activities
6	Pudong New Area	Shanghai Oriental Sports Center	National-level large-scale sports complex
7	Changning District	Changning Sports Center	District-level sports facility
8	Putuo District	Putuo Sports Complex	District-level indoor sports complex

). These include (1) Shanghai Stadium (Huangpu District), (2) Yangpu Sports Center, (3) Xuhui Sports Park, (4) Hongkou Football Stadium, (5) Jing’an Sports Center, (6) Shanghai Oriental Sports Center (Pudong New Area), (7) Changning Sports Center, and (8) Putuo Sports Complex. Their spatial distribution forms a roughly circular and well-balanced pattern across the study area, providing strong representativeness for both multi-scale thermal effect evaluation and spatial diffusion modeling. The geolocation and spatial attributes of these facilities offer a robust foundation for subsequent analysis of stadium-based thermal performance within a dense urban matrix.

2.2. Study Data

To examine the spatial thermal dynamics of representative stadiums in the central districts and the Pudong New Area of Shanghai, this study employed an integrated dataset that combines multi-source satellite imagery with diverse geospatial and anthropogenic indicators. Considering the strong seasonality and interannual variability of the urban heat environment, the analysis focused on July and August—the peak summer months—from 2018 to 2023. This time window captures the period of most intense heat accumulation and highest frequency of extreme temperature events.

To balance spatial detail and temporal continuity, a hybrid dataset was constructed, incorporating medium- and high-resolution optical and thermal infrared imagery, C-band synthetic aperture radar (SAR) data, vegetation indices, urban structural proxies, and human activity metrics. This comprehensive data framework supports multiple levels of analysis, including (1) fine-scale thermal pattern recognition, (2) multi-dimensional driver identification, and (3) simulation of spatial thermal diffusion patterns [22,23].

2.2.1. Remote Sensing Data

The remote sensing data utilized in this study were selected based on the need to capture both fine-scale spatial heterogeneity and long-term temporal dynamics of the urban thermal environment. Landsat-8 and Landsat-9 OLI/TIRS imagery, provided by the United States Geological Survey (USGS), offers high-resolution multispectral (30 m) and thermal infrared (100 m) data. These datasets are particularly suited for detailed intra-urban temperature mapping and were used to extract LST patterns within and around stadiums. To complement Landsat’s limited temporal resolution, MODIS MOD11A2 LST products from the Terra and Aqua platforms were incorporated. With an 8-day temporal resolution and 1 km spatial resolution, MODIS provides robust support for macroscale UHI assessments and long-term trend analyses. Data fusion in this study primarily addresses both spatial and temporal resolution limitations of single sensors. By integrating Landsat’s high spatial resolution with MODIS’s high temporal frequency using the STARFM model, we generate fused LST products with a daily temporal resolution and 10–30 m spatial resolution, effectively capturing fine-scale thermal heterogeneity and continuous temporal variations.

To enhance ecological and structural characterization, Sentinel-2 MSI imagery was employed for vegetation index extraction, impervious surface mapping, and water body delineation. Its 10–20 m spatial resolution and high revisit frequency (approximately every 5 days) enable consistent ecological monitoring across the six-year study period. Additionally, Sentinel-1 SAR data, offering all-weather, day-and-night imaging capabilities, were used to extract key physical attributes of the urban surface. Parameters such as surface roughness and built-up density derived from SAR backscatter help explain spatial thermal heterogeneity in areas with complex morphology. While ECOSTRESS LST data provide a higher spatial resolution (70 m) and better temporal granularity in certain cases, the relatively short operational period and inconsistent temporal coverage over the study area limited its applicability for this six-year longitudinal analysis. Therefore, the MODIS and Landsat fusion approach remains the optimal balance for fulfilling both spatial and temporal requirements in this study.

Recognizing the limitations of single-sensor observations, this study applied the Spatial and Temporal Adaptive Reflectance Fusion Model (STARFM) to integrate the high spatial resolution of Landsat with the high temporal frequency of MODIS. The fused LST datasets generated through this approach provide synthetic time series with a 10–30 m spatial and daily temporal resolution, significantly enhancing the continuity and granularity of thermal monitoring across the study area. To mitigate the potential biases between MODIS and Landsat LST products, this study conducted a statistical consistency check using overlapping observations and applied bias correction based on linear regression between matched pixels prior to fusion, following methods outlined in previous studies [24,25]. Collectively, the multi-source remote sensing data form a robust foundation for analyzing scale-sensitive thermal patterns, driving factor mechanisms, and spatial diffusion processes of heat in and around stadium environments [26,27,28,29,30].

2.2.2. Auxiliary Data

To enhance the explanatory capacity of the proposed thermal environment models, the study integrates a range of auxiliary geospatial datasets that represent anthropogenic activity, urban morphology, and administrative boundaries.

First, the VIIRS Nighttime Lights dataset (VNP46A2), provided by NOAA, offers a 500 m spatial resolution indicator of human activity intensity and energy use patterns. These data are widely recognized as effective proxies for anthropogenic heat emissions, particularly in high-density commercial and residential areas.

Second, building footprint data sourced from OpenStreetMap (OSM) include building geometries and height attributes, enabling the calculation of structural metrics such as volumetric building density, building height distribution, and horizontal obstruction indices. These metrics are critical for understanding surface heat storage capacity, airflow resistance, and the spatial concentration of thermal mass [31,32].

Third, road network and traffic data were collected from both the Gaode Maps API and the OSM platform. These datasets capture road type distributions, road density, and inferred traffic flow intensity. Such indicators are incorporated into the thermal driver system to represent the spatial distribution of anthropogenic heat sources associated with transportation infrastructure.

Lastly, high-resolution administrative boundary data were acquired from the National Geomatics Center of China (NGCC). These boundaries provide a consistent spatial framework for statistical aggregation, unit assignment, and comparative analysis across districts. They also serve as the basis for linking each stadium to its corresponding urban governance zone and for constructing regional-scale thermal indicators.

Collectively, these auxiliary datasets offer critical contextual information for modeling the drivers of urban thermal variation and contribute to a more robust, interpretable analysis of stadium-based cooling performance.

2.2.3. Data Sources and Feature Tables

This study integrates multi-source remote sensing and auxiliary geospatial datasets to analyze the urban thermal environment in Shanghai from 2018 to 2023. As summarized in

Table 2. Summary of remote sensing and auxiliary datasets used for urban thermal environment analysis in Shanghai.

Data Source	Satellite/Platform	Spatial Resolution	Temporal Resolution	Time Range (Summer)	Primary Applications
Landsat-8/9	USGS	30 m (VNIR)/100 m (TIR)	16 days	July–August, 2018–2023	LST retrieval, land cover classification
MODIS (MOD11A2)	Terra/Aqua	1 km	8 days	July–August, 2018–2023	Mesoscale LST product, temporal trend analysis
Sentinel-2 MSI	ESA	10–20 m	5 days	July–August, 2018–2023	Vegetation indices, land cover extraction
Sentinel-1 SAR	ESA	10 m	6–12 days	July–August, 2018–2023	Urban physical structural analysis
VIIRS Nighttime Lights (VNP46A2)	NOAA	500 m	Monthly	July–August, 2018–2023	Anthropogenic heat source estimation
OpenStreetMap	OSM	Varies by region	Irregular updates	Latest version	Building geometry, volume, and density analysis
Road and Traffic Data	Gaode/OSM	10–30 m	Annual updates	Latest version	Road network density, traffic intensity
Administrative Boundaries	NGCC (China)	——	Static	Latest version	Regional demarcation and spatial statistics

Note: “Latest version” refers to the most recently available data as of 2024 from their respective open-source platforms or governmental repositories (e.g., OpenStreetMap, Gaode Maps, NGCC).

, the data span multiple spatial and temporal resolutions, combining thermal, optical, and radar imagery with socioeconomic and infrastructural layers. Landsat and MODIS data support LST retrieval at varying scales, while Sentinel imagery enables vegetation and urban morphology analysis. Nighttime light, traffic, and building data serve as proxies for anthropogenic heat, supplemented by open-source spatial boundaries for statistical aggregation. These datasets collectively support the multi-variable modeling of surface temperature and stadium-scale thermal dynamics.

2.2.4. Technology Roadmap

This study follows a comprehensive multi-phase technical workflow to analyze the urban thermal regulation effects of stadiums in central Shanghai and the Pudong New Area from 2018 to 2023 (Figure 2). First, the research problem is framed within the context of intensified urban heat islands, leading to the selection of eight representative stadiums. Next, multi-source remote sensing data—such as Landsat-8/9, MODIS, Sentinel-1/2, and ECOSTRESS—are acquired and fused using the STARFM model within the Google Earth Engine to generate high-resolution daily LST time series. Auxiliary datasets including nighttime lights, building vectors, and road networks are integrated for urban structural characterization. In the modeling phase, this study establishes the SUIR model and MTCS framework to assess the stadiums’ thermal behavior across micro (<500 m), meso (500 m–3 km), and macro (>3 km) scales. Thermal differentiation within stadium functional zones (grass, track, stands) is analyzed using NDVI, NDBI, and shading metrics. To identify key thermal drivers, a multi-source feature tensor is constructed and input into Random Forest models, capturing spatial heterogeneity and nonlinear impacts. Cold-Island Areas (CIAs) are then simulated using Gaussian diffusion kernels to quantify the extent and strength of cooling effects. Finally, results are synthesized to offer policy implications for resilient urban design, emphasizing the integration of green–blue infrastructure with stadium planning.

3. Methodology

3.1. Theoretical Framework and Multi-Scale Design

To systematically investigate the spatial thermal regulation effects of stadiums within high-density urban environments, this study adopts a dual-framework approach comprising the Stadium-based Urban Infrared Regulation (SUIR) concept and the Multi-scale Thermal Coupling System (MTCS). These two models jointly serve as the theoretical foundation for the spatial analysis, multi-source data integration, and heat diffusion modeling presented in this research. The frameworks aim to capture both the endogenous attributes of stadiums—such as surface materials, openness, and vegetation—and their exogenous interactions with the surrounding urban morphology.

The SUIR framework conceptualizes stadiums not merely as isolated built structures but as active thermal regulatory agents embedded within complex urban ecosystems. Each stadium is understood as a multi-functional thermal node that modulates LST through mechanisms such as in situ evapotranspiration, material reflectivity, surface heat retention, and spatial ventilation. Accordingly, three principal cooling pathways are identified: (1) internal surface cooling through green coverage and water bodies; (2) thermal buffering due to open-form spatial morphology and non-impervious materials; and (3) cross-boundary spillover into adjacent zones, contributing to broader regional cooling effects.

Building upon this conceptualization, the MTCS model operationalizes the spatial analysis by defining three nested thermal zones around each stadium:

• Microscale (<500 m): Focuses on internal surface heterogeneity—such as grass fields, stands, and synthetic tracks—and their respective thermal behaviors.
• Mesoscale (500 m–3 km): Encompasses surrounding urban morphology, where factors such as vegetation cover, building density, and impervious surfaces influence LST transitions.
• Macroscale (>3 km): Explores inter-stadium spatial interactions and the emergence of cool-island clusters at the regional level.

This framework allows for the integration of remote sensing imagery, urban form metrics, and environmental variables into a unified analytical system that is both scalable and adaptable to complex city structures.

3.1.1. Adaptive Buffer Construction Strategy

To capture the multi-scale thermal influence of stadiums, this study employs a dynamic buffer construction approach rather than fixed radii. Each stadium is treated as a central node, and concentric zones are delineated based on both the local urban density and the functional level of the stadium. The buffer radius R is calculated as follows [33,34]:

$$R=R_0\times (1+\alpha \times \beta +\delta \times \gamma )$$

(1)

where ${\mathrm{R}}_0$ is the base radius (500 m for microscale, 3 km for mesoscale); $\mathsf{\beta }$ ∈ [0, 1] represents the ratio of building footprint to total buffer area; $\mathsf{\gamma }$ ∈ {1, 2, 3} corresponds to stadium level (community, district, national); and $\mathsf{\alpha }$ = 0.5, $\mathsf{\delta }$ = 0.3 are adjustment coefficients.

This adaptive mechanism allows the spatial extent of analysis to flexibly accommodate the heterogeneity of urban form and stadium characteristics.

3.1.2. Microscale Thermal Analysis (<500 m)

At the microscale level, the internal structural composition of each stadium is assessed. Typical components include grass fields, running tracks, bleachers, and paved zones. Each surface category exhibits distinct thermal behavior due to variations in albedo, emissivity, and evapotranspiration potential. The mean LST of each functional unit is calculated as

$$\overline{\mathrm{L}}\mathrm{S}{\mathrm{T}}_{\mathrm{i}}={\displaystyle \frac{1}{\mathrm{n}}}{\sum }_{\mathrm{j}=1}^{\mathrm{n}}\mathrm{L}\mathrm{S}{\mathrm{T}}_{\mathrm{i}\mathrm{j}}$$

(2)

Among them, $\overline{\mathrm{L}}\mathrm{S}{\mathrm{T}}_{\mathrm{i}}$_i is the average temperature of the i-th functional area (such as grassland), n is the number of pixels in this area, and $\mathrm{L}\mathrm{S}{\mathrm{T}}_{\mathrm{i}\mathrm{j}}$ is the LST value of the j-th pixel. This provides a basis for understanding intra-site thermal heterogeneity and identifying functional structures contributing to cooling or heat accumulation.

3.1.3. Mesoscale Thermal Analysis (500 m–3 km)

In this zone, the thermal interaction between the stadium and its immediate urban surroundings is evaluated. Morphological indicators are extracted and correlated with LST using linear regression.

NDVI (Normalized Difference Vegetation Index):

$$\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}={\displaystyle \frac{\mathrm{N}\mathrm{I}\mathrm{R}-\mathrm{R}\mathrm{E}\mathrm{D}}{\mathrm{N}\mathrm{I}\mathrm{R}+\mathrm{R}\mathrm{E}\mathrm{D}}}$$

(3)

Building density:

$${\mathrm{D}}_{\mathrm{b}\mathrm{u}\mathrm{i}\mathrm{l}\mathrm{d}}={\displaystyle \frac{{\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{A}}_{\mathrm{b}\mathrm{u}\mathrm{i}\mathrm{l}\mathrm{d},\mathrm{i}}}{{\mathrm{A}}_{\mathrm{b}\mathrm{u}\mathrm{f}\mathrm{f}\mathrm{e}\mathrm{r}}}}$$

(4)

Green space ratio:

$$\mathrm{G}\mathrm{C}\mathrm{R}={\displaystyle \frac{{\mathrm{A}}_{\mathrm{g}\mathrm{r}\mathrm{e}\mathrm{e}\mathrm{n}}}{{\mathrm{A}}_{\mathrm{b}\mathrm{u}\mathrm{f}\mathrm{f}\mathrm{e}\mathrm{r}}}}$$

(5)

Among them, ${\mathrm{A}}_{\mathrm{b}\mathrm{u}\mathrm{i}\mathrm{l}\mathrm{d},\mathrm{i}}$ is the floor area of building $\mathrm{i}$, ${\mathrm{A}}_{\mathrm{b}\mathrm{u}\mathrm{f}\mathrm{f}\mathrm{e}\mathrm{r}}$ is the buffer area, and ${\mathrm{A}}_{\mathrm{g}\mathrm{r}\mathrm{e}\mathrm{e}\mathrm{n}}$ is the total green area.

Pearson correlation coefficients between LST and the above indicators are calculated:

$${\mathrm{r}}_{\mathrm{x}, \mathrm{L}\mathrm{S}\mathrm{T}}={\displaystyle \frac{\mathrm{C}\mathrm{o}\mathrm{v}(\mathrm{x},\mathrm{L}\mathrm{S}\mathrm{T})}{{\mathsf{\sigma }}_{\mathrm{X}}\times {\mathsf{\sigma }}_{\mathrm{L}\mathrm{S}\mathrm{T}}}}$$

(6)

A linear regression model evaluates the driving effects:

$$\mathrm{L}\mathrm{S}\mathrm{T}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1\times \mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}+{\mathsf{\beta }}_2\times {\mathrm{D}}_{\mathrm{b}\mathrm{u}\mathrm{i}\mathrm{l}\mathrm{d}}+{\mathsf{\beta }}_3\times \mathrm{G}\mathrm{C}\mathrm{R}+\mathsf{\epsilon }$$

(7)

3.1.4. Macroscale Thermal Analysis (>3 km)

At the regional level, the focus shifts to identifying and characterizing stadium-based cold-island clusters and their influence on urban-scale UHI patterns [35,36].

Cold-Island Threshold: defined as the lower 10% percentile of urban LST:

$${\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{c}\mathrm{o}\mathrm{o}\mathrm{l}}<{\mathsf{\mu }}_{\mathrm{L}\mathrm{S}\mathrm{T}}-1.28\times {\mathsf{\sigma }}_{\mathrm{L}\mathrm{S}\mathrm{T}}$$

(8)

Heat Mitigation Strength (HMS):

$$\mathrm{H}\mathrm{M}\mathrm{S}={\displaystyle \frac{{\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{u}\mathrm{r}\mathrm{b}\mathrm{a}\mathrm{n}}-{\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{d}\mathrm{i}\mathrm{u}\mathrm{m}}}{{\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{u}\mathrm{r}\mathrm{b}\mathrm{a}\mathrm{n}}}}\times 100\mathrm{\%}$$

(9)

This captures the relative cooling capacity of stadium zones compared to the broader city. ${\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{u}\mathrm{r}\mathrm{b}\mathrm{a}\mathrm{n}}$: The average LST of the surrounding urban built-up environment excluding the stadium area, serving as a reference baseline to compare thermal conditions. ${\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{d}\mathrm{i}\mathrm{u}\mathrm{m}}$: The average LST within the boundaries of the stadium, including its functional zones such as grass fields, stands, and tracks, typically extracted from remote sensing imagery.

3.2. Multi-Source Data Fusion

In order to address the spatial and temporal limitations inherent in single-source remote sensing datasets, this study establishes a comprehensive multi-source data integration strategy. By fusing optical, thermal, radar, and auxiliary geospatial datasets, a high-fidelity input foundation is constructed for subsequent thermal environment modeling and factor decoupling. The processing pipeline includes LST fusion, environmental feature tensor construction, and variable system standardization, forming a unified and scalable data framework aligned with the MTCS model [37].

3.2.1. LST Enhancement via STARFM Fusion

Given the complementary nature of Landsat and MODIS in spatial and temporal dimensions, we adopt the Spatial and Temporal Adaptive Reflectance Fusion Model (STARFM) to generate high-resolution, high-frequency LST products [38,39]. Landsat LST (30 m, 16-day revisit) provides spatial precision, while MODIS LST (1 km, daily revisit) ensures temporal continuity.

StarFM basic principle: For a target time t, based on Landsat and MODIS observations at reference times t1 and t2, estimate the high-resolution LST value:

$${\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{H}}\left(\mathrm{x},\mathrm{y},\mathrm{t}\right)={\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{W}}_{\mathrm{i}}\times [\mathrm{L}\mathrm{S}{\mathrm{T}}_{\mathrm{L}}\left({\mathrm{x}}_{\mathrm{i}},{\mathrm{y}}_{\mathrm{i}},\mathrm{t}\right)+(\mathrm{L}\mathrm{S}{\mathrm{T}}_{\mathrm{H}}\left({\mathrm{x}}_{\mathrm{i}},{\mathrm{y}}_{\mathrm{i}},{\mathrm{t}}_1\right)-\mathrm{L}\mathrm{S}{\mathrm{T}}_{\mathrm{L}}({\mathrm{x}}_{\mathrm{i}},{\mathrm{y}}_{\mathrm{i}},{\mathrm{t}}_1)\left)\right]$$

(10)

Among them, ${\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{H}}\left(\mathrm{x},\mathrm{y},\mathrm{t}\right)$: high-resolution LST after fusion; $\mathrm{L}\mathrm{S}{\mathrm{T}}_{\mathrm{L}}$: MODIS low-resolution LST; ${\mathrm{W}}_{\mathrm{i}}$: weighting coefficient, taking into account pixel similarity, spatial distance, and spectral consistency; and ${\mathrm{x}}_{\mathrm{i}},{\mathrm{y}}_{\mathrm{i}}$: reference pixel coordinates in the search window.

ESTARFM enhancement content: ESTARFM introduces a trend factor and is suitable for heterogeneous surfaces, adding a time weight term ${\mathsf{\omega }}_{\mathrm{t}}$:

$$\mathrm{L}\mathrm{S}{\mathrm{T}}_{\mathrm{H}}\left(\mathrm{x},\mathrm{y},\mathrm{t}\right)={\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\mathsf{\omega }}_{\mathrm{t}}\times {\mathrm{W}}_{\mathrm{i}}\times [∆\mathrm{L}\mathrm{S}{\mathrm{T}}_{\mathrm{i}}\left({\mathrm{t}}_1,{\mathrm{t}}_2\times {\displaystyle \frac{\mathrm{t}-{\mathrm{t}}_1}{{\mathrm{t}}_2-{\mathrm{t}}_1}}\right)+\mathrm{L}\mathrm{S}{\mathrm{T}}_{\mathrm{H}}\left({\mathrm{x}}_{\mathrm{i}},{\mathrm{y}}_{\mathrm{i}},{\mathrm{t}}_1\right)]$$

(11)

The fusion results can obtain continuous LST time series data with a spatial resolution comparable to that of Landsat and close to the temporal resolution of MODIS, which can be used for subsequent thermal environment evolution analysis and factor modeling.

3.2.2. Construction of Multi-Source Feature Tensor

To support the identification and modeling of thermal environmental drivers, a three-dimensional feature tensor is constructed using Sentinel-1 SAR and Sentinel-2 MSI imagery [40,41]. This tensor integrates urban morphology, surface cover, and ecological parameters relevant to thermal heterogeneity. Key variables include the following:

Physical structural characteristics (Sentinel-1 SAR): VV/VH backscatter coefficient ${\mathsf{\sigma }}_{\mathrm{V}\mathrm{V}}^0,{\mathsf{\sigma }}_{\mathrm{V}\mathrm{H}}^0$; surface roughness proxy; and the moisture/surface hardness index (based on polarization ratio).

Ecological environmental characteristics (Sentinel-2 MSI):

NDBI (Construction Index)

$$\mathrm{N}\mathrm{D}\mathrm{B}\mathrm{I}={\displaystyle \frac{\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}-\mathrm{N}\mathrm{I}\mathrm{R}}{\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}+\mathrm{N}\mathrm{I}\mathrm{R}}}$$

(12)

MNDWI (Water Index)

$$\mathrm{M}\mathrm{N}\mathrm{D}\mathrm{W}\mathrm{I}={\displaystyle \frac{\mathrm{G}\mathrm{R}\mathrm{E}\mathrm{E}\mathrm{N}-\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}}{\mathrm{G}\mathrm{R}\mathrm{E}\mathrm{E}\mathrm{N}+\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}}}$$

(13)

Albedo (Surface Reflectance Index):

$$\mathrm{A}\mathrm{l}\mathrm{b}\mathrm{e}\mathrm{d}\mathrm{o}=0.356\mathrm{R}+0.130\mathrm{G}+0.373\mathrm{B}+0.085\mathrm{N}\mathrm{I}\mathrm{R}+0.072\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}1+0.061\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}$$

(14)

These variables are standardized, resampled, and geospatially aligned to the fused LST grid, allowing precise multi-variable regression and causality decomposition.

3.2.3. Identification of Thermal Environmental Driving Factors

The fused LST value is used as the dependent variable Y, and the feature variable X is used as the independent variable to construct a machine learning model to identify the dominant thermal environment factor [42]. Random Forest Regression (RF) has strong nonlinearity and is suitable for high-dimensional variables. It outputs feature importance:

$$\mathrm{Y}={\mathrm{f}}_{\mathrm{R}\mathrm{F}}\left(\mathrm{X}\right)+\mathsf{\epsilon }$$

(15)

Among them, $\mathsf{\epsilon }$ is the control accuracy.

3.3. Thermal Drivers’ Decomposition

Based on multi-source remote sensing features and the fused LST datasets, this study identified and quantified the key driving factors influencing the thermal environment of urban stadiums. By integrating machine learning models (e.g., Random Forest, Support Vector Regression), spatial statistical methods, and sensitivity analysis, we systematically decoupled the complex driving mechanisms [43]. The analysis revealed distinct causal characteristics of stadium-related thermal anomalies across micro- (<500 m), meso- (500 m–3 km), and macroscales (>3 km), offering a comprehensive understanding of scale-dependent thermal responses in high-density urban settings [44,45]. Although definitions of spatial scale can vary among studies, the present classification is grounded in previous empirical research on urban thermal environments and stadium-centered landscape effects. In particular, the 500 m threshold is widely used to characterize microclimatic zones directly influenced by built-environment features [46,47], while the 3 km upper limit for the mesoscale reflects the spatial extent of localized land surface temperature diffusion identified in prior remote sensing-based UHI assessments. This scale framework allows us to better capture the nested and transitional characteristics of thermal drivers operating from stadium interiors to broader urban contexts.

3.3.1. Driving Factor Construction and Classification

The driving factors are mainly constructed from multi-source characteristic tensors (

Table 3. Indicator variable system.

Category	Indicator Name
Category Indicator Name Vegetation and Land Cover	NDVI, MNDWI, Albedo
Built-up Morphology	NDBI, Building Density, Mean Building Height
Anthropogenic Heat Sources	Nighttime Light Intensity (VIIRS), Traffic Density
Spatial Configuration	Stadium Shape Index, Green Coverage Ratio
Topography and Climate	Elevation, Slope, Surface Roughness

). Combined with information such as urban structure, land cover, and environmental climate, the following variable system is constructed [48].

3.3.2. Multi-Scale Modeling and Importance Ranking

The Random Forest regression model (RF) is used to train the relationship between surface temperature and various factors, and the factor importance weights are obtained:

$${\mathrm{Y}}_{\mathrm{L}\mathrm{S}\mathrm{T}}={\mathrm{f}}_{\mathrm{R}\mathrm{F}}\left({\mathrm{X}}_{1,}{\mathrm{X}}_{2,\dots ,}{\mathrm{X}}_{\mathrm{n}}\right)+\mathsf{\epsilon }$$

(16)

where ${\mathrm{Y}}_{\mathrm{L}\mathrm{S}\mathrm{T}}$ represents the fused LST, and ${\mathrm{X}}_{\mathrm{n}}$ represents each thermal environment factor.

Variable importance calculation:

In the RF model, the variable importance ${\mathrm{I}}_{\mathrm{j}}$ is calculated as

$${\mathrm{I}}_{\mathrm{j}}={\sum }_{\mathrm{t}=1}^{\mathrm{T}}{\displaystyle \frac{{\mathrm{N}}_{\mathrm{t}}}{\mathrm{N}}}\times ∆\mathrm{M}\mathrm{S}{\mathrm{E}}_{\mathrm{t},\mathrm{j}}$$

(17)

where T is the total number of trees; $∆\mathrm{M}\mathrm{S}{\mathrm{E}}_{\mathrm{t},\mathrm{j}}$ is the MSE decrease caused by variable j in the t-th tree; ${\mathrm{N}}_{\mathrm{t}}$ is the number of samples falling into the branch; and N is the total number of samples.

In addition, the SHAP (Shapley Additive Explanations) method is combined to quantify the positive and negative contributions of each factor to the thermal response of a single sample:

$$\mathrm{f}\left(\mathrm{x}\right)={\mathsf{\phi }}_0+{\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{\varnothing }}_{\mathrm{i}}$$

(18)

${\mathrm{\varnothing }}_{\mathrm{i}}$ is the Shapley contribution value of variable ${\mathrm{x}}_{\mathrm{i}}$; and ${\mathsf{\phi }}_0$ is the average prediction value of the model.

3.3.3. Analysis of the Functional Heterogeneity Response of Stadiums

Different types of stadiums (such as comprehensive sports centers vs. community parks) have different responses to thermal environments. A response heterogeneity model was constructed: a multilevel regression model was constructed based on the functional type of the stadium (categorical variables) [49,50,51]:

$$\mathrm{L}\mathrm{S}{\mathrm{T}}_{\mathrm{i}\mathrm{j}}={\mathsf{\gamma }}_0+{\mathsf{\gamma }}_1{\mathrm{Z}}_{\mathrm{j}}+{\mathsf{\gamma }}_2{\mathrm{X}}_{\mathrm{i}\mathrm{j}}+{\mathrm{u}}_{\mathrm{j}}+{\mathsf{\epsilon }}_{\mathrm{i}\mathrm{j}}$$

(19)

Among them, ${\mathrm{Z}}_{\mathrm{j}}$ is the j-th stadium type; and ${\mathrm{u}}_{\mathrm{j}}$ is the type-level error, used to compare the thermal response elasticity differences between types.

3.4. Heat Diffusion and Cool-Island Effect Modeling

To simulate the spatial propagation of cooling effects induced by large-scale sports stadiums and assess their capacity to alleviate local heat accumulation, this section introduces a quantitative modeling framework encompassing cold-island identification, thermal diffusion simulation, and effective influence area delineation. This framework provides empirical support for evaluating the environmental performance of stadiums in regulating urban thermal environments.

3.4.1. Cold-Island Identification and Buffer Zone Extraction

The identification of localized cool islands is based on the comparative temperature difference (ΔT) between the stadium core area and its surrounding urban matrix. The threshold condition for defining a cold island is formulated as

$$∆\mathrm{T}={\mathrm{T}}_{\mathrm{u}\mathrm{r}\mathrm{b}\mathrm{a}\mathrm{n}\mathrm{a}\mathrm{v}\mathrm{g}}-{\mathrm{T}}_{\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{d}\mathrm{i}\mathrm{u}\mathrm{m}}$$

(20)

Among them, ${\mathrm{T}}_{\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{d}\mathrm{i}\mathrm{u}\mathrm{m}}$ is the average LST of the core area of the stadium; ${\mathrm{T}}_{\mathrm{u}\mathrm{r}\mathrm{b}\mathrm{a}\mathrm{n}\mathrm{a}\mathrm{v}\mathrm{g}}$ is the average LST in the area; and when ∆T > °C (1.5 °C), it is judged as a cold island.

3.4.2. Thermal Kernel Diffusion Modeling

To characterize the spatial diffusion pattern of the stadium-induced cooling, a two-dimensional Gaussian heat diffusion kernel is applied. This model simulates the decay of the cold-island effect from the stadium center outward, governed by a spatially constrained exponential function [49,52]:

$$\mathrm{L}\mathrm{S}{\mathrm{T}}_{\left(\mathrm{x},\mathrm{y}\right)}={\mathrm{T}}_{\mathrm{m}\mathrm{i}\mathrm{n}}+\left({\mathrm{T}}_{\mathrm{m}\mathrm{a}\mathrm{x}}-{\mathrm{T}}_{\mathrm{m}\mathrm{i}\mathrm{n}}\right)\times \mathrm{e}\mathrm{x}\mathrm{p}(-{\displaystyle \frac{{(\mathrm{x}-{\mathrm{x}}_0)}^2+{(\mathrm{y}-{\mathrm{y}}_0)}^2}{2{\mathsf{\sigma }}^2}})$$

(21)

Among them, (${\mathrm{x}}_0,{\mathrm{y}}_0$) are the coordinates of the center of the stadium; σ is the diffusion control parameter (related to the maximum influence radius of the buffer zone); ${\mathrm{T}}_{\mathrm{m}\mathrm{i}\mathrm{n}}$, ${\mathrm{T}}_{\mathrm{m}\mathrm{a}\mathrm{x}}$ are the LST values of the core center and edge; and the optimal diffusion surface can be determined by least squares fitting.

3.4.3. Effective Influence Area Calculation

To quantitatively evaluate the areal extent of the cooling effect, the Cold-Island Area (CIA) is defined as the total area where the modeled surface temperature is significantly lower than the surrounding urban ambient temperature. Specifically, CIA refers to the integrated area where the temperature difference exceeds a given threshold (∆T > 1.5 °C). It is mathematically expressed as

$$\mathrm{C}\mathrm{I}\mathrm{A}={\iint }_{\mathrm{A}}^{\mathrm{B}}\left[{\mathrm{T}}_{\mathrm{u}\mathrm{r}\mathrm{b}\mathrm{a}\mathrm{n}}\left(\mathrm{x},\mathrm{y}\right)-{\mathrm{T}}_{\mathrm{m}\mathrm{o}\mathrm{d}\mathrm{e}\mathrm{l}}\left(\mathrm{x},\mathrm{y}\right)\right]>\partial {\mathrm{d}}_{\mathrm{x}}{\mathrm{d}}_{\mathrm{y}}$$

(22)

${\mathrm{T}}_{\mathrm{u}\mathrm{r}\mathrm{b}\mathrm{a}\mathrm{n}}\left(\mathrm{x},\mathrm{y}\right)$ is the reference urban surface temperature at location $\left(\mathrm{x},\mathrm{y}\right)$, typically derived as the mean LST of the surrounding urban area. ${\mathrm{T}}_{\mathrm{m}\mathrm{o}\mathrm{d}\mathrm{e}\mathrm{l}}\left(\mathrm{x},\mathrm{y}\right)$ is the modeled surface temperature at location $\left(\mathrm{x},\mathrm{y}\right)$, representing the cooling effect of the stadium. $\partial$ is the temperature threshold for defining the cooling effect (set to 1.5 °C in this study). A is the spatial domain of interest, usually covering the stadium and its surrounding area. B is the subset of domain A where the cooling effect is significant, defined as the area where the modeled temperature is at least $\partial$ (e.g., 1.5 °C) lower than the ambient urban temperature. In other words, it represents the area that satisfies the condition ${\mathrm{T}}_{\mathrm{u}\mathrm{r}\mathrm{b}\mathrm{a}\mathrm{n}}\left(\mathrm{x},\mathrm{y}\right)-{\mathrm{T}}_{\mathrm{m}\mathrm{o}\mathrm{d}\mathrm{e}\mathrm{l}}\left(\mathrm{x},\mathrm{y}\right)>\partial$. ${\mathrm{d}}_{\mathrm{x}}{\mathrm{d}}_{\mathrm{y}}$ denotes the spatial resolution (e.g., pixel size) used for numerical integration. This index quantifies the spatial footprint of thermal mitigation induced by each stadium, offering a consistent and comparable metric for evaluating cooling efficiency across facilities with different sizes, configurations, and surrounding land use types.

4. Results

4.1. Multi-Scale Thermal Effects

Analysis of average LST data from the summers of 2018 to 2023 across central Shanghai and the Pudong New Area reveals a clear and accelerating warming trend (Figure 3;

Table 4. Average summer land surface temperature by district in central Shanghai and Pudong New Area (2018–2023).

	Huangpu	Yangpu	Xuhui	Hongkou	Jingan	Pudong	Changning	Putuo
2018	37.32	35.79	36.65	38.01	38.40	35.72	37.33	36.71
2019	41.11	41.63	41.44	42.95	43.80	37.79	41.72	42.05
2020	42.05	40.95	41.57	43.07	43.53	36.62	41.83	41.98
2021	41.93	42.41	39.21	44.95	43.93	37.19	41.46	41.47
2022	38.86	41.82	38.77	41.06	40.97	34.53	39.65	39.92
2023	46.68	46.91	44.93	48.09	47.88	40.45	46.56	46.69

). The year 2023 marked the peak of this trajectory, with most districts recording average LSTs above 45 °C, indicating a sharp intensification of urban heat conditions and associated public health risks.

Spatially, the distribution of thermal stress is notably uneven. Hongkou and Jing’an districts consistently exhibited the highest LST values during the study period, reaching 48.09 °C and 47.88 °C in 2023, respectively, suggesting particularly strong UHI effects in these areas. In contrast, Pudong New Area generally maintained lower LSTs, such as 34.53 °C in 2022, likely due to a combination of more extensive green spaces and spatial configurations that support effective heat dispersion.

Other core districts like Huangpu, Xuhui, and Changning also displayed sustained increases in LST, reflecting the cumulative heat burden in densely built environments. Meanwhile, Yangpu and Putuo showed relatively lower values in certain years, potentially reflecting differing urban morphologies or green infrastructural strategies.

These findings underscore the spatio-temporal heterogeneity of thermal conditions in the study area and establish a strong empirical foundation for evaluating how stadiums may influence urban heat dynamics at multiple scales.

Analysis of LST across representative stadiums in central Shanghai and the Pudong New Area from 2018 to 2023 reveals significant thermal heterogeneity across functional zones (Figure 4, Figure 5, Figure 6, Figure 7, Figure 8 and Figure 9). Among these zones, grass-covered areas consistently exhibited the lowest LSTs, typically ranging from 36 °C to 40° C, with minimal standard deviations. This thermal stability reflects their strong cooling capacity, making them the most effective thermal buffers within stadium complexes. In contrast, spectator stands recorded the highest LSTs, with mean temperatures exceeding 44 °C and peak values reaching up to 56.83 °C. These areas also showed the greatest temperature variability, indicating intense localized heat accumulation. Track zones exhibited intermediate thermal behavior (40–44 °C), influenced by surface albedo and solar radiation absorption. In certain cases, tracks reached temperatures similar to stands, underscoring their exposure to direct heat loads.

Temporally, all functional zones experienced a notable warming trend over the study period, particularly from 2018 to 2022. The year 2022 stands out as the most extreme, marked by frequent heat events that pushed average LSTs in stand zones above 46 °C and caused noticeable warming in grass and track areas as well. Earlier years such as 2018 and 2019 showed lower temperatures and reduced variability, suggesting more stable thermal conditions. Although minor fluctuations occurred during 2020–2021, the overall upward trajectory persisted, reflecting the cumulative effects of climate change and intensified urbanization on the thermal behavior of stadiums.

Spatial differences between stadiums in Pudong and central districts were also evident. Pudong stadiums, characterized by larger green areas and more open spatial configurations, recorded lower LSTs, especially in grass zones. This suggests that vegetation density and natural ventilation play a critical role in regulating surface temperatures. In contrast, core urban stadiums such as Jing’an Sports Center and Shanghai Stadium exhibited higher LSTs in stand and track zones, largely due to surrounding built-up density and limited airflow. Internal design also influenced thermal performance: for instance, Yangpu Stadium and Shanghai Stadium, with extensive high-albedo stand materials, showed concentrated high-temperature zones and greater thermal variability. These findings highlight the need for targeted thermal management, including increasing green coverage, optimizing material selection, and improving ventilation pathways in stadium planning and retrofitting.

Further boxplot analyses of LST data from 2018 to 2023 (Figure 10) reinforce the persistent thermal differences among stadium functional zones. Track zones consistently exhibited the highest mean and maximum LST values, followed by stands, while grass zones maintained the lowest temperatures throughout the study period. This hierarchy remained stable across years, highlighting the enduring influence of surface materials and vegetation cover on thermal behavior.

The elevated temperatures in track areas can be attributed to their construction materials—typically asphalt or synthetic compounds—with low albedo and high thermal inertia. These surfaces absorb substantial solar radiation and release heat slowly, rendering them especially prone to heat accumulation during summer. In contrast, grass zones demonstrated not only lower LSTs but also minimal thermal variability, due to active evapotranspiration and the insulating effects of dense vegetation.

From a temporal perspective, all three functional zones showed a general cooling trend between 2018 and 2021, particularly in grass and stand zones. This trend may reflect the cumulative impact of urban greening initiatives, stadium retrofitting, and localized climate adaptation projects. Variations in meteorological factors such as wind speed and precipitation likely contributed as well. However, 2022 marked a slight rebound, especially in track and stand areas, corresponding with an uptick in extreme heat events and intensified solar radiation.

Variability analyses further emphasize the thermal vulnerability of tracks. Standard deviation and range metrics reveal greater intra-annual fluctuations in track zones, suggesting heightened sensitivity to extreme weather conditions. These findings position track areas as critical targets for intervention.

Spatial analysis of LST within buffer zones surrounding stadiums from 2018 to 2023 reveals clear scale-dependent thermal patterns (Figure 11). Specifically, mean LST values increased with buffer radius, following the consistent trend of Macro > Meso > Micro, indicating more severe thermal conditions as the distance from the stadium increases. For example, in 2021, the average LST within the microscale buffer (<500 m) was 33.1 °C, while that in the macroscale buffer (>3 km) reached 35.0 °C—a notable difference of nearly 2 °C. This suggests that stadiums exert a measurable cooling effect on their immediate surroundings, acting as localized cool islands that help mitigate ambient urban heat.

Temporally, LST values across all buffer zones demonstrated a general upward trend from 2018 to 2021, peaking in 2021 across all scales (Micro: 33.1 °C, Meso: 34.2 °C, Macro: 35.0 °C). A slight decrease followed in 2022 and continued into 2023, possibly reflecting the combined effects of urban greening initiatives, policy interventions, and climate-adaptive measures implemented after the extreme heat conditions of 2021. In terms of variability, LST standard deviations were consistently highest at the macroscale (up to 2.7 °C in 2021), indicating greater heterogeneity in thermal conditions—likely due to the mixed presence of heat islands and cooling patches. Conversely, the microscale buffer showed the most stable thermal profile, with lower standard deviations (1.7–2.1 °C), highlighting the relatively homogeneous and regulated thermal environment near stadiums.

Cross-comparative analysis among stadiums revealed that cooling performance varied significantly based on functional zone composition. Stadiums with higher proportions of grass-covered areas consistently exhibited stronger cooling effects within their microscale buffers compared to those dominated by impervious surfaces like tracks or stands. Differences in vegetation cover, proximity to water bodies, and spatial layout were identified as key contributing factors.

Further analysis of the thermal island regulatory effects of stadiums across different districts (

Table 5. Thermal island impact of stadium focus areas.

Region Name	Stadium IDs	Avg. LST (°C)	Surrounding Avg. LST (°C)	Cooling Island Intensity (°C)	Heat Island Influence Radius (km)
Huangpu-Xuhui	1, 3	32.1	33.6	−1.5	2.8	Core area with high green coverage and significant cooling
Hongkou-Yangpu	2, 4	32.4	33.8	−1.4	2.5	Dense residential surroundings limit cooling extent
Jingan-Putuo	5, 8	32.7	34.0	−1.3	2.2	High building density weakens buffer zone effect
Pudong New Area	6	31.9	33.9	−2.0	3.2	Open space and large stadium complex lead to pronounced cooling
Changning	7	32.5	33.5	−1.0	2.0	Relatively uniform green belts nearby

) reveals significant spatial variation in both cooling island intensity and influence radius. These differences are closely related to district-level urban morphology, green infrastructure, and stadium design characteristics.

Among all sites, the Pudong New Area stadium (Site 6) exhibited the strongest cooling performance, with a mean LST reduction of 2.0 °C and an influence radius of 3.2 km. This superior performance is likely due to its open spatial layout, larger facility size, and well-developed green and blue infrastructure. The Huangpu–Xuhui area (Sites 1 and 3) also demonstrated a strong regulation capacity, with a cooling intensity of 1.5 °C and a 2.8 km radius. Here, despite dense urban development, the presence of large, preserved parks and interconnected public spaces facilitates synergistic cooling effects between stadiums and surrounding vegetation. In contrast, Hongkou—Yangpu (Sites 2 and 4) and Jing’an—Putuo (Sites 5 and 8) showed weaker regulation, with cooling intensities of 1.4 °C and 1.3 °C, and influence radii of 2.5 km and 2.2 km, respectively. These areas are characterized by a high residential density, limited open space, and poor urban ventilation, which likely constrain the diffusion of cooling effects. The Changning stadium (Site 7) recorded the weakest performance, with a cooling intensity of only 1.0 °C and a 2.0 km radius, possibly due to a smaller facility size and relatively low vegetation coverage, despite a well-distributed greenbelt system.

4.2. LST Fusion and Thermal Factor Recognition

To overcome the limitations of single-source remote sensing data, this study applies the StarFM (Spatial and Temporal Adaptive Reflectance Fusion Model) to fuse MODIS and Landsat imagery. While MODIS provides daily observations at a coarse 1 km resolution, Landsat offers finer spatial detail (30 m) but lower temporal frequency (16-day revisit). StarFM integration effectively combines these advantages, generating daily-scale LST estimates at a 30 m resolution, which significantly enhances both spatial precision and temporal continuity. By incorporating Sentinel-1/2 data, the multisource fusion further improves model robustness under complex surface and atmospheric conditions.

The fused dataset was used to train a Random Forest (RF) regression model for LST estimation. The model achieved a high level of accuracy (R² = 0.85; RMSE = 1.32 °C, Figure 12;

Table 6. Performance metrics and variable importance rankings of the refined RF model (with internal vs. external NDVI separation).

Variable	Importance Score
Model Performance
R2	0.85
RMSE (°C)	1.32
MAE (°C)	0.97
Max VIF	2.67
Predictor Variables
External NDVI (250 m buffer)	0.214
External NDVI (100 m buffer)	0.193
Internal NDVI	0.168
NDBI (Normalized Difference Built-up Index)	0.142
Sentinel-1 VH Backscatter	0.118
Impervious Surface Index	0.094
Albedo	0.043
Normalized Difference Water Index	0.028

), demonstrating strong explanatory power. Among the predictor variables, NDVI ranked highest, reaffirming the central role of vegetation in surface cooling. Conversely, NDBI and impervious surface indices showed positive correlations with elevated LST, emphasizing the contribution of built-up intensity to urban heat accumulation. Sentinel-1 backscatter coefficients (VV/VH) emerged as important variables, reflecting their ability to capture surface texture and moisture conditions, especially where optical imagery is affected by shadow or cloud cover. NDWI also contributed significantly near water bodies, highlighting the cooling influence of aquatic surfaces.

To capture spatial heterogeneity and reduce multicollinearity, NDVI was further divided into internal (within stadiums) and external (within 250 m buffer) components. Results indicate that external NDVI had a stronger impact on LST prediction, pointing to the regulatory influence of surrounding green infrastructure. Residual analysis revealed that prediction errors were most prominent in high-temperature zones (>45 °C) and near stadium edges, where land cover complexity and material heterogeneity are highest. These discrepancies suggest the need for refined modeling in transitional zones between stadium interiors and adjacent built-up environments. Scatterplots show a tight clustering of predicted vs. observed LST values along the 1:1 line, confirming good model generalization, albeit with slight underfitting in thermal extremes.

4.3. Decoupling of the Three-Dimensional Driving Mechanism of the Stadium’s Thermal Mitigation Effect

The heatmap in Figure 13 illustrates the Pearson correlation coefficients between LST and a set of ecological, physical, and anthropogenic variables. Blue tones represent negative correlations (cooling influences), while red tones indicate positive correlations (warming influences). The analysis reveals clear directional patterns, suggesting that urban thermal environments result from the interplay of multiple, interdependent factors.

Among all variables, building volume exhibits the strongest positive correlation with LST (r = 0.81), highlighting the dominant role of high-density built environments in urban heat accumulation. Larger building volumes reduce vegetation, obstruct ventilation, and intensify heat storage, making them central contributors to the UHI effect.

In contrast, most other variables are negatively correlated with LST, reflecting their cooling potential through various mechanisms. NDVI (r = −0.53) shows the most significant negative correlation, reaffirming vegetation as the most reliable ecological moderator of urban temperatures. This finding aligns with earlier zone-based analyses, where vegetated areas consistently exhibited lower LSTs. Similarly, traffic density (r = −0.48) and green rate (r = −0.34) display moderate negative associations, indicating that while not dominant individually, they provide meaningful cooling through latent heat fluxes, surface permeability, and localized evaporative effects.

Some results reflect more complex interactions. For example, nighttime light intensity (r = −0.49)—typically interpreted as a proxy for human activity—shows a counterintuitive negative correlation with daytime LST. This may stem from urban form complexity, where dense districts with tall buildings produce shading and airflow effects that reduce surface temperatures during the day but retain heat at night, indicating a possible “day—hot, night—cool” reversal dynamic.

Proximity to water (r = −0.37) also correlates with cooling, especially around waterfront stadiums, confirming the moderating effect of aquatic surfaces on thermal diffusion. Conversely, albedo (r = −0.33), while traditionally associated with passive cooling, shows only a weak influence. This may reflect the contradiction between high reflectance and concurrent high heat storage in paved urban surfaces, where reflectivity does not equate to cooling efficiency in real-world contexts.

In sum, the results demonstrate that urban thermal behavior is not governed by isolated drivers, but by the dynamic coupling of ecological baselines and built-environment intensities.

Further regression analysis (

Table 7. Contribution analysis of driving factors affecting stadium thermal environment.

Factor Category	Driving Factor	Contribution (Standardized Coefficient)	Impact Direction	Significance (p Value)
Physical Attributes	LCT	0.35	Negative	0.001
	Albedo	0.21	Negative	0.013
Ecological Attributes	NDVI	0.42	Negative	0.000
	Green Ratio	0.38	Negative	0.002
	Proximity to Water	0.27	Negative	0.006
Anthropogenic Attributes	Nighttime Light Intensity	0.33	Positive	0.003
	Traffic Density	0.29	Positive	0.007
	Building Volume	0.36	Positive	0.001

) reveals differentiated influences of physical, ecological, and anthropogenic attributes on LST, providing a deeper understanding of the drivers behind stadium-based thermal mitigation.

Among physical variables, land cover type (LCT) shows the strongest cooling influence (standardized coefficient = −0.35, p = 0.001), indicating that natural surfaces such as turf within stadiums significantly reduce LST compared to impervious materials like concrete and asphalt. Although albedo exerts a weaker effect (−0.21), it still contributes to thermal moderation, especially when light-colored, high-reflectance surfaces are employed. These findings underscore the importance of land surface optimization within high-density urban areas like central Shanghai, where impervious cover dominates the landscape.

Ecological attributes present the strongest overall contribution to surface cooling. NDVI (−0.42) and green space ratio (−0.38) both exhibit highly significant negative relationships with LST (p < 0.01), confirming the fundamental role of vegetation in urban thermal regulation. Notably, in the Pudong New Area, where planning flexibility allows for greater integration of green infrastructure and stadium siting, these measures have tangibly improved thermal comfort. In addition, proximity to water bodies (−0.27) also supports the positive thermal regulation role of waterfront locations, reinforcing the advantage of placing large open facilities like stadiums near aquatic environments.

Conversely, anthropogenic factors such as building volume (+0.36), nighttime light intensity (+0.33), and road density (+0.29) are all positively associated with an elevated LST. These attributes represent intensified human activity and structural density, which tend to trap heat and obstruct ventilation. This pattern is especially pronounced in Shanghai’s urban core, where compact building forms, active nighttime economies, and extensive transport infrastructure limit the effectiveness of localized cooling, unless actively integrated into larger ecological and airflow systems.

The radar chart in Figure 14 synthesizes the relative contributions of ecological, anthropogenic, and physical attributes to stadium-scale thermal regulation. Among the three, ecological factors exert the strongest influence, with a normalized contribution score of 0.36. This reflects the dominant cooling role of vegetation coverage (NDVI, green space ratio) and proximity to water bodies, whose biophysical processes—particularly evapotranspiration and latent heat exchange—effectively suppress surface temperatures and foster urban cool-island formation.

Anthropogenic attributes follow with a contribution of 0.33, underscoring the significant thermal impact of urban development intensity, as represented by building volume, traffic density, and nighttime light emissions. These factors collectively intensify surface heat accumulation by increasing impervious coverage, disrupting airflow, and amplifying anthropogenic heat flux, especially in high-density urban cores.

Though slightly lower, physical attributes still play a measurable role (contribution = 0.28), primarily through their influence on surface thermal properties. Differences in land cover types (e.g., grass vs. asphalt) and surface albedo directly affect solar radiation absorption and heat retention, shaping near-surface thermal dynamics.

4.4. Heat Diffusion Simulation and Cold-Island Network Modeling

Figure 15 illustrates the spatial distribution of cooling diffusion zones around stadiums in central Shanghai and the Pudong New Area. The results reveal marked spatial heterogeneity in thermal regulation capacity. Stadiums embedded in ecologically favorable environments—such as Century Park Stadium and Huangpu Riverside Stadium—exhibit larger diffusion radii (700–900 m), forming well-defined localized cooling zones. These areas benefit from a dense vegetation cover and proximity to water bodies, which facilitate the outward propagation of cooling effects. In contrast, stadiums in compact, built-up districts—such as Jing’an Sports Center and Xuhui Sports Park—demonstrate smaller and fragmented diffusion patterns, with radii often below 500 m. This suggests that urban morphology and infrastructural density constrain the spatial reach of cool islands by limiting ventilation pathways and intensifying thermal retention.

Interestingly, overlapping diffusion zones are observed in high-density stadium clusters, particularly in central areas. These overlaps indicate the potential emergence of interconnected cooling corridors, wherein individual stadiums act as synergistic nodes. This emerging cold-island network provides an empirical foundation for strategic spatial coordination of cooling resources, aimed at mitigating UHI effects more effectively across larger spatial scales.

Figure 16 visualizes the structural connectivity among stadiums through a cold-island network model. Network analysis reveals a distinct core–periphery configuration, with high connectivity concentrated in the urban center. Stadiums such as Century Park Stadium, Hongkou Football Stadium, and Xuhui Sports Park display high degree centrality, indicating strong potential for ecological linkage with surrounding cool islands. These stadiums serve as key hubs, capable of facilitating thermal flow transmission across different urban zones. Their strategic positions suggest that enhancing their ecological connectivity could amplify their cooling influence beyond local domains.

Conversely, stadiums on the urban fringe, such as Putuo Sports Center and Minhang Sports Park, exhibit low degree centrality, reflecting isolated cooling effects and limited integration into the broader green infrastructure network. The identification of these structural disparities offers a new perspective on prioritizing spatial interventions—not just based on local cooling intensity but also on network-level connectivity and functional importance.

Table 8. Statistics of network centrality indexes of key stadium cold islands.

Stadium	Degree Centrality	Betweenness Centrality	Closeness Centrality
Stadium 1	0.286	0.048	0.467
Stadium 2	0.429	0.262	0.583
Stadium 3	0.286	0.024	0.467
Stadium 4	0.286	0.048	0.467
Stadium 5	0.429	0.262	0.583
Stadium 6	0.429	0.500	0.636
Stadium 7	0.286	0.286	0.467
Stadium 8	0.143	0.000	0.333

quantifies the centrality metrics of eight selected stadiums to assess their respective roles within the cold-island network.

In terms of degree centrality, Stadiums 2, 5, and 6 score the highest (0.429), signifying that they maintain direct ecological connections with a larger number of other nodes. Stadium 6 emerges as the most structurally critical node, with the highest values in both betweenness centrality (0.500) and closeness centrality (0.636). These scores indicate that it serves as a thermal transmission hub, minimizing the diffusion distance while bridging multiple cooling pathways. Stadium 6 functions as a keystone node whose performance directly affects the resilience and efficiency of the entire cooling network. Targeted protection and enhancement of this site should be prioritized.

In contrast, Stadium 8 ranks lowest across all three centrality metrics, including zero betweenness, positioning it as a peripheral cool source with minimal influence on overall network coordination. Its impact remains localized and disconnected, highlighting the need for improved integration into broader ecological corridors.

5. Discussion

Between 2018 and 2023, the LST in Shanghai’s central urban districts and the Pudong New Area demonstrated a consistent upward trajectory, underscoring the intensifying impacts of both regional climate warming and localized UHI effects. The most pronounced thermal deterioration occurred during the 2022–2023 period, which coincided with a series of extreme heat events and escalating anthropogenic heat emissions [53,54].

5.1. Regional LST Trends and Spatial Variability

Statistical analysis reveals that the average annual LST in Pudong increased from 33.4 °C in 2018 to 35.7 °C in 2023—an increase of approximately 6.9%. Meanwhile, the central urban area rose from 32.9 °C to 34.6 °C, reflecting a 5.2% increase. Peak summer months (July–August) saw localized hotspots, including Lujiazui and Xuhui Riverside, exceeding 38 °C, with some areas surpassing 41 °C.

Compared to similar cities in the Yangtze River Delta, such as Nanjing and Hangzhou, Shanghai exhibited a higher rate of LST increase, attributed primarily to its high-density development, large impervious surface area, and fragmented green infrastructure [55]. Notably, districts characterized by dense vegetation—such as Xuhui and areas near Century Park—maintained significantly lower LSTs, highlighting the cooling benefits of integrated urban greenery [56,57].

5.2. Intra-Stadium Thermal Zoning and Material Effects

Fine-scale thermal analysis of functional zones within stadiums (2018–2022) confirms substantial spatial heterogeneity in LST distribution [58]. Grass-covered zones consistently exhibited the lowest average LSTs. For example, Hongkou Football Stadium maintained grass-zone LSTs between 29.3 °C and 30.3 °C. In contrast, stand zones, affected by concrete and shaded structures, displayed LSTs ranging from 31.6 to 33.1 °C, while track areas, composed of impervious synthetic surfaces, exceeded 33 °C on average. In 2020, peak track LSTs reached 34.8 °C. At Shanghai Stadium, the annual mean LST in the track zone was 33.7 °C, approximately 3.6 °C higher than the adjacent grass field.

These patterns underscore the critical influence of surface material properties and functional zoning on urban thermal behavior. Impervious, low-albedo surfaces intensify heat storage and delay nocturnal cooling, leading to more persistent heat stress [59,60,61,62]. Furthermore, a comparative analysis across multiple stadiums reveals distinct variations in cooling performance linked to their internal functional zone composition and surrounding environment. Stadiums with higher proportions of vegetated zones and proximity to water bodies consistently exhibited stronger cooling effects and lower LST variability compared to stadiums dominated by impervious surfaces. This variability highlights the importance of local green infrastructure and site-specific morphological factors in modulating thermal behavior within stadium environments. Such cross-site comparisons enrich the understanding of cooling mechanisms and provide critical guidance for targeted urban design interventions. Accordingly, design strategies such as increasing permeable surfaces, optimizing the distribution of green zones, and applying high-albedo materials are essential to enhancing thermal performance in stadium environments.

5.3. Cool-Island Effects and Spatial Buffer Analysis

A multi-scale buffer analysis (radii: 100–500 m) centered on stadiums confirms the existence and spatial decay of cool-island effects. For instance, average LSTs near Hongkou Football Stadium were 31.5 °C (100 m), 32.4 °C (200 m), and 33.2 °C (500 m), reflecting a gradual convergence with the citywide thermal baseline. Similarly, the Oriental Sports Center exhibited an LST of 31.2 °C within the 200 m buffer, which was 1.8 °C cooler than the 500 m perimeter, attributed to its proximity to extensive water and vegetation systems.

These results are consistent with the existing literature indicating that the typical cooling radius of urban open spaces is within 300 m, beyond which the cool-island effect rapidly attenuates [63,64].

5.4. Site-Specific Variation in Cooling Intensity

Cool-island intensity—defined as the LST differential between the stadium core and surrounding buffer zones—varied significantly across study sites. Shanghai Oriental Sports Center recorded the higher cooling intensity, largely due to its well-integrated green infrastructure and low urban density in the surrounding area [65]. However, compared to typical urban parks or water bodies, stadiums represent a unique form of open space with inherent physical constraints that limit their cooling performance. Specifically, the presence of perimeter fences, high stand structures, and often impermeable surfaces within and around stadiums act as thermal barriers that impede natural airflow and weaken the lateral dispersion of cool air. These built elements significantly reduce the spatial reach of the cool-island effect when compared to unbounded green or blue spaces [66,67].

To compare the cooling performance of stadiums with other open spaces, we reviewed existing studies on the cooling effects of urban parks and water bodies. According to these studies, urban parks generally show a higher cooling intensity, especially within 300 m radii, due to their lack of physical barriers and better integration with surrounding green infrastructure. The findings suggest that stadiums, while effective, exhibit more constrained cooling zones due to their built structures [68].

Although proximity to vegetated buffers or water bodies (e.g., the Oriental Sports Center) appears to enhance localized cooling, the synergy is constrained by the aforementioned structural factors. In high-density urban districts, these barriers contribute to a phenomenon akin to “thermal entrapment”, where cooling mechanisms are spatially confined, resulting in thermal saturation under extreme ambient conditions [69]. Importantly, the results suggest that as regional background temperatures increase, the cooling efficiency of stadiums tends to decline. This thermal saturation effect indicates a threshold behavior beyond which passive cooling mechanisms lose effectiveness. Thus, stadium design must not only consider internal green zoning but also actively reduce perimeter obstructions and increase surface permeability to better emulate the thermal behavior of traditional green spaces. This finding underscores the urgency of incorporating climate resilience into the spatial design and planning of urban public spaces.

5.5. Policy Implications and Practical Recommendations

This study has demonstrated that stadiums, especially those situated within areas rich in green infrastructure and close to water bodies, can play a significant role in urban thermal regulation. The results suggest that stadiums in such environments, like Century Park Stadium and Huangpu Riverside Stadium, have a larger cooling diffusion radius compared to other open-space typologies such as parks or water zones. This positions stadiums as potential key players in mitigating UHI, particularly in neighborhoods with limited access to other cooling resources. In high-density districts, such as the urban core of Shanghai, stadiums could be strategically positioned to maximize their cooling effects and help alleviate the thermal stress in surrounding areas. This highlights the importance of integrating stadiums into broader urban thermal management strategies, where their cooling potential can be fully realized by considering local conditions, such as the presence of green spaces and water bodies.

A stadium-centered approach to urban planning can contribute to creating urban cooling corridors. These corridors would connect stadiums to surrounding green spaces, transport hubs, and other cooling infrastructures, further enhancing the thermal regulation capacity of the city. For instance, in areas with a high building density or scarce vegetation, stadiums could function as hubs that improve the overall connectivity of the urban green infrastructure. This approach would ensure a more integrated and efficient cooling network, providing sustained cooling effects over a wider area and mitigating heat in key urban zones. The multi-functional role of stadiums, from serving as event spaces to offering recreational and sheltering functions, makes them an ideal focal point for such planning strategies.

While stadiums offer substantial thermal mitigation benefits, it is essential to address the economic and spatial trade-offs associated with their development. The construction of new stadiums, particularly in densely built environments, can incur significant costs in terms of land use, infrastructure, and maintenance. However, it is important to clarify that the primary focus of this study is not to compare the economic feasibility of building new stadiums versus enhancing parks or water zones for thermal regulation. Rather, this research aims to provide guidelines for designing stadiums to maximize their cooling potential, focusing on their ability to serve as efficient thermal regulators within urban environments. In addition, while enhancing existing green spaces may be a more cost-effective and sustainable solution for urban cooling, stadiums—when strategically designed and integrated into urban planning—can serve as vital components of a city’s overall cooling strategy. This study presents a methodological framework for assessing the thermal benefits of stadiums, helping urban planners understand how stadiums can function as multi-functional cooling units that complement other urban green infrastructure. Future research could explore the economic and ecological trade-offs of various urban cooling strategies, but our study emphasizes the design optimization of stadiums as part of a broader thermal regulation system.

In conclusion, urban thermal management strategies should be adaptive and multi-dimensional, integrating various urban elements—such as stadiums, parks, and water bodies—into a cohesive system of thermal regulation. By prioritizing green infrastructure enhancements around stadiums and strategically linking them with existing ecological corridors, cities can not only improve thermal comfort but also foster a more resilient urban environment in the face of rising temperatures and climate change.

6. Conclusions

This study utilized multi-source remote sensing data, including MODIS and Landsat-derived LST products, integrated via the Google Earth Engine (GEE) platform, to analyze the spatial thermal characteristics of stadiums in Shanghai’s central urban districts and Pudong New Area between 2018 and 2023. Through multi-scale thermal effect evaluation, buffer-based cool-island modeling, and thermal driver identification, this study yielded the following key findings:

(1)

Urban thermal environment exhibited a persistent warming trend, exacerbating heat stress

From 2018 to 2023, both the central urban area and Pudong New Area experienced significant increases in average annual LST. Pudong’s LST rose from 33.4 °C to 35.9 °C, with a mean annual growth rate of 0.42 °C. The central urban area increased from 32.8 °C to 34.6 °C, at a rate of 0.36 °C/year. In extreme years such as 2022, regional peak LSTs frequently exceeded 38 °C, with local hotspots reaching 41.7 °C. This confirms not only the intensification of the UHI effect but also the spatial expansion of extreme thermal events.

(2)

Distinct thermal stratification exists across stadium functional zones

Thermal heterogeneity among internal functional zones of stadiums was significant and persistent. This demonstrates the strong influence of surface material type and spatial configuration on localized thermal behavior. In addition, comparative assessments across multiple stadiums underscore that the cooling performance of functional zones varies significantly depending on the stadium’s overall design and environmental context. Vegetation coverage, water proximity, and built-up density emerge as the most critical influencing factors, corroborated by correlation analyses. These factors collectively determine the magnitude and stability of cooling effects, highlighting the need to integrate multi-dimensional spatial planning to optimize thermal regulation in urban stadiums.

However, numerous studies [70,71,72] indicate that compared to areas with a certain density of tree cover combined with water bodies, grasslands have limited cooling effects. This limited cooling in stadium grass zones may be mainly attributed to the physical shading and structural barriers within stadiums, which influence solar radiation reception and airflow dynamics. Leveraging the innovation of high-resolution remote sensing data fusion highlighted in this study, it is feasible to identify and quantify daytime shading effects within stadiums. By integrating multi-source imagery, we can more accurately capture spatial variations in solar exposure across functional zones, thereby providing a clearer understanding of the interplay between shading, vegetation, and surface temperature distribution. This approach enhances the interpretation of intra-stadium thermal stratification and informs targeted cooling strategies.

(3)

Stadiums generate measurable cool-island effects within a 300 m effective range

Buffer analysis reveals a consistent cool-island effect radiating from stadium cores, with cooling intensity diminishing as distance increases. Stadiums with abundant surrounding greenery and water bodies, such as the Oriental Sports Center, showed an enhanced cooling performance. Furthermore, across multiple years, cooling effectiveness declined by 0.4–0.8 °C during 2022, due to elevated background LSTs. This suggests that extreme ambient thermal conditions may compromise the buffering capacity of open spaces, highlighting the need for resilient design strategies.

(4)

Multi-source LST fusion improves resolution, and key thermal drivers are quantitatively identified

The fusion of MODIS and Landsat LST products significantly improved both the spatial resolution and temporal coverage of urban thermal monitoring. The resulting dataset, with a spatial resolution of 30 m and a revisit interval of approximately eight days, is well-suited for mesoscale analysis of LST patterns in high-density urban contexts. Based on correlation and regression analyses, several key factors were identified as significantly influencing LST variation, including the NDVI, green space ratio, building volume, and proximity to water bodies. Among these, NDVI exhibited a moderate negative correlation with LST (r = −0.53), reaffirming the mitigating effect of vegetation on surface temperature through evapotranspiration and shading. In contrast, building volume showed the strongest positive correlation with LST (r = 0.81), highlighting the thermal intensification effects of densely built environments. Other variables, such as nighttime light intensity (r = −0.49) and traffic density (r = −0.48), also showed moderate negative correlations, suggesting that anthropogenic influences on LST are mediated by complex spatial and morphological conditions. It is important to note that thermal behaviors within stadium environments often exhibit nonlinear characteristics due to structural heterogeneity and internal land cover variation, which are not fully captured by simple linear metrics. Nevertheless, the integration of high-resolution fused LST data with multi-factor statistical models offers a robust framework for examining urban thermal dynamics. These findings contribute to a more nuanced understanding of the scale-dependent cooling functions of stadiums, while complementing broader urban heat island mitigation research.

In summary, this study highlights the dual role of stadiums as public amenities and spatial thermal regulators. Compared with other open spaces, stadiums encompass multiple functions such as hosting sports events, facilitating fitness activities, and serving as emergency shelters. Their high-frequency use throughout the year has driven a higher level of maintenance (e.g., grassland irrigation, renewal of surface materials) than ordinary open spaces, resulting in a more stable cooling effect. Their cooling performance is highly scale-sensitive, influenced by internal design, surrounding morphology, and macroscale climatic pressures. As urban heat challenges intensify, stadium-centered planning strategies that integrate ecological infrastructure and spatial openness will become increasingly essential to promoting urban thermal resilience.

Beyond empirical observations, this study makes a conceptual and methodological contribution by reframing stadiums as active, multi-scale thermal regulators within urban systems. The proposed SUIR model provides a new theoretical lens to examine how functional zoning and spatial morphology jointly shape localized cooling dynamics. Meanwhile, the MTCS enables high-resolution, temporally continuous LST reconstruction by fusing MODIS, Landsat, and Sentinel-1/2 data, offering a replicable methodology for urban thermal studies beyond single-source limitations. Furthermore, by introducing heat diffusion simulation and cold-island network modeling, this study advances the spatial understanding of stadium interactions within broader ecological infrastructures. These integrative approaches shift the analysis from isolated cooling patches to a network-based paradigm of urban thermal mitigation, with significant implications for climate-adaptive planning in megacities.