Cooling Efficiency of Urban Green Spaces Across Functional Zones: Mitigating Heat Island Effects Through Spatial Configuration

Abstract

With the acceleration of urbanization, the urban heat island effect has garnered increasing attention. However, few studies have explored the differential impacts of urban green spaces on the UHI across various urban functional zones (UFZs). This study takes Xiamen Island as the research object and selects nine representative landscape pattern indices to characterize the spatial patterns of UGS in each urban functional zone. Through Pearson correlation analysis, four landscape indices—largest patch index (LPI), mean patch area (AREA_MN), area-weighted average shape index (SHAPE_AM), and aggregation index (AI)—were chosen to reveal the varying influences of UGS spatial patterns on the UHI in different urban functional zones. These four landscape indices reflect aspects such as area, shape complexity, density size, and variation, as well as the aggregation of UGS. To address the spatial autocorrelation of variables, a spatial regression model was established. Given that the parameters of the spatial lag model outperformed those of the spatial error model, the spatial lag model was selected. Key findings reveal that the cooling efficiency of UGS varies across UFZs. In urban residential zones (URZs), UGS with complex shapes significantly enhances cooling, as indicated by a negative correlation between SHAPE_AM and LST (β = −0.446, p < 0.05). In urban village zones (UVZs), larger green patches have a stronger cooling effect, with AREA_MN showing a significant negative correlation with LST (β = −1.772, p < 0.05). The results indicate that UGS in different urban functional zones plays distinct roles in mitigating the UHI, with its cooling effects being associated with the spatial patterns of UGS. Therefore, it is recommended to adopt differentiated planning strategies for UGS in various urban functional zones to contribute to a more sustainable and thermally comfortable urban environment.

1. Introduction

Over half of the global population resides in urban areas, with projections estimating that this figure will reach 70% by 2050 [1]. The rapid growth of urban populations and the swift aggregation of human activities in cities have accelerated changes in land use and land cover, leading to significant differences between urban and rural areas in terms of anthropogenic emissions and surface thermal characteristics. Urban areas exhibit notably higher air and surface temperatures compared to surrounding rural regions, a phenomenon known as the urban heat island (UHI) effect [2].

The UHI effect has been observed worldwide, particularly in densely populated metropolises [3,4]. This phenomenon contributes to increased consumption of cooling energy and water resources during hotter seasons [5]. It also exacerbates air and water pollution, alterations in biological communities [6], and thermal health risks for urban inhabitants [7]. Given the significant threats posed by the UHI effect to human health and sustainable urban development, there has been a surge in interest among researchers, city managers, and the public in mitigating urban heat [8,9].

Impervious surfaces, water bodies, and vegetation are key indicators affecting the UHI index in urban landscapes [10,11]. Impervious surfaces enhance surface sensible heat flux, leading to increased surface temperatures [12], while the high heat capacity of water bodies and the evaporative cooling of vegetation can reduce surface temperatures [13]. Therefore, the rational configuration of landscapes has become one of the effective measures to alleviate the UHI effect [14]. Strategies to improve the urban thermal environment include optimizing both 2D and 3D urban morphology [15], utilizing reflective surfaces, and enhancing blue–green infrastructure [16].

Extensive research has demonstrated the pivotal role of urban green spaces (UGSs) in mitigating UHI through shading, evapotranspiration, and airflow regulation [17]. However, the cooling efficiency of UGS is highly heterogeneous, influenced by factors such as spatial configuration, patch characteristics, and urban functional zoning [18]. Recent studies emphasize that UGS cooling effects are not uniform across urban landscapes. For instance, Zhou et al. (2017) identified that tree canopy configuration significantly impacts microclimate regulation, with clustered trees in residential areas providing greater cooling than dispersed arrangements [19]. Similarly, Li et al. (2020) revealed that UGS shape complexity (e.g., fractal edges) enhances energy exchange between vegetation and built-up areas, thereby reducing localized temperatures [20]. Urban green spaces (UGSs) have garnered particular attention due to their widespread distribution in cities and their significant role in mitigating urban heat. Increased UGS coverage can reduce urban average surface temperatures by 4.73 °C, while optimizing UGS spatial distribution can reduce urban average surface temperatures by 1.06 °C [21]. As green space coverage increases, surface temperatures decrease linearly, but temperatures can vary by up to 8 °C within the same coverage rate [22]. Green spaces directly contribute to lowering the temperature of the surrounding environment through transpiration and shading [23]. The impact of green space patch size and shape on surface temperature has become a recent research focus [24]. Diverse types and compositions of urban green spaces have a stronger cooling effect than those comprised of simple structures [25]. Addressing the urban heat island issue from the perspective of landscape patches, especially the types, configurations, compositions, and shapes of green and blue patches, is crucial [26]. Enhancing the connectivity of green spaces is a more effective cooling measure than altering the size or shape of existing green spaces [27]. Despite these advances, most studies have focused on aggregated UGS metrics (e.g., total area or coverage) while overlooking the interplay between UGS spatial patterns and urban functional zones—a critical determinant of thermal dynamics.

Urban functional zones (UFZs), defined as regions with distinct socioeconomic activities and land use characteristics, exhibit unique thermal behaviors due to variations in surface materials, building density, and anthropogenic heat emissions [28]. Zhang et al. (2021) further highlighted that UGS in commercial zones requires specific spatial configurations (e.g., connectivity) to counteract heat generated by high human activity [29]. Nevertheless, existing research predominantly employs traditional regression models, neglecting spatial autocorrelation inherent in thermal data, which risks biased estimations and incomplete insights [30].

A critical gap persists in understanding how UGS spatial patterns—quantified through landscape metrics—interact with UFZs to modulate UHI. While Liu et al. (2023) explored seasonal variations in UGS cooling effects [31], their analysis lacked functional zone-specific recommendations. Similarly, Ke et al. (2021) identified variance in UGS impacts across zones but did not integrate spatial regression techniques to address autocorrelation [32]. These limitations underscore the need for a spatially explicit framework that links UGS configuration to UFZ-specific thermal outcomes.

This study addresses these gaps by investigating the differential impacts of UGS spatial patterns on UHI across five UFZs in Xiamen Island, China. Utilizing high-resolution remote sensing data and spatial regression models (spatial lag model, SLM), we quantify how landscape metrics modulate LST in residential, commercial, industrial, and public service zones. Our work advances current knowledge in three key aspects: (1) We provide empirical evidence on how UGS cooling mechanisms vary across UFZs, challenging the “one-size-fits-all” approach in urban planning. (2) By adopting SLM, we account for spatial dependencies in thermal data, enhancing model accuracy and policy relevance. (3) We dissect the roles of patch size, shape, and aggregation—metrics often analyzed in isolation—to offer holistic design guidelines. This approach not only refines UHI mitigation strategies but also establishes a replicable methodology for cities facing similar climatic and urbanization pressures.

2. Methods

2.1. Study Area

Xiamen City is located on the southeast coast of Fujian Province, at approximately 117°53′~118°26′ E and 24°23′~24°54′ N, and is an important node in the economic zone on the west coast of the Taiwan Strait and the Maritime Silk Road. The terrain is mainly composed of coastal plains, plateaus, and hills, with the terrain sloping from the northwest mountainous area to the southeast coastal area. Located in the global subtropical monsoon climate zone, it is also influenced by the oceanic climate, with warm and humid winters and hot and humid summers. The annual average temperature is around 21 degrees Celsius, with the lowest average temperature in January and the highest average temperature in August. The average annual rainfall is around 1200 mm, with the highest rainfall from May to August. The total land area of the region is 1699.39 km², of which Xiamen Island has a land area of 132.5 km². Xiamen Island, as the political and economic epicenter of Xiamen City, is not only highly urbanized with an advanced road system, but also notably exhibits a pronounced urban heat island effect (Figure 1). Being a key city in the southwestern Fujian economic zone, Xiamen has experienced rapid urbanization and significant landscape transformations over the past decades, which have exacerbated the urban heat island effect (UHI) and led to substantial thermal environmental stress [33,34]. The objective of this study is to devise a planning framework for the thermal environment of Xiamen based on landscape pattern metrics of urban green space. By incorporating digital elevation model (DEM) topographic maps and considering practical research requirements, areas such as mountains, water bodies, airports, and docks, which are infeasible for landscape pattern regulation, have been excluded.

2.2. Model Methods

2.2.1. Functional Urban Zoning

In this study, we obtained land use and cover data for Xiamen Island by interpreting high-resolution imagery from Google Earth. These images, with a spatial resolution of 0.14 m, were captured from 10 March to 14 November 2018. This period was chosen due to its clear weather conditions and minimal cloud cover, which are crucial for accurate land surface temperature (LST) data. This date also falls within the autumn season, a period of significant transition for the urban heat island (UHI) effect, allowing for a representative analysis of UHI dynamics. Additionally, the data quality and availability on this date made it suitable for the study. Based on the administrative boundaries of Xiamen Island, we precisely stitched and clipped the imagery. Subsequently, through meticulous visual interpretation, we exhaustively classified the land use and cover features depicted in the imagery. To ensure the accuracy of the classification, we randomly selected 200 sample points for accuracy assessment, which revealed a classification accuracy of 96%. In identifying urban functional zones (UFZs), this study particularly focused on urban road networks that may impede local heat dissipation, a concept originating from the research by Sun et al. in 2013 [35]. Considering the actual conditions and research objectives, we divided the interpretation results of remote sensing images according to socioeconomic functions and adjusted them in conjunction with digital elevation model (DEM) topographic maps, excluding mountainous areas, water bodies, and airports and docks, which do not participate in landscape pattern regulation. According to the research by Wang et al. (2025), land use types in Xiamen Island are categorized into five distinct urban functional zones: urban residential zones (URZs), urban village zones (UVZs), commercial zones (COZs), municipal public facilities zones (MUZs), and industrial and warehouse zones (IWZs) [36], Urban functional zones and urban green spaces are shown in Figure 2.

2.2.2. Temperature Inversion

This study used Landsat 8 images taken on 22 September 2019, with cloud cover below 5% and no cloud obstruction in the study area. The Landsat 8 image consists of 11 spectral bands, including 1 panchromatic band (15 m~15 m) and 9 multispectral bands (30 m~30 m) obtained by the Operational Land Imager (OLI). The two thermal infrared bands (100 m~100 m) were obtained by the Thermal Infrared Sensor (TIRS). Using an atmospheric correction algorithm for surface temperature inversion, the image was preprocessed with radiation correction and geometric correction, with an accuracy of >0.5 pixels. The basic principle of the atmospheric correction method is as follows: First, estimate the impact of the atmosphere on surface thermal radiation, and then subtract this part of the atmospheric impact from the total amount of thermal radiation observed by satellite sensors to obtain the intensity of surface thermal radiation. Then, convert this intensity of thermal radiation into the corresponding surface temperature.

The thermal infrared radiation brightness value L λ received by the satellite sensor consists of three parts: atmospheric upward radiation brightness L ↑; the true radiation brightness on the ground, which reaches the energy of the satellite sensor after passing through the atmosphere; and the energy reflected by downward radiation from the atmosphere upon reaching the ground. The expression for the thermal infrared radiation brightness value L λ received by satellite sensors can be written as (the radiation transfer equation):

L λ = [ε B (T_S) + (1 − ε) L ↓] τ + L ↑(1)

T_S = K₂/ln (K₁/B (T_S) + 1)(2)

In the formula, B (T_S) is the blackbody thermal radiation brightness; TS is the surface temperature; L ↑ represents the atmospheric upwelling radiation brightness; L ↓ is the atmospheric downward radiation brightness, and K₁ and K₂ are constants. The thermal infrared band L ↑ of Landsat 8 is 1.25 W/(m ² · SR · μ m), L ↓ is 2.37 W/(m ² · SR · μ m), K₁ is 774.89, and K₂ is 1321.08.

2.2.3. Explanatory Variables

Landscape metrics are the key variables for explaining variations in surface temperature. To circumvent potential multicollinearity issues arising from redundant indicators, we selected eight metrics to describe the spatial pattern of urban green spaces (UGSs): number of patches (NP), patch density (PD), largest patch index (LPI), landscape shape index (LSI), mean patch area (AREA_MN), area-weighted mean patch shape index (SHAPE_AM), mean Euclidean nearest-neighbor distance (ENN_MN), and aggregation index (AI) (Table 1). NP is a simple measure of the extent of subdivision or fragmentation of the patch type. PD, as a fundamental metric of landscape structure, reveals the degree of landscape fragmentation, which is particularly significant for applications across landscapes of varying scales. LPI quantifies the percentage of total landscape area comprised by the largest patch. LSI is a standardized metric used to measure the complexity of landscape edges. AREA_MN, as a patch size metric, reflects the average size of patches, assessing the overall average condition of UGS within each study unit. SHAPE_AM, as a shape metric, evaluates the complexity of the spatial configuration of the landscape within study units. ENN_MN measures the distance between UGS patches within each study unit. Finally, AI is an important metric for measuring the degree of aggregation of patches of the same type within a landscape, which can be used to analyze landscape connectivity and fragmentation. The calculation of landscape pattern metrics was conducted using Fragstats v4.2 software.

2.2.4. Spatial Autocorrelation

Everything is related to everything else, but near things are more related than distant things [37]. Spatial autocorrelation analysis is used to examine the statistical distribution patterns of spatial data, and its results depend on the spatial distribution of the data, differing from traditional quantitative statistics. Spatial autocorrelation includes global and local spatial autocorrelation. Global spatial autocorrelation analysis is used to determine the existence of spatial autocorrelation in the study area; local spatial autocorrelation analysis identifies the local regions where spatial autocorrelation phenomena occur [38]. Global autocorrelation is achieved through spatial autocorrelation indices, the most famous of which is the Global Moran’s I [39]. GeoDa (v1.22) is used for spatial autocorrelation analysis, including the creation of spatial weight matrices, obtaining Moran’s I (testing its significance level through Monte Carlo simulation methods), and LISA cluster maps [40]. Global Moran’s I and LISA represent global and local spatial autocorrelation, respectively.

2.2.5. Spatial Regression Models

The most fundamental spatial regression models are the spatial lag model (SLM), which accounts for spatial autocorrelation of variables, and the spatial error model (SEM), which considers the correlation of random errors [30]. The SLM equation is as follows:

(3)$\mathrm{Y}=\mathsf{\rho }\mathrm{W}\mathrm{Y}+\mathrm{X}\mathsf{\beta }+\mathsf{\epsilon }$

In this equation, Y is the dependent variable, X is the independent variable, W is the spatial weight matrix, ρ is the coefficient of the spatial lag term WY, $\mathsf{\beta }$ is the parameter vector, and $\mathsf{\epsilon }$ is the residual. In this model, the dependent variable of the study area is not only related to the independent variables of the area itself, but is also influenced by the dependent variables of adjacent areas.

The SEM equation is as follows:

(4)$\mathrm{Y}=\mathrm{X}\mathsf{\beta }+\mathsf{\epsilon }$

(5)$\mathsf{\epsilon }=\mathsf{\lambda }\mathrm{W}\mathsf{\epsilon }+\mathsf{\mu }$

In these equations, Y is the dependent variable, X is the independent variable, $\mathsf{\beta }$ is the parameter vector, W is the spatial weight matrix, $\mathsf{\lambda }$ is the coefficient of the spatial error term, and $\mathsf{\mu }$ is the regression residual vector. In this model, the dependent variable of the study area is not only related to the independent variables of the area itself, but also to the independent and dependent variables of adjacent areas.

To determine between the spatial lag model and the spatial error model, one can compare the two Lagrange multiplier (LM) tests and their robustness (Robust) [41]. If the Lagrange multiplier (lag) is statistically more significant than the Lagrange multiplier (error), and the robust LM (lag) is significant while the robust LM (error) is not, then the spatial lag model should be chosen. Conversely, if the Lagrange multiplier (error) is statistically more significant than the Lagrange multiplier (lag), and the robust LM (error) is significant while the robust LM (lag) is not, then the spatial error model is more appropriate.

3. Results

3.1. The Spatial Characteristics of LST

The land surface temperature (LST) in the study area is categorized into four classes using the Jenks natural breaks method: low temperature (30.18~34.93 °C), medium temperature (34.93~36.67 °C), high temperature (36.67~38.44 °C), and extremely high temperature (38.44~44.12 °C). Industrial and storage land in the northwest direction has higher temperatures, while parks, green spaces, and areas around water bodies have lower temperatures. The medium- and high-temperature zones account for 71.7% of the study area, with the low- and extremely high-temperature zones each comprising about 14% of the entire study area (Figure 3). Within each functional zone, the characteristics of temperature zoning differ; the area proportion of low-temperature zones gradually decreases, while that of extremely high-temperature zones gradually increases. The area proportion of low-temperature zones is the largest in the urban residential zone (URZ), accounting for 30.2% of the total area of URZ, and the area proportion of extremely high-temperature zones is the smallest in URZ, accounting for 1.4% of the total area of URZ. The area proportion of low-temperature zones is the smallest in the industrial and warehouse zone (IWZ), accounting for 0.4% of the total area of IWZ, and the area proportion of extremely high-temperature zones is the largest in IWZ, accounting for 30.9% of the total area of IWZ. The medium- and high-temperature zones remain the dominant areas in each functional zone (Figure 4).

Figure 5 presents the Local Indicators of Spatial Association (LISA) cluster map for land surface temperature (LST) in the study area. The map highlights significant local spatial correlation patterns, primarily categorized into high–high and low–low clusters. High–high clusters (indicated in red) are areas where high LST values are surrounded by other high LST regions. These clusters are predominantly located in the northwest part of the study area, where industrial and storage lands are concentrated. This spatial pattern suggests that these regions experience elevated temperatures due to concentrated industrial activities and reduced green spaces. Conversely, low–low clusters (indicated in blue) represent areas with low LST values surrounded by other low LST regions. These clusters are mainly found around parks, green spaces, and water bodies in the southern part of the study area. The presence of these clusters indicates that these regions benefit from the cooling effects of green spaces and water bodies, resulting in lower surface temperatures. The spatial distribution of LST clusters reveals a clear divide between the northwest and southeast parts of the study area. The northwest exhibits a high–high cluster pattern, reflecting the impact of urbanization and industrial activities on the urban heat island (UHI) effect. In contrast, the southeast shows a low–low cluster pattern, highlighting the role of green spaces and water bodies in mitigating the UHI effect.

3.2. Landscape Metrics Analysis

In the detailed analysis of landscape metrics across five urban functional zones, URZ exhibited the highest values for NP (3434), PD (920.940), LSI (73.485), and SHAPE_AM (1.803). Both NP and PD showed a clear decreasing trend from URZ to UVZ, COZ, MUZ, and IWZ. UVZ ranked second in NP (1227), PD (329.060), LPI (0.928), and LSI (73.485), indicating relatively high landscape fragmentation and density. COZ displayed moderate values for most landscape metrics, generally falling at the middle level among the five functional zones. This suggests a balanced spatial configuration of green spaces in COZ. MUZ had the highest LPI (1.550), indicating high landscape connectivity. However, LSI (19.747) and SHAPE_AM (1.580) were relatively low, suggesting simpler or more regular patch shapes. Additionally, MUZ exhibited higher AREA_MN (0.093), ENN_MN (85.972), and AI (67.579), reflecting its spatial distribution characteristics. IWZ had the lowest NP (138) and PD (37.009), indicating low green space fragmentation and density. Conversely, IWZ had the highest ENN_MN (151.696) and AI (68.250), suggesting widely dispersed patches with low spatial connectivity. This pattern may reflect ecological fragmentation issues but also potential ecological isolation advantages (Table 2).

3.2.1. Spatial Autocorrelation Analysis

Within the entire study area, the Moran’s I value for land surface temperature (LST) was 0.279, indicating a positive spatial autocorrelation in the distribution of LST. Monte Carlo simulation experiments revealed that the spatial autocorrelation of LST was significant at the 0.001 level, with a z-value of 23.981. This result underscores a pronounced clustering pattern in the spatial distribution of LST. Furthermore, the Moran’s I values for all individual urban functional zones were greater than zero, providing additional evidence for the positive autocorrelation in the distribution of LST.

3.2.2. Pearson Correlation Analysis

Through Pearson correlation analysis, we can clearly observe significant correlations between different landscape metrics and between LST and these landscape metrics. As shown in Figure 6, the correlations between LST and LPI, AREA_MN, SHAPE_AM, and AI are statistically significant at the 0.01 level. LPI is an indicator of the dominant patch types in the landscape, and its significant correlation with LST reveals the potential key role that dominant patch types may play in regulating the urban thermal environment. Furthermore, AREA_MN and SHAPE_AM are indicators reflecting patch area and shape complexity, and their significant correlations further highlight the potential importance of urban green space patterns in mitigating the urban heat island effect. The aggregation index (AI) examines the connectivity between patches of various landscape types; the smaller the value, the more fragmented the landscape.

By selecting the four landscape indices (LPI, AREA_MN, SHAPE_AM, AI) based on ecological significance, statistical significance, and practical relevance, we ensured that the indices used in our study provided a comprehensive and robust representation of the spatial patterns of UGS and their impact on the UHI effect. This selection process enhances the methodological transparency and strengthens the validity of our findings. Therefore, selecting these four landscape indices to study the crucial role of urban green space (UGS) patterns in regulating the urban heat island effect lays a foundation for subsequent spatial regression analysis.

3.2.3. Spatial Regression Model Selection

The p-value associated with Moran’s I is 0.0000, which at the 1% significance level strongly rejects the null hypothesis that “there is no spatial correlation in the residuals”, indicating significant spatial correlation in the residuals of the ordinary least squares (OLS) regression model. Although both the spatial lag (LM lag) and spatial error (LM error) models show significance, the robust spatial lag model is more significant than the spatial error model, leading to the selection of the spatial lag model (SLM).

To determine the most appropriate model, we conducted tests for the Lagrange multiplier (LM) and Robust Lagrange multiplier (robust LM) indicators [42]. Table 3 presents the results of the LM and robust LM tests for spatial correlation. The data from the table show that both the lag and error test statistics of the LM are highly significant, and the values for the spatial lag model (SLM) are greater than those for the spatial error model (SEM), indicating a strong spatial dependence in this context, making SLM preferable to SEM. Further robust LM tests for SLM and SEM reveal that the robust LM for SLM is significantly different, while that for SEM is not, suggesting that SLM is more suitable than SEM. In summary, the results from both LM and robust LM tests indicate that SLM is more appropriate for this study than SEM. Therefore, we selected the SLM to further explore indicators for mitigating the urban heat island effect (Table 3).

3.2.4. Spatial Regression Analysis

The results from the spatial lag model reveal that the impact of urban green spaces (UGS) on the urban heat island (UHI) effect varies across the entire study area and among the five urban functional zones. Overall, the proportion of urban green spaces plays a significant role in mitigating UHI and exhibits a negative correlation with LST, which is conducive to alleviating UHI. Specifically, the correlation between LST and UGS landscape metrics differs significantly across functional zones. In URZ, the shape complexity of UGS significantly affects LST; in UVZ, the mean patch area is significantly positively correlated with LST, with a correlation coefficient of −1.722; in MUZ, the proportion of UGS and patch connectivity play an important role in UHI mitigation. The proportion of UGS has a relatively stable and significant negative impact on surface temperature across the entire study area and public service areas. Regardless of the urban functional zones, the impact of LPI on surface temperature is not significant (Table 4).

To ensure the validity of our spatial regression models, we conducted diagnostics using a scatter plot of the residuals. The residual plot (Figure 7) displays the distribution of residuals relative to the predicted values. The residuals are randomly scattered around zero without any discernible pattern, indicating that the model assumptions are met. This plot reveals a somewhat random distribution of points, and these diagnostics confirm that our models are robust and that the fundamental assumptions of our regression analysis are largely satisfied.

4. Discussion

4.1. Differences in the Impact of Urban Green Spaces on Urban Heat Islands

Our study emphasizes the differential role of urban green spaces (UGSs) in mitigating the urban heat island (UHI) effect across various urban functional zones. The composition and spatial layout of UGS significantly influence the urban thermal environment [43]. Optimizing the spatial configuration of existing UGS is an effective method to alleviate the urban thermal environment. UGS exerts different climate regulation functions based on diverse landscape characteristics, such as diversity and configuration [31]. The negative correlation between urban green spaces and land surface temperature (LST) indicates that the higher the coverage, fragmentation, and richness of UGS types, the more pronounced the cooling effect [44]. Increasing the urban green space ratio (UGSR) is an effective approach to alleviate the urban thermal environment [9].

The landscape configuration and composition of UGS can effectively mitigate the urban heat island effect [45]. UGS can effectively reduce environmental temperatures through shading and transpiration effects [46]. Liu et al. (2024) found that when the urban green space ratio (UGSR) was in the range of 20% to 30%, a significant correlation was presented with the highest correlation coefficient of AREA_MN, which reached 0.267 (p < 0.01) [44]. This is consistent with the findings in this study that AREA_MN in the urban village zone (UVZ) is significantly positively correlated with LST. The negative correlation of AREA_MN results indicates that larger UGS patch sizes have better cooling effects.

The shape index SHAPE_AM, representing the shape of UGS, shows a significant negative correlation because the increased complexity of UGS shape can promote energy exchange between vegetation and impervious areas, thereby reducing surface temperatures [32]. Complex UGS patterns are beneficial in mitigating UHI [9]. The significant negative correlation between UHI and SHAPE_MN is as expected, as urban green spaces with higher SHAPE_MN can provide more shading and increase energy exchange between green spaces and impervious surfaces, thereby reducing surface temperatures [20].

In the study area and all individual urban functional zones, LPI was not found to have a significant impact on the urban heat island index. This finding is consistent with the studies of Liu et al. (2013) [44]. Ren et al. (2025) found that the largest patch index (LPI) exhibits a more complex relationship, transitioning from a negative to a positive correlation after surpassing a threshold value of 30. This suggests that there are intricate interactions occurring within the urban environment [47].

In MUZ, AI has a significant positive impact on the urban heat island. This is consistent with the conclusion of Yu et al. (2020), that large patch gaps are not conducive to reducing UHI [48]. However, individual urban functional zones other than the public service zone are not affected by this. Increasing the aggregation of green space patches can enhance the cooling effect of urban green areas [44]. Isolated green space conditions are negatively related to the reduction in LST, indicating that isolated green spaces lack effective connections with the surrounding environment, thus failing to effectively promote the dispersion of heat and reduce surface temperatures.

The results indicate that differentiated urban green space planning strategies are necessary in specific urban functional zones to mitigate the urban heat island effect. The results show that increasing the proportion of urban green space ratio is the most effective method to alleviate urban heat stress. Therefore, the focus of UGS planning should be on increasing urban green spaces through planting more trees, which has also been suggested in previous studies [13]. Hence, the emphasis of UGS planning should be on augmenting urban green spaces via the planting of additional trees. Therefore, when conducting UGS planning, the landscape characteristics of UGS must be considered.

4.2. The Suitability of the Methodology

Choosing the appropriate methods and models is key to analyzing the relationship between the urban heat island effect and its influencing factors. Traditional regression methods, such as ordinary least squares (OLS) regression, cannot account for spatial autocorrelation due to the spatial independence of the dependent variable, rendering the model invalid and the results misleading. While geographically weighted regression (GWR) and multiscale geographically weighted regression (MGWR) are powerful tools for spatial analysis, they were not included in this study due to model complexity, data requirements, and interpretability. GWR and MGWR models are computationally intensive and require a substantial amount of data to estimate local parameters effectively. Given the scope of our study and the available resources, we opted for spatial regression models that provided a balance between complexity and interpretability. Our study aimed to assess the overall relationship between urban green spaces (UGSs) and the urban heat island (UHI) effect across different urban functional zones, and the spatial regression models we used were more suitable for this purpose. We prioritized models that provided clear and actionable insights, making them more appropriate for our research questions and data. Since spatial regression can effectively handle spatially dependent data (e.g., LST), it is necessary to select spatial regression when studying how to mitigate the urban heat island effect. At the same time, attention should be paid to whether the spatial lag model (SLM) or the spatial error model (SEM) is more suitable for specific case studies, as SLM is more effective when the dependent variable is influenced by nearby locations in the study area, while SEM is more applicable when the model’s error terms have spatial correlation [49]. Therefore, spatial regression is a promising method for addressing spatial correlation issues and is of great help in urban thermal environment research.

4.3. Planning Strategies for Urban Functional Zones

To operationalize the findings of this study, we propose actionable planning strategies tailored to the distinct characteristics of each urban functional zone (UFZ) in Xiamen Island. These recommendations integrate landscape metrics, spatial regression insights, and best practices from global case studies to enhance climate resilience and livability. In urban residential zones (URZs), designing urban green spaces (UGSs) with irregular shapes can maximize edge effects and energy exchange between green and built-up areas, such as pocket parks and linear green corridors with fractal edges to improve microclimate regulation. In commercial zones (COZs), improving the aggregation and connectivity of UGS patches can counteract heat generated by high human activity, such as developing green networks linking rooftop gardens, street trees, and plazas [50]. In industrial and warehouse zones (IWZs), increasing the mean patch area by creating centralized green buffers around industrial clusters can isolate heat sources and reduce particulate matter dispersion. Underutilized industrial lands can be converted into ecological parks with native vegetation to enhance biodiversity and cooling efficiency. In municipal public facilities zones (MUZs), distributing small UGS patches near hospitals, schools, and government buildings can ensure equitable access while preventing thermal inertia. “Cool routes” with shaded walkways and water features can be designed to connect public facilities, leveraging evaporative cooling effects. In urban village zones (UVZs), protecting large, contiguous UGS patches can maintain cooling benefits, with restrictions on infill development that fragments green areas [51]. Local communities can be engaged to co-design green spaces, integrating traditional knowledge with modern ecological practices. By tailoring UGS configurations to UFZ-specific thermal behaviors, cities like Xiamen can achieve dual goals of climate resilience and social equity. These strategies not only address the urban heat island (UHI) effect, but also advance global agendas for sustainable urbanization.

4.4. Limitations and Future Prospects

This study focuses on the spatial configuration of UGS and its impact on the urban heat island (UHI) effect. However, it is important to note that socioeconomic factors play a significant role in the accessibility and distribution of UGS. Previous studies have shown that UGS accessibility and distribution often vary significantly across different socioeconomic groups, raising concerns about environmental justice and social equity [52]. Low-income areas often lack sufficient green spaces, making them more vulnerable to the UHI effect and related health risks. Therefore, future research should incorporate socioeconomic data to explore the relationship between UGS distribution, socioeconomic factors, and the UHI effect. This would provide a more comprehensive understanding of the issue and help policymakers develop more equitable and effective urban planning strategies [53].

Although this study has thoroughly explored the differentiated impacts of urban green spaces (UGSs) on mitigating the urban heat island (UHI) effect across different urban functional zones, several limitations exist, providing directions for future research. Firstly, due to data constraints, this study only analyzed summer imagery and did not fully consider the potential impacts of seasonal vegetation changes on land surface temperature (LST), which may limit the comprehensiveness of the results to some extent. Secondly, this study did not include nighttime LST data because the used Landsat data only cover daytime conditions. Moreover, satellite sensors mainly observe horizontal surfaces and have limited observation capabilities for vertical walls, which significantly affect the microclimate of streets in densely built areas, thus making the representation of urban vertical structures through satellite remote sensing insufficient.

We recognize the potential of machine learning-based approaches, including deep learning, for predictive modeling of urban temperature dynamics. Machine learning techniques can effectively handle complex, non-linear relationships, and large datasets, making them suitable for urban temperature prediction. For example, deep learning models such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs) have been successfully applied to analyze satellite imagery and time-series temperature data, providing accurate predictions of urban temperature patterns. Future research could explore the integration of machine learning models with high-resolution thermal data to enhance the predictive capabilities and spatial resolution of urban temperature dynamics. This would not only improve the accuracy of temperature predictions but also provide more detailed insights into the factors influencing urban heat distribution.

Future research can be expanded in multiple aspects: First, integrating multi-temporal daytime and nighttime thermal data, such as medium-resolution ASTER data and low-resolution MODIS data, to comprehensively explore the diurnal relationship between the urban form and LST. Second, extending the research to cities with different climatic conditions to verify the universality of the results of this study. Third, combining the characteristics of specific vegetation types (such as trees, shrubs, and grasses) to deeply analyze their contributions to the cooling effects of UGS, providing more targeted suggestions for urban green space planning.

5. Conclusions

This study demonstrates that the cooling effects of urban green spaces (UGSs) on the urban heat island (UHI) effect are shaped by their spatial configuration within distinct urban functional zones. Spatial regression analysis revealed that in urban residential zones, the shape complexity of UGS (SHAPE_AM) significantly reduced land surface temperature (LST) (β = −0.446, p < 0.05), supporting the design of irregularly shaped green spaces to enhance edge effects. In urban village zones, a larger mean patch area (AREA_MN) exhibited a stronger cooling effect (β = −1.772, p < 0.05), emphasizing the importance of preserving contiguous green patches. Conversely, aggregated green spaces in industrial zones (aggregation index, AI: β = −0.012, p < 0.05) showed weaker cooling performance, suggesting fragmented distributions may be more effective. Notably, the largest patch index (LPI) displayed no significant correlation with LST (p > 0.05), challenging the prioritization of single large UGS patches for UHI mitigation.

These findings advocate for differentiated planning strategies: (1) prioritizing irregularly shaped green corridors in residential areas to maximize edge interactions; (2) preserving existing large green patches in urban villages; and (3) implementing distributed green belts in industrial zones to counteract localized heat sources. The insignificance of LPI implies that urban planners should focus on optimizing connectivity and geometric complexity of UGS rather than expanding singular large parks. Methodologically, the spatial lag model effectively addressed clustered thermal patterns (Moran’s I = 0.279, p < 0.001), underscoring the necessity of spatially explicit frameworks. These insights provide actionable guidance for subtropical coastal cities while highlighting the need for context-specific adaptations in global urban planning.

Author Contributions

Conceptualization, Y.R.; Methodology, Y.W.; Software, Y.W.; Formal analysis, Y.W.; Investigation, Y.W. and Y.R.; Data curation, Y.W.; Writing—original draft, Y.W.; Writing—review & editing, Y.R.; Supervision, Y.R.; Project administration, Y.R.; Funding acquisition, Y.R. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The data presented in this study are available on request from the corresponding author.

Acknowledgments

We thank the United States Geological Survey (https://www.usgs.gov, accessed on 22 September 2019) and the Geospatial Data Cloud site (http://www.gscloud.cn, accessed on 25 September 2019) for providing Landsat imagery free of charge. We are also grateful to reviewers for their helpful comments and suggestions for improving the manuscript.

Conflicts of Interest

The authors declare no conflicts of interest.

Footnotes

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Figures and Tables

Figure 1. Map of the study area.

Figure 2. Urban functional zones and urban green space in the study area.

Figure 3. Spatial distribution of LST in the study area.

Figure 4. The proportion of LST categories in each urban functional zone.

Figure 5. Aggregation map of LST in urban functional zones.

Figure 6. Pearson correlation coefficients for LST and landscape metrics.

Figure 7. Residual plot for spatial lag model analysis of urban heat island effect.

Table 1

Landscape metrics selected in the study.

Indicators	Abbreviation	Description	Formula	Unit
Number of Patches	NP	The number of patches in the landscape	NP = N	None
Patch density	PD	The density of patches of UGS patches	PD = ${\displaystyle \frac{N}{A}}($10,000)(100)	Number per 100 ha
Largest Patch Index	LPI	The percentage of the landscape comprised by the largest patch.	LPI = ${\displaystyle \frac{\mathrm{m}\mathrm{a}\mathrm{x}\left(a_{ij}\right)}{A}}$(100)	Percent
Landscape Shape Index	LSI	A standardized measure of total edge or edge density that adjusts for the size of the landscape.	LSI = ${\displaystyle \frac{0.25{\sum }_{k=1}^m{e}_{ik}^{*}}{\sqrt{A}}}$	None
Mean patch size	AREA_MN	The average size of UGS patches	MN = ${\displaystyle \frac{{\sum }_{j=1}^n{x}_{ij}}{n_i}}$	None
Area-weighted mean shape index	SHAPE-AM	The area-weighted average shape index of UGS patches	AM = ${\sum }_{j=1}^n\left[x_{ij}\left({\displaystyle \frac{a_{ij}}{{\sum }_{j=1}^n{a}_{ij}}}\right)\right]$	None
Mean Euclidean nearest neighbor distance	ENN_MN	The ENN weighted by the relative area of UGS patches	ENN_MN = ${\sum }_{i=1}^m{\sum }_{j=1}^n{h}_{ij}$	Meters
Aggregation Index	AI	The number of like adjacencies involving the corresponding class, divided by the maximum possible number of like adjacencies involving the corresponding class	AI = $\left[{\displaystyle \frac{g_{ii}}{max-g_{ii}}}\right](100)$	Percent

where N = total number of patches in the landscape, A = total landscape area (m²), includes any internal background present. $a_{ij}$ = area (m²) of patch ij. $e_{ik}^{*}$ = total length (m) of edge in landscape between patch types (classes) i and k; includes the entire landscape boundary and some or all background edge segments involving class i. $h_{ij}$ = distance (m) from patch ij to nearest neighboring patch of the same type (class), based on patch edge-to-edge distance, computed from cell center to cell center. $g_{ii}$ = number of like adjacencies (joins) between pixels of patch type (class) i based on the single-count method. $max-g_{ii}$ = maximum number of like adjacencies (joins) between pixels of patch type (class) I based on the single-count method.

Table 2

Values of landscape metrics of five urban functional zones.

TYPE	NP	PD	LPI	LSI	AREA_MN	SHAPE_AM	ENN_MN	AI
URZ	3434	920.940	0.839	73.485	0.063	1.803	28.385	50.299
UVZ	1227	329.060	0.928	36.387	0.057	1.616	41.006	56.852
COZ	672	180.219	0.523	30.849	0.059	1.719	57.717	51.371
MUZ	361	96.814	1.550	19.474	0.093	1.580	85.972	67.579
IWZ	138	37.009	0.499	12.378	0.099	1.575	151.696	68.250

Table 3

Spatial regression model.

TEST	MI/DF	VALUE	PROB
Moran’s I (error)	0.2608	22.3916	0.0000
Lagrange Multiplier (lag)	1	415.9703	0.0000
Robust LM (lag)	1	4.7843	0.0287
Lagrange Multiplier (error)	1	414.0992	0.0000
Robust LM (error)	1	2.9132	0.0879
Lagrange Multiplier (SARMA)	2	418.8834	0.0000

Table 4

Spatial lag model regression modeling results.

	Overall	Residential	Village	Commercial	Utilities	Industrial
Variable	Coefficient	Coefficient	Coefficient	Coefficient	Coefficient	Coefficient
LPI	0.688	−0.798	3.021	−1.423	−1.752	1.685
AREA_MN	−0.436	1.017	−1.772 *	−1.181	0.417	−7.508
SHAPE_AM	−0.446	−0.153 *	−0.917	−0.211	−0.819	0.811
AI	−0.002	−0.004	−0.014	0.009	−0.017 *	−0.012
UGS_prop	−2.630 *	1.050	−3.195	−0.052	−3.345 *	−4.636
R-squared	0.378	0.313	0.535	0.597	0.532	0.578
AIC	1474.79	645.76	313.99	367.148	252.74	117.57

* indicates p < 0.05.

References

1. Birch, E.L.; Wachter, S.M. Chapter 1. World Urbanization: The Critical Issue of the Twenty-First Century. Global Urbanization; Eugenie, L.B.; Susan, M.W. University of Pennsylvania Press: Philadelphia, PA, USA, 2011; pp. 3-23.

2. Oke, T.R. The Energetic Basis of the Urban Heat-Island. Q. J. R. Meteorol. Soc.; 1982; 108, pp. 1-24. [DOI: https://dx.doi.org/10.1002/qj.49710845502]

3. Ding, W.; Chen, H. Investigating the microclimate impacts of blue–green space development in the urban–rural fringe using the WRF-UCM model. Urban Clim.; 2024; 54, 101865. [DOI: https://dx.doi.org/10.1016/j.uclim.2024.101865]

4. Wang, X.; Li, B.; Liu, Y.; Yang, Y.; Fu, X.; Shen, R.; Xu, W.; Yao, L. Trends and attributions of the long-term thermal comfort across the urban–rural gradient in major Chinese cities. Appl. Geogr.; 2024; 164, 103221. [DOI: https://dx.doi.org/10.1016/j.apgeog.2024.103221]

5. Lin, Z.; Xu, H.; Yao, X.; Yang, C.; Ye, D. How does urban thermal environmental factors impact diurnal cycle of land surface temperature? A multi-dimensional and multi-granularity perspective. Sustain. Cities Soc.; 2024; 101, 105190. [DOI: https://dx.doi.org/10.1016/j.scs.2024.105190]

6. White, M.A.; Nemani, R.R.; Thornton, P.E.; Running, S.W. Satellite Evidence of Phenological Differences Between Urbanized and Rural Areas of the Eastern United States Deciduous Broadleaf Forest. Ecosystems; 2002; 5, pp. 260-273. [DOI: https://dx.doi.org/10.1007/s10021-001-0070-8]

7. Poumadère, M.; Mays, C.; Le Mer, S.; Blong, R. The 2003 Heat Wave in France: Dangerous Climate Change Here and Now. Risk Anal.; 2005; 25, pp. 1483-1494. [DOI: https://dx.doi.org/10.1111/j.1539-6924.2005.00694.x]

8. Cui, P.; Xv, D.; Tang, J.; Lu, J.; Wu, Y. Assessing the effects of urban green spaces metrics and spatial structure on LST and carbon sinks in Harbin, a cold region city in China. Sustain. Cities Soc.; 2024; 113, 105659. [DOI: https://dx.doi.org/10.1016/j.scs.2024.105659]

9. Yao, L.; Li, T.; Xu, M.; Xu, Y. How the landscape features of urban green space impact seasonal land surface temperatures at a city-block-scale: An urban heat island study in Beijing, China. Urban For. Urban Green.; 2020; 52, 126704. [DOI: https://dx.doi.org/10.1016/j.ufug.2020.126704]

10. Guo, L.; Liu, R.; Men, C.; Wang, Q.; Miao, Y.; Zhang, Y. Quantifying and simulating landscape composition and pattern impacts on land surface temperature: A decadal study of the rapidly urbanizing city of Beijing, China. Sci. Total Environ.; 2019; 654, pp. 430-440. [DOI: https://dx.doi.org/10.1016/j.scitotenv.2018.11.108]

11. Wu, Z.; Wang, Y.; Ren, Y. Optimizing green space-building landscape characteristics of key urban functional zones for comprehensive thermal environment mitigation. Landsc. Urban Plan; 2025; 257, 105314. [DOI: https://dx.doi.org/10.1016/j.landurbplan.2025.105314]

12. Kuang, W.; Yang, T.; Liu, A.; Zhang, C.; Lu, D.; Chi, W. An EcoCity model for regulating urban land cover structure and thermal environment: Taking Beijing as an example. Sci. China Earth Sci.; 2017; 60, pp. 1098-1109. [DOI: https://dx.doi.org/10.1007/s11430-016-9032-9]

13. Dai, Z.; Guldmann, J.-M.; Hu, Y. Spatial regression models of park and land-use impacts on the urban heat island in central Beijing. Sci. Total Environ.; 2018; 626, pp. 1136-1147. [DOI: https://dx.doi.org/10.1016/j.scitotenv.2018.01.165] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/29898520]

14. Guo, F.; Hu, D.; Schlink, U. A comprehensive metric scheme for characterizing the heterogeneity of urban thermal landscapes: A case study of 14-year evaluation in Beijing. Ecol. Indic; 2024; 166, 112268. [DOI: https://dx.doi.org/10.1016/j.ecolind.2024.112268]

15. Wu, W.; Guo, F.; Elze, S.; Knopp, J.; Banzhaf, E. Deciphering the effects of 2D/3D urban morphology on diurnal cooling efficiency of urban green space. Build. Environ.; 2024; 266, 112047. [DOI: https://dx.doi.org/10.1016/j.buildenv.2024.112047]

16. Zhang, H.; Kang, M.-Y.; Guan, Z.-R.; Zhou, R.; Zhao, A.-L.; Wu, W.-J.; Yang, H.-R. Assessing the role of urban green infrastructure in mitigating summertime Urban Heat Island (UHI) effect in metropolitan Shanghai, China. Sustain. Cities Soc.; 2024; 112, 105605. [DOI: https://dx.doi.org/10.1016/j.scs.2024.105605]

17. Zhou, W.; Huang, G.; Cadenasso, M.L. Does spatial configuration matter? Understanding the effects of land cover pattern on land surface temperature in urban landscapes. Landsc. Urban Plan; 2011; 102, pp. 54-63. [DOI: https://dx.doi.org/10.1016/j.landurbplan.2011.03.009]

18. Bowler, D.E.; Buyung-Ali, L.; Knight, T.M.; Pullin, A.S. Urban greening to cool towns and cities: A systematic review of the empirical evidence. Landsc. Urban Plan; 2010; 97, pp. 147-155. [DOI: https://dx.doi.org/10.1016/j.landurbplan.2010.05.006]

19. Zhou, W.; Wang, J.; Cadenasso, M.L. Effects of the spatial configuration of trees on urban heat mitigation: A comparative study. Remote Sens. Environ.; 2017; 195, pp. 1-12. [DOI: https://dx.doi.org/10.1016/j.rse.2017.03.043]

20. Li, X.; Zhou, W. Optimizing urban greenspace spatial pattern to mitigate urban heat island effects: Extending understanding from local to the city scale. Urban For. Urban Green.; 2019; 41, pp. 255-263. [DOI: https://dx.doi.org/10.1016/j.ufug.2019.04.008]

21. Xu, C.; Huang, Q.; Haase, D.; Dong, Q.; Teng, Y.; Su, M.; Yang, Z. Cooling Effect of Green Spaces on Urban Heat Island in a Chinese Megacity: Increasing Coverage versus Optimizing Spatial Distribution. Environ. Sci. Technol.; 2024; 58, pp. 5811-5820. [DOI: https://dx.doi.org/10.1021/acs.est.3c11048]

22. Wu, Z.; Zhou, Y.; Ren, Y. Green space-building integration for Urban Heat Island mitigation: Insights from Beijing’s fifth ring road district. Sustain. Cities Soc.; 2024; 116, 105917. [DOI: https://dx.doi.org/10.1016/j.scs.2024.105917]

23. Al Shawabkeh, R.; AlHaddad, M.; Al-Fugara, A.k.; Al-Hawwari, L.; Al-Hawwari, M.I.; Omoush, A.; Arar, M. Modeling the impact of urban land cover features and changes on the land surface temperature (LST): The case of Jordan. Ain Shams Eng. J.; 2024; 15, 102359. [DOI: https://dx.doi.org/10.1016/j.asej.2023.102359]

24. Cao, X.; Onishi, A.; Chen, J.; Imura, H. Quantifying the cool island intensity of urban parks using ASTER and IKONOS data. Landsc. Urban Plan; 2010; 96, pp. 224-231. [DOI: https://dx.doi.org/10.1016/j.landurbplan.2010.03.008]

25. Li, X.; Zhou, W.; Ouyang, Z.; Xu, W.; Zheng, H. Spatial pattern of greenspace affects land surface temperature: Evidence from the heavily urbanized Beijing metropolitan area, China. Landsc. Ecol.; 2012; 27, pp. 887-898. [DOI: https://dx.doi.org/10.1007/s10980-012-9731-6]

26. Fan, H.; Yu, Z.; Yang, G.; Liu, T.Y.; Liu, T.Y.; Hung, C.H.; Vejre, H. How to cool hot-humid (Asian) cities with urban trees? An optimal landscape size perspective. Agric. For. Meteorol.; 2019; 265, pp. 338-348. [DOI: https://dx.doi.org/10.1016/j.agrformet.2018.11.027]

27. Shih, W. Greenspace patterns and the mitigation of land surface temperature in Taipei metropolis. Habitat Int.; 2017; 60, pp. 69-80. [DOI: https://dx.doi.org/10.1016/j.habitatint.2016.12.006]

28. Chen, Y.; Yang, J.; Yang, R.; Xiao, X.; Xia, J. Contribution of urban functional zones to the spatial distribution of urban thermal environment. Build. Environ.; 2022; 216, 109000. [DOI: https://dx.doi.org/10.1016/j.buildenv.2022.109000]

29. Zhang, X.; Li, X.; Liu, H.; Song, Y.; Gao, M. Research on urban thermal environment differentiation based on functional zones-take four districts of Dalian as an example. Urban Clim.; 2024; 58, 102152. [DOI: https://dx.doi.org/10.1016/j.uclim.2024.102152]

30. Anselin, L. GIS research infrastructure for spatial analysis of real estate markets. J. Hous. Res.; 1998; 9, pp. 113-133. [DOI: https://dx.doi.org/10.1080/10835547.1998.12091930]

31. Liu, F.F.; Yan, W.J.; Kong, F.H.; Yin, H.W.; Ban, Y.L.; Xu, W.B. A review on the urban green space cooling effect based on field measurement of air temperature. Ying Yong Sheng Tai Xue Bao = J. Appl. Ecol.; 2017; 28, pp. 1387-1396.

32. Ke, X.L.; Men, H.L.; Zhou, T.; Li, Z.Y.; Zhu, F.K. Variance of the impact of urban green space on the urban heat island effect among different urban functional zones: A case study in Wuhan. Urban For. Urban Green.; 2021; 62, 127159. [DOI: https://dx.doi.org/10.1016/j.ufug.2021.127159]

33. Dong, J.; Lin, M.X.; Zuo, J.; Lin, T.; Liu, J.K.; Sun, C.G.; Luo, J.C. Quantitative study on the cooling effect of green roofs in a high-density urban Area-A case study of Xiamen, China. J. Clean. Prod.; 2020; 255, 120152. [DOI: https://dx.doi.org/10.1016/j.jclepro.2020.120152]

34. Wenqian, J.; Guoyu, R.; Fengjun, J.; Jiajun, H.; Panfeng, Z. Spatial-temporal characteristics of the urban heat island effect in Xiamen, China. Urban Clim.; 2023; 52, 101725. [DOI: https://dx.doi.org/10.1016/j.uclim.2023.101725]

35. Sun, R.; Lü, Y.; Chen, L.; Yang, L.; Chen, A. Assessing the stability of annual temperatures for different urban functional zones. Build. Environ.; 2013; 65, pp. 90-98. [DOI: https://dx.doi.org/10.1016/j.buildenv.2013.04.001]

36. Wang, Y.; Ren, Y.; Zheng, X.; Wu, Z. The Impact of Building and Green Space Combination on Urban Thermal Environment Based on Three-Dimensional Landscape Index. Sustainability; 2025; 17, 241. [DOI: https://dx.doi.org/10.3390/su17010241]

37. Tobler, W.R. A Computer Movie Simulating Urban Growth in the Detroit Region. Econ. Geogr.; 1970; 46, pp. 234-240. [DOI: https://dx.doi.org/10.2307/143141]

38. Getis, A. Reflections on spatial autocorrelation. Reg. Sci. Urban Econ.; 2007; 37, pp. 491-496. [DOI: https://dx.doi.org/10.1016/j.regsciurbeco.2007.04.005]

39. Moran, P.A. Notes on continuous stochastic phenomena. Biometrika; 1950; 37, pp. 17-23. [DOI: https://dx.doi.org/10.1093/biomet/37.1-2.17]

40. Anselin, L. Exploring Spatial Data with GeoDa TM: A Workbook; Center for Spatially Integrated Social Science: Santa Barbara, CA, USA, 2005.

41. Anselin, L.; Florax, R.J. Small sample properties of tests for spatial dependence in regression models: Some further results. New Directions in Spatial Econometrics; Springer: Berlin/Heidelberg, Germany, 1995; pp. 21-74.

42. Yang, X.; Jin, W. GIS-based spatial regression and prediction of water quality in river networks: A case study in Iowa. J. Env. Manag.; 2010; 91, pp. 1943-1951. [DOI: https://dx.doi.org/10.1016/j.jenvman.2010.04.011]

43. Yuan, B.; Zhou, L.; Dang, X.; Sun, D.; Hu, F.; Mu, H. Separate and combined effects of 3D building features and urban green space on land surface temperature. J. Environ. Manag.; 2021; 295, 113116. [DOI: https://dx.doi.org/10.1016/j.jenvman.2021.113116]

44. Liu, J.; Wu, J.; Yang, Y.; Zhang, B.; Yin, L. Exploring the spatiotemporal impacts of urban green space patterns on the core area of urban heat island. Ecol. Indic.; 2024; 166, 112254. [DOI: https://dx.doi.org/10.1016/j.ecolind.2024.112254]

45. Zhang, Y.; Murray, A.T.; Turner, B.L. Optimizing green space locations to reduce daytime and nighttime urban heat island effects in Phoenix, Arizona. Landsc. Urban Plan; 2017; 165, pp. 162-171. [DOI: https://dx.doi.org/10.1016/j.landurbplan.2017.04.009]

46. Wu, Y.; Patuano, A.; Mashhoodi, B.; Lenzholzer, S.; Acred, A.; Narvaez Zertuche, L. How small green spaces cool urban neighbourhoods: Optimising distribution, size and shape. Landsc. Urban Plan; 2025; 253, 105224. [DOI: https://dx.doi.org/10.1016/j.landurbplan.2024.105224]

47. Ren, S.; Huang, Z.; Yin, G.; Yan, X.; Dong, Q.; Qi, J.; Zheng, J.; Bao, Y.; Zhang, S. Spatially-optimized greenspace for more effective urban heat mitigation: Insights from regional cooling heterogeneity via explainable machine learning. Landsc. Urban Plan; 2025; 256, 105296. [DOI: https://dx.doi.org/10.1016/j.landurbplan.2025.105296]

48. Yu, S.; Chen, Z.; Yu, B.; Wang, L.; Wu, B.; Wu, J.; Zhao, F. Exploring the relationship between 2D/3D landscape pattern and land surface temperature based on explainable eXtreme Gradient Boosting tree: A case study of Shanghai, China. Sci. Total Environ.; 2020; 725, 138229. [DOI: https://dx.doi.org/10.1016/j.scitotenv.2020.138229]

49. Zhang, L.; Ma, Z.; Guo, L. An Evaluation of Spatial Autocorrelation and Heterogeneity in the Residuals of Six Regression Models. For. Sci.; 2009; 55, pp. 533-548. [DOI: https://dx.doi.org/10.1093/forestscience/55.6.533]

50. Pantaloni, M.; Botticini, F.; Marinelli, G. Urban Green Space Accessibility as a Strategic Tool to Design Green Infrastructure and Pedestrian Way. Proceedings of the New Challenges for Sustainable Urban Mobility: Volume II; Brescia, Italy, 6–8 September 2023; Springer: Cham, Switzerland, 2024; pp. 71-81.

51. Mostafazadeh, R.; Alaei, N.; Mirchooli, F.; Hussain, S. Changes in urban green space configuration and connectivity using spatial graph-based metrics in Ardabil developing city, Iran. Environ. Monit. Assess.; 2024; 196, 778. [DOI: https://dx.doi.org/10.1007/s10661-024-12922-6]

52. Derdouri, A.; Murayama, Y.; Morimoto, T.; Wang, R.; Haji Mirza Aghasi, N. Urban green space in transition: A cross-continental perspective from eight Global North and South cities. Landsc. Urban Plan; 2025; 253, 105220. [DOI: https://dx.doi.org/10.1016/j.landurbplan.2024.105220]

53. Bressane, A.; Loureiro, A.I.S. Political and socioeconomic drivers of urban green space management: Insights from a case study in a Metropolitan Region of an Emerging Economy. Urban For. Urban Green.; 2025; 104, 128591. [DOI: https://dx.doi.org/10.1016/j.ufug.2024.128591]