Correlation Analysis between Urban Green Space and Land Surface Temperature from the Perspective of Spatial Heterogeneity: A Case Study within the Sixth Ring Road of Beijing

Abstract

Urban greening has been widely regarded as the most effective, lasting, and economical strategy for alleviating the effects of urban heat islands (UHIs). Previous studies on the cooling effect of urban green spaces (UGSs) tend to analyze the correlation between landscape metrics and land-surface temperature (LST) based on a global parameter estimation, while ignoring urban heterogeneity and autocorrelation. This study focuses on the sixth ring road of Beijing and uses Landsat 8 imagery to retrieve the LST and extract the position of UGSs. We propose a new approach to optimize the selection of landscape metrics, to identify the least and most effective metrics to establish a geographically weighted regression (GWR) model, and to plot the distribution of local regression coefficients to investigate the spatially heterogeneous cooling effects of greenspaces. The effect of UGS landscape metrics on the LST differs according to spatial location; the method enhances our understanding of the effects of UGS spatial configuration on UHIs and better guides the planning and construction of future UGSs.

1. Introduction

Accelerated urbanization has led to drastic changes in the types of land cover, as well as the physical and chemical properties of soil and water. Consequently, the processes of water vapor and energy exchanges between soil and the atmosphere have shifted, thereby affecting the urban thermal environment system [1]. In turn, urban thermal environment changes have affected the regional climate, urban hydrology, air quality, and many urban ecological processes, such as material metabolism and the energy cycle. All the above have caused a series of safety problems regarding the ecological environment, thereby seriously affecting the normal production and life of urban residents; they constitute the bottleneck restricting the sustainable development of cities. Urban green space (UGS) can change the air flow and heat exchange through the transpiration of vegetation leaves and solar radiation blockage; they form urban “cold island” centers, thereby improving the urban thermal environment [2,3]. Therefore, as living urban infrastructures, UGS is among the most effective, lasting, and economical strategies for mitigating the effects of urban heat islands (UHIs) [4,5].

Many scholars have conducted in-depth research on the mitigation of the effects of UHIs through UGSs. Research on the urban thermal environment mainly focuses on the atmospheric temperature (AT) and land surface temperature (LST) [6]. These parameters, albeit numerically different, correlate well with each other [7] and can both express the spatial distribution of the urban thermal environment well. AT data are mainly obtained through in situ monitoring, which can have high temporal resolution; however, spatial coverage can be poor due to the usually small number of monitoring stations. Conversely, research on LST is mainly based on LST inversion, based on a thermal infrared band of remotely sensed imagery. This procedure is more time efficient and labor saving than AT data acquisition, and can effectively and comprehensively detect LST characteristics. Surface urban heat islands (SUHIs) based on remote sensing have also become the main focus of current UHI research [8]. Previous research mainly focused on understanding the relationships among land surface characteristics, vegetation abundance, and the LST [9,10]. In recent years, due to the advantages of quantitative remote sensing with regard to landscape patterns, remote sensing studies have paid special attention to the influence of spatial UGS patterns on the LST [11,12,13].

According to the landscape ecology theory, diverse UGS landscape features can be classified as composition and configuration characteristics [14]. The former refers to the abundance and variety of land cover types, including UGS area, type, vegetation quality, and the latter refers to the spatial layout of land cover types, where the UGS shape and distribution determine the UGS configuration [1,15,16]. Through remote sensing, the relationship of each metric with the LST can be revealed. Previous comprehensive research on the impact of UGSs on the LST preferred regular grids or concentric buffers as analysis units [14,17,18,19], while performing simple linear regressions and curve fitting on the landscape metrics and LST of each analysis unit and dividing to obtain a global correlation coefficient [20]. The results they produced were based on the overall correlation of the overall study area, yielding only an overall correlation coefficient. However, urban landscapes are heterogeneous, and the different composition and configuration of the landscape in each area of the city affects ecological processes. For example, it has been demonstrated that urban heterogeneity affects urban ecological resilience and sustainability [21,22]. Urban heterogeneity may lead to different UGS cooling effects in different spatial units [18,23]. Furthermore, some scholars have previously concluded that the cooling effect of UGS shapes in different cities has the opposite effect. For example, Zhou et al. (2017) found a better cooling effect for complex shapes in Sacramento, but they also found a better cooling effect for simple UGS shapes in Baltimore; they attributed this contradictory finding to the different background climatic conditions [24]. Their study likewise calculated a global correlation coefficient for the entire city. Therefore, we analyzed the cooling effect of UGS in different areas of a city from the perspective of urban heterogeneity. We obtained positive and negative results for different areas of a city by calculating the local correlation between UGS and the LST in different areas, which helps us to enhance our understanding regarding the effects of UGS spatial configuration on UHIs.

In addition, related studies have also shown that spatial autocorrelation may affect the relationships of landscape metrics with the LST [24,25]. Geographically weighted regression (GWR) is a new method for modeling spatial heterogeneity [26]; based on the GWR model, analyzing the relationship between UGSs and the LST in different regions becomes feasible. Although the application of the GWR model is common, we applied the GWR model to analyze the correlation between the LST and UGS from the perspective of spatial heterogeneity based on Landsat data. We combined the cold- and hot-spot analyses to visualize the cooling effect of spatial inhomogeneity of UGS and to obtain a good explanation. This series of methods constitutes a new approach for studying the correlation between UGS and LST.

In this study, we use Landsat 8 imagery for UGS extraction from within the sixth ring road of Beijing and retrieve the LST based on its thermal infrared band. On the macroscopic scale, from the perspective of spatial autocorrelation and heterogeneity, we applied the GWR method to comprehensively consider the UGS cold source intensity and the different UGS cooling effects under different ecological and artificial environments in order to better guide UGS planning and construction.

2. Methodology

2.1. Study Area

Beijing (39.4–41.6° N, 115.7–117.4° E) is a mega city and one of China’s national central cities, i.e., a national political, economic, and cultural center. It was classified as an Alpha+ city (i.e., a highly integrated city fulfilling advanced service needs) by the globalization and world cities research network (GaWC) in 2020. By 2020, the built-up area of Beijing had reached 1469 km² with a permanent population of more than 21 million. The GDP was 3537.1 billion yuan, the annual household electricity consumption for urban and rural residential was 20,494.47 million kWh, the total gas supply of natural gas was 19,243.47 million m³, the total gas supply of liquefied petroleum gas was 434,286 tons, and the amount of anthropogenic heat sources was immense. Therefore, the effect of UHIs has become exceptional. In recent years, in order to become an international first-class harmonious and livable city, Beijing has paid considerable attention to ecological and greening constructions, with the greening rate of urban areas reaching 48.5% in 2020. Our study area is the urban area within the sixth ring road of Beijing, which reflects the rapid urbanization of Beijing. This area is also the most densely populated area of the city, and with its greening construction reaching certain benchmarks, it is an ideal area for analyzing the role of UGSs on UHI mitigation.

2.2. Data Sources

From the USGS website (https://earthexplorer.usgs.gov), accessed on 20 March 2022, we downloaded a Landsat 8 OLI/TIRS image of Beijing (path: 123; row: 32; cloudiness: 0.2%; resolution: 30 m). It was acquired on 20 September 2020 and was used to extract UGS patches, as well as to perform LST inversion using its band 10. Due to the low resolution of Landsat 8 images, the UGS extraction is coarse and prohibits the precise determination of land cover types, such as forests, shrubs, grasslands, and croplands. In this study, such land cover types were included in the UGSs when extracting them. For this reason, we also utilized the 10-m land cover data for 2020 published by the ESA (https://viewer.esa-worldcover.org, accessed on 20 March 2022), which include a subdivision of the land cover into 11 types, such as forest, shrubland, grassland, cropland, built-up, bare vegetation, water bodies, etc., with an official accuracy of 74.4% [27]. To ensure higher classification accuracy, we used visual interpretation for correction, and used its more detailed land-cover classification (i.e., forest, shrubland, grassland, and cropland) for proving the rationale behind categorizing all forest, shrubland, grassland, and cropland patches as UGS to establish a GWR model.

Based on the convenience of obtaining Landsat data, we mainly study the extraction of UGS using Landsat data, while the 10-m land cover data for 2020 published by the ESA is mainly used to verify and analyze the general pattern of the different cooling effects of forests, shrubs, and grasses in UGSs. Although it has finer classification accuracy, it only has data for 1 year, 2020. Therefore using Landsat data can extend our approach into other years and can ensure data homogeneity based on Landsat data for LST retrieval and extraction of UGS patches.

2.3. Extraction of UGS Patches

After preprocessing, i.e., applying spatial and atmospheric corrections, since the original Landsat 8 image was of a single band, it was necessary to combine multiple single-band images into a multi-band color image to improve the accuracy of land classification; however, there was considerable redundant information between bands. To reduce mutual interference, bands with large amounts of information were selected for information extraction, while the correlation between them was the smallest. In this study, we used a principal component analysis, a band correlation analysis, and a normalized difference vegetation index (NDVI) calculation based on ENVI5.3. Subsequently, we determined the optimal band combination according to the correlation analysis of various band data, including each band, each principal component, and the NDVI to increase image interpretation accuracy.

Table 1. Contribution of each principal component of the data source image.

Principal Component	Eigenvalues	Contribution Rate	Cumulative Contribution Rate
PC1	3,490,585.81	90.61%	90.61%
PC2	311,692.2	8.09%	98.70%
PC3	41,592.29	1.08%	99.78%
PC4	5169.37	0.13%	99.91%
PC5	2478.64	0.06%	99.97%
PC6	744.11	0.02%	99.99%
PC7	262.93	0.01%	100.00%

shows that the principal component 1 (PC1) in the image had the largest amount of information; therefore, 7 bands of the image, as well as the generated bands NDVI and PC1, were selected for correlation analysis and we determined the optimal band combination. The optimum index factor (OIF) proposed by Chavez in the United States to determine the most informative 3-band combination was used as basis for selecting the band combination [28,29,30]. The formula is as follows:

$$OIF={\displaystyle \sum _{i=1}^3{S}_i}/{\displaystyle \sum _{j=1}^3{R}_{ij}},$$

(1)

where S_i is the standard deviation of band i, and R_ij is the correlation coefficient between band i and band j (see

Table 2. Correlation coefficient matrix among the bands.

Correlation	Band 1	Band 2	Band 3	Band 4	Band 5	Band 6	Band 7	ndvi	pc1
Band 1	1.00
Band 2	0.99	1.00
Band 3	0.98	0.98	1.00
Band 4	0.96	0.97	0.98	1.00
Band 5	0.71	0.66	0.76	0.67	1.00
Band 6	0.82	0.79	0.88	0.83	0.95	1.00
Band 7	0.90	0.90	0.95	0.93	0.82	0.95	1.00
NDVI	0.43	0.36	0.48	0.36	0.89	0.76	0.57	1.00
PC1	0.83	0.80	0.88	0.82	0.97	0.99	0.93	0.80	1.00
Standard deviation	371.84	345.28	422.13	438.95	1395.22	935.90	633.74	0.35	1868.31

). The larger the OIF, the greater the amount of information contained in the 3-band combination, the higher the independence between bands, and the smaller the information redundancy.

Finally, bands 4, 5, and PC1 were selected as the optimal band combination. In order to greatly improve interpretation accuracy, the images were fused with the Landsat 8 panchromatic band after band combination (resolution: 15 m). The fused image is shown in Figure 1; the image clarity is greatly improved, fully meeting the classification requirements.

The supervised classification tool based on ENVI5.3 was used for classifying the processed images. As already mentioned, because of the low Landsat 8 resolution, we divided land cover into 3 types (impervious surfaces, water bodies, and UGS) which included forests, shrubs, grasslands, and croplands. Based on the macroscopic scale analysis, the 3 land cover types met the research requirements (see Section 3.3.1 for the rationality of this classification). The classification results are shown in Figure 2a. In this study, urban forests, shrubs, grasslands, and croplands were all classified as UGS, and the kappa coefficient was 0.96, thereby meeting the research requirements.

2.4. LST Retrieval

The LST is the radiative skin temperature of the land derived from infrared radiation [31]. Bands 10 and 11 of Landsat 8 are thermal infrared bands and can be used for LST retrieval. The method has been fully verified and used by scholars and experts internationally. In this paper, we selected band 10. First, the digital number (DN) of band 10 was converted to spectral radiation; second, the spectral radiation was converted to the brightness temperature of the sensor using Plank’s law; and third, the surface brightness temperature was converted into the actual surface temperature by estimating the surface radiance. We refer to the literature for details on the specific inversion method [20,32]. The LST map is shown in Figure 2b.

2.5. Calculation of Landscape Metrics Based on the Moving Window Method

According to previous studies, the cooling effect of UGS has a scale effect [33] and is influenced by landscape metrics, such as UGS area, shape, and distribution. Here, we calculated the UGS landscape metrics based on the moving window method of FRAGSTATS v4.2. As the cooling effect of UGS has a scale effect, the choice of moving window size is crucial. If the window is too small, large UGS patches will be cut, thereby affecting the estimation of the cooling effect of UGS; if the window is too large, the cooling effect of UGS patches will also be greatly weakened, resulting in insufficient samples [34].

We therefore selected the percentage of UGS (PLAND) index in FRAGSTATS v4.2 as the basis for window size selection. As the image resolution was 30 m, a multiple of 30 was selected for window size according to the software characteristics, thereby ensuring calculation accuracy. According to UGS extraction, the average area of UGS patches was 62,032 m² (i.e., approximately 249 × 249 m), while 99.8% of UGS patches were smaller than the 1800 × 1800 m window size. Therefore, we selected the window sizes of 300 × 300, 600 × 600, 900 × 900, 1200 × 1200, 1500 × 1500, and 1800 × 1800 m, and the final window sizes were determined based on the correlation analysis results of their respective PLANDs and LSTs. Figure 3 shows that the 1500 × 1500 m window size exhibited a better fit between PLAND and LST, which better reflected the cooling effect of UGS at this scale, while the sample was sufficient. The number of samples in the 1800 × 1800 m window was slightly smaller.

2.6. Statistical Analysis Using Moran’s I and Getis-Ord Gi*

The purpose of using Moran’s I and Getis-Ord Gi* was to find the spatial location of cold and hot spots in the study area and to evaluate the spatial LST distribution pattern. Both indices were realized by the geographic statistical analysis extension module of ArcGIS [35].

Moran’s I is the most commonly used index for analyzing spatial autocorrelation. The formula is as follows:

$$I=\frac{n{\displaystyle {\sum }_{i=1}^n{\displaystyle {\sum }_{j=1}^n{w}_{ij}|x_i-\overline{x}|\left|x_j-\overline{x}\right|}}}{{\displaystyle {\sum }_{i=1}^n{\displaystyle {\sum }_{j=1}^n{w}_{ij}{\displaystyle {\sum }_{i=1}^n{\left(x_i-\overline{x}\right)}^2}}}},$$

(2)

where I is Moran’s I, n is the number of 1500 × 1500 m windows, and x_i and x_j are the average temperatures of the i and j windows, respectively. $\stackrel{-}{x}$ is the average temperature of all windows, and w_ij is the weight of window i relative to window j. The inverse distance weighted method is used here to generate.

Getis-Ord Gi* is a local spatial autocorrelation index based on the total distance matrix for detecting the locations of high or low values of each element clustering in space and has strict statistical test standards. The formula is as follows:

$$G{i}^{\ast }=\frac{{\displaystyle \sum _j^n{w}_{ij}{x}_j}}{{\displaystyle \sum _j^n{x}_j}},$$

(3)

where w_ij is the spatial weight matrix between window i and j, x_j is the average temperature of window j, and n is the total number of windows. Using the statistical results of Getis-ord Gi*, the region can be divided into 3 different temperature zones: hot spot, cold spot, and insignificant regions.

2.7. GWR Analysis Based on LST and UGS

Landscape composition and configuration play important roles regarding the spatial LST distribution in an area; however, due to different spatial planning contexts and ecological backgrounds in cities, the influence of the UGS landscape composition and configuration on the LST cooling effect can vary in different areas [20,36]. Therefore, we used the GWR method for investigating the degree of influence of each UGS landscape metric on the cooling effect in different locations within the study area.

GWR is a local regression technique that considers spatial heterogeneity. The model has the form:

$${\mathrm{Y}}_i={\beta }_0(u_{\mathrm{i}},v_{\mathrm{i}})+{\displaystyle \sum _{i=1}^k{\beta }_j}(u_{\mathrm{i}},v_{\mathrm{i}})X_{\mathrm{ji}}+{\epsilon }_i$$

(4)

where Y_i denotes the LST, X_ji represents the landscape metrics, k is the total number of spatial units involved in the analysis, ε_i denotes the random error term, (u_i, v_i) indicates the spatial location of sample i, β₀ (u_i, v_i) is the intercept at the location i, and β_j (u_i, v_i) is the local estimated coefficient of the independent variable X_ji. The Gaussian function was used to determine the weight, and the Akaike information criterion (AIC) method was used to determine the optimal bandwidth [37].

Based on FRAGSTATS v4.2, we used the 1500 × 1500 m windows and screened them out by the above moving window analysis to analyze the UGS landscape pattern within the sixth ring road of Beijing. Each window is an analysis unit represented in the above formula.

To better and more accurately explore the comprehensive influences of LST and various UGS landscape metrics, we selected as many commonly used landscape metrics from the class level as possible to characterize the UGS landscape pattern. Landscape metrics are selected as shown in Figure 4a. These landscape metrics covered all aspects of green landscape pattern, and the LST was used as the dependent variable for establishing a GWR model. Before the regression analysis, each variable was normalized to eliminate the influence of dimension on the regression analysis [20]. However, as shown in Figure 4a, each landscape metric was usually correlated with the others, and the inclusion of various landscape metrics in the model led to multicollinearity. This is also the drawback of selecting landscape metrics in past studies, where there is no standard method, but the landscape metrics are selected empirically for correlation analysis separately, which leads to high collinearity among the selected indices, resulting in some meaningless conclusions while obtaining consistent results. Therefore, we used the RStudio regsubsets function and selected the best subsets of calculated regressions to filter the landscape metrics. As shown in Figure 4b, when landscape metrics were 8, the adjusted R² of the subset regression equation would maximize. Nevertheless, in order to minimize the number of selected landscape metrics, we assumed that 4 landscape metrics were sufficient for the adjusted R² to approximate the maximum, as choosing more landscape metrics would be of little significance. The final selected landscape metrics were PLAND, LSI, SHAPE_MN, and DIVITION. PLAND represented landscape composition, and the other metrics represented landscape configuration.

After screening the metrics, the ordinary least squares regression (OLS) model was used for further verifying the collinearity and significance of the selected landscape metrics based on ArcGIS. OLS is the most widely used regression model and a suitable starting point for all spatial regression models. After analysis, DIVITION, i.e., a metric with high collinearity, was excluded, while PLAND, SHAPE_MN, and LSI were selected. PLAND represents the UGS proportion in the 1500 × 1500 m window and can characterize the UGS landscape composition in the study area. SHAPE_MN represents the shape complexity of regional UGS patches. LSI is similar to SHAPE_MN, albeit representing the complexity of landscape shape constructed by all UGS patches within the window, and it represents the patch shape irregularity of the overall landscape. In Fragstasts4.2, it is a patch aggregation metric, reflecting the degree of aggregation of patches. These 3 metrics were able to represent the spatial UGS pattern characteristics. In addition, to better explain the interpretation ability of UGS landscape metrics on LST and to avoid the influence of the water-cooling effect, the PLAND of water bodies was also added as an independent variable. All the selected independent variables are shown in

Table 3. Description of the landscape metrics used for the independent variables of the GWR model.

Metric	Abbreviation	Formula	Description
Percentage of landscape	PLAND	$PLAND=\frac{{{\displaystyle {\sum }_{\mathrm{i}=1}^{n}a}}_i}{A}$	PLAND1 is the proportion of UGS patch area within an analysis unit. PLAND2 is the proportion of water body patch area within an analysis unit.
Mean shape index	SHAPE_MN	$SHAPE\_MN=\frac{1}{N}\times {\displaystyle \sum _{i=1}^n\frac{0.25P_i}{\sqrt{a_i}}}$	Average shape index of the UGS patches within an analysis unit.
Landscape shape index	LSI	$LSI=\frac{0.25E}{\sqrt{A}}$	Shape index of all UGS patches boundaries within an analysis unit.

a_i: area of patch i; p_i: perimeter of patch i; E: perimeter of all UGS patches within an analysis unit; A: total area of the analysis unit; N: number of UGS patches.

. Finally, a GWR analysis was conducted based on ArcGIS, thereby obtaining the spatial distribution of local regression coefficients for each landscape metric (see Section 3.4). As shown in

Table 4. Estimation parameters of the GWR and OLS model.

Models	GWR	OLS
AICc	715.07	1396.81
R2	0.9079	0.769436
Adjusted R2	0.8921	0.768519

, by comparison with OLS models based on global parameter estimations, the GWR model can effectively deal with spatial non-stationary phenomena in regression analysis and can better quantify the relationships between landscape patterns and LST, having higher explanatory power and reliability [38,39].

3. Results and Analysis

3.1. UGS Statistics

The classification results of the UGS extraction are shown in Figure 2a. After extracting the UGS patches and taking the ring road as the statistical unit, we counted the area and number of green patches in each ring road range to analyze UGS distribution characteristics.

As shown in

Table 5. Distribution of UGS in Beijing within different ring roads.

Ring Road	Maximum Patch Area(ha)	Mean Patch Area (ha)	Number of Patches
2nd ring	200.6	1.193640539	686
3rd ring	127.4	1.237739804	1265
4th ring	147.5	1.495361226	1771
5th ring	1424	3.699887817	3073
6th ring	11500	8.194246931	8927

, the number of UGS patches gradually increases from within the second to within the sixth ring road of Beijing; additionally, the average patch size gradually becomes larger. In general, buildings within the fourth ring road are dense, the ecological land is insufficient, and the average patch area is not large; however, there are several large parks (i.e., green patches), such as the Yuyuantan and Zizhuyuan parks. Large green patches, such as the summer palace and the old summer palace, are distributed between the fourth and the fifth ring road. In addition, Beijing has built the first greenbelt in this region, which uses evacuated and vacated spaces to plant large-scale greenery; hence, there are many large-scale UGSs in this range. In the sixth ring road region, there is relatively more ecological land for afforestation. Beijing has planned the second greenbelt at this location, coupled with a plain afforestation project; hence, large-scale green patches are more distributed. In addition, the largest UGS patches (i.e., at the Xishan Mountain) are also within this range.

3.2. Spatial LST Distribution

We applied Moran’s I to prove the existence of LST clusters and used Getis-Ord GI* for identifying the locations of hot spots and cold spots. Moran’s I statistics are shown in

Table 6. Moran’s I statistics.

Moran’s I	Z-Score	Distribution Type
0.727096	32.197042	Clustering distribution

The Z-score of Moran’s I is greater than the critical value of 2.58 (p < 0.01), which meets the significance level test.

, indicating the existence of significant spatial LST clustering distribution characteristics within the sixth ring road of Beijing. Furthermore, we used Getis-Ord GI* to find the cold and hot LST spots and to analyze the spatial LST distribution. Cold/hot LST spot distribution is shown in Figure 2c. The hot spots are mainly distributed within the fifth ring road and especially in the middle (i.e., within the second ring road) and southern areas. The hot spots outside the fifth ring road are mainly in the Beijing Shunyi airport area. The cold spots are mainly located in the Xishan Mountain and northern areas outside the fifth ring road. The identification of cold- and hot-spot areas helps to clarify the role of UGS on LST, to find effective measures for UHI mitigation, and to reasonably configure the UGS landscape pattern.

Taking the ring road as the statistical unit, as shown in Figure 5a, the LST gradually decreases from the central region (i.e., within the second ring road) toward the outer region. The third and fourth ring road regions have basically the same urban environmental background with small differences regarding the LST, while the fifth and sixth ring road regions exhibit a significant decrease in the average LST as the UGS patches increase and become larger.

3.3. Analysis of the Cooling Effect of Internal UGS Patches

UGSs have a significant cooling effect on UHIs, with vegetation abundance and structure, as well as the shade within their patches being important influences. The NDVI are considered as a key biophysical parameter in thermal remote sensing analysis. Previous studies have shown that the NDVI is a good indicator of surface vegetation cover. Understanding the relationship between LST and vegetation indices can elucidate the role of vegetation cover in the urban thermal environment [40]. In addition, the cold island effect of each patch depends on its size, perimeter, and shape [6,25].

According to the extraction of UGS patches, the area of each patch varies greatly, and the number ratio of each size patch is extremely uneven, while small UGS patches account for the main part. In order to more accurately analyze the influence of different sizes of UGS patches on the cooling effect, we used a stratified sampling method for selecting UGS patches and ensuring that the samples contain each area size of the patches, while the number of patches in each area is approximately the same. Eventually, 1080 UGS patches were selected for establishing quantitative relationships between relevant indicators and LST from the perspective of the internal characteristics of the patches themselves.

3.3.1. Relationships of the NDVI with the Cooling Intensity within UGS Patches

To better understand the effect of vegetation cover on UHIs, NDVI is a key biophysical indicator for characterizing vegetation conditions. We applied regression and Pearson correlation analyses for estimating the relationship between LST and NDVI. The correlation analysis was first conducted over the whole area, i.e., not limited within the UGS patches, while excluding the water bodies to avoid their interference. To ensure a more accurate analysis of the correlation between NDVI and LST and to circumvent the influence of scale factors, we used a pixel-to-pixel method based on ArcGIS. The results showed that the correlation coefficient was −0.674, indicating that LST has a significant negative correlation with NDVI, which is consistent with previous research [41] and confirms the important influence of vegetation cover on the cooling effect.

Subsequently, based on the selected green patches, ArcGIS and SPSS software were used for extracting the average NDVI values in each green patch and establishing the correlation models with the average LST in each patch. Pearson’s correlation coefficient was −0.381(p < 0.01), where it is evident that the correlation of the average NDVI with LST in green patches were basically consistent with the above conclusion, but the correlation coefficient was significantly smaller. The implication is that the cooling intensity within a UGS patch still depends on its vegetation quality. However, given that UGS patches are themselves cold sources, the actual gaps between the NDVI values in most UGS patches were not large and the correlations between the NDVI and LST were not high, but the quality of vegetation in UGS patches was still a factor affecting the intensity of cooling.

The differences between NDVI values in most UGS patches were small, and the correlations between vegetation type and quality of LST were not high. This conclusion also supported our decision to classify forest, shrubland, grassland, and cropland as UGSs during extraction. To verify the above conclusions, we utilized the 10-m resolution land-use data of ESA and the statistics regarding the average LST and area of each category. The results are shown in Figure 5b, where it is evident that there are small differences in the average LSTs of forests, shrubs, grasslands, and croplands; however, the overall cooling effect still adheres to the following pattern: forest > shrub > cropland > grass.

3.3.2. Relationships between the Characteristics of UGS Patches and the Cooling Intensity

The size, perimeter, and complexity of UGS patches are important characteristics to consider; therefore, we analyzed the area, perimeter, and shape characteristics of the extracted UGS patches using FRAGSTATS v4.2. The AREA index was chosen for the size calculation, representing the area of the patch; the PERIM index was chosen for the perimeter calculation, representing the edge length of the patch and the contact surface for heat exchange with the surrounding area; and the PARA and SHAPE indices were chosen for the complexity calculation, where SHAPE measured the shape complexity of the patch by calculating the deviation of its shape from a circle or a square of the same area. The results are shown in

Table 7. Correlation between metrics of UGS patch and LST (1080 samples).

	Pearson’s Correlation Coefficient	LST	AREA	PERIM	GYRATE	PARA	SHAPE
LST	r	1	−0.197 **	−0.204 **	−0.311 **	0.415 **	−0.287 **
AREA	r	−0.197 **	1	0.994 **	0.874 **	−0.226 **	0.808 **
PERIM	r	−0.204 **	0.994 **	1	0.893 **	−0.237 **	0.849 **
PARA	r	0.415 **	−0.226 **	−0.237 **	−0.431 **	1	−0.337 **
SHAPE	r	−0.287 **	0.808 **	0.849 **	0.933 **	−0.337 **	1

** p < 0.01.

. The LST was weakly negatively correlated with the area and perimeter, with correlation coefficients of only −0.197 and −0.204, respectively. The negative correlation of SHAPE with LST was relatively higher, with a correlation coefficient of −0.287, while PARA (i.e., the perimeter-to-area ratio) was significantly positively correlated with LST, with a correlation coefficient of 0.415. Through detailed comparisons, we found that patches with high SHAPE values were all large patches. Large patch areas had generally more complex shapes, while small patch areas had more regular shapes. PARA exhibited the opposite behavior, where large PARA values were characteristic of small and narrow patches. Therefore, large patches had high SHAPE values and low LSTs, and therefore SHAPE was negatively correlated with LST. On the contrary, high PARA values were characteristic of small and narrow patches, which were more vulnerable to external thermal environment interferences with high LSTs. Therefore, PARA was significantly positively correlated with LST. Overall, it is evident that the patch shape metrics were more affected by the patch size, while the correlation between AREA and LST was weaker.

To further elucidate these results, we made a scatter plot analysis of LST and AREA. The results are shown in Figure 6. The scatter plot shows that with increasing area, the LST stabilizes, i.e., when the green patch reaches a certain area threshold, the cooling effect inside the green patch gradually weakens, and finally cools to a stable value. To find this area threshold, we used the RStudio to fit the scatter plot by segments and found it to be equal to 11.632 ha (R² = 0.193, p > 0.01). Although the R² value was not high, it still indicated that the internal LST would not change significantly when the UGS area was larger than a threshold. However, the area threshold cannot be specified because the internal characteristics of each UGS patch differ. These conclusions would help to explain the distribution characteristics of regression coefficients in Figure 8.

3.4. Correlation Analysis of LST and UGS Landscape Pattern at a Regional Scale

The cooling effect of UGS patches is not only influenced by their own characteristics, but also by the compound influence of the surrounding environment and adjacent patches. The UGS pattern is a key landscape element affecting the urban thermal environment, and its landscape metrics are closely related to the magnitude and extent of UGS cooling, which restricts its cooling effect [36]. We analyzed its correlation at a regional scale of 1500 × 1500 m and revealed its spatial differentiation characteristics to better guide future UGS construction. The spatial distribution of local regression coefficients based on the GWR model is shown in Figure 8. In numerical terms, the regression coefficient indicates the magnitude of the effect of the independent variable on the dependent variable [37].

We combined the regression coefficients and analyzed the contribution of each landscape metric to the LST. The results are shown in Figure 7. PLAND indicates that the proportion of UGS patches in the region is negatively correlated with LST and has the greatest impact on LST cooling (i.e., −57.96%), followed by that of LSI (i.e., −14.76%). In addition, SHAPE_MN was positively correlated with LST, contributing 11.27%.

Combined with the local regression coefficient distribution map, we analyzed the spatial distribution characteristics of the effect of each landscape metric on LST in the study area. The local regression coefficient distribution map can visualize the heterogeneity of the cooling effect of UGSs.

As shown in Figure 8a, PLAND can characterize the landscape composition of UGS in the study area. Within the fifth ring road of Beijing, except for a few large parks, the general UGS patch area is small and the PLAND value is low. Between the fifth and sixth ring roads the second green partition has been built; the area of UGS patches is large, and the PLAND value is high. As there are obvious differences between the area and distribution characteristics of UGS patches within the fifth ring road and those between the fifth and sixth ring roads, the influence of PLAND on LST is also different in these two regions. In the first area (i.e., within the fifth ring road), the PLAND values are generally small, but the cooling effect on LST is more obvious. The layout of the negative effect of PLAND on LST is basically consistent with the spatial layout of hot LST spots. The area of UGS patches is larger (i.e., higher PLAND values), and the cooling effect is more obvious. Other areas within the fifth ring road with small negative impact on the PLAND values of UGS patches are mainly the areas with large green patches and water bodies, such as the summer palace and the Olympic Forest Park, where the overall LST is low. In the second area, (i.e., between the fifth and sixth ring roads), the PLAND values are relatively high in most regions; however, except for large, integrated green patches in the western mountains and areas with high LST values in the Shunyi airport and its surrounding areas, the cooling effect on LST is not as pronounced as that in the first area. This is consistent with the conclusion that the cooling effect of the patches is weakened when the green patches are larger than a certain area threshold. Areas with a more obvious cooling effect between the fifth and sixth ring roads are also those with higher LSTs.

The SHAPE_MN represents the shape complexity of regional UGS patches. According to the observations, except for the huge and complete UGS patches in the western mountains area outside the sixth ring road, the smaller the UGS patches are, the lower the shape complexity (i.e., lower SHAPE_MN values) of the patches. Among the larger patches, the shape complexity of irregular patches is higher, while the shape complexity of regular green patches is relatively low. Combined with Figure 8b, the effect of SHAPE_MN on LST is mainly positive, i.e., the larger the SHAPE_MN, the more complex the shapes of UGS patches, and the higher the average LST. This can also be explained by landscape ecology, i.e., the more complex the shape of the patch is, the more extensive its contact is with the outside world, involving higher energy transfer and poorer internal preservation [42,43]. The range of local regression coefficients for negative effects is only −0.07–0, which is basically negligible and has a small distribution range. The effect of SHAPE_MN on LST is consistent with the effect of PLAND, and its positive effect on LST is also distributed in the high LST areas and adjacent areas, which are basically consistent with the hot spot areas. Conversely, within the second ring road, although it is a hot area, this part of the UGS patch is small and reduced, and the cooling effect is poor, so the positive effect is relatively low. The positive effect of SHAPE_MN on LST is relatively low between the fifth and sixth ring roads because the UGS patches in this area are generally large, and the cooling effect has basically reached the threshold, so the effect of patch shape complexity on surface temperature is relatively reduced.

The LSI represents the complexity of the landscape shape constructed by all UGS patches in the 1500 × 1500 m window. It is a patch aggregation metric in FRAGSTATS v4.2 which is similar to the SHAPE_MN distribution and represents the irregularity of the patch shape from the overall landscape level but can reflect the degree of patch aggregation. According to the observations, the LSI distribution within the sixth ring of Beijing is opposite to the SHAPE_MN distribution. The UGS patches within the fifth ring road are small, complex, and dense, and the landscape shape complexity is high (i.e., LSI is high). Between the fifth and sixth ring road, the UGS patches are generally large, occupying almost the entire window (1500 × 1500 m). The overall landscape complexity is relatively low (i.e., LSI is low), but there are also areas with slightly smaller UGS patches with complex and dense shapes. The effect of LSI on LST is negative, i.e., the denser the green patches in the region are, the greater the LSI is, and the greater the contribution to the cooling effect is. However, the negative effect of LSI is relatively low in the region with larger patches and lower LST. It is evident in Figure 8c that the regions with large LSI contribution to cooling are southwest and southeast of the fifth ring road and north of Shunyi Airport, which are all regions with a slightly high LST and high LSI. The regions within the second ring road are also regions with large negative impact of LSI on LST, but they are not as good as the above regions because the green patches within the second ring road are small. This indicates that the cooling effect of small and dense green patches is not as good as that of slightly large and dense green patches.

4. Discussion

4.1. Different Impact of UGS Landscape Metrics on LST in Different Spatial Locations

Previous studies tended to establish quantitative relationships between the LST and landscape metrics based on a global parameter estimation, with little consideration of spatial heterogeneity and autocorrelations [37]. In this paper, we first obtained global parameter estimates using the OLS method without considering spatial heterogeneity, and then used the GWR method to obtain local parameter estimates by considering the different degrees of influence of UGS landscape metrics on LST in different spatial locations. Finally, we compared the fitting effects of the two methods, retrieving an OLS-adjusted R² of 0.77 and a GWR-adjusted R² of 0.89. We obtained a better fitting effect with the GWR regression method, indicating that the degree of influence of UGS landscape metrics on LST is different in different spatial locations. Current research on the relationship between UGS landscape pattern and LST is confusing; the consensus is that large-scale patches have the best cooling effect, but the ideal shape and distribution pattern are still controversial [6,44]. In this paper, we concluded that simple UGS shapes have better cooling effects, which is consistent with some previous findings [43]. However, there are also studies such as Yu et al. (2018), who concluded that a more complex UGS shape has a stronger cooling effect [44]; the landscape metrics they chose to represent the UGS shape with LSI, found that LSI was negatively correlated with LST, and determined that more complex patches have better cooling effect. The conclusion that LSI is negatively correlated with LST is consistent with this paper; however, SHAPE is more representative of the shape characteristics of individual patches and more descriptive when analyzing the cooling effect of the UGS patch shape. In addition, as mentioned above, different cities have reached opposite conclusions in past studies [24].

In this paper, the local regression coefficient distribution map was obtained using the GWR method, which can clearly reveal the influence of the spatial UGS pattern in each region on LST. The influence of the landscape metrics of each area of a city on LST has both positive and negative effects, thereby elucidating the cooling effect of UGS landscape configuration and composition on LST, deepening our understanding of the relationships of UGS composition and configuration with LST, while also being conducive to UGS planning and construction. In addition, through exploring the reasons why the impacts of certain regional landscape metrics of a city on LST are positive or negative, it is worth considering whether we can address why different studies have different conclusions.

4.2. Hot LST Spots Are the Focus of Future UGS Construction

According to the results, the hot spots in Beijing are mainly concentrated within the fifth ring road, and especially in the middle of the fifth ring road (i.e., within the second ring road) and the southern area. According to Figure 8, although the green patches in hot spots are generally small, the UGS landscape metrics in the hot spots have a relatively strong impact on the cooling effect of LST. This shows that UGS planning and construction in hot spots has a more obvious effect on cooling. Furthermore, it can be seen that greenery areas have the highest contribution to the cooling effect, followed by the denser and more regular shaped UGS patches. Therefore, the most prominent initiative for mitigating the effects of UHIs is to build centralized and large park areas. However, hot spots tend to have intensive land construction and ecological land is insufficient, so more attention should be paid to small- and medium-sized UGSs. This could supplement community parks and street garden parks, making use of every bit of space for greenery. It could improve government institutions, residential areas, roads, and other ancillary UGSs in order to increase UGS patch density. Regularly shaped green patches are most appropriate. In addition, water is also important, as it affects the cold island effect. According to statistics, the average LST of the green patch around the water body in the study area is 27.68 °C, and the minimum LST is 23.86 °C, which is also the lowest LST of the green patch in the whole six ring road area. It is evident that the water body has obvious significance for the cooling effect of UGSs [43]. Therefore, according to the actual UGS situation, water features can be appropriately added to achieve a combination of forest and water, which can better play the role of UGS patches in alleviating the effects of UHIs.

4.3. The Approach of Visualizing the Cooling Effect of Spatial Inhomogeneity of UGS to Help Guide the Construction of UGSs

As shown in Figure 8, according to the above analysis, PLAND has a greater contribution to the cooling effect. The proportion of UGS patches in the landscape area is large, and the regional average LST is low; however, when UGS patches are large enough, their cooling effect weakens. The contribution of SHAPE_MN and LSI to the cooling effect is much lower than that of PLAND. The cooling effect of regular UGS patches is better than that of complex shapes. The denser the patches are, the better the cooling effect is. Using the GWR method, the influence of the regional UGS landscape pattern on LST can be visualized, so that the regional UGS landscape composition and configuration can be more clearly and intuitively analyzed with respect to their impact on LST. Furthermore, regarding local urban greening construction needs, it is evident in the local regression coefficient distribution map that ecological land is limited and the UGS patches cannot be too large, especially in the built-up area. Combined with the hot spots of urban thermal environment, PLAND can clarify the influence of UGSs on the cooling effect in high LST areas to determine whether it can further enhance the greening area and quality. SHAPE_MN and LSI provide input regarding UGS construction, particularly regarding aggregation and dispersion situations, respectively.

4.4. Further Research on the Effect of UGS on Cooling the Surroundings

The LST is mainly influenced by the land surface, and the interaction and cumulative effects of ecological and artificial elements such as UGS, water bodies, and the built environment all contribute to the cooling effect. The sky-view factor, street canyon aspect ratio, and direction are important variables that can affect the cooling distribution [45,46]. This paper only analyzes the cooling effect of vegetation and water on the average surface temperature at their individual as well as regional scales; it does not analyze the influence of urban building height, density, and architectural form on surface temperature, or the combined effect of limiting the cooling range of UGS. However, this paper is based on Landsat 8 data, and the resolution of 30 m is not high enough for studying the urban street-scale LST. Furthermore, the resolution of the Landsat 8 thermal sensor is only 100 m which is insufficient for analyzing the effect of buildings on LST. Higher resolution satellite images and on-site measurements are considered to be better data sources, but this study is based on the accessibility and ease of analysis of Landsat 8 data to improve analytical applicability and provide guidance for future UGS planning and construction. We believe that the current research on the relationship between neighborhood building form on UHIs and the cooling range of UGSs in cities is still mainly appropriate to be conducted at the microscopic scale through in situ measurements and microclimate model simulations. Conversely, in research on the cooling effect of UGS on the surrounding areas, many scholars generally use a buffer zone analysis to determine the maximum cooling range, which homogenizes and simplifies the cooling effect of UGS on the surrounding environment. This ignores the complex surface features around UGS patches and the compound effect of adjacent UGS patches on their cooling effect [47], which is insufficient and needs further research.

5. Conclusions

In this paper, we proposed a new approach to investigate the spatially heterogeneous cooling effects of UGSs. Based on the analysis, we determined a 1500 × 1500 m grid as the analysis cell and optimized the selection of landscape metrics to establish the GWR model.

We firstly analyzed the impact of the patches’ own characteristics on the cooling effect, and then analyzed it at the regional scale. The conclusions are as follows: (1) the areas with strong influence of each landscape metrics on LST are basically consistent with the hot LST spots; (2) the cooling effect of various UGSs is according to the following pattern: forest > shrub > cropland >grass; (3) PLAND has the largest contribution to the cooling effect of LST, followed by LSI, and SHAPE_MN shows a positive correlation; (4) the internal LST would not change significantly if the UGS patch area was larger than a certain range; and (5) the cooling effect of regular UGS patches is better than that of complex shapes, and the denser the patches are, the better the cooling effect is.

These conclusions would help to explain the distribution characteristics of regression coefficients generated by the GWR model. The distribution of regression coefficients could visualize the cooling effect of spatial heterogeneity of UGS. The approach can clearly indicate the degree of influence of the spatial pattern of UGS on LST in each part of the study area, which can enhance the understanding on the effects of spatial configuration of UGSs on UHIs and better guide the future planning and construction of UGSs.