How Plot Spatial Morphology Drives Surface Thermal Environment: A Spatial and Temporal Analysis of Nanjing Main City

Abstract

Rapid urban development has changed urban substrate conditions, greatly affecting urban ecology and heating urban environment. Mitigating urban temperature rises by optimizing urban morphology is considered a promising approach; most studies ignore spatial and temporal heterogeneity. This study analyzes how plot spatial form influences urban thermal environment in the main Nanjing area from 2001, 2006, 2011, 2016, and 2021, based on geographically weighted regression models (spatio-temporal- and multi-scale). Results show that: 1. The formation of geothermal heat islands matches the direction of urban expansion, mainly due to changes in land substrate; 2. the spatio-temporal model performs best, indicating that urban morphology and surface thermal environment have obvious spatio-temporal heterogeneity; obvious scale differences exist in each index influencing the heat island effect; and 3. floor area ratio (FAR) and building density (BD) negatively and positively correlate with surface thermal conditions, with gradually increasing effect, respectively. Normalized difference vegetation index (NDVI) and distance from the nearest water body (Dis_W) negatively and positively correlate with surface thermal conditions separately; good ecological infrastructure reduces surface temperatures but shows a gradually weakening effect. Proximity to roads is associated with warmer thermal environment. This study elucidates how urban form influences surface thermal environments and suggests measures to reduce surface temperatures in the main urban Nanjing area.

1. Introduction

In rapid urbanization, urban populations are growing, development boundaries are gradually expanding, and the original natural land surface is gradually being replaced by the built environment, leading to an increasing urban heat island (UHI) effect [1]. Previous studies have shown that urban heat islands can contribute to global warming [2], increase urban energy demand [3], increase number of heatwave days [4], trigger respiratory, circulatory, cardiovascular, and cerebrovascular diseases in sensitive populations [5], and also lead to degradation of air quality, deterioration of the water environment, and disturbance of biological habitats [6,7,8].

The urban heat island effect refers to the situation whereby the atmospheric and surface temperature are higher than those of the surrounding suburbs or rural areas in rapid urbanization and industrialization [9,10]. In recent years, scholars have proposed the concept of an urban thermal environment based on the urban heat island effect. Comparing the two concepts, the common point is that the characterization factors are both surface and atmospheric temperatures. However, the measures of the urban thermal environment are related to various factors such as meteorological conditions [11], land cover classification [12,13], human activities [14], and urban morphology [15,16]. In terms of population activity, artificial elements with a high albedo, high solar absorption and high heat capacity reduce heat flux; anthropogenic heat such as traffic and air conditioning systems also increases the heat accumulation in the area [4], and the urban heat island effect is, to some extent, a complete reflection and manifestation of the urban thermal environment. Research into urban thermal environments is divided between the urban surface layer, urban canopy layer, and urban boundary layer [17]. The primary data source on the urban canopy and boundary layer is atmospheric temperature data, which have the advantage of temporal continuity. However, such data may also have low accuracy of scale conversion and present difficulties when analyzing the characteristics of large urban plan layouts [18,19]. Surface temperature data are the primary data source on the urban surface layer [20]. Rao first proposed remote sensing data in 1972 to study urban heat island effects [21]. Remote sensing technology advancements have substantially improved the data and approaches available for analyzing the surface thermal environment and are currently the main source for analyses of metropolitan thermal environments [22].

There is an extensive literature spanning the formation mechanisms, simulation, and mitigation measures on urban heat islands. The contributions of green areas, conversion of natural areas to other land uses, urban morphology, and artificial heat sources have been analyzed in detail. By studying the relationship between cold islands and green areas in Nanjing City, Kong et al. concluded that areas with extensive forest vegetation cover have a good cooling effect [23], because trees are one of the most important vegetation types, and their height relative to lower vegetation types (e.g., grassland and cultivated land) generally produces stronger transpiration and cooling effects [24]. The wind environment also affects surface thermal temperature, with the most potent cooling effect when the grid size is 25 m [25]. While He et al. suggested that the regional morphological characteristics of compact high-rise grid areas substantially improve ventilation and can mitigate UHI [26]. Yang et al. found a positive correlation between building density and wind speed at building densities below 0.6, thus affecting the urban thermal environment [27]. Molina studied changes in land cover classification over 20 years in Kennedy and reported a strong positive correlation between land surface temperature (LST) and the proportion of impervious surface [28]. The relationship between two-dimensional indicators (such as patch density and landscape shape indices) and urban heat islands have been examined [29,30,31]. More recently, studies have examined relationships between UHI and three-dimensional (3D) building parameters, such as the sky view factor and building height [32,33]. Abbas et al. suggest that roads contribute significantly to UHIs, and that urban heatwaves can be alleviated by changing the structure and materials of pavements [34]. Urban form is one of the most essential factors in the UHI effect, and even 3D architectural form has a more significant influence than other factors, such as surface coverage [35].

Among the existing literature, the conclusions on urban morphology and thermal environment are inconsistent, owing to the influence of experimental methods, research objects, research scale, artificial environment, and natural scale. For example, in the choice of study scale, urban morphology has a small effect on UHIs at the national or supra-national scales [36], whereas it has a large effect in block or regional studies [30]. Such differences can also arise in the models used to analyze the urban thermal environment, and studies have displayed that the results can differ between time series models and cross-sectional models, which can lead to negative or positive correlations between urban morphology indicators and UHI effects [37]. Most scholars have applied ordinary least squares (OLS), geographically weighted regression (GWR), and other methods to analyze urban morphology and thermal environment [38,39]. OLS estimates the dependent variable as a linear function of multiple predictor variables, ignoring spatial variation in coefficients. GWR allows coefficients to vary spatially, but its bandwidth is fixed for each observation [40]. Both models cannot measure the relationship between time variation and spatial distribution between variables.

Studies have shown that the population density in China is concentrated in the southeastern urban agglomerations, where this climate is comfortable and tends to continue to increase [41]. Moreover, the southeastern urban agglomeration is the region of the country most severely exposed to heat islands during the summertime [42]. At present, global warming and abnormally high temperatures in the Northern Hemisphere summer of 2022 exacerbated the deterioration of the urban thermal environment, highlighting the urgent need for urban planning to actively explore more effective environmental optimization and control strategies to cope with such changes. The present study analyzes the spatial characteristics associated with the thermal environment in the main urban area of Nanjing during 2001, 2006, 2011, 2016, and 2021. The most representative spatial morphological indicators of land parcels were selected, and GTWR (geographically and temporally weighted regression [43,44] and MGWR (multiscale geographically weighted regression) [45,46] were used to analyze the spatial and temporal relationships and driving mechanisms between the spatial morphology of land parcels and the surface thermal environment. We also propose corresponding policy measures to improve the thermal environment, enhance the quality of urban development and the living standards of citizens, and provide corresponding theoretical references for the optimization of urban spatial forms.

2. Materials and Methods

2.1. Study Area

Nanjing is the capital city of Jiangsu Province, China, located in the western part of the Yangtze River Delta (latitude 31°14′–32°37′ N, longitude 118°22–119°14′ E; see Figure 1). Annual average temperature is approximately 16.3° Celsius; annual precipitation is approximately 1150 mm. It is a subtropical monsoon climate characterized by hot and rainy summers, low temperatures, and little rain in winter. Nanjing is a typical city in this climate zone, ranking 6th out of 15 cities in the region with an urban population of more than 2 million. Specifically, it represents the situation in towns with hot summers and cold winters. Nanjing’s climate and topography make it a “furnace city,” which is of some research significance.

The main city of Nanjing covers an area of 809.97 km² and contains two parts: Jiangnan main city and Jiangbei new city. The main city of Jiangnan consists of “one core and six areas”: the core area of the old city, Xianlin District, Tiebei District, Qilin District, Hexi District, Chengnan District, and Dongshan District. The new main city of Jiangbei is composed of “one core and two pieces”, specifically the core area of Jiangbei, the Dafang district, and the Sanqiao district (Nanjing Bureau of Planning and Natural Resources, 2021).

Key to districts: GD = Gudu Cultural core; JB = Jiangbei core; DC = Dachang; TB = Tiebei; XL = Xianlin; QL = Qilin; DS = Dongshan; CN = Chengnan; HX = Hexi; SQ = Sanqiao.

2.2. Date Sources

This study makes use of satellite pictures from the United States Geological Survey (USGS; https://www.usgs.gov/ accessed on 12 May 2022) website, namely Landsat-5 TM images from 2001, 2006, and 2011, and Landsat-8 OLI images from 2016 and 2021 (for a total of 20 years). In six (for TM) or eight (for OLI) spectral bands at the visible and shortwave wavelengths, the pictures have a spatial resolution of 30 × 30 m. Other vector data from Open Street Maps (OSM; https://www.openstreetmap.org/ accessed on 15 May 2022) include administrative borders, buildings, road networks, and water bodies.

Data preprocessing mainly includes inversion of surface temperature and land cover from remote sensing images and the correction and verification of vector data. After FLAASH atmospheric correction, radiometric calibration, band combination, image mosaic, and clipping using ENVI 5.3 software, the atmospheric correction method was used to invert the surface temperature and obtain surface temperature data and NDVI (normalized difference vegetation index), which were then decoded using a support vector machine classification method (SVM) [47] to obtain urban land data for each year in the main urban area of Nanjing. The land cover types include urban pervious surfaces, impervious surfaces, and water. The gamma coefficient was set to 1/6 for the Landsat TM images and 1/8 for the Landsat OLI images. The penalty parameter was set to 100. A confused matrix was used to confirm the classification accuracy. The human–computer interaction method was used to check and correct the vector data with reference to Google Earth historical satellite images to ensure more accurate interpretation. The road network and building vector data were intersected using ArcGIS 10.5 software, FAR (floor area ratio) and BD (building density) were calculated, Dis_R and Dis_W were calculated using Euclidean distance analysis (Figure 2).

2.3. Research Methods

This study employs a three-stage process: (i) classification based on preprocessed heat island effect classes and surface cover types; (ii) determining relationships between urban morphology and surface thermal environment using OLS, GTWR, and MGWR, and evaluating model performance; and (iii) analyzing the model outputs to identify the mechanisms driving each type of influence.

2.3.1. Land Surface Temperature Inversion

Using Landsat imagery, the surface temperatures for the five years of the study area were inverted based on the atmospheric correction method [48] to establish the radiative transfer equation:

$$L\lambda =\left[\epsilon \times B\left(T_s\right)+\left(1-\epsilon \right)\times L\downarrow \right]\times \tau +L\uparrow$$

(1)

$L_{\lambda }$ is the brightness of the thermal infrared radiation received by the satellite sensor (W $\times {\mathrm{m}}^{-2}\times {\mathrm{sr}}^{-1}\times {\mathsf{\mu }\mathrm{m}}^{-1}$), $ϵ$ is surface emissivity, $T_s$ is the real surface temperature (K), B($T_s$) is the brightness of blackbody radiation (W $\times {\mathrm{m}}^{-2}\times {\mathrm{sr}}^{-1}\times {\mathsf{\mu }\mathrm{m}}^{-1}$), $\tau$ is the atmospheric transmittance in the thermal infrared segment, and $L_{\uparrow }$ and $L_{\downarrow }$ are the atmospheric downward and upward radiation (W $\times {\mathrm{m}}^{-2}\times {\mathrm{sr}}^{-1}\times {\mathsf{\mu }\mathrm{m}}^{-1}$), respectively. This equation can be derived from the temperature $T_s$ of the black body in the thermal infrared band radiation brightness B($T_s$), according to Planck’s formula function, and can be obtained from the temperature $T_s$ as:

$$T_s=\frac{K2}{\ln \left(\frac{K1}{B\left(T_s\right)}+1\right)}$$

(2)

$K_1$ and $K_2$ were obtained from image header files. Atmospheric profile parameters were obtained from the NASA website (https://atmcorr.gsfc.nasa.gov/ accessed on 20 May 2022) and were inverted for five years of remotely sensed imagery. Owing to the large time span of the images selected for the study and the differences in meteorological conditions, direct comparison was not possible. The surface heat field variability index (UTFVI) was calculated from the surface temperature [49] and the threshold method was used to classify the HI into six classes corresponding to the UHI effect classes (

Table 1. Heat Island Classification.

Hot Field Variability Index (HI)	Heat Island Effect Level
HI ≤ 0	No heat island effect
0 < HI ≤ 0.05	Low heat island effect
0.05 < HI ≤ 0.1	Medium heat island effect
0.1 < HI ≤ 0.15	Sub-high heat island effect
0.15 < HI ≤ 0.2	High heat island effect
0.2 < HI	Ultra-high heat island effect

) as shown in Equation (3) [50]:

$$HI=\frac{T_{LST}-T_{mean}}{T_{mean}}$$

(3)

2.3.2. Selection of Spatial Morphological Indicators for Plots

Random points were generated in ArcGIS 10.5 based on the vector boundary of the main urban area of Nanjing, and the nearest small spacing was limited to 200 m to ensure that the random points were as scattered as possible and covered a larger area. The indicators of each point were extracted to random points, anomalies located at the boundary were manually checked, and 3000 valid random points were obtained for each year of the study (Figure 3).

Urban morphology is a relatively macroscopic concept, defined as the spatial configuration of fixed elements [51]. However, it is more often expressed as a two-dimensional plane of the whole city and the basic morphological pattern. As this study investigates surface thermal environmental differences and impacts within the city, additional indicators expressing micro-environmental differences are needed, namely spatial morphological indicators reflecting site differences. Therefore, this study adopt the indicators that were used in previous studies and employs BD, FAR, NDVI, Dis_R, and Dis_W, which are five more representative indicators to portray the city from three perspectives of building morphology, ecological infrastructure, and transportation system, collectively called the spatial morphology indicators of the site [52]. Details are presented in

Table 2. Classification and definition of spatial morphological indicators for land parcels.

Plot Spatial Form	Name	Definition
Architectural form	BD	The ratio of building footprint to floor area within a certain area
	FAR	Ratio of gross floor area to net site area within a certain area
Ecological infrastructure	NDVI	The difference between the reflected values in the near-infrared band versus that in the red band is > the sum of the two. This is an important metric of plant growth and nutritional information
	Dis_W	Distance to the nearest water body
Transportation system	Dis_R	Distance to nearest road

.

2.3.3. Selection and Evaluation of Regression Models

(1)

Ordinary least squares

OLS regression is a common statistical method for analyzing the interdependence of the explanatory and explanatory variables and is calculated as follows:

$$y_i={\beta }_0+\sum {}_{k=1}^p{\beta }_k{X}_{ik}+{\epsilon }_i$$

(4)

where $y_i$ is surface temperature at point i, ${\beta }_0$ is the model intercept, p is the number of selected morphologies, $X_{ik}$ is the value of the kth morphological indicator at point i, ${\beta }_k$ is the regression coefficient of the kth indicator, and ${\epsilon }_i$ is the random error. OLS is a non-spatial model; it calculates the average fit of the coefficients globally but cannot reflect spatial variations due to differences in the geographic environment.

(2)

Spatio-temporal geographically weighted regression model

In 2010, Huang et al. extended the GWR model to include a temporal dimension, termed the GTWR [43]. By incorporating both spatial and temporal coordinates, GTWR produces a local weighting matrix that simultaneously reflects both spatial and temporal heterogeneity. Traditional geographically weighted regression analysis does not include a time dimension and the studied object’s fixed spatial coordinate data. The GTWR model, in contrast, necessitates that the object under study has distinct spatial coordinates at distinct points in time; more significant coordinate overlap yields results that are more akin to those of a linear regression analysis with time as the study, which is better able to describe the spatio-temporal relationship between the explanatory variables and the dependent variable. The formula is as follows:

$$y_i={\beta }_0\left(u_i,v_i,t_i\right)+\sum _k^m{\beta }_k\left(u_i,v_i,t_i\right)X_{ik}+{\epsilon }_i$$

(5)

($u_i$,$v_i$) denotes the latitude and longitude coordinates of the ith sample point, ($t_i$) denotes the observation time, ${\beta }_k$($u_i$,$v_i$,$t_i$) is the regression coefficient of the kth independent variable of the ith sample point, ${\epsilon }_i$ is the random error of the ith sample point, and ${\beta }_0$($u_i$,$v_i$,$t_i$) is the regression coefficient of the ith sample point, which is determined by the spatial location of sample point i and the time decision.

(3)

Multiscale Geographically Weighted Regression (MGWR).

Fotheringham and colleagues created MGWR [39]. The local coefficient variation between the explanatory variables and each explanatory variable at the same scale limits both GWR and GTWR. However, the conditional correlations between the explanatory factors and other explanatory variables are subject to change at various geographic scales in MGWR. According to the examination of the link between the surface thermal environment and each variable, the regression coefficients of MGWR are based on local regression results, and each explanatory variable has the maximum bandwidth accessible. The formula used is as follows:

$$y_i={\beta }_0\left(u_i,v_i\right)+\sum _k^m{\beta }_{bwk}\left(u_i,v_i\right)X_{ik}+{\epsilon }_i$$

(6)

where ($u_i,v_i$) denotes the coordinates of the sample points, and ${\beta }_{bwk}$ denotes the explanatory variable K for the regression coefficients. $bwk$ is the bandwidth, and ${\epsilon }_i$ is the error term of the model.

The regression analysis in the present study utilized ArcGIS 10.5, GTWR, and MGWR2.1 software.

3. Results

3.1. Land Cover Patterns and Changes

The SVM shows that the classification Kappa coefficients for 2001, 2006, 2011, 2016, and 2021 are 83.6%, 84.5%, 87.3%, 93.0%, and 93.4%, respectively, thereby meeting the requirements for evaluating changes in land cover.

From a temporal perspective, the impervious surface area in the main urban area of Nanjing increased from 328.31 km² to 604.03 km² in 20 years, an increase of 257.72 km², while the permeable surface area decreased from 454.09 km² to 188.55 km², a decrease of 266.54 km², with a large amount of permeable surface being replaced by impervious surface (see

Table 3. Land cover transfer matrix (km²).

2001	2021
	Water	Impervious Surface	Pervious Surface	Total
Water	9.22	13.76	4.59	27.57
Impervious surface	3.76	294.19	30.36	328.31
Pervious surface	4.41	296.08	153.60	454.09
Total	17.39	604.03	188.55	809.97

). Between 2001 and 2006, the impervious surface increased by 148.34 km², and the city was in a high growth phase. Between 2006 and 2016, the impervious surface increased by 124.01 km², and the city’s growth rate returned to a medium level. Between 2016 and 2021, the city’s growth rate entered a low growth mode, and the impervious surface only increased by 3.37% during this period (Figure 4).

From a spatial perspective, the central area of Nanjing’s main city has mostly stayed the same in the past 20 years. However, the impervious areas of the surrounding Xianlin, Qilin, Dachang, Sanqiao, Hexi, and Dongshan districts have increased dramatically. In 2001–2006, the impervious areas of all districts in the central city expanded. However, mainly in Hexi, Chengnan, and Dongshan districts, the center of gravity of urban development shifted to the southeast. In 2006–2011, the impervious areas of Kirin, Chengnan, and Dachang districts expanded rapidly, and the center of gravity of the city shifted to the north. In 2011–2021, the expansion of impervious surface in the main urban area of Jiangnan gradually saturated, while the establishment of the Jiangbei National New Area caused the center of gravity of urban development to continue to shift to the north. Dachang, Jiangbei, and Sanqiao districts expanded rapidly, with many previous surfaces being replaced (Figure 5).

3.2. Spatial and Temporal Distribution Characteristics of Heat Islands

The spatial distribution of heat islands in the main urban area of Nanjing varies greatly from 2001 to 2021, mainly characterized as “strong in the north and weak in the south, and migrating from the center to the periphery”, i.e., a trend of expansion outward from the center (Figure 6). In 2001, urban heat islands were mainly concentrated in the Tiebei and Dachang high-tech districts and were closely related to the density of industrial plants in the area and the large amount of heat generated by industrial production; heat islands were sporadically distributed in the Hexi and Dongshan districts; and were distributed sporadically in other areas due to their low levels of urbanization. After 20 years of urbanization, the temperature zones in Nanjing also changed significantly.

Analyzed from a temporal perspective, the average summer surface temperature of the central city increased from 29.54 degrees in 2001 to 34.67 degrees in 2021, and the surface temperature of Nanjing increased significantly. Moreover, 51.69 km² of UCI area decreased during the 20 years, 179.25 km² of Low UHI area decreased and almost disappeared, and 41.35 km² of High UHI and Ultra UHI area increased(

Table 4. Transfer matrix of heat island effect.

2001	2021
	UCI	Low UHI	Medium UHI	Sub-High UHI	High UHI	Ultra UHI	Total
UCI	137.55	1.24	76.19	47.53	16.76	11.13	290.39
Low UHI	47.78	0.02	62.21	45.30	16.63	8.64	180.58
Medium UHI	28.93	0.03	54.77	44.22	16.27	7.70	151.93
Sub-high UHI	13.58	0.03	30.85	33.89	15.15	6.50	100.01
High UHI	6.29	0.02	14.14	16.69	10.44	6.15	53.73
Ultra UHI	4.56	0.00	7.70	8.03	6.01	7.04	33.34
Total	238.69	1.33	245.86	195.66	81.26	47.16	809.97

). From 2001 to 2011, the UCI and Low UHI area continued to decrease, and the total area of UH increased in general. From 2011 to 2021, the area of UCI decreased. Low UHI in the heat island area continued to decrease and nearly disappeared, Medium UHI and Sub-High increased in general, and High UHI increased in general, Sub-High overall increase, and High UHI and Ultra UHI area decreased (Figure 7).

From a spatial perspective, the heat island effect decreases from high to low-level circles. In 2001, Tiebei and Dachang High-tech Zones were the main concentrations of urban heat islands. From 2001 to 2006, the heat island area was concentrated in the central city and spread to the city’s south, and the wedge-shaped cold island area disappeared. From 2006 to 2021, the Ultra UHI in the central city continued to decrease, while the Ultra UHI in Dongshan, Tiebei, and Dazhan increased, and the UCI area in the northern part of the city also decreased. However, the UCI area in Hexi and Kirin increased slightly. There is a gradual “north to south strength and migration from the center to the periphery,” i.e., a trend of expansion from the center outwards (Figure 6).

Industrial land has a significant impact on heat islands; however, with industrial transformation and upgrading in recent years, industrial land within the main urban area has contracted, and so the impact of industrial production on the surface thermal environment has decreased. ArcGIS was used to extract centers of gravity for the urban heat island and urban impervious surface categories, and showed that both shifted in the same directions in 2001, 2006, 2011, 2016, and 2021: from west to east and then northwest (see Figure 8). This demonstrates an important link between alteration of the urban subsurface and the occurrence of heat islands, and that urban heat islands are closely related to urban development processes [53].

3.3. Spatial Autocorrelation Results

Geoda 9.5 software was used to calculate the global Moran’s index of thermal environment autocorrelation for Nanjing Main City in 2001, 2006, 2011, 2016, and 2021 (see

Table 5. Global Moran I-value, p-value, and Z-value.

Year	Moran’s I	Z Score	p Value
2001	0.453	42.83	0.001 ***
2006	0.519	48.86	0.001 ***
2011	0.350	33.71	0.001 ***
2016	0.465	43.76	0.001 ***
2021	0.396	36.54	0.001 ***

Note: *** indicates that the parameter passed the significance test at the 1% level.

). The global Moran’s indices for thermal environment were all > zero, p-values were all < 0.05, and z-values were all > 2.58. The data for each year showed high spatial autocorrelation, making them suitable for subsequent analyses. Areas with strong heat island effects showed significant spatial clustering, as did cold islands. The local Moran’s index was used to produce scatter plots of the Moran’s I index for each study year (2001, 2006, 2011, 2016, and 2021).

As can be seen from Figure 9, the H-H aggregation district in the central area of the main city gradually shifted to the Dachang district, while the L-L aggregation district was located in the central area of the main city and Sanqiao and Xianlin districts, with significant points gradually decreasing. The H-L and L-H aggregation districts were scattered and not aggregated.

3.4. GTWR Model Estimation and Result Analysis

The independent variables were tested for multicollinearity to avoid estimation bias caused by interaction effects. As shown in

Table 6. Covariance test of urban morphology variables.

	NDVI	FAR	BD	Dis_W	Dis_R
VIF	1.155	2.489	2.450	1.062	1.152

, the variance inflation factor (VIF) of each variable was less than 5.0, indicating that there was no multicollinearity among the variables. In addition, unit root tests were conducted to check the stability of the variables and avoid pseudo-regression problems. The results (see

Table 7. Panel root unit test results.

	LLC		IPS		ADF		Conclusion
	Statistical Quantities	p-Value	Statistical Quantities	p-Value	Statistical Quantities	p-Value
LST	−2.2 × 102	0.000 ***	−15.1603	0.000 ***	65.0602	0.000 ***	Smooth
NDVI	−1.5 × 102	0.000 ***	−30.9100	0.000 ***	55.0895	0.000 ***	Smooth
Dis_W	−1.7 × 102	0.000 ***	−32.3390	0.000 ***	56.4783	0.000 ***	Smooth
Dis_R	−1.8 × 103	0.000 ***	−35.0039	0.000 ***	58.7999	0.000 ***	Smooth
BD	−1.1 × 104	0.000 ***	−40.3368	0.000 ***	67.1552	0.000 ***	Smooth
FAR	−1.8 × 105	0.000 ***	−25.6697	0.000 ***	56.4691	0.000 ***	Smooth

Note: *** indicates that the parameter passed the significance test at the 1% level.

) confirmed that all variables were stationary [54].

The model was computed using ArcGIS10.5 software and GTWR plug-in, and the results were compared with the OLS model. Referring to the OLS model results, R² and Adj.R² were 0.3028 and 0.3020, respectively, and the AICc was 55,383.2. The R² values of the GTWR model is shown in

Table 8. GTWR regression model parameters.

Dimensions	Explanatory Variable	Max	Minimum	Average Value
Architectural form	FAR	0.388	−0.977	−0.353
	BD	12.889	1.442	8.013
Ecological Infrastructure	NDVI	−1.922	−7.845	−4.376
	Dis_W	0.091	−0.064	0.029
Transportation System	Dis_R	0.048	−0.454	−0.258
R²	0.8464
Adj.R²	0.8422
AICc	41614.9

: It can be seen that the R² and Adj.R² values of the GTWR model are significantly higher (0.8422) and AICc (41,614.9) is significantly lower than in the OLS model. These results indicate that the GTWR model performs much better than the traditional OLS model. Multiple time-series analyses using panel data provide a new perspective on the time-varying dynamics of the study.

From Table 8, the average effects of the five selected elements indicate that Dis_W, FAR, and Dis_R have a bidirectional effect on the surface temperature. For surface temperature, NDVI was negatively correlated whereas BD was positively correlated. Statistical analyses of the regression data show that the five variables have very different effects on surface temperature, with the largest effects seen for FAR and BD. For each unit increase in BD and FAR, surface temperature increases and decreases by 0.08 °C and 0.35 °C, respectively. The effects of these variables on the spatial and temporal heterogeneity of surface temperature are further analyzed below.

3.4.1. Influence of NDVI

The mean regression coefficient of NDVI on urban surface temperature was −4.376, indicating that NDVI has a negative effect on urban surface temperature, and the larger the absolute value of NDVI, the greater the surface temperature reduction effect. Vegetation and green spaces can be very effective for carbon sequestration and oxygen release to mitigate UHI effects. However, from a temporal perspective, the mean value of the regression coefficient of NDVI changed from −5.620 to −3.228 during 2001–2021, indicating that the beneficial effect of NDVI in lowering urban surface temperature gradually diminished. The mean regression coefficients from 2001 to 2011 show a slow rise and fall overall, and the absolute value of the mean regression coefficients from 2011 to 2021 decreases rapidly. This is because the permeable surface area within the main city of Nanjing has decreased by nearly 60% in 20 years; green areas still have a cooling effect as cool islands of the city, but their range of influence is greatly reduced, which causes the average regression coefficient of NDVI to increase in the corresponding years.

Based on the spatial distribution characteristics (see Figure 10), the spatial distribution of NDVI mean regression coefficients in 20 decades show a decreasing trend, with the southeast as the core. In 2001, the strong significant areas were located in the Qilin, Xianlin, and Dachang districts; the core areas of the old city and part of the Hexi districts were weakly significant; and the significance of other districts was average. By 2021, the spatial distribution characteristics of strong and weak significance were the same, but the absolute value of the NDVI mean regression index decreased. This is because in 2001 the Qilin, Xianlin, and Dachang districts had abundant natural resources, with 76.3%, 75.8%, and 57.9% permeable surface area, respectively, and the NDVI level was relatively high. However, over the following 20 years the core areas of the Jiangbei, Hexi, and Dongshan districts developed rapidly, permeable surface area decreased, and the surface thermal environment was affected by NDVI at a relatively low level. The Qilin and Xianlin districts have also undergone rapid development; large forests and mountains still have a cooling effect, but their ranges of influence are significantly reduced. Taking Zijinshan Forest Park in the core area of the old city as an example: A buffer zone of 1–4 km range was established in 1 km units and the average temperature inside each buffer zone was calculated. The cooling intensity of each buffer zone in 2001 was 1.02 °C, 0.50 °C, 0.22 °C, and 0.10 °C, respectively, whereas in 2021, these had declined to 0.61 °C, 0.43 °C, 0.16 °C, and 0.06 °C, respectively. A cooling effect was therefore still evident in 2021, but its impact was reduced.

3.4.2. Influence of FAR

The GTWR model showed that FAR had a negative effect on urban surface temperature, with a mean regression coefficient of −0.353. The mean regression coefficient changed from −0.285 to −0.681 in 20 decades, indicating a gradual increase in this negative effect of FAR on urban surface temperature. Although the increase in plot ratio brings more floor space and higher energy consumption, built-up areas with a high plot ratio tend to form a street valley pattern that provides functions such as shading and blocking sunlight, and tend to form a ventilation corridor to promote heat exchange between the city and its surroundings, and so the increase in plot ratio can effectively reduce surface temperature.

In terms of spatial and temporal distribution (see Figure 10), the average regression coefficient of FAR changed from a high center and low periphery to a low center and high periphery layout over 20 years; however, the overall degree of impact was significantly higher, and the change was closely related to the key areas of urban construction and the change in FAR in each area. The Jiangbei core districts and Xianlin, Dongshan, Dachang, Qilin, and Hexi districts have been the major construction areas of Nanjing’s urban outreach expansion over the past 20 years, and their volume ratios have seen the most rapid increases, which corresponds to LST. Only three areas within the main city have special characteristics. The volume ratio of the core area of the old city has hardly increased, but the average temperature has increased by approximately 6.0 °C, so it can be seen that the main factor causing the temperature increase is not the increase in volume ratio (Figure 11). The smallest temperature rise in the Tiebei district was related to the transformation of industrial enterprises in the area. In contrast, the Hexi area was influenced by government policies, and the construction of a high-standard, modernized Hexi New Town began in the south-central part of the Hexi area in 2011, and a large number of surfaces were converted from permeable to high-volume impermeable surfaces, both of which had a negative effect on the surface thermal environment. Therefore, although volume ratio increased most in the Hexi area, the average temperature only rose by about 3.2 °C, and the absolute value of the FAR regression coefficient versus surface thermal environment was low.

3.4.3. Influence of Building Density

The main reason for the effect of building density on surface temperature is that building roofs have much lower capacity for heat storage compared with green areas and water surfaces of the same area; with solar irradiation, this leads to warmer surface temperatures in areas with greater building density. The average regression coefficient of BD on urban surface temperature in Nanjing main city was 8.013, and increased from 6.307 to 9.637 between 2001 and 2021, indicating that the positive correlation between BD and urban surface temperature has strengthened over time.

During the 20-year study period, the regression coefficient of Nanjing’s central urban area showed a trend from high to low but then rose again (see Figure 10), while the regression coefficient for the peripheral areas of the main city showed a consistent increase, the pattern of which is closely related to shantytown renovation, urban outreach expansion, and internal filling. For example, prior to 2011, the Chengnan District had 30.67 km² of shantytowns, accounting for 39.6% of the total area. Subsequently, all new developments were required to comply with building densities specified in the “Jiangsu Province Urban Planning Management Technical Regulations” and other relevant regulations. The overall building density of the district has decreased from 18.1% to 15.2%, while the area of open space such as green space within each site has increased, and the influence of building density on surface temperature has decreased. With the completion of Nanjing South Station in 2011 and the relocation of Daxiao Airport in 2015, the area entered a stage of infill development; the original permeable ground was gradually replaced by various construction areas; over a period of 10 years, impermeable ground cover increased by 4.33 km² (from 78.24% to 83.84%), and the influence of building density on surface temperature increased.

3.4.4. Influence of Road Proximity

The average regression coefficient of Dis_R for urban surface temperature was −0.258, indicating that Dis_R had a negative impact on urban surface temperature (Table 8). The average regression coefficient in 20 decades decreased from −0.173 to −0.344, showing an increasing trend followed by a decreasing trend. This shows that the negative impact of Dis_R on urban surface temperature gradually decreases and increases. This is because dark-colored road surfaces absorb and then release large amounts of solar heat energy during daytime, and the use of motorized vehicles also generates a large amount of artificial heat, which can raise the surface temperature of the road and the surrounding area.

Total road length in Nanjing main city increased from 4308 km to 5033 km over 20 decades (an increase of 0.17 times), and motor vehicle ownership in Nanjing increased from 315,000 to 2,936,000 from 2006 to 2021 (an increase of nearly 10 times). All of these factors contribute to the gradual increase in the influence of Dis_R factors on the surface thermal environment in surrounding areas.

From a spatial perspective (see Figure 10), the mean regression coefficient of Dis_R changes from a pattern of high center and low periphery to a pattern of high periphery and low center, and the pattern is closely related to the increase in road network density and motor vehicle ownership. In general, the high road network density in the western area of the river, the core area of the old city, and the southern area of the city showed small changes (Figure 12), with early building age, short distances between buildings and the road boundary line, and good road greening, which in combination have a small increase in temperature relative to other areas. The peripheral areas of the main city, such as Dachang, Xianlin, Qilin, Dongshan, and Sanqiao districts, have a dense network of elevated roadways, and new roads are wide with little green coverage, which has a greater negative impact on the surface thermal environment in the surrounding areas.

3.4.5. Influence of Water Proximity

Dis_W has a positive effect on urban surface temperature, with a mean regression coefficient of 0.029. The mean regression coefficient decreases from 0.052 to 0.006 across the 2001–2021 study period, indicating that the positive effect of Dis_W on urban surface temperature gradually decreases and the negative effect increases. The reason for the effect of Dis_W on the urban surface temperature is that the specific heat capacity of water is relatively large, which tends to form partial cold island. The local cold islands promote energy exchange. However, after 20 years of development, the impermeable surface area within the central city has grown, and the area of water bodies has been greatly reduced, thereby weakening the influence of water bodies on surface temperature. Second, the increase in buildings located close to water bodies may inhibit water–land circulation, such that temperatures are actually lower with greater distance from water bodies.

Spatially, the average regression coefficient of Dis_W gradually shifts from a high center surrounded by a low layout to a low center surrounded by a high layout (see Figure 10). According to the heat island layout map, the heat island effect was significant in the central urban area in 2001, and the natural cold islands such as the Yangtze River and Xuanwu Lake played a cooling role; therefore, the farther away from the water body, the higher the temperature. With urban development and construction, the heat island migrated to the surrounding urban area, resulting in the gradual and significant cooling effect of the water body on the surrounding urban area. A large number of lakes or paddy fields in the Qilin and Sanqiao districts have been replaced by impermeable surfaces; therefore, the cooling effect of water bodies decreases year by year or even appears negative.

3.5. MGWR Model Estimation and Result Analysis

Prior to conducting MGWR regression analysis, the selected spatial morphological indicators of the plots in 2021 were screened; the Dis_W regression coefficient was found to be insignificant in OLS regression analysis and was therefore excluded.

Table 9. Comparison of OLS, GWR, and MGWR models.

Regression Model	AICc	R²	Adj.R²
OLS	8701	0.348	0.346
GWR	8328	0.569	0.496
MGWR	4330	0.661	0.609

shows that the MGWR regression results are more accurate than those of OLS with GWR. The results of the 2021 MGWR are shown in

Table 10. MGWR regression result.

Explanatory Variables	Bandwidth	Max	Middle Value	Minimum	Average Value
NDVI	43	0.486	−0.401	−1.074	−0.368
BD	159	0.673	0.362	0.041	0.365
FAR	2999	−0.214	−0.221	−0.228	−0.221
Dis_R	441	0.090	−0.123	−0.343	−0.129

and Figure 13, and the bandwidth of each indicator follows the sequence: NDVI < BD < FAR < Dis_R. Variable bandwidths provide a more accurate and comprehensive perspective for exploring the relationships between spatial data.

The bandwidth of NDVI is 43, accounting for 2.2% of the total sample size; therefore, the regression scale is approximately 18 km², which has considerable spatial heterogeneity, and the regression coefficient ranges from −0.216 to −0.672, with an average regression coefficient of −0.368, which indicates that an increase in surface vegetation can reduce surface temperature. From a spatial perspective, the influence was greater in areas with lush surface vegetation and less pronounced in areas with less surface vegetation. Higher NDVI values correspond to lower regression coefficients; for example, large mountainous areas such as Zijin Mountain, Mufu Mountain, and large parks in urban areas. In contrast, lower NDVI values correspond to higher regression coefficients, for example, Xinjiekou and Dachang districts in the city center. From an urban macroscopic perspective, urban forests lose less water and have more stable heat fluxes than grasslands, which is more conducive to mitigating heat waves and other effects of climate change [55]. From a micro perspective, the results indicate that the area threshold for cooling effects by green space was 0.848 hm² in high-density areas within the main city of Nanjing, and 0.384 hm² in low-density areas [56]. To determine the extent of NDVI influence, the study used R (version 4.1.1) to conduct smoothing spline analysis of NDVI and temperature, and the results showed significant changes in temperature reduction with surface vegetation cover greater than 60% (Figure 14). In the process of urban development, protecting natural ecological elements and planning artificial ecological elements above a certain scale in areas with a high level of urbanization can preserve more complete ecosystems, and ecological corridors can effectively improve the capacity and scope of the cold-island effect associated with green spaces.

The bandwidth of FAR is 2999, accounting for 99.9% of the total sample size, with a regression scale of 807 km², weak spatial heterogeneity, which is globally significant, and a mean coefficient of −0.221, which confirms that the negative effect of FAR on the global is significant. Spatially, the effect of floor area ratio on the surface thermal environment was more pronounced at locations with more intensive site development and less evident at less developed sites, but the difference was not significant. For example, the regression coefficients were small in the more developed areas of the Xinjiekou and Jiangbei core areas, and generally higher in the southern part of the city that has a low volume ratio, but the difference was only 0.009. The gentle decline in the fitted curve also reflects the global nature of the effect of volume ratio.

The bandwidth of BD is 159, accounting for 6.25% of the total sample size, and the regression scale is approximately 51 km², which indicates large spatial heterogeneity. BD regression coefficients range from 0.041 to 0.673 and averaged 0.365, indicating that BD has a highly significant positive effect on surface temperature. Analyzed from a spatial perspective, the higher the BD, the larger the regression coefficient, and the lower the BD, the smaller the regression coefficient. The corresponding regression coefficients are higher in the core area of the old city and the Dachang district, which have higher densities. The regression coefficients of the Dongshan district and Jiangbei core areas, which are under development, are generally lower. Combined with the fitted curve, building density in the newly built areas should be limited to less than 40% as much as possible in order to avoid a strong heat island effect. In addition, in the context of stock development, increasing building heights may provide another means of mitigating the heat island effect that results from high-density development.

The bandwidth of Dis_R is 441, accounting for 14.7% of the total sample size, and the regression scale is approximately 119 km², indicating large spatial heterogeneity. The regression coefficients range from −0.343 to 0.090, averaging −0.129, indicating that Dis_R negatively affects surface temperature reduction. From the spatial analysis, this indicator is closely related to road network density and population activities. The absolute value of the regression coefficient of Dis_R is smaller. The degree of influence is lower in the northern area of the core area of the old city and the eastern area of the Chengnan district, where the roads were mostly built during 1919–1949, and the road greening conditions were good; the higher the density of the road network, the larger the absolute value of the regression coefficient. In addition, the spatial distribution of the area with a larger absolute value of regression coefficient shows a strong relationship with the urban expressway system; the correlation between the two was 56.7%, which was significantly higher than that in other areas, which is closely related to the width of the expressway road, less greenery, and greater vehicle numbers.

4. Discussion

This study uses Landsat-5 TM and Landsat-8 OLI remote sensing data to obtain the urban surface temperature in the main urban area of Nanjing. A series of urban morphological indicators are selected for a calculation to investigate the relationship between the thermal environment and the spatial morphology of land parcels in the main urban area of Nanjing. In addition, this study fully considers the relationship between the spatial morphology of land parcels and the surface thermal environment from the perspective of time and space, which is essential for optimizing urban development and improving the human living environment.

4.1. Differences from Related Studies

Compared with previous studies, this study has the following characteristics: 1. Current research on the urban surface thermal environment and urban morphology is mainly concerned with the measurement of the surface thermal environment, the influence of vegetation on the thermal environment, the correlation analysis of indicators and the thermal environment, and the influence of microscopic building morphology on the thermal environment. This study is based on quantifying the spatial morphology of plots and the condition of the surface thermal environment at the urban scale, exploring the strength of different scales and variables in the same urban system, and providing a reference point for studying their spatial effects. 2. Existing studies have focused on the effects of different indicators of urban morphology on the surface thermal environment and have made cross-sectional comparisons. In contrast, this study has made a longitudinal comparison between the spatial morphology of plots and the surface thermal environment at different levels. The data are analyzed from a long time series perspective to explain the dynamic patterns of change, thus providing predictions and recommendations for future changes.

4.2. Spatial and Temporal Driving Mechanisms of Plot Spatial Patterns on the Surface Thermal Environment

From the research process, it is clear that the urban heat island effect is becoming increasingly evident as urban development and expansion increase the impermeable surface of the city and that this law exists objectively and clearly. Further research into the impact of urban land use on the urban thermal environment includes elements such as green areas and water surfaces as sources of cooling and industry and roads as sources of heat, and elements such as floor area ratio, building density, and building height within the site. In the case of Nanjing’s central city, the various elements exert different influences in different spaces and at different times.

Industrial land has a significant influence on the heat island effect. However, as the city develops and industrial land in the central urban area decreases, the influence of this factor weakens and is not included in the quantitative analysis of this study. Green space, water bodies, and volume ratio are negatively correlated with the thermal environment, while building density, and roads are positively correlated. Over 20 years, the intensity of the five indicators on the thermal environment of Nanjing’s central city showed a BD > NDVI > FAR > Dis_R > Dis_W. On the spatial scale, the influence of FAR is significantly more substantial than the other indicators; Dis_R is second and higher than BD; NDVI is the lowest. In contrast, NDVI > BD > FAR > Dis_R and Dis_W are insignificant. The results obtained in this study are consistent with the findings of existing studies. However, there are differences in the scope of the study and the selection of elements, which deepen and refine existing studies.

The floor area ratio will be discussed among the various influencing factors. Generally speaking, an increase in floor area ratio means more floor space, more production and living activities, more energy consumption and heat release on the same building site, and a more pronounced heat island effect in the area. However, the actual study results show that there is not a simple linear relationship between increased plot ratio and elevated surface thermal environment. Under conditions where the site plot ratio is less than 2.0, the surface thermal environment rises as the plot ratio increases. In contrast, under conditions where the plot ratio of the site is more significant than 2.0, the surface temperature decreases instead as the plot ratio increases. The results of this study are closely related to the relevant urban planning and building design technical regulations. Taking the Jiangsu Province Urban planning Management Technology Stipulation as an example, a floor area ratio of around 2.0 indicates multi-story buildings of no more than 24 m building height (residential buildings generally do not exceed 35 m). Under conditions within the upper limit of 2.0, an increase in floor area ratio does not generally bring about drastic changes in other elements. However, it increases total building energy consumption, heat release, and surface temperature. The building density decreases as the number of building stories increases and the plot ratio increases further. The number of building stories increases from multi-story to high-rise, corresponding to an increase in plot ratio of about one time, but the building density of residential buildings decrease from 28% to 20%, and the building density of office buildings decreases from 45% to 35%. The building density of commercial buildings also decreases by 5%. The decrease in building density and the increase in the open space ratio often means an increase in the green space ratio. At the same time, with higher floor area ratios and lower building densities, meaning higher building heights, taller buildings can leave huge shadows on the ground surface, causing a drop in surface temperatures. The increase in the green space ratio and building shadows reduce the surface temperature. In contrast, the increase in energy consumption due to the increase in floor area ratio has a positive effect on the surface temperature, and the two superimpose on each other to bring about an adverse effect of floor area ratio on the thermal environment of the surface.

4.3. Suggestions for Optimizing the Spatial Form of Urban Plots

• In developing countries, the urbanization process continues, and the outwards expansion and internal filling of cities have been ongoing for a long time. Although concepts and measures such as sponge cities can add impervious surfaces, there is still a trend of increasing impervious surfaces and reducing green areas and water systems within the main city. However, considering the actual experience of human use and the relevant requirements of planning for green space, the combination of large and small green spaces can form a more beneficial urban green space system, while retaining and building large green spaces in the city as much as possible. As opposed to that, incorporating green spaces can produce better outcomes when balancing human use and cooling the surface thermal environment. Additionally, increasing the amount of green space on each site is crucial. Additionally, adding extra green space to each site and connecting the roads can reduce surface temperatures.
• In terms of building form, the construction of low-density, high-rise buildings provides an important means of slowing increases in surface temperature; however, taller buildings are associated with higher construction and operational costs and greater energy consumption. Furthermore, high-rise developments can suffer a range of other problems, and their larger spatial scale can result in a less appealing urban environment. Therefore, further discussion is required on how to strike an appropriate balance between building density, volume ratio, and human comfort. As China faces a situation of growing population and declining land availability, “medium density” urban forms may provide solutions to various urban challenges, namely medium development intensity, more open space, and mid-rise building heights, which can not only enrich the spatial form but also improve the livability of cities, relieve pressure on land, and alleviate heat island effects. Secondly, the color of the building roof affects the efficiency of the absorption of solar energy, increasing the height increases the reflective surface and frequency of reflection of sunlight from the building, and using materials with high solar reflectivity to increase the albedo can greatly improve the effectiveness of mitigating the thermal environment.
• The center city’s road network is crowded in terms of traffic, and anthropogenic heat sources have a significant impact, which has a significant impact on the thermal environment nearby. The expressway’s breadth, in particular, significantly affects the surface temperature environment. In particular, the width of the expressway has a greater impact on surface thermal environment. In the context of promoting pedestrian-friendly development, a mobility planning model of “dense road network, narrow roads, and small neighborhoods” can promote the use of walking and public transportation, compared with the traditional neighborhood scale. Although the road itself increasingly contributes to higher temperatures, low-carbon mobility options such as walking, cycling, and public transportation have the potential to account for a greater proportion of residents’ trips, thereby reducing the energy consumption of transportation. In addition, relatively narrow roads, which are easier to cover, are conducive to improving the quality of the surface thermal environment.

4.4. Limitations of the Study and Further Perspectives

The present study has its shortcomings. For example, the deeper driving mechanisms of the spatial morphology of the plot on influencing the thermal environment of the surface need to be further investigated. In terms of the three-dimensional morphological arrangement of buildings within the plot, many scholars have used software simulations to obtain an optimal solution for reducing building energy consumption and surface temperature. Further research is needed to discuss the mechanisms of interaction between elements such as building density, building height, and green space ratio concerning floor area ratio, as well as the improvement of the urban thermal environment concerning the actual use and psychological perception of people. Secondly, the study has only been conducted for the main urban area of Nanjing due to the difficulty of obtaining long time series data. Future studies will build on this foundation to conduct comparative studies for more cities to explore the more prominent effects of plot spatial form on the surface thermal.

5. Conclusions

This study adopts a temporal and geographically weighted regression model (GTWR) and a multi-scale geographically weighted regression model (MGWR) to explore the driving mechanisms of urban land use spatial patterns on the urban thermal environment from both temporal and spatial perspectives. The study takes the central city of Nanjing as the study area. The five-time points of 2001, 2006, 2011, 2016, and 2021 as the study years to obtain the surface temperature of the study area and analyze the characteristics of its thermal environment. Using random points as the essential elements to calculate the surface temperature and related indicators of building morphology, ecological infrastructure, and transportation facilities in the area, to derive the laws of their effects on surface temperature, the results show that:

• Between 2001 and 2021, the impervious surface area of Nanjing’s central city increased by 84.5%, and the area of solid and powerful heat island effect increased by 47.5%, with the center of gravity of the spatial distribution of heat islands changing in the same direction as the center of gravity of the impervious surface of the city. The spatial center of gravity of the heat island distribution and the center of gravity of the urban impervious surface change in the same direction.
• In terms of driving mechanisms, the spatial pattern of urban sites has an important impact on the surface thermal environment. Specifically, NDVI, or green space, has a negative effect on the surface thermal environment, where green space with vegetation cover greater than 0.6 significantly affects the reduction of surface temperature in surrounding areas. BD, or rising building density, has a positive effect on the surface thermal environment, to a greater extent, in the periphery of the main city than in the central city. When building density is greater than 40%, it significantly affects the increase in surface temperature in surrounding areas. FAR is the opposite. In areas with a development base, an increase in the volumetric ratio will reduce surface temperatures, but in the construction of new areas, an increase in the volumetric ratio will lead to an increase in surface temperatures. The underlying driving mechanism should be related to an increase in impermeable surfaces and a decrease in green space. When the plot ratio is between 0 and 2, the surface temperature shows a slight increase; for 2–4, the change is stable; for more than four, the surface temperature decreases slightly; the surface temperature is also directly proportional to Dis_W. The surface temperature is inversely proportional to the distance from Dis_R. The surface temperature gradually decreases with increasing distance between 0 and 500 m, and the maximum influence distance is 1000 m.