Investigation into the Mechanism of the Impact of Sunlight Exposure Area of Urban Artificial Structures and Human Activities on Land Surface Temperature Based on Point of Interest Data

Abstract

With rapid urbanization, the urban heat island (UHI) effect has intensified, posing challenges to human health and ecosystems. This study explores the impact of sunlight exposure areas of artificial structures and human activities on land surface temperature (LST) in Hefei and Xuzhou, using Landsat 9 data, Google imagery, nighttime light data, and Point of Interest (POI) data. Building shadow distributions and urban road surface areas were derived, and geospatial analysis methods were applied to assess their impact on LST. The results indicate that the sunlight exposure areas of roofs and roads are the primary factors affecting LST, with a more pronounced effect in Xuzhou, while anthropogenic heat plays a more prominent role in Hefei. The influence of sunlight exposure on building facades is relatively weak, and population density shows a limited impact on LST. The geographical detector model reveals that interactions between roof and road sunlight exposure and anthropogenic heat are key drivers of LST increases. Based on these findings, urban planning should focus on optimizing building layouts and heights, enhancing greening on roofs and roads, and reducing the sunlight exposure areas of artificial structures. Additionally, strategically utilizing building shadows and minimizing anthropogenic heat emissions can help lower local temperatures and improve the urban thermal environment.

1. Introduction

With the acceleration of global urbanization, issues related to the urban thermal environment have become increasingly prominent, particularly the urban heat island (UHI) effect, which significantly impacts human activities and daily life [1]. The UHI effect not only leads to increased energy consumption during the summer but also exacerbates health risks for urban residents, particularly heat-related conditions such as cardiovascular diseases and heatstroke [2]. Additionally, rising temperatures threaten the stability of urban ecosystems, affecting biodiversity and the health of vegetation within cities. Therefore, investigating the causes of urban thermal environment intensification is of great practical importance. As a key indicator of the urban thermal environment, land surface temperature (LST) is influenced by a variety of factors, including impervious surfaces, urban green spaces, water bodies, and anthropogenic heat emissions [3]. Exploring the relationship between these factors and LST can provide a scientific basis for mitigating the UHI effect and optimizing urban design.

Existing studies have shown that building morphology, density, and spatial distribution directly influence the diurnal and nocturnal variation in LST. The relationship between building density and LST varies considerably across different urban functional zones [4]. For instance, research in Xi’an found that areas with high building density, such as commercial and industrial zones, and high impermeability of materials, tend to have higher LST compared to green spaces or low-density residential areas. During the day, building roofs and facades become the primary absorbers of solar radiation, affecting the distribution of heat within the city [5]. Similarly, in the rapidly urbanizing city of Doha, the expansion of buildings and impervious surfaces has substantially increased LST. Building height and density restrict heat dissipation, particularly at night, when materials like concrete and asphalt store large amounts of heat, further exacerbating the deterioration of the urban thermal environment [6]. Additionally, a study in Hangzhou found that both the 2D and 3D characteristics of buildings notably affect the diurnal and nocturnal variations in LST. Two-dimensional features such as building coverage ratio and impervious surface area, and 3D features such as building height and sky view factor (SVF), have differential impacts across various zones, with LST rising especially at night due to heat retention [7].

The synergistic effect of buildings and vegetation plays a crucial role in mitigating urban thermal environments. Studies have shown that the combination of buildings and green spaces can effectively regulate local temperatures. In highly urbanized areas, an increase in the Largest Patch Index (LPI) of buildings exacerbates the heat island effect, while well-planned green spaces can reduce local temperatures during the day and decrease the impact of heat storage on LST at night [8]. Furthermore, the interaction between building height and vegetation has a notable influence on both daytime and nighttime LST, especially in densely built-up core areas, where this effect is more pronounced [9].

At the spatial scale, building height and density are closely associated with the spatial heterogeneity of LST. Studies have shown that the “urban canyon effect” in densely built areas restricts air circulation, leading to increased heat accumulation and exacerbating the urban heat island effect at night [10]. Meanwhile, the 2D and 3D characteristics of buildings, such as the spatial distribution of sunlight exposure on roofs and facades, contribute to variations in LST distribution across different functional zones [10,11,12]. The application of the geographical detector model (GDM) has revealed a complex relationship between building height, roof area, and green space coverage, demonstrating that the influence of these factors on LST varies across different urban areas and time periods [11].

Although existing studies have extensively explored the 3D characteristics of urban buildings, focusing primarily on building height and overall facade area, several underexplored issues remain. First, while building facade area is considered a major factor influencing land surface temperature, in reality, only part of the facade receives direct solar radiation at any given time. This partial exposure characteristic may contribute to the urban thermal environment in a more complex way than the assumed total facade area. However, this aspect is often overlooked in many studies, which may result in less precise assessments of the relationship between building facades and LST. Second, the shadowing effect of buildings on surrounding structures and roads has received relatively little attention in existing studies. Building shadows not only reduce temperatures in shaded areas but also alter the mechanisms of heat absorption and dissipation in these zones. Research has shown that temperatures in shaded areas, such as building shadows or beneath tree canopies, are typically 5 °C to 10 °C cooler than in sunlit areas at noon, with tree canopy cover reducing temperatures by around 2 °C [13]. This is because shaded areas receive less direct solar radiation, thereby reducing heat accumulation. As a result, rooftops or facades in shaded areas cannot be considered positive contributors to land surface temperature. However, this important mechanism is often inadequately considered in existing models and analyses, which may lead to biased evaluations of the urban thermal environment, particularly in high-density built areas.

Finally, although most studies have concentrated on the role of buildings in the urban heat island effect, relatively little attention has been paid to roads, which are also a critical factor. Roads occupy a large proportion of the urban surface, and their materials, such as asphalt and concrete, possess high thermal capacities, absorbing substantial heat during the day and releasing it gradually at night, thereby affecting the urban thermal environment. However, due to their linear spatial distribution, many studies tend to classify roads as part of the impervious surface [2], or only consider their length as a linear feature [3], rarely investigating the independent role of road surface area in the urban thermal environment. This approach may underestimate the actual contribution of roads, especially in high-traffic areas where their impact on land surface temperature is more pronounced.

In recent years, Point of Interest (POI) data have been widely used to identify urban functional zones and analyze urban spatial structures. This data source offers several advantages, including frequent updates, wide spatial coverage, high accuracy, and a large, comprehensive data volume [14,15,16]. Additionally, by integrating multi-source data, such as remote sensing images and mobile signal data, POI data effectively support the identification of spatial distribution characteristics of urban functional areas [17]. This enables the more precise extraction of building shadows as well as sunlight exposure areas for both buildings and roads, greatly enhancing the accuracy of urban thermal environment studies. Accordingly, this study uses POI data and spatial analysis techniques to further examine the impact of sunlight exposure areas for buildings and roads on LST, addressing existing research gaps related to the shadowing effect of buildings and the independent role of roads. This approach provides theoretical support for more precise strategies in regulating the urban thermal environment.

This study aims to leverage multi-source big data to extract and analyze high-density building and road areas in Hefei and Xuzhou. Using geospatial analysis techniques, it evaluates how sunlight exposure areas of building roofs, facades, and roads affect the urban thermal environment, as well as the roles of population density and anthropogenic heat flux (AHF) in influencing LST. Ultimately, this research seeks to reveal the mechanisms by which sunlight exposure areas of urban artificial structures and human activities impact LST. We hope that this study will provide new perspectives and methods for urban planning, helping to alleviate the urban heat island effect and optimize the urban thermal environment.

2. Materials and Methods

2.1. Study Area and Data Sources

2.1.1. Study Area

This study selects Hefei City and Xuzhou City as the research areas to explore the impact of sunlight exposure areas and human activities on LST. These two cities are located in Anhui and Jiangsu Province, respectively, both in the economically developed central–eastern region of China, with notable locational advantages. Hefei, the capital of Anhui Province, is situated between the Yangtze and Huai Rivers, near Chaohu Lake, and connected to the Yangtze River via the Nanfei River, making it an important transportation hub in the region. Xuzhou, located in northern Jiangsu Province, lies at the intersection of the eastern coastal economic belt and the North China economic region, offering a strategic geographic position connecting East China and North China and serving as a major transportation hub in northern Jiangsu. Both Hefei and Xuzhou experience distinct seasons, with average annual temperatures of 16 °C and 14 °C, respectively. Hefei has a subtropical monsoon climate, while Xuzhou has a warm temperate monsoon climate. With rapid economic development and accelerated urbanization in recent years, the scale of urban construction in both cities has expanded considerably, intensifying the urban heat island effect.

These two cities were chosen not only for their unique geographic locations and climatic characteristics but also for the wealth of geospatial big data accumulated during the urbanization process, providing a rich research foundation and ideal experimental conditions for studying the effects of sunlight exposure areas and human activities on LST. In addition, Hefei and Xuzhou differ markedly in terms of city size, population, and industrial structure. As a provincial capital, Hefei has a more active economy with a focus on high-tech industries and a higher population density, whereas Xuzhou is primarily centered around traditional manufacturing, with relatively higher building density and road coverage. These differences provide valuable research conditions for analyzing the diverse impacts of artificial structures and human activities on LST under different urban development models, helping to reveal variations in urban thermal environment effects across different types of cities.

Since the extracted building POI data only cover the core urban areas (with missing data for the suburbs), the specific scope of the study area was further delineated based on the administrative boundaries of the main urban districts and major roads, as shown in Figure 1. The two study areas will hereafter be referred to as Hefei and Xuzhou, respectively.

2.1.2. Data Sources

This study collected various types of urban geospatial POI data, as detailed in Figure 2 and

Table 1. POI and remote sensing data information for the two study areas.

Data Type	Data Name	Acquisition Time (UTC + 8)	Data Source
POI Data	Building Rooftop Contour and Building Height Data	September 2021 (Hefei) July 2023 (Xuzhou)	Baidu Map Open Platform [19]
	Road Data	September 2021 (Hefei) July 2023 (Xuzhou)
	Population Spatial Heatmap	16 June 2022, 11:00 AM (Hefei) 5 July 2023, 11:00 AM (Xuzhou)
Satellite Data	Landsat 9 Imagery	Scene ID: LC09_L1TP_121038_20220616_20230411_02_T1 16 June 2022, 10:43:10 AM (Hefei) Scene ID: LC09_L1TP_121036_20230705_20230705_02_T1 5 July 2023, 10:42:16 AM (Xuzhou)	United States Geological Survey (USGS) [20]
	VIIRS/NPP NTL Data	Sense ID: SVDNB_npp_20220601-20220627_75N060E_vcmcfg_v10_c202207141300 June 2022 (Hefei) Sense ID: SVDNB_npp_20230701-20230726_75N060E_vcmcfg_v10_c202308111200 July 2023 (Xuzhou)	Earth Observation Group (EOG) [21]
	Google Earth Imagery	19 September 2022 (Hefei) 20 November 2023 (Xuzhou)	Google Earth [22]

. First, Baidu Maps’ urban building vector data (including building footprint outlines and heights) were used to calculate the sunlight exposure areas and shadow distribution of buildings. Baidu Maps’ urban road linear vector data were combined with high-resolution imagery from Google Earth to extract polygonal vector data for roads at various levels within the cities. Baidu’s population heatmap data were utilized to estimate the spatial distribution density of the population in real time. Regarding remote sensing data, Landsat 9 data were used to retrieve LST, while the Visible Infrared Imaging Radiometer Suite/National Polar-orbiting Partnership Nighttime Lights (VIIRS/NPP NTL) data were used to estimate AHF in the cities. Additionally, ground-based meteorological observation data were collected from the China Meteorological Data Service Center [18] to assist in the LST retrieval (

Table 2. Ground observation data at the satellite overpass time and geographic data of the study area.

Study Area	Date	Weather	Dew Point Temperature (°C)	Latitude	Average Elevation (m)
Hefei	16 June 2022	Sunny and cloudless	16.00	31.85°	28.4
Xuzhou	5 July 2023	Sunny and cloudless	20.90	34.24°	39.8

).

Since the selected datasets have different spatial resolutions, the two study areas, Hefei and Xuzhou, were divided into 300 × 300 m grid cells to ensure the accuracy and comparability of the analysis results, as shown in Figure 1b,c. This grid division allows for the integration of original data with varying spatial resolutions into a consistent spatial scale. Once the calculations for all indicators were completed, the data were extracted and aggregated based on this unified grid system, ensuring that all datasets were analyzed at the same spatial resolution. This approach not only eliminates errors that might arise from differences in spatial resolution but also improves the comparability between different datasets.

2.2. LST Retrieval

Landsat 9 is the latest in the Landsat series of Earth-observing satellites, equipped with two advanced sensors: the Operational Land Imager 2 (OLI-2) and the Thermal Infrared Sensor 2 (TIRS-2). The OLI-2 sensor captures data across nine visible, near-infrared, and shortwave infrared bands with a spatial resolution of 30 m, along with a panchromatic band at a 15 m resolution. TIRS-2, on the other hand, collects thermal data in two thermal infrared bands at a spatial resolution of 100 m, which enables detailed thermal measurements of the Earth’s surface. This high-resolution multispectral configuration allows for Landsat 9 to provide comprehensive and accurate land surface temperature data, making it particularly valuable for environmental monitoring and climate studies. Although Landsat 9 level 2 data include a surface temperature (ST) product, the use of data from the Advanced Spaceborne Thermal Emission and Reflection Radiometer Global Emissivity Database (ASTER GED), which contains gaps, leads to missing surface temperature values in certain regions. Directly using the ST product would introduce substantial errors in this study. Therefore, we employed a split-window algorithm based on the Landsat 9 TIR-2 band to retrieve LST [23].

$$T_s={\displaystyle \frac{c_2{\lambda }_{10}^{-1}}{\ln \left(\frac{c_1{\lambda }_{10}^{-5}}{B_{10}\left(T_s\right)}+1\right)}}$$

(1)

$$B_{10}\left(T_s\right)=A_0{L}_{10}+A_1{L}_{11}+A_2$$

(2)

where T_s is the land surface temperature, and c₁ = 1.19104 × 10⁸ W·μm⁴·m⁻²·sr⁻¹ and c₂ = 1.43877 × 10⁴ μm·K are the first and second radiation constants of Planck’s law, respectively. λ₁₀ = 10.8372 μm is the effective wavelength of the TIR band 10. L_10/11 are the radiance values for TIR band 10 or 11, respectively. The intermediate variables A₀, A₁, and A₂ can be calculated using the following equations.

$$A_0={\displaystyle \frac{D_{11}}{C_{10}{D}_{11}-C_{11}{D}_{10}}}$$

(3)

$$A_1={\displaystyle \frac{-D_{10}}{k\left(C_{10}{D}_{11}-C_{11}{D}_{10}\right)}}$$

(4)

$$A_2={\displaystyle \frac{b{D}_{10}\left(C_{11}+D_{11}\right)}{k\left(C_{10}{D}_{11}-C_{11}{D}_{10}\right)}}$$

(5)

$$C_{10/11}={\epsilon }_{10/11}{\tau }_{10/11}$$

(6)

$$D_{10/11}=\left(1-{\tau }_{10/11}\right)\left[\left(1-{\epsilon }_{10/11}\right){\tau }_{10/11}{\phi }_{10/11}+1\right]$$

(7)

where b and k are radiometric conversion parameters, which can be calculated using Equations (8) and (9). τ_10/11 and φ_10/11 are atmospheric parameters for TIR band 10 or 11, respectively, and can be calculated using Equations (10) and (11).

$$k={\displaystyle \frac{c_1^2{\lambda }_{10}^{-4}{\lambda }_{11}^{-6}{\left(c_1{\lambda }_{10}^{-5}{L}_{10}^{-1}+1\right)}^{({\lambda }_{10}{\lambda }_{11}^{-1}-1)}}{{[{\left(c_1{\lambda }_{10}^{-5}{L}_{10}^{-1}+1\right)}^{\left({\lambda }_{10}{\lambda }_{11}^{-1}\right)}-1]}^2{L}_{10}^2}}$$

(8)

$$b={\displaystyle \frac{c_1{\lambda }_{11}^{-5}}{{\left(c_1{\lambda }_{10}^{-5}{L}_{10}^{-1}+1\right)}^{\left({\lambda }_{10}{\lambda }_{11}^{-1}\right)}-1}}-k{L}_{10}$$

(9)

$${\tau }_{10/11}=a_{0}w+a_1$$

(10)

$${\phi }_{10/11}=a_2\ln \left(w\right)+a_3$$

(11)

where λ₁₁ = 12.0253 μm is the effective wavelength of TIR band 11; w is the atmospheric water vapor content, which can be calculated based on meteorological observation data and empirical formulas [24], as shown in Equations (12) and (13). a_x (x = 0 to 3) are fitting coefficients, and their values for TIR band 10 or 11 are listed in

Table 3. The fitting coefficients between atmospheric parameters and water content for Landsat 9 TIR-2 band 10 and 11.

Coefficients	Landsat 9 TIR Band 10	Landsat 9 TIR Band 11
a0	−0.0523	−0.0531
a1	0.9495	0.8315
a2	1.4073	0.6079
a3	1.1641	0.4856

.

$$w=\left[0.2-{\displaystyle \frac{0.066}{{\left(\psi -33\right)}^2+4.41}}\right]e_v+0.04\mathit{exp}\left(0.6H_E\right)-{\displaystyle \frac{0.05}{{\left(\psi -25\right)}^2+0.25}} (\psi <33°)$$

(12)

$$w=\left[0.17+{\displaystyle \frac{0.066}{{\left(\psi -33\right)}^2+4.41}}\right]e_v+0.03\mathit{exp}\left(-1.39{H_E}^2+2.74H_E+0.15\right) (\psi \ge 33°)$$

(13)

$$e_v=0.6112exp\left({\displaystyle \frac{17.67T_d}{T_d+243.5}}\right)$$

(14)

where H_E is the elevation of the study area, ψ is the latitude of the center of the study area, and e_v is the atmospheric water vapor pressure, which can be calculated from the dew point temperature T_d in the meteorological observation data at the satellite overpass time.

The land surface emissivity ε_10/11 for TIR band 10 or 11 can be calculated using the fractional vegetation coverage (FVC), the emissivity of pure vegetated surfaces ε_10/11,v, and the emissivity of pure bare soil surfaces ε_10/11,s, as shown in the following equations.

$${\epsilon }_{10/11}=FVC\cdot {\epsilon }_{10/11,v}+\left(1-FVC\right)\cdot {\epsilon }_{10/11,s}$$

(15)

$$FVC={\left({\displaystyle \frac{NDVI-NDV{I}_s}{NDV{I}_v-NDV{I}_s}}\right)}^2$$

(16)

where NDVI is the normalized difference vegetation index calculated from Landsat 9 OLI-2 band 4 and band 5; NDVI_v = 0.86 and NDVI_s = 0.2 are the NDVI values for pure vegetated and bare soil pixels, respectively [23]. The values of ε_10/11,v and ε_10/11,s can be calculated based on the land surface emissivity of ASTER GEDV3 dataset’s band 13 and 14 [23,25], with the values shown in

Table 4. The emissivity of pure vegetation and bare soil surfaces for Landsat 9 TIR-2 band 10 and 11.

	Landsat 9 TIR Band 10	Landsat 9 TIR Band 11
Pure Vegetation Emissivity	0.9816	0.9843
Pure Soil Emissivity	0.9741	0.9787

.

2.3. Building Shadow Extraction

The position and area of building shadows on the surface are influenced not only by the geometric shape of the buildings themselves but also by the relative position of the sun in the sky, which can be represented by the solar elevation angle (h) and azimuth angle (A_z). As shown in Figure 3, h refers to the angle between the sun’s rays and their projection on the horizontal plane. A_z refers to the angle between the projection of the sun’s rays on the horizontal plane and the local meridian. A_z is measured with due south as 0°, with angles measured negatively from south to east and north, and positively from south to west and north. When the sun rises directly in the east, the azimuth angle is −90°, and when it sets directly in the west, the azimuth angle is 90°. Since the solar azimuth angle simulated in ArcGIS 10.8 takes due north as 0° (A_z,N), and the satellite overpass time is during the morning local time, the complementary angle of the solar azimuth (180° − A_z) was calculated as part of the process. The equations used are shown below [26].

$$A_{z,N}={180}^{°}-A_z$$

(17)

$$A_z=arccos\left[\right(\mathit{sin}\psi \sin h-\sin \delta )/(\cos \psi \cos h\left)\right]$$

(18)

$$\delta =0.006918-0.399912\cos \beta +0.070257\sin \beta -0.006758\cos 2\beta +0.000907\sin 2\beta -0.002697\cos 3\beta +0.00148\sin 3\beta$$

(19)

$$h=\mathrm{a}\mathrm{r}\mathrm{c}\mathrm{s}\mathrm{i}\mathrm{n}(\sin \psi \sin \delta +\cos \psi \cos \delta \cos \omega )$$

(20)

$$\omega ={\displaystyle \frac{\pi }{12}}(ST-12)$$

(21)

$$ST=t_b+{\displaystyle \frac{(l_b-l_h)\times 4+229.183\eta }{60}}$$

(22)

$$\eta =0.000043+0.002061\mathit{cos}\beta -0.032040\mathit{sin}\beta -0.014974\mathit{cos}2\beta -0.040685\mathit{sin}2\beta$$

(23)

$$\beta =2\pi {\displaystyle \frac{D_n-1}{365}}$$

(24)

where h is the solar elevation angle; δ is the solar declination; β is the day angle; ω is the hour angle; ST is the solar time; η is the time difference; l_b and l_h are the longitudes of the local standard time zone (Beijing) and the local longitudes (Hefei and Xuzhou), respectively; t_b is the Beijing time (UTC + 8) of the satellite overpass; and D_n is the day number of the satellite overpass. The required date, time, and geographic parameters for the calculation are shown in

Table 5. Geographic data of the study area and satellite overpass time data (with all units, including time, converted to rad).

Study Area	Beijing Longitude (lb)	Local Longitude (lh)	Latitude (ψ)	Beijing Time (tb)	Day Number (Dn)
Hefei	2.0944	2.0469	0.5559	10.7197	167
Xuzhou		2.0460	0.5976	10.7044	186

; all angles and time units were converted into radians (rad). After calculating the solar azimuth angle and solar elevation angle, combined with the building footprint data of Hefei City with height attributes, ArcGIS can be used to simulate the spatial distribution of building shadows at the time of the satellite overpass.

2.4. Sunlight Exposure Area Extraction

Due to the height differences and complex spatial layouts of buildings in urban environments, building shadows often cover parts of surrounding buildings or roads, preventing these areas from receiving direct solar radiation. This shading effect causes the temperatures in shadowed areas to be notably lower than those in areas directly exposed to sunlight. Therefore, in analyzing the urban heat island effect, the portions of buildings and roads covered by shadows are not considered positive contributing factors. To accurately assess the sunlight exposure areas of buildings and roads, this study employed a spatial analysis method based on ArcGIS, with the methodology flow shown in Figure 4.

First, using the Aspect tool and Raster Calculator in ArcGIS, the solar azimuth calculated in Section 2.3 is input to extract the shaded and sunlit parts of the building projections. Then, by multiplying the shaded and sunlit parts with the building heights, the heights of the shaded and sunlit facades are obtained. Next, the Hillshade tool in ArcGIS is used to input data such as the shaded facade height, solar azimuth, and solar elevation angle, resulting in the spatial distribution of building shadows within the study area. Finally, by overlaying the building shadow data with the spatial data of the sunlit facades, roofs, and roads, the sunlight exposure areas of building facades, roofs, and roads are determined, as shown in Figure 5.

2.5. AHF Estimation

In the study of AHF, accurately assessing the urban thermal environment depends on the choice of model. Traditional estimation methods typically rely on direct energy consumption data, which are difficult to obtain and constrained by time and space limitations. NTL data, by capturing the distribution of artificial light sources on the surface, can effectively reflect the intensity of human activities, such as population concentration, industrial production, and traffic flows. These activities are commonly associated with substantial heat emissions, making NTL a valuable tool for evaluating the spatial distribution of AHF. The high resolution of NTL enables the precise identification of heat emission characteristics across different urban areas, providing essential support for the study and management of the urban thermal environment [27].

The use of single NTL data still has limitations in capturing the complexity of the urban thermal environment. However, the principle of combining NTL data with the NDVI to estimate AHF has been widely discussed and applied in urban thermal environment research [28]. This study employs a refined anthropogenic heat flux (RAHF) model, which integrates nighttime light intensity and NDVI, and introduces the human settlement index (HSI) to more comprehensively reflect the characteristics of AHF in urban areas. This approach not only provides a higher fitting accuracy (R² = 0.989) but also captures the complexity of the urban thermal environment more effectively, offering more precise data support for studying the urban heat island effect [29].

$$AHF=48.287{HSI}^2-17.716HSI+2.541$$

(25)

$$HSI={\displaystyle \frac{(1-{NDVI}_{max})+{NTL}_n}{\left(1-{NTL}_n\right)+{NDVI}_{max}+{NTL}_n\times {NDVI}_{max}}}$$

(26)

2.6. Spatial Autocorrelation Analysis

In this study, global and local spatial correlation analyses [30] were conducted to demonstrate the spatial relationship between LST and variables such as building roof sunlight exposure area, sunlight exposure area of sun-facing building facades, road sunlight exposure area, population density, and AHF. This approach helps identify the interaction effects between these variables and LST in urban spaces. Global spatial autocorrelation analysis is primarily used to analyze the overall spatial distribution pattern of variables in the study area, determining whether there is clustering or dispersion, with Moran’s I commonly used as the indicator. A Moran’s I value close to 1 indicates a clustering effect, a value close to −1 suggests a dispersal effect, and a value close to 0 implies no significant spatial autocorrelation.

Local spatial autocorrelation analysis is a statistical method used to detect local correlations between two variables in geographic space, revealing their spatial co-variation. Through a LISA (Local Indicators of Spatial Association) map, the spatial distribution of these regions can be visually displayed, showing spatial patterns such as high-value clusters (High–High), low-value clusters (Low–Low), low values surrounded by high values (Low–High), and high values surrounded by low values (High–Low).

2.7. Geographical Detector

The geographical detector is a statistical tool used to quantify the impact of different factors on a phenomenon, such as LST, in geographic space [31]. Its basic principle is that if an independent variable X significantly influences a dependent variable Y, then these two variables should exhibit similar spatial distribution characteristics. By comparing the global variance with the local variance of each factor, the geographical detector can evaluate the explanatory power of each factor on spatial differentiation. The core of the factor detector is to calculate the q value, which assesses the explanatory power of a single factor on the spatial heterogeneity of LST. Additionally, the interaction detector is used to evaluate whether the combined effect of two factors on LST is enhanced or weakened. By analyzing the interaction q value, it is possible to determine whether the two factors mutually enhance each other (increase in q value), suppress each other (decrease in q value), or act independently. The interaction detector provides a richer explanatory capability for studying complex geographical phenomena.

3. Results

3.1. LST Retrieval Results

The LST retrieval results for Hefei and Xuzhou are shown in Figure 6a and Figure 6b, respectively. The results indicate that the average LST for Hefei and Xuzhou is 312.3 K and 309.5 K, respectively. High-temperature areas in both cities are primarily concentrated in the urban core and densely built-up regions, with some hot zones also observed in the outskirts, especially in factory and warehouse districts. In Hefei, the high-temperature range is mostly between 313 K and 322 K, while in Xuzhou, it is concentrated between 312 K and 327 K.

In addition to the urban core, high-temperature zones are also concentrated in the peripheral industrial and warehouse areas. These elevated temperatures are often closely linked to the materials and structures of factory buildings. Many of the factory and warehouse buildings are constructed with metal materials (such as steel panels or metal roofs) or concrete which have high heat capacity and thermal conductivity, enabling them to rapidly absorb and retain solar radiation, resulting in elevated surface temperatures. Moreover, these areas often lack greenery, leading to poor heat dissipation, which further exacerbates the rise in temperature. Consequently, these zones exhibit notably high daytime temperatures, contributing significantly to the UHI effect.

Low-temperature areas are distributed not only in the vegetated areas on the outskirts of the cities or near water bodies but also in landscaped parks or greenbelts within urban areas, where temperatures are typically below 304 K, showing a clear cooling effect. By overlaying the LST with Baidu population heatmaps, building and road POI data, and nighttime light data, it can be observed that regions with high LST often align with areas of high population density, dense urban development, and strong nighttime light intensity, suggesting that these factors may influence the spatial distribution of surface temperature.

3.2. Correlation Between LST and All Factors

The average LST, total roof sunlight exposure area, total facade sunlight exposure area, total road sunlight exposure area, population heatmap values, and average AHF values were extracted for each grid in Hefei and Xuzhou using a fishnet method. The abbreviations and descriptions of all influencing factors are shown in

Table 6. Abbreviations and definitions of all factors.

Factor Abbreviation	Description
Roof	Sunlight exposure area of building roofs
Facade	Sunlight exposure area of building facade
Road	Sunlight exposure area of roads
Population	Population density
AHF	Anthropogenic heat flux

. After normalizing all the indicators (as shown in Figure 7), a matrix scatter plot was used to compare the correlation between the various factors, as shown in Figure 8.

In both cities, roof sunlight exposure showed a strong correlation with LST, with coefficients of 0.560 in Hefei and 0.589 in Xuzhou. This indicates that roofs, as primary absorbers of solar energy, play a substantial role in elevating urban temperatures, especially in high-density areas where greater roof exposure leads to higher local temperatures. Roads also displayed a notable correlation with LST, with coefficients of 0.397 in Hefei and 0.519 in Xuzhou, suggesting that materials like asphalt and concrete, which retain heat during the day and release it slowly at night, impact temperatures in surrounding areas. The slightly stronger correlation in Xuzhou may reflect differences in road layout and materials.

For AHF, the correlation with LST was 0.519 in Hefei and 0.424 in Xuzhou, underscoring the influence of human activities such as traffic, industrial processes, and air conditioning systems. In Hefei, AHF ranked second in correlation with LST, likely due to higher economic activity, whereas in Xuzhou, AHF ranked third, with roads exerting a more substantial influence.

Facade sunlight exposure showed a moderate correlation with LST, with coefficients of 0.301 in Hefei and 0.398 in Xuzhou. Although facades have more dispersed sunlight exposure than roofs, they still contribute to local temperatures, especially in densely built areas with taller buildings. Population density had the lowest correlation with LST in both cities (0.238 in Hefei and 0.311 in Xuzhou), likely influencing LST indirectly through building density, road networks, and traffic flow rather than directly affecting temperatures.

Further analysis revealed a strong correlation between roof and facade (0.586 in Hefei and 0.615 in Xuzhou), indicating that sunlight exposure on roofs and facades often increases simultaneously in high-density areas. The correlation between road and population was moderate in both cities (0.377 in Xuzhou and 0.383 in Hefei), suggesting that denser populations are typically associated with greater road coverage. Additionally, the correlation between road and AHF was 0.467 in Hefei and 0.394 in Xuzhou, indicating the connection between road traffic and heat emissions. Finally, the correlation between facade and AHF (e.g., 0.461 in Xuzhou) implies that high AHF levels often coincide with areas featuring large facades, such as commercial zones with high human activity, contributing to localized heat accumulation.

3.3. Spatial Correlation Between LST and All Factors

As shown in Figure 9, the bivariate spatial correlation analysis revealed distinct clustering patterns between LST and each influencing factor in both Hefei and Xuzhou. In Hefei, AHF exhibited the strongest bivariate spatial correlation with LST, with a Moran’s I value of 0.492, indicating that areas with high AHF are closely associated with elevated LST values. Roof sunlight exposure also demonstrated a high degree of spatial association with LST, with a Moran’s I value of 0.434, suggesting that roof areas with high sunlight exposure are primarily clustered in regions with increased LST, particularly in the city center.

In Xuzhou, roof displayed the highest bivariate spatial correlation with LST among all factors, with a Moran’s I value of 0.430, reflecting a strong spatial linkage between roof sunlight exposure and elevated LST in densely developed urban areas. AHF in Xuzhou showed a moderate spatial correlation with LST, with a Moran’s I value of 0.394, indicating a more dispersed spatial relationship compared to the clustering observed between LST and roof or road sunlight exposure.

For both cities, road demonstrated a pronounced bivariate spatial correlation with LST. In Hefei, Moran’s I value for the correlation between road and LST was 0.359, while in Xuzhou, it was 0.376, suggesting that major traffic arteries and commercial zones with high road sunlight exposure are spatially clustered with higher LST areas. The spatial correlation between facade sunlight exposure and LST was moderate, with Moran’s I values of 0.335 in Xuzhou and 0.283 in Hefei, indicating a relatively dispersed spatial association with LST. Population density exhibited the weakest spatial correlation with LST among the factors, with Moran’s I values of 0.214 in Hefei and 0.251 in Xuzhou, reflecting minimal spatial alignment between population density and LST patterns.

3.4. Impact Weight of Each Factor on LST

The geographical detector model was employed to assess the independent explanatory power of various factors on LST, as well as their interactions. When combined with the spatial correlation analysis from the previous section, this approach provides a deeper understanding of how different factors interact to influence surface temperature across urban areas.

As shown in

Table 7. Factor detector results between all impact factors and LST.

	Hefei		Xuzhou
	q Statistic	p Value	q Statistic	p Value
Roof	0.314	0.000	0.343	0.000
Facade	0.166	0.000	0.174	0.000
Road	0.233	0.000	0.302	0.000
Population	0.063	0.000	0.126	0.000
AHF	0.392	0.000	0.218	0.000

, the factor detector results indicate that in Hefei, AHF has the strongest independent explanatory power on LST, with a coefficient of 0.392. Roof sunlight exposure follows, with an explanatory power coefficient of 0.314, highlighting the role of roofs in absorbing solar radiation and contributing to the UHI effect. Road sunlight exposure ranks third, with a coefficient of 0.233, indicating its contribution to LST in the urban thermal environment. Facade sunlight exposure has a moderate explanatory power, with a coefficient of 0.166, while population density has the weakest explanatory power in Hefei, with a coefficient of only 0.063, suggesting its limited direct impact on LST.

In Xuzhou, roof sunlight exposure has the strongest independent explanatory power, with a coefficient of 0.343, highlighting its influence on LST distribution. Road ranks second, with a coefficient of 0.302, further confirming its importance in Xuzhou’s thermal environment. In contrast, AHF shows a relatively weaker explanatory power in Xuzhou, with a coefficient of 0.218, indicating that while anthropogenic heat contributes to LST, its effect is less pronounced compared to that of roofs and roads. Facade sunlight exposure has a coefficient of 0.174, higher than that of population density, which is 0.126. Similarly to Hefei, this indicates that facades still play a notable role in densely built areas, while population density remains the factor with the weakest explanatory power in Xuzhou.

As shown in Figure 10, the interaction detector results indicate that in Hefei, the interaction between roof and AHF had the greatest impact on LST, with an interaction coefficient of 0.537. This finding suggests that in areas with dense building clusters and concentrated human activities, these factors jointly intensify LST. Following that, the interaction between road and AHF ranked second, with a coefficient of 0.447, further supporting the importance of roads and AHF in Hefei’s thermal environment.

In Xuzhou, the interaction detector results show that the interaction between roof and road had the most pronounced impact on LST, with an interaction coefficient of 0.515. This finding indicates that the combined effect of building roofs and roads is the primary driver of LST increase in Xuzhou. In terms of AHF, it exhibits weaker independent explanatory power and lower interaction coefficients with other factors, which reflects Xuzhou’s relatively smaller economic scale, traditional industrial structure, and lower urban development intensity compared to Hefei. Xuzhou’s industries are still largely focused on traditional manufacturing, with a smaller share of high-tech industries. Consequently, economic activity levels and industrial density are lower than those in Hefei, resulting in a less pronounced impact of AHF on LST.

4. Discussion

In this study, we conducted an in-depth analysis of the correlation and spatial autocorrelation between LST and factors such as roof, road, facade, population, and AHF. We also compared our results with other studies. It is worth noting that we excluded buildings and roads covered by shadows from the positive influencing factors on LST in our study. This approach further improved the accuracy of the results, as shaded areas have a minimal or even cooling effect on LST. Therefore, excluding these areas allowed us to provide a more realistic explanation of the heat distribution pattern, which is particularly applicable to the analysis of building roofs and roads, as these areas are often shaded, affecting their heat absorption and release differently than sunlit areas.

Our findings underscore that in both cities, roof, AHF, and road are the top three factors with the highest weight in influencing LST. Although the influence of AHF is weaker in Xuzhou compared to Hefei, these factors, whether analyzed independently or through interaction detection, all exhibit a notable impact on LST, underscoring the critical roles of roofs, roads, and anthropogenic heat in shaping the urban thermal environment. Additionally, while the influence of facade is less pronounced than roof, road, and AHF, it still has a stronger association with LST than population, indicating that sunlight exposure on building facades acts as a secondary contributor to urban heat patterns. Population, on the other hand, shows the weakest explanatory power for LST in both cities, suggesting that it indirectly affects surface temperature through its correlation with other urban factors such as building density and traffic.

In our study, Moran’s I of building roof sunlight exposure area in Hefei (0.434) and its independent explanatory power (0.314) ranked second among all factors. In Xuzhou, Moran’s I (0.430) and independent explanatory power (0.343) of building roof sunlight exposure area ranked first, indicating that roof sunlight exposure played a dominant role in increasing LST. Similarly, Deng et al. [32] found that building roofs and densely developed areas exhibit high thermal conductivity (10–60 W/m·K at 25 °C), allowing them to heat up quickly, with LST considerably higher than water surfaces, roads, and permeable surfaces. Our study further supports this conclusion. Additionally, a study on Austin found that by greening 3.2% of total building roofs, the average LST was reduced by 1.96 °C [33]. This further validates the importance of building roofs in influencing LST, particularly in mitigating the urban heat island effect by optimizing building characteristics and sunlight exposure areas.

In our study, road sunlight exposure had a notable impact on LST. In Xuzhou, the independent explanatory power (q value) of road sunlight exposure was 0.302, ranking second, while in Hefei, it was 0.233, ranking third. This suggests that roads play a key role in the thermal environment of both cities, prominently affecting surface temperatures through their widespread presence. Similarly, Liu et al. [34] found that roads impact LST considerably, with highways and low-grade roads raising LST within 180 m and 150 m, respectively, and interchanges and intersections affecting LST within 300 and 150 m. Liu’s study further highlighted a linear increase in LST with road length, consistent with our findings and underscoring the impact of road expansion on LST. Unlike traditional studies on roads [34,35], which often treat roads as linear features, we directly extracted road area data, providing a more accurate measure of their contribution to LST. This approach emphasizes the importance of roads in our study, underscoring their role in the urban heat island effect.

Some studies have reported a high degree of consistency between the spatial patterns of AHF and LST [36], with industrial parks, commercial centers, and transportation hubs having the highest AHF values (AHF > 30) [37], while residential areas in older districts have relatively lower AHF values [38]. This indicates that high economic activity zones in city centers usually show a higher impact of AHF on LST. Our study also found that AHF had a greater impact on LST in Hefei, with higher economic activity and industrialization compared to Xuzhou, further confirming that AHF is an important driver of LST in economically active regions.

In our study, the spatial correlation between building facade sunlight exposure and LST was relatively weak, and its explanatory power ranked second to last among all factors. This may be because Landsat sensors primarily receive electromagnetic waves from the direction perpendicular to the ground, capturing little thermal radiation from building facades, resulting in a relatively weak direct impact on LST. Future research could explore the influence of building facades on urban thermal environments using drones equipped with thermal infrared sensors.

The positive effect of population density on LST has been demonstrated by many studies [39]. While our results are consistent with this, they also show that population density had the lowest explanatory power among all factors. Zheng et al. [40] also pointed out that, compared to population density, impervious surface and green space indicators played a more dominant role in affecting surface temperature. Therefore, population density tends to influence LST indirectly through other urban elements, such as building density, road networks, and AHF, rather than being a primary driving factor.

5. Conclusions

This study enhances the understanding of the UHI effect by precisely identifying the sunlight exposure areas of urban structures, uncovering the distinct impacts of building roof and facade exposure, road exposure, population density, and AHF on LST. The findings indicate that roof and road sunlight exposure are the primary drivers of LST in both cities. AHF had a stronger influence in Hefei, characterized by higher economic activity, while in Xuzhou, roof and road exposure played a more significant role. The impact of building facade exposure on LST was minor, with population density contributing the least. Factor interaction analysis showed that in Hefei, the combined effect of roof exposure and AHF was most influential, whereas in Xuzhou, the interaction between roof and road exposure primarily drove LST increases.

Therefore, to effectively mitigate the urban heat island effect, it is recommended to optimize building layouts and heights in urban planning, increase greenery on building roofs and roads, and reduce the impact of road sunlight exposure on LST. Additionally, making efficient use of building shadow effects and taking measures to reduce AHF emissions will help lower local LST and enhance thermal comfort in the urban environment.

Future research could incorporate advanced technologies such as drone-based thermal infrared imaging to obtain more accurate 3D building shapes and temperature characteristics. This approach would enable a deeper investigation into the influence of building structures and roads on LST across diverse urban layouts and climatic conditions, offering more tailored strategies to mitigate the urban heat island effect.