Analysis of the Spatial and Temporal Evolution Characteristics and Driving Forces of the Surface Thermal Environment in Lanzhou City

Abstract

The urban heat island effect is becoming more and more serious due to the continuous expansion of cities in China, and improving the urban thermal environment is an important prerequisite for creating a livable city. Based on the Landsat TM images of 2001 and 2011 and Landsat TIRS images of 2021, this study investigated the spatial and temporal distribution and evolution characteristics of the urban thermal environment in Lanzhou City in the last 20 years by using the radiation equation conduction method to invert the surface temperature of Lanzhou City. The results show the following: (1) The radiation range of the medium temperature zone in Lanzhou City spreads in an “extended” style from 2001 to 2021 until it covers the whole main urban area, the secondary high and low temperatures decrease, and the corresponding medium temperature zone increases over a large area. (2) The average temperatures of Lanzhou urban area and its arable land, forest, grassland, and water area within the urban area reaches 10–25 °C, and the heat island area decreases by 9.56% in 20 years, with the high-temperature zone decreasing by 42.32%. (3) The proportion of water bodies and the proportion of impermeable surfaces are dominant factors in the spatial differentiation of surface temperature, and the interaction and synergy of various influencing factors affect the spatial differentiation of surface temperature.

1. Introduction

Cities are the hubs of modern social, cultural, and economic activities. The physical process of urbanization gives cities unique morphological and thermodynamic properties, in which the original natural and human landscapes are replaced by impermeable surfaces such as buildings and roads. Cities are also centers of resource use, with high levels of energy and water consumption [1]. This produces high anthropogenic heat fluxes [2]. The city is also a center of resource use, with high energy and water consumption, resulting in anthropogenic heat fluxes, greenhouse gas, and aerosol emissions [3,4]. These characteristics and activities create a unique urban climate [5,6,7,8,9,10]. By mid-century, 68% of the world’s population is expected to live in urban areas [11]. The growth of urban populations increases the demand for services, housing, and utilities, and changes in land cover and infrastructure characteristics regulate local and regional weather and climate through biophysical and biogeochemical processes [12,13,14,15,16,17].

Most studies on local urban weather and climate disturbances have focused on urban heat islands (UHI) [18,19,20], one of the oldest and most studied topics in urban climatology, dating back to the work of Howard [21]. Since this pioneering work, scholars from different countries and regions have studied hundreds of cities, focusing on the heat island effect, using various methods. Founda et al. [22] found that the summer daytime heat island intensity in the city of Athens increased at a significant rate of 0.8 °C/10a. Sachindra et al. [23] concluded that Melbourne’s nighttime heat island intensity showed a significant increase in the period 1952–2010, showing a significant upward trend during the period 1952–2010. Torres–Valcarcel et al. [24] performed a detailed analysis of temperature changes in Puerto Rico over the last century and showed that urbanization has increased minimum, maximum, and average temperatures in the warmest areas by 0.5 °C and increased temperatures in the most coolest areas by 2 °C, and pointed to the future use of remote sensing data to estimate surface temperatures as a way to improve the accuracy of meteorological data in spatial distribution, which is an urgent and pressing issue. Xiong Ying et al. [25] used principal component analysis to explore thermal environment causal mechanisms and driving factors, and pointed out that adding an analysis of influencing factors such as building density and volume ratio can make the study of urban habitat, thermal environment effect, influencing factors and formation mechanisms more comprehensive. Lin Rongping et al. [26] used the principal component multiple regression method to explore coastal areas. The results of the study showed that human activities are an essential cause of urban heat islands. George et al. [27] used MODIS surface temperature data from 2007 to 2013 to explore the spatial and temporal distribution pattern of the thermal landscape in Beijing, and found that changes in the spatial and temporal pattern of the thermal environment in the ecological connotation development area are most sensitive to the migration of the surface’s thermal landscape plenum in Beijing.

In summary, UHI includes the traditionally known canopy heat islands canopy urban heat islands (CUHI) [28] and surface urban heat islands surface urban heat islands (SUHI) [29]. CUHI is measured by in situ measurements from weather stations, which are generally built in the suburbs around cities and often do not cover the entire urban environment [30,31]. Due to the spatial continuity of satellite-derived observations, it is easier and more accurate to study the intra-urban variation in the surface urban heat islands’ (SUHI’s) effect, so this paper uses remote sensing images to invert the surface temperature to obtain the spatial and temporal thermal environment at the study area scale and observe the SUHI.

As a typical representative of western river valley cities, the main built-up area of Lanzhou City is formed and developed in the river valley. The east–west river valley topography constrains the main urban area. The urban geographical space is narrow, and the urban construction land is developed on the flat land and the secondary terraces of the river valley. The air mobility in the central urban area of Lanzhou City is poor, and there are frequent industrial waste gas emissions. The urban heat island effect is evident due to the influence of the urban geographical environment and human factors. Along with the city’s expansion, the urban heat island gradually expands, and rapid urbanization causes increasingly major environmental problems such as urban heat islands, which increases the human and material resources that must be invested in counteracting this urban pathological climate environment. The study of the urban thermal environment in this region is of great significance for urban planning and ecological environment construction. Previous studies on Lanzhou City have focused on CUHI [32,33], and there are fewer studies on SUHI. It is not easy to quantitatively analyze the magnitude of the influence produced by the influencing factors in previous studies on SUHI, and the interaction of multiple influencing factors cannot be analyzed. Due to the existence of the inverse thermosphere, the atmospheric diffusion conditions in Lanzhou City are terrible. Therefore, as a river-valley-type city with two mountains sandwiched by a river, the surface heat island effect in Lanzhou City is more typical and evident than in other places. Based on this, this paper takes the main urban area of Lanzhou City as the study area. It selects the frontal area index (FAI), building height (BH), building density (BD), water percentage (WP), impervious surface percentage (ISP), and fractional vegetation cover (FVC) as the driving factors to investigate the influence of urban buildings and land cover types on the urban thermal environment to provide a reference for ecological environment planning during urban planning.

2. Overview of the Study Area and Data Methodology

2.1. Overview of the Study Area

Lanzhou is located in northwest China, between 35°34′ and 37°07′ N latitude and 102°35′ and 104°34′ E longitude, in the transition zone between monsoon and non-monsoon climate zones, and has a typical, temperate, semi-arid climate with significant diurnal temperature differences, cold winters, and hotter summers. As the only provincial capital city with the Yellow River passing through the city, Lanzhou has a typical thermal environment effect. Statistics from 1980–2019 show that the number of high-temperature days (the highest temperature greater than 35 °C) in 2000–2009 was 19d, and the total number of high-temperature days in 2010–2019 was 23d, while the number of high-temperature days in 1980–1999 was only 3d, which shows that the thermal environment effect in Lanzhou City became more intense after 2000. A real problem faced during the development of Lanzhou City is the need to alleviate the thermal environment effect in urban summer and improve the comfort of the human living environment.

By the end of 2021, Lanzhou City will have five districts and three counties under its jurisdiction, and the city’s resident population is 4,384,300, with an urbanization rate of 83.56%. In this paper, the central urban area where An Ning District, Cheng Guan District, Qilihe District, and Xigu District are located was selected as the study area, as shown in Figure 1.

2.2. Data Acquisition and Processing

The selected images were imaged during the strongest surface reflectivity of the year, from June to September, including Landsat5_TM images with transit times of 11:18 AM on 13 July 2001 and 11:26 AM on 27 July 2011, and Landsat8_OLI/TIRS images with transit times of 11:37 AM on 4 July, 11:38 AM on 7 August, and 11:38 AM on 8 September 2021, 11:38 AM on 4 July, 11:38 AM on 7 August, and 11:38 AM on 8 September 2021, and Landsat8_OLI/TIRS images. The data were downloaded from the geospatial data cloud (www.gscloud.cn, accessed on 13 October 2022), with strip numbers and row numbers of 131/35 and 130/35, and a spatial resolution of 30 m × 30 m. The remote sensing images of the study area were imaged during the hot summer, with clear weather and less than 5% clouds, and the image quality was good. The images were pre-processed with ENVI 5.3 software for geometric correction, radiometric calibration, atmospheric correction, and study area cropping. Land cover data were obtained from the world’s first 10 m resolution global land cover dataset, FROM-GLC10 [34], developed by Prof. Gong Peng’s team at Tsinghua University, and land cover types were classified as forest, farmland, grassland, water body, and impermeable layer. The building vector data were obtained from Baidu Map, containing information on spatial location, building outlines, and building levels. They were subjected to various forms of data processing such as coordinate projection conversion, height attribute anomaly deletion, and isolated building elements’ removal in non-core areas.

2.3. Research Methodology

2.3.1. Surface Temperature Inversion and Classification

Based on the atmospheric correction method, the thermal infrared band was selected to invert the surface temperature of the central city of Lanzhou, the thermal infrared band of Landsat5_TM image is the TM6 band, and the thermal infrared band of Landsat8_OLI/TIRS is TIRS10 band. In the atmospheric correction method, also known as the radiative transfer equation method, the intensity of thermal radiation obtained at the satellite sensor can be expressed based on the atmospheric correction method. The thermal infrared band was selected to invert the surface temperature of the central city of Lanzhou, and the thermal infrared band of Landsat5_TM image was the TM6 band. The thermal infrared band of Landsat8_OLI/TIRS was the TIRS10 band. In the atmospheric correction method, also known as the radiative transfer equation method, the intensity of thermal radiation obtained at the satellite sensor can be expressed as [35]:

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

(1)

where L_λ is the thermal radiation intensity (W·m⁻²·sr⁻¹·μm⁻¹), which is a known value and can be calculated from the image DN value; $\epsilon$ is the surface-specific emissivity, which is calculated by dividing the image into water, buildings, and vegetation and using the hybrid image element decomposition method [36]. B(T_s) is the blackbody thermal radiation intensity expressed by Planck function; T_s is the surface temperature (K); τ is the atmospheric transmittance in a thermal infrared band; L_↑ and L_↓ are the atmospheric uplinks and downlink thermal radiation intensity, respectively; τ, L_↑, and L_↓ were obtained from NASA: (http://atmcorr.gsfc.nasa.gov/, accessed on 20 October 2022) by entering the information in the image header file. Therefore, if the surface specific emissivity $\epsilon$ is known, we can solve B(T_s) from Equation (1), and then use Equation (2) [35] to solve the surface temperature:

$$T_s=K_2/\ln \left[1+K_1/B\left(T_s\right)\right]$$

(2)

where T_s is the surface temperature (K); for TM6 band, K₁ = 607.76 (W·m⁻²·sr⁻¹·μm⁻¹), K₂ = 1260.56 K, for TIRS10 band, K₁ = 774.89 (W·m⁻²·sr⁻¹·μm⁻¹), K₂ = 1321.08 K.

When comparing the urban thermal environment at different times, because of the different imaging times of remote sensing images, the maximum and minimum values of the reflected surface temperature vary greatly, and there are also abnormal temperature values. This is not convenient for direct comparison and analysis, so the images need to be normalized. The mean-standard deviation method [37] was used to classify the surface thermal environment levels in the study area, and the surface thermal environment levels were classified into five categories: low-temperature zone, sub-low-temperature zone, medium-temperature zone, sub-high-temperature zone, and high-temperature zone, as

Table 1. Hierarchies of land surface temperature obtained by the mean-standard deviation method.

Surface Thermal Environment Class	Classification Criteria
Low-Temperature Zone	LSTi < M − 1.5 × S
Sub-low temperature zone	M − 1.5 × S < LSTi ≤ M − 0.5 × S
Medium temperature zone	M − 0.5 × S < LSTi ≤ M + 0.5×S
Sub-heat zone	M + 0.5 × S < LSTi ≤ M + 1.5 × S
High-Temperature Zone	LSTi > M + 1.5 × S

Note: LST_i is the LST value of image I in the study area, M is the mean value of LST, and S is the standard deviation of LST.

shows.

2.3.2. Acquisition of Surface Information

Firstly, the central city of Lanzhou was divided into 120 m × 120 m square research grids with the help of ArcGIS10.2 software, and then the LST and influence factor indicators within each grid were quantified. The spatial quantification equation is shown below:

(1)

Frontal Area Index (FAI)

The frontal area index is the ratio of the windward area of the building to the total area of the plot [38]. Projected frontal area refers to the total surface area of the building facing the incoming wind direction minus the area blocked by the upwind building, i.e., the area of the windward side of the building directly exposed to the incoming wind. Based on the definition of surface dimensions by morphometry, the windward area of a building varies with the actual wind direction. The formula for FAI [39] is the following:

$$\lambda f(Z)={\displaystyle \sum _{i=1}^{16}\lambda f(Z,\theta )\times \mathrm{P}\theta ,\mathrm{i}}$$

(3)

where λf(Z) is the frontal area index of a specific direction, Pθ,i is the frequency of the specified wind direction, and 16 was used as the wind direction value in this study. In order to measure and simulate the frontal area index (FAI), we needed to obtain the height of the building, use GIS for statistical analysis, calculate the height of the building, and use C# programming to realize the single wind direction Then, we calculated the area of the windward side of the building and the area ratio of the grid in the grid, and weighed the frontal area index of the grid according to the wind frequency of different wind directions.

(2)

Building Height (BH)

The average building height reflects the average building height information. The calculation formula [40] is the following:

$$H=\frac{{\displaystyle {\sum }_{i=1}^n{H}_i\times A_i}}{{\displaystyle {\sum }_{i=1}^n{A}_i}}$$

(4)

where H is the average height of the buildings in the statistical unit; H_i is the height of the ith building; A_i is the floor area of the ith building.

(3)

Building Density (BD)

Building density reflects the density of buildings in the horizontal direction. The formula [41] is as follows:

$$D=\frac{{\displaystyle {\sum }_{i=1}^n{A}_i}}{A}$$

(5)

where D is the average building density in the statistical unit; A_i is the floor area of the ith building; A is the area of the statistical unit.

(4)

Water Percentage (WP) and [42] Impervious Surface Percentage (ISP)

$$D=\frac{S_{\mathrm{i}}}{S}$$

(6)

where D is the percentage of statistical unit (including the percentage of WP and ISP); S is the area of statistical unit, and S_i is the area of WP (ISP) in the ith statistical unit.

(5)

Fractional Vegetation Cover (FVC)

FVC was calculated using the image dichotomous model, and the formula [43] was the following:

$$FVC=\left(NDVI-NDV{I}_s\right)/\left(NDV{I}_v-NDV{I}_s\right)$$

(7)

where NDVI is the normalized vegetation index; NDVI_v is the NDVI value when the area is fully covered by vegetation; NDVI_s is the NDVI value when the area is not covered by vegetation [44].

2.3.3. Pearson’s Correlation Analysis

In this paper, Pearson’s simple correlation coefficient was chosen to measure the closeness of the correlation between the variable factors, which is an important indicator of whether two variables are correlated, generally expressed as r. This is defined as the quotient of the covariance and standard deviation of variable X and variable Y [45]. The specific formula is as follows:

$$r=\frac{{\displaystyle \sum _{i=1}^n(X_i-\overline{X})(Y_i-\overline{Y})}}{\sqrt{{\displaystyle \sum _{i=1}^n(X_i-\overline{X}{)}^2}}\sqrt{{\displaystyle \sum _{i=1}^n(Y_i-\overline{Y}{)}^2}}}$$

(8)

where n is the number of samples of variables X and Y; X_i and Y_i denote the sample data of variables X and Y; $\overline{X}$ and $\overline{Y}$ denote the sample means of variables X and Y; r takes values between [–1, 1]. The larger its absolute value, the stronger the correlation.

2.3.4. Geodetectors

Geodetectors [46] is a new statistical method to detect spatial anisotropy and reveal the driving factors behind this, which can detect both numerical and qualitative data without the assumption of linearity. The q values are calculated as follows:

$$q=1-\frac{{\displaystyle {\sum }_{h=1}^L{N}_h{\sigma }_h^2}}{N{\sigma }^2}$$

(9)

where q is the explanatory power of the geographic factor; L is the classification or partition of the dependent variable LST or geographic factor X; h is the partition variable (h = 1,…, L); q takes values in the range of [0, 1], and a larger value indicates the stronger influence of geographic factor X on the spatial partitioning of LST.

The interactions between factors were classified into five types based on q values: nonlinearly attenuated, single-factor nonlinearly attenuated, two-factor enhanced, independent, and nonlinearly enhanced, as

Table 2. Types of interactions and judgment criteria.

Interaction Type	Judgment Criteria
Non-linear weakening	q(X1 ∩ X2) < Min(q(X1), q(X2))
Single-factor nonlinear attenuation	Min(q(X1), q(X2)) < q(X1 ∩ X2) < Max(q(X1), q(X2))
Two-factor enhancement	q(X1 ∩ X2) > Max(q(X1), q(X2))
Independent	q(X1 ∩ X2) = q(X1) + q(X2)
Non-linear enhancement	q(X1 ∩ X2) > q(X1)+q(X2)

shows.

3. Results and Analysis

3.1. Spatial Distribution Characteristics of the Thermal Environment

According to the aforementioned surface temperature inversion calculation method (Equations (1) and (2)), the surface temperatures of Lanzhou city center in 2001, 2011, and 2021 were obtained by ArcGIS10.2 software (Figure 2). The corresponding temperature ranges were 15.60–45.87 °C, 20.16–46.39 °C, and 21.67–51.92 °C, and the mean value increased from 32.26 °C in 2001 to 39.08 °C in 2011, and further increased to 39.66 °C in 2021. It shows that the temperature in Lanzhou City continues to increase. From the comparative analysis of Figure 2 and Figure 3, it can be seen that the spatial and temporal evolution characteristics of surface temperature are closely related to the land use type of the city and the surface temperature increases after the land use type is changed to construction land.

In 2001, the average temperature in the central city of Lanzhou was 32.26 °C, and the high-temperature areas were concentrated in Tumendun, southwest of Dunhuang Road Street, most of Shajing Yi Street, south of Yanjiaping Street, and west of West Station Street. The highest temperature in the urban area reached 45.87 °C, and the average temperatures of cultivated land, forest land, grassland, and water were 34.21 °C, 33.55 °C, 38.21 °C, and 26.71 °C, respectively, with differences of 11.66 °C, 12.32 °C, 7.65 °C and 19.13 °C from the highest temperature in the urban area. In 2011, the range of high temperature decreased compared to 2001, but the average temperature in the urban area increased by 6.83 °C compared with 2001. The average temperature in the study area exceeded 39 °C, the highest temperature in some areas increased to 46.39 °C, and the average temperature of cultivated land, forest land, grassland, and water in the urban area increased to 38.03 °C, 37.18 °C, 41.70 °C, and 30.55 °C. By 2021, the heat island effect in the central city of Lanzhou will be further increased.

In addition to Renshoushan Forest Park, Pengjiaping Central Park, and Lanzhou City Botanical Garden, as well as the Yellow River, Yintan, Matan, Yantan, and Tanjinzi Original Ecological Wetland Park, other streets had a high temperature or sub-high temperature. The highest temperature reached 48.92 °C, while the average temperature of cultivated land, forest land, and grass also showed a small increase to 39.82 °C, 37.24 °C, and 42.54 °C. The average temperature of the water area decreased slightly to 28.93 °C, and the differences between the highest temperature in the city were 9.10 °C, 11.67 °C, 3.38 °C, and 19.99 °C, respectively.

On the whole, the average temperature difference between Lanzhou city and its non-construction land from 2001 to 2021 was 10.00–20.00 °C. The heat island phenomenon spread in a “pie” style, and the expansion of high and low temperatures synchronized with the expansion of construction land in the river-valley-type city, with “two mountains and one river” in recent years. The construction land in the river valley was the first to become the heat island coverage area, and the other surface types also showed different degrees of warming. The rest of the surface types also showed different degrees of warming.

3.2. Temporal Evolution Characteristics of the Thermal Environment

The surface temperature area statistics of different years and levels are shown in

Table 3. The area statistics of the level area of land-surface temperature in Lanzhou center city from 2001 to 2021 (km²).

Time	Low-Temperature Zone	Sub-Low-Temperature Zone	Medium-Temperature Zone	Sub-Heat Zone	High-Temperature Zone
2001	13.367	31.577	104.591	57.500	6.012
2011	15.129	24.819	108.746	61.691	2.662
2021	13.852	20.993	120.895	53.976	3.468

. From Table 3 and Figure 4, the thermal environment of the urban area is shown to be constantly changing from the sub-low-temperature zone and sub-high-temperature zone to the medium-temperature zone, with a phenomenon of gathering in the middle. The area ratio of the low-temperature zone increased from 6.27% in 2001 to 6.50% in 2021. The area ratio of the medium temperature zone increased from 49.09% in 2001 to 56.75% in 2021, which is a significant increase. The proportion of sub-cold-, sub-high-, and high-temperature zones showed a decreasing trend, from 14.82%, 26.99%, and 2.82% in 2001 to 9.85%, 25.34%, and 1.63% in 2021, respectively.

During the 20-year period from 2001 to 2021, the heat island area (sub-high-temperature zone and high-temperature zone) decreased from 63.51 km² in 2001 to 57.44 km² in 2021, accounting for 24.44% of the main urban area of Lanzhou, and the heat island area decreased by 6.61 km². Among them, the high-temperature zone changed the fastest, and the growth rate in 2021 was about 0.5 times that in 2001 (42.32% decrease). The growth rates of the low-temperature zone, sub-low-temperature zone, medium-temperature zone, and sub-high-temperature zone were 3.63%, −33.52%, 15.59%, and −6.13%, respectively.

3.3. Analysis of Spatial and Temporal Changes in the Thermal Environment

The five classes of LST according to the mean-standard deviation were analyzed in the order of 1~5, from low temperature to high temperature, and if the difference <0, this means the temperature class is decreasing; if the difference >0, this means the temperature class is increasing. The results (Figure 5) show that, from 2001 to 2011, the areas with increasing temperature levels were located in the Yintan, Matan, and Yanbei streets, related to the encroachment of forest land and farmland due to urban construction. In particular, the forested area of Yintan Road, Kongjiaya, and Anningbao Street decreased from 4.31 km² in 2001 to 0.17 km² in 2011, and the surface temperature increased by 12.27 °C; the cultivated area of Matan decreased from 2.07 km² in 2001 to 0.59 km² in 2011, the forested area decreased from 0.61 km² to 0.18 km², and the surface temperature increased by 8.18 °C. The area of forest land in Yanbei Street decreased from 1.37 km² to 0.1km², the area of cultivated land decreased from 5.27 km² to 1.19 km², and the temperature increased by 11.55 °C. The areas where the temperature level increased from 2011 to 2021 were more scattered than those from 2001 to 2011, mainly in Pengjiaping, Anningbao, and Qingbaishi Street. In 2011, the cultivated area of Pengjiaping was 6.42 km², and in 2021, this decreased to 1.23 km², the forested area decreased from 1.45 km² to 0.73 km², and the grassland area decreased from 1.18 km² to 0.44 km². The surface temperature increased by 7.72 °C, the cultivated area of Anningbao street decreased from 9.29 km² to 4.35 km², and the surface temperature increased by 6.62 °C. Arable land in the northwestern part of Xiuchuan Street and the northern part of Chenping Street on both sides of the Yellow River will disappear by 2021; arable land in Qingbaishi Street and the northeastern part of Yanbei Street and the northern part of Donggang Street will decrease from 6.38 km² in 2011 to 3.13km² in 2021. The woodland area will decrease from 0.96 km² to 0.33 km², the grassland area will decrease from 0.72 km² to 0.27 km², and the surface temperature will increase by 9.40 °C. By 2021, the arable land, woodland, and grassland were distributed in the eastern part of Qingbaishi Street, and the arable land and woodland and grassland in Yanbei Street and Donggang Street were replaced by construction land. In general, the distribution characteristics of a temperature increase or decrease from 2001 to 2021 are generally consistent with those from 2011 to 2021.

3.4. Spatial Distribution of Thermal Environment Impact Factors

As shown in Figure 6, a 120 m grid was used as the statistical unit to quantify the influencing factors in the central city of Lanzhou using the ArcGIS spatial statistical analysis tool. The spatial distribution of the frontal area index in Figure 6a showed a gradual increase from the eastern part of the An Ning district to the southwestern part of the Chengguan district, with the highest FAI values in commercial and high-rise residential areas generally above 3.68, and the lowest FAI values in industrial parks and low-rise residential areas below 0.56. Figure 6b shows the trend of gradually decreasing building height from the southwest of Chengguan District to the surrounding area, with the building height of most multi-story residential areas below 29.5 m, and the building height of industrial parks and old residential areas generally below 11.14 m. The layout of high-rise building groups is concentrated in Jiuquan Road, Zhangye Road, and Gaolan Road streets in the southwest of Chengguan District. Figure 6c shows the density of buildings near high-rise residential areas, large green areas, and water surfaces to be below 15%; the density of buildings in new multi-story residential areas is 15–65%; the density of buildings in industrial parks, old residential areas, and urban commercial areas is generally above 65%. Figure 6d shows that the spatial distribution of water resources is extremely unbalanced, with the Yellow River running through the city, except for most areas where the percentage of water body area is less than 4.32%. Figure 6e shows the spatial distribution of the percentage of impervious surface area, with 37.33% of the areas having an impervious surface area of more than 90%, reflecting the high level of urbanization in Lanzhou. As can be seen from Figure 6f, the vegetation mainly covers the periphery of the built-up area of the city, with 5.07% of the areas having a coverage above 80% including high vegetation coverage areas such as Lanzhou Botanical Garden, Renshoushan Forest Park, Yintan Wetland Park, Baitashan Park, Tanjianzi Original Wetland Park, and Wuquanshan Park, while the vegetation coverage in built-up areas is generally lower than 16.21%.

3.5. Correlation Analysis and Regression Analysis of Thermal Environment Influence Factors

3.5.1. Correlation Analysis

To quantitatively analyze the influence of each influencing factor on the urban thermal environment, 406,340 random points were generated in the region for different influencing factors, the values of each influencing factor and the surface temperature values corresponding to each random point were obtained, and the correlation coefficients between the 6 influencing factors and the surface temperature were obtained after removing the abnormal values (

Table 4. Pearson correlation between influencing factors and surface temperature.

Projects	Frontal Area Index	Building Height	Building Density	Water Percentage	Impervious Surface Percentage	Fractional Vegetation Cover
Pearson correlation coefficient	0.130	0.017	0.391	−0.229	0.366	−0.440
Significance level	0.01	0.01	0.01	0.01	0.01	0.01

).

The absolute magnitude of the coefficients represents the degree of influence of the influencing factors on the urban thermal environment. As shown in Table 4, among the above-mentioned factors, building density, the percentage of impervious surface, frontal area index, and building height have positive effects on the surface temperature, and the correlation coefficient values decreased to 0.39, 0.37, 0.13, and 0.02, respectively, representing that building density and the impervious surface in the city are the main factors that lead to the rise in urban thermal environment and heat island effects in the city. The greater the proportion of both in the city, the more heat sources there are in the city, the lower the heat dissipation capacity, and the higher the thermal environment temperature. The factors that have a negative correlation with surface temperature are vegetation cover and water bodies, with correlation coefficients of −0.44 and −0.23, both of which are negative, representing that the above factors can reduce urban surface temperature. The higher the proportion of these in the city, the more significant the cooling and humidifying benefits of vegetation and water bodies, and the more they can alleviate the urban heat island effect.

3.5.2. Regression Analysis

The linear fit model of individual influencing factors and surface temperature (Figure 7) further elucidates the degree of influence of each influencing factor on the surface temperature. At the grid scale of 120 m × 120 m, each 10% increase in building density and impervious surface percentage can increase the surface temperature by 0.51 °C and 0.28 °C in summer. Among them, the cooling effect of water bodies is the most obvious, with each 10% increase in the proportion of water bodies decreasing the surface temperature by 0.98 °C. Although the frontal area index also has a warming effect, this effect is not obvious.

Considering that the values of each influencing factor and surface temperature are continuous random variables, a multiple linear regression model was further used to investigate the differences in the degree of contribution of each influencing factor to the change in surface temperature, and the four factors with the greatest influence on surface temperature, i.e., building density, percentage of water bodies, percentage of impervious surface and vegetation cover, were selected to construct a multiple linear regression equation. Finally, the multivariate linear equation was obtained for the main urban area of Lanzhou City in 2021. The multivariate linear function model of the urban thermal environment in Lanzhou city in the summer of 2021 is shown in Equation (10).

$$T=36.58+3.29BD-3.30FVC-7.11WP-0.20ISP$$

(10)

where T is the summer surface temperature of Lanzhou City; the definition domain of each element is [0, 1]. The model has R² = 0.29, Durbin–Waston = 1.83, and the variance inflation factor VIF << 5 for each element, which indicates that the model regression fits well and is statistically significant. Moreover, the four selected elements can explain 28.6% of the surface temperature variance. Meanwhile, each group of observations is independent, and there is no multicollinearity. The P-P diagram shown in Figure 8 was drawn according to the relationship between the cumulative proportion of variables and the cumulative proportion of the specified distribution, in which the points are close to a straight line and fall on or near the diagonal, indicating that the data obey normal distribution.

3.6. Driving Force Analysis of LST

In this paper, the natural breakpoint method was used to classify the six factors affecting surface temperature, and the geodetector model was introduced to calculate the explanatory power of each influencing factor on surface temperature LST (

Table 5. LST impact factor detection results.

Geographical Factors	Explanatory Power q	p
Frontal area index	0.03	0.00
Building height	0.03	0.00
Building density	0.14	0.00
Water percentage	0.50	0.00
Impervious Surface Percentage	0.22	0.00
Fractional Vegetation Cover	0.04	0.00

).

3.6.1. Factor Detection Analysis

The detection results (Table 5) show that the explanation of LST by each influencing factor in descending order is the percentage of water bodies (0.50) > the percentage of impervious surface (0.22) > building density (0.14) > vegetation cover (0.04) > frontal area index (0.03) > building height (0.03), and all six factors passed the 0.01 significance level test. Among them, the explanatory power of water body percentage and impervious surface percentage on LST is greater than 0.2, which is the main driving factor, and the factor detection q-statistic of water body percentage reaches 0.50, reflecting that water body percentage has the greatest influence on the spatial variation in surface temperature in the main urban area of Lanzhou. The explanatory power of building density is greater than 0.05, which is the secondary driver, indicating that it plays a more important role in the spatial variation in surface temperature.

3.6.2. Interaction Analysis of Factors

According to the interaction detection results (Figure 9), different interaction types of influencing factors have two types of interactions on the surface temperature of the main urban area of Lanzhou City, namely two-factor enhancement and non-linear enhancement, among which the interaction of two-factor enhancement accounts for most of the interactions; that is, in most cases, the interaction of influencing factors on the surface temperature of the main urban area of Lanzhou City has a greater influence than that of a single influencing factor. ISP∩FVC and WP∩FVC showed nonlinear enhancement, and the interaction results of the remaining factors showed two-factor enhancement. The explanatory power of the interaction between FVC and each factor is significantly higher than that of the single factor, which indicates that FVC is indirectly acting on LST.

4. Discussion

This paper uses Landsat TM image data from 2001 and 2011. Landsat TIRS image data in 2021 were used to invert the surface temperature at three time points in the central city of Lanzhou in the summer using the radiation equation conduction method as well as to analyze the spatial and temporal evolution of LST and the influence of FAI, BH, BD, WP, ISP, and FVC on the surface temperature distribution in the summer, plus the extent of the influence of LST on summer surface temperature distribution. From 2001 to 2021, the significant expansion of built-up land and the shrinkage of cultivated, forested and grassland areas were fundamental reasons for the increase in the average surface temperature.

Green areas and water bodies have a good improvement effect on the thermal environment in summer and can improve the comfort of the human living environment. This paper concludes that WP and FVC significantly negatively correlate with surface temperature. In contrast, ISP has a significant positive correlation with surface temperature, which is consistent with the findings of Yuanyuan Chen [42] and Xinbo Song [41]. This paper also points out that FAI, BH, and BD have a significant positive correlation on the surface temperature at the three-dimensional level, which is consistent with the findings of Feng Zhangxian [47] and Sun Jiabin [48] that surface temperature is positively correlated with the two-dimensional building index. For Lanzhou, a northwestern inland city with a dry climate in summer, to improve the comfort of the human living environment, the planting of deciduous broad-leaved trees such as acacia and willow trees should be increased to reduce the area of impervious surfaces, for example, the laying of non-essential impervious surfaces such as tiles and cement can be reduced during the construction of public recreational facilities such as parks, and sparsely wooded grassland parks filled with natural scenery should be built. When carrying out urban planning, by reasonably laying out water bodies and vegetation, areas with cooling effects are spread to a larger area, thus improving the summer thermal environment. For example, in Chongqing, a typical mountainous city, green areas and water bodies have an excellent mitigation effect on the heat island phenomenon [49]. In contrast, water bodies have less influence on the surrounding thermal environment, and green areas have more influence on the surrounding thermal environment.

The construction of ecological civilization is a 1000-year plan for the sustainable development of the Chinese nation. Lanzhou, as the only capital city of the Yellow River passing through the city, has the geographical advantage of “two mountains sandwiched by a river” and has the natural conditions needed to build a city of landscape. To create a city of landscape and build the ecological pearl of the Yellow River, it is more important to include green areas and water bodies in the construction of urban ecological environment. Against, the background of global warming, high temperatures and heat waves are frequent. In the context of global warming and high-temperature heat waves, improving the surface thermal environment in cities such as the mainland positively impacts sustainable urban development. Since building data for 2001 and 2011 were challenging to obtain, it was impossible to explore the influence of building dynamics on surface temperature over the past 20 years. Different urban building types may also have different degrees of influence on surface temperature. Due to data limitations, this study failed to classify urban building types in Lanzhou into local climatic zones; thus, it is missing an exploration of the degree of influence of different types of urban buildings on surface temperature, and, in future studies, different local types of urban built-up areas could be considered for further exploration.

5. Conclusions

(1)

With the development of the city, the heat island area of Lanzhou City was reduced by 6.608 km², and the average surface temperature increased by 7.41 °C.

(2)

The city’s expansion between 2001 and 2011 increased the temperature level in Yin Tan, Ma tan, and Yan Bei streets; the re-expansion of urban construction land between 2011 and 2021 increased the temperature level in Pengjiaping, An Ningbao, and Qing Baishi streets.

(3)

FVC, BD, ISP, and WP have the strongest correlation with the urban thermal environment, on which the multiple linear regression equations were constructed with coefficients of −3.30, 3.29, −0.20, and −7.11, respectively, and a constant of 36.58. The magnitude of the absolute value of the regression coefficient represents the degree of influence of the independent variables on the surface temperature, with WP having the most significant influence on the surface temperature and ISP having the slightest influence on the surface temperature.

(4)

Among the influencing factors, WP and ISP are the main drivers of LST spatial distribution. The interaction detection results show that the interactions between FVC and WP and FVC and ISP are nonlinearly enhanced, and the remaining factors are two-way enhanced by two interactions.

In this study, we only selected the drivers for 2021 to analyze the impact of each influencing factor on the thermal environment. Future studies can analyze the spatial and temporal evolution of the drivers and explore their driving mechanisms in detail.