Remote Sensing Monitoring and Multidimensional Impact Factor Analysis of Urban Heat Island Effect in Zhengzhou City

Abstract

In the 21st century, the rapid urbanization process has led to increasingly severe urban heat island effects and other urban thermal environment issues, posing significant challenges to urban planning and environmental management. This study focuses on Zhengzhou, China, utilizing Landsat remote sensing imagery data from five key years between 2000 and 2020. By applying atmospheric correction methods, we accurately retrieved the land surface temperature (LST). The study employed a gravity center migration model to track the spatial changes of heat island patches and used the geographical detector method to quantitatively analyze the combined impact of surface characteristics, meteorological conditions, and socio-economic factors on the urban heat island effect. Results show that the LST in Zhengzhou exhibits a fluctuating growth trend, closely related to the expansion of built-up areas and urban planning. High-temperature zones are mainly concentrated in built-up areas, while low-temperature zones are primarily found in areas covered by water bodies and vegetation. Notably, the Normalized Difference Built-up Index (NDBI) and the Normalized Difference Vegetation Index (NDVI) are the two most significant factors influencing the spatial distribution of land surface temperature, with explanatory power reaching 42.7% and 41.3%, respectively. As urban development enters a stable stage, government environmental management measures have played a positive role in mitigating the urban heat island effect. This study not only provides a scientific basis for understanding the spatiotemporal changes in land surface temperature in Zhengzhou but also offers new technical support for urban planning and management, helping to alleviate the urban heat island effect and improve the living environment quality for urban residents.

1. Introduction

Since the beginning of the 21st century, the rapid development of urbanization has led to significant changes in urban land use structures, with vegetation, bare soil, and water bodies gradually transforming into built-up areas. This transformation has altered the nature of urban surfaces and generated a significant thermal effect known as the “urban heat island (UHI)” [1]. The urban heat island effect not only changes local surface temperatures and causes urban microclimate variations but also leads to the accumulation of atmospheric pollutants, adversely affecting human health [2,3,4]. Therefore, in-depth research on urban thermal environments and the heat island effect is of great significance for urban planning, heat island mitigation, and improving living environment quality for people.

The urban heat island effect can be categorized into the surface heat island effect and atmospheric heat island effect [5]. Surface temperature and atmospheric temperature are important indicators of these two heat island effects. Surface temperature has an indirect but significant effect on air temperature; however, the change in air temperature is usually smaller than the change in surface temperature in the same area. In order to study the urban heat island effect, surface temperature is chosen as the main indicator in this paper to represent the surface urban effect. Monitoring urban heat islands using the LST primarily includes three methods: field measurements, data modeling, and remote sensing data retrieval [6]. Among them, field measurement refers to the observation of urban heat island phenomena at fixed stations or mobile devices. Data modeling refers to the simulation of the urban heat island effect by establishing a mathematical–physical model on the basis of field measurement data [7]. Remote sensing data retrieval refers to the inversion of the surface temperature by obtaining the reflection characteristics of the absorption of solar radiation according to features in different bands to produce the corresponding radiation values through remote sensing [8]. With the advantages of wide coverage and easy accessibility, remote sensing data retrieval has been increasingly used by scholars to study the urban heat island effect [9,10,11]. Some researchers have found that land use changes are a significant factor influencing the urban heat island effect. Zhongjian Shen et al. found that built-up areas and cultivated land are the main contributors to temperature increases, while forest land and water bodies act as temperature reducers by integrating landscape indices and examining the correlation between different land-use types and changes in land surface temperature [12]. R. Estoque et al. utilized Landsat data to monitor and study the heat island intensity in mountain cities, recommending the consideration of both landscape planning and measures for the formation and mitigation of urban heat islands [13]. Furthermore, some researchers have found that the increase in impervious surfaces leads to higher LST, thereby exacerbating the urban heat island effect [14,15,16]. Morabito et al. investigated the impact of urban building surfaces on LSTs in Italian cities using linear regression analysis, finding a significant linear relationship between the two [17]. Yuexiang Wang et al. retrieved LSTs from remote sensing imagery and extracted impervious surfaces, using standard deviation ellipses and landscape index methods to analyze the direction and type of urban expansion in Nanjing [18]. Can Zhang et al. obtained the main land-use types in the central urban area of Handan using multi-temporal Landsat imagery and supervised classification methods, and analyzed the relationship between impervious surfaces and the urban heat island effect by combining spatial analysis methods such as the center migration model and standard deviation ellipse, showing that the expansion direction of the heat island is basically consistent with the expansion direction of impervious surfaces, significantly impacting the heat island effect [19]. Ullah et al. used Landsat remote sensing imagery, the CAMarkov model, and the Pearson correlation coefficient to analyze the impact of land use changes on the LST in Tianjin, and predicted the spatial distribution pattern and changes in LST from 2020 to 2050, finding that the LST in built-up areas is rising and is expected to increase in the future [20]. Assaf et al. used a Bayesian network model to identify winter NDVI and summer NDVI as two key factors influencing UHI [21]. Cetin et al. explored the relationship between LST, UHI, NDVI, and NDBI by studying the evolution of LST and UHI effects, finding a strong negative correlation between UHI and NDVI and a strong positive correlation between UHI and NDBI [22]. Chubarova et al. investigated the diurnal, seasonal, and annual variations in Surface Urban Heat Island Intensity (SUHII) and land surface temperature (LST) across three cities in India (Lucknow, Kolkata, and Pune) over the past two decades, discovering that an increase in Aerosol Optical Depth (AOD) led to a decrease in temperature and surface solar radiation, resulting in a cooling effect in urban areas [23]. Kudzai Shaun Mpakairi and Justice Muvengwi found that nighttime lights, as a representation of human activities, contributed more to summer nighttime surface urban heat islands [24].

Regarding the study of the spatiotemporal characteristics of urban thermal environments and their influencing factors, Yunfan Song et al. analyzed the spatiotemporal variations of the urban heat island effect in Chengdu during the summer using MODIS Land Surface Temperature (LST) products, which are derived from the Moderate Resolution Imaging Spectroradiometer onboard the Terra and Aqua satellites. MODIS LST provides global surface temperature measurements at a spatial resolution of 1 km and a temporal resolution of twice daily (daytime and nighttime), making it a valuable tool for studying diurnal temperature variations. They discovered significant diurnal variations in the heat island effect and found a negative correlation between vegetation coverage and thermal field intensity, with the correlation being more pronounced during the daytime [25]. Zhang et al., focusing on Xuzhou, utilized four phases of Landsat 8 remote sensing data and surface observation data from 2014 to 2020, employing an improved single-window algorithm for LST retrieval. They calculated and analyzed the relationship between landscape pattern indices and LSTs using moving window methods and Moran’s index, finding spatial correlations between landscape pattern indices and LSTs at various patch levels [26]. Liang et al. examined the Beijing–Tianjin–Hebei urban agglomeration, using Landsat remote sensing data and urban construction land data to analyze the dynamic relationship between landscape information and land surface heat island intensity from 2010 to 2018 in five cities through a spatial regression model. They found that the relationship between landscape information and heat island intensity varied with city size and that urban expansion led to an increase in heat island intensity [27]. However, the influencing factors of urban thermal environments are complex, and a single factor cannot fully describe the spatiotemporal changes in urban thermal environments or explain their influencing factors. Existing studies mainly focus on the spatial and temporal distribution characteristics of the urban heat island and its influencing factors under a single factor, while the in-depth exploration of the spatial and temporal changes of the urban heat island effect and its formation mechanism under the comprehensive consideration of multiple factors, such as surface characteristics, meteorological conditions, and socio-economic factors, is still insufficient.

Therefore, this study takes Zhengzhou as an example, selecting Landsat data from the key development years of 2000, 2004, 2009, 2013, and 2020. Using atmospheric correction methods to retrieve LST, the study combines land use data and the spatiotemporal characteristics of thermal landscapes. By integrating the gravity center migration model and geographical detector, the formation mechanisms of the urban heat island effect are analyzed from surface characteristics, meteorological, and socio-economic factors. This aims to provide new technical and practical support for urban planning and management, thereby better addressing and mitigating the urban heat island effect.

2. Materials and Methods

2.1. Study Area

This study selects the central urban area of Zhengzhou, Henan Province, as the research area (including the Zhongyuan District, Erqi District, Guancheng District, Jinshui District, and Huiji District), as shown in Figure 1. The region is located between 113°37′ E to 114°14′ E and 34°36′ N to 34°46′ N, with a total area of approximately 825.66 km². The study area lies in the Huang–Huai–Hai alluvial plain and has a temperate continental monsoon climate characterized by distinct seasons, with hot and rainy summers. The annual average temperature is around 16.5 °C, and the annual average rainfall is approximately 676.7 mm.

2.2. Data

2.2.1. Remote Sensing Data

In this study, based on the Google Earth Engine (GEE) platform, Landsat series images with good imaging quality for the summer months from 2000 to 2020, as well as nighttime light data, were selected to reflect the extent and intensity of human activities. During the data acquisition process, the reacquired remote sensing images were filtered according to the vector boundaries and time scales of the study area, and then necessary preprocessing operations, cloud masking, geometric correction, and image clipping.

2.2.2. Auxiliary Data

Auxiliary data include sample data selected to achieve land use classification of remote sensing images using the random forest algorithm, 1 km spatial resolution population density data (https://www.worldpop.org/, accessed on 16 June 2022), and road density data calculated from road network data (http://openstreetmap.org/, accessed on 16 June 2022).

2.3. Land Surface Temperature Retrieval

The Atmospheric Correction Method, also known as the Radiative Transfer Equation (RTE) method, is the most fundamental approach for LST retrieval [28,29,30]. Its principle is based on Planck’s law, where the radiance values of a blackbody at temperature Ts in the thermal infrared band are obtained. The true surface temperature Ts is then calculated using the inverse function of Planck’s formula. The calculation process for the Atmospheric Correction Method is as follows:

(1) Calculate the radiance $L_{\lambda }$ in the thermal infrared band, with units of W/(m²·sr·μm) using Equation (1).

$$L_{\lambda }=gain\times DN+offset$$

(1)

where gain and offset are the gain and offset values.

(2) Calculate the radiance $\mathrm{B}\left(\mathrm{T}\mathrm{s}\right)$ at blackbody temperature Ts, with units in Kelvin (K), using Equation (2). The intermediate parameter $\epsilon$ represents the surface emissivity, which is calculated based on the Fractional Vegetation Cover (FVC). $L\uparrow$ and $L\downarrow$ are the upwelling and downwelling atmospheric radiance values, respectively, with units of W/(m²·sr·μm). $\tau$ represents the atmospheric transmittance. Since real-time atmospheric profile data for the study area are difficult to obtain and calculate, this study uses standard atmospheric profile data to calculate $\tau$, $L\uparrow$, and $L\downarrow$. These calculations are performed by inputting the Landsat image acquisition time and central coordinates into the atmospheric parameter calculation website (https://atmcorr.gsfc.nasa.gov/, accessed on 14 June 2022). The specific radiative parameters queried based on the five periods of imagery are shown in

Table 1. Atmospheric radiation parameters in the study area.

Data	$\mathit{\tau }$	$\mathit{L}\uparrow$	$\mathit{L}\downarrow$
16 June 2000	0.63	2.84	4.49
30 August 2004	0.69	2.43	3.86
12 August 2009	0.76	2.10	3.42
4 June 2013	0.80	1.76	2.95
26 August 2020	0.68	2.80	4.43

.

$$\mathrm{B}\left(\mathrm{T}\mathrm{s}\right)={\displaystyle \frac{L_{\lambda }-L\uparrow -\tau \times \left(1-\epsilon \right)\times L\downarrow }{\epsilon ·\tau }}$$

(2)

where $\epsilon$ is the surface emissivity, L↑ and L↓ are the atmospheric upwelling and downwelling radiance values, and $\tau$ is the atmospheric transmittance.

(3) Calculate the land surface temperature Ts in degrees Celsius (°C) using Equation (3).

$$\mathrm{T}\mathrm{s}={\displaystyle \frac{K_2}{\mathrm{l}\mathrm{n}(K_1/B(Ts)+1)}}-273.15$$

(3)

where K₁ and K₂ are calibration constants specific to different sensors. For Landsat 5 TM Band 6: K₁ = 607.76 W/(m²·sr·μm), K₂ = 1260.56 K; for Landsat 8 TIRS Band 10: K₁ = 774.89 W/(m²·sr·μm), K₂ = 1321.08 K.

2.4. Heat Island Intensity Classification

Due to the inconsistency in the image acquisition times, it is challenging to perform quantification. Previous studies have demonstrated that the mean-standard deviation method is more effective than other classification methods in capturing the spatial distribution and temperature variation details of urban heat islands [31]. In this study, surface temperature was normalized, and classifications were made based on different combinations of the mean and standard deviation of surface temperature. Seven levels were identified: strong cool island areas (SCIAs), moderately strong cool island areas (MSCIAs), weak cool island areas (WCIAs), normal areas (NAs), weak heat island areas (WHIAs), moderately strong heat island areas (MSHIAs), and strong heat island areas (SHIAs). The classification standards are shown in the

Table 2. Temperature classification standard.

Classification Level	Criteria
strong heat island areas (SHIAs)	$\mathrm{T}\ge T_m+2.5\mathsf{\sigma }$
moderately strong heat island areas (WSHIAs)	$T_m+1.5\mathsf{\sigma }\le \mathrm{T}<T_m+2.5\mathsf{\sigma }$
weak heat island areas (WHIAs)	$T_m+0.5\mathsf{\sigma }\le \mathrm{T}<T_m+1.5\mathsf{\sigma }$
normal areas (NAs)	$T_m-0.5\mathsf{\sigma }\le \mathrm{T}<T_m+0.5\mathsf{\sigma }$
weak cool island areas (WCIAs)	$T_m-1.5\mathsf{\sigma }\le \mathrm{T}<T_m-0.5\mathsf{\sigma }$
moderately strong cool island areas (MSCIAs)	$T_m-2.5\mathsf{\sigma }\le \mathrm{T}<T_m-1.5\mathsf{\sigma }$
strong cool island areas (SCIAs)	$\mathrm{T}<T_m-2.5\mathsf{\sigma }$

Note: In the table, $T_m$ represents the mean of the normalized surface temperature and $\mathsf{\sigma }$ represents the standard deviation of the normalized surface temperature.

.

2.5. Center of Gravity Shift Model

The center of gravity shift model is used to describe the spatial evolution of the urban heat island effect, which is specifically reflected in the migration of the center of gravity of urban heat island patches in different time periods [32,33]. By analyzing the movement distance and direction of the center of gravity, the dynamic change characteristics of urban heat islands and their influencing factors can be revealed. In this study, the centers of gravity of urban heat island patches in different time periods were extracted by analyzing the land surface temperature (LST) classification results. Meanwhile, the centers of gravity of built-up areas were extracted based on the land use classification results. The center of gravity of built-up areas is the spatial gravity of densely built-up areas, which is calculated using the building classification results from remote sensing imagery. The coordinates of the center of gravity are used to calculate the distance and angle of center of gravity migration within each administrative area to reveal the impact of built-up area expansion on the urban heat island effect.

(1)

Gravity shift distance model

The gravity migration distance model quantitatively describes the distance of gravity changes between urban heat island areas over different periods. The model is as follows:

$$\mathrm{D}=\sqrt{{(X_{t+1}-X_t)}^2+{(Y_{t+1}-Y_t)}^2}$$

(4)

where D is the gravity migration distance of the urban heat island patch or built-up area, X_t and Y_t are the coordinates of the gravity in year t, and X_t+1 and Y_t+1 are the coordinates of the gravity in year t + 1.

(2)

Gravity shift angle model

The gravity shift angle model quantitatively describes the change in direction of the gravity of urban heat island patches over different time periods. The model is as follows:

$$\mathsf{\theta }=\arctan \left[\left(X_{t+1}-X_t\right)/{(Y}_{t+1}-Y_t)\right]$$

(5)

where $\mathsf{\theta }$ is the gravity migration angle of the urban heat island patch or built-up area, measured clockwise from the north direction to the line formed by the coordinates. The angle ranges from 0° to 360°. X_t and Y_t are the coordinates of the gravity in year t, and X_t+1 and Y_t+1 are the coordinates of the gravity in year t + 1.

2.6. Exploration of Factors Influencing Urban Heat Island Intensity

The urban heat island effect is typically formed by the combined action of multiple influencing factors. Based on the availability of existing literature and data on factors and using the 2020 LST inversion results as a basis, ten factors were selected for analysis, encompassing land surface characteristics (F), meteorological conditions (W), and socio-economic factors (S). Basic information on these influencing factors is summarized in

Table 3. Factors affecting the urban heat island effect.

Factor Categories	Factors Influencing	Abbreviation
surface characteristics	Normalized Difference Vegetation Index	NDVI
	Soil-Adjusted Vegetation Index	SAVI
	Normalized Built-up Index	NDBI
	Modified Normalized Difference Water Index	MNDWI
	Land Use Land Cover	LULC
	Elevation	ELEVATION
meteorology	Aerosol Optical Depth	AOD
social economy	Nighttime light	NL
	Population density	PD
	Road density	RD

.

Geographic detector analysis is employed to describe the contribution of each factor to urban heat island effects.

$$\mathrm{U}\mathrm{H}\mathrm{I}\mathrm{I}=f\left(F,W,S\right)$$

(6)

where f is a function representing the complex relationship between urban heat island intensity (UHII) and land surface characteristics (F), meteorological conditions (W), and socio-economic factors (S).

As a new statistical model, a geographic detector can measure the spatial heterogeneity of LST and explore the underlying influencing factors. Unlike traditional models, it can analyze the relationships between variables without assuming linearity or dealing with collinearity issues [34,35]. The factor detection function is used to detect spatial variations in LST and the degree to which various factors influence it. In Equation (6), q represents the explanatory power of the factors on LST. A higher q value indicates stronger explanatory power, while a lower value indicates weaker influence. The calculation method is shown in the following equation:

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

(7)

In the equation, q values range from [0, 1]. Here, h = 1, …, L, represents the number of influencing factors, which is 10 in this experiment. N_h denotes the number of sample points for variable X in category h, and N is the total number of grids. ${\sigma }_h^2$ and ${\sigma }^2$ are the variances of LST within category h and across all grids, respectively.

3. Results

3.1. Temporal and Spatial Evolution of Urban Thermal Environment

Based on five periods of Landsat images, using the RTE method, summer LSTs were retrieved for the study area on 16 June 2000; 30 August 2004; 12 August 2009; 4 June 2013; and 26 August 2020. The inversion results are shown in the left subplot of Figure 2. In the figure, the blue areas represent regions with lower LSTs, primarily located near rivers, lakes, and other water bodies. Areas transitioning from blue to red indicate increasing LSTs, with red indicating higher temperatures predominantly in urbanized areas, showing a patchy distribution.

On the right side of Figure 2, we also added time-series curves of LSTs and solar radiation (SR) for each year and calculated the correlation between these two variables. By analyzing the data for the years 2000, 2004, 2009, 2013, and 2020, we observe a significant correlation between LSTs and solar radiation with a correlation coefficient around 0.72. As the temperature in 2000 was indeed more severe, the correlation coefficient is 0.53 This indicates that solar radiation has a significant effect on the change in surface temperature, and the trend of LSTs in different years is closely related to the change in solar radiation intensity. Specifically, higher solar radiation is usually associated with higher LST values, which further supports the relationship between the formation of the urban heat island effect and solar radiation intensity.

Based on the inversion results, the average LST in the study area was 38.97 °C in 2000, the highest among the five periods analyzed. The lowest temperature was 22.24 °C, mainly located north of the Yellow River in the Huiji District. The highest temperature, 51.98 °C, was found near the farmland and southwestern mountainous areas of the study area. Observations from Landsat images during the same period indicated that croplands had generally low vegetation cover, and soils near the Yellow River contained a significant amount of sediment. Sparse vegetation cover on mountains exposed bare rocks, contributing to a higher LST in 2000. In comparison to 2000, the LST noticeably decreased in 2004, with an average temperature of 31.31 °C, the lowest among the five periods. The lowest temperature was 13.24 °C, still located in water bodies within the study area, while the highest temperature was 50.70 °C, predominantly concentrated in the central urban area. By 2009, the LST had slightly increased compared to 2000 and 2004, but both the average temperature and standard deviation did not exceed the levels observed in 2000. With the increase in urban development, the LST values showed a rise, although the number of high-temperature areas decreased, showing a more concentrated trend. In 2013, the average land surface temperature was close to that of 2000 at 38.31 °C. The highest and lowest temperatures, along with the standard deviation, were higher than those in 2004, indicating greater variability in LST values in 2013, though not reaching the dispersion observed in 2000. Overall, there was an expansion in the range of LST values. By 2020, the average and standard deviation of the LST had decreased by approximately 10% compared to 2013. The dispersion trend of the LST weakened, with a reduction in the number of high-temperature areas leading to a decrease in the average temperature. However, there was a polarization trend in high and low temperature values.

3.2. Urban Heat Island Effect Spatiotemporal Evolution

Based on the inversion results for 2000, 2004, 2009, 2013, and 2020, the intensity of the heat island effect was categorized into seven classes according to the category classification criteria in Table 2. These include strong cool island areas (SCIAs), moderately strong cool island areas (MSCIAs), weak cool island areas (WCIAs), normal areas (NAs), weak heat island areas (WHIAs), moderately strong heat island areas (WSHIAs), and strong heat island areas (SHIAs).

The spatial distribution of thermal landscapes over the five-year period is illustrated in the left subplot of Figure 3. In the year 2000, the central urban area of the study region primarily exhibited normal heat island intensity levels, while areas below normal were mainly distributed in the Jinshui District and Huiji District. Areas above normal were found predominantly to the east of the Jinshui District, along the Yellow River in the northern part of the Huiji District, and in most parts of the Erqi District and Guancheng District. Overall, the thermal landscape showed a pattern of higher intensities towards the edges and lower intensities in the middle. After 2004, the study region displayed evident urban heat island effects, except in the Huiji District, where areas above normal heat island levels predominated in other administrative districts. The distribution pattern of thermal landscapes dispersed from the center outward, with thermal patches away from the city center appearing scattered. During this period, vegetation and water bodies played a role in temperature reduction, forming cooling patches to alleviate heat environmental issues. By 2009, areas above normal heat island levels still centered mainly around the urban core, but the spatial distribution shifted from clustering to dispersion, expanding towards the Guancheng District and Erqi District, with relatively minor changes in temperature classification in other areas. In 2013 and 2020, the areas above normal heat island levels gradually expanded, primarily consisting of normal and weak heat island areas, while strong heat island areas clustered in isolated patches. The distribution pattern of thermal landscapes evolved from central clustering to a global spread, with a development trend leaning towards the Guancheng District.

Figure 3 also includes the annual average LST and solar radiation data for both the study area and rural regions. Although the yearly variations are minimal, the data consistently show that urban areas experience higher surface temperatures and greater solar radiation compared to rural areas, particularly during the summer. This pattern highlights the impact of urbanization on local climate, with the urban areas showing more intense heat and higher solar radiation due to the concentration of buildings, impermeable surfaces, and reduced vegetation.

This consistent difference between urban and rural areas underscores the significance of urbanization in exacerbating the urban heat island effect. The presence of vegetation and water bodies continues to play a crucial role in moderating temperatures by creating cooling patches. This combined analysis of thermal landscapes, LSTs, and solar radiation data provides a comprehensive understanding of the evolving spatial patterns of the urban heat island effect in Zhengzhou over the 20-year study period.

Considering the different development speeds and planning directions of each district, the areas of thermal landscape grades were statistically analyzed based on administrative divisions for each period to reflect the spatial distribution of urban heat islands. This is shown in Figure 4.

In 2000, thermal patches in Zhengzhou were mainly distributed in the southern urban areas of the Erqi District and Guancheng District, while cooler patches were concentrated in the Huiji District, Zhongyuan District, and Jinshui District. Thermal patches in the Erqi District and Guancheng District exceeded 50%, primarily classified as normal and weak heat island areas. In contrast, the Huiji District, Jinshui District, and Zhongyuan District were predominantly classified as normal and weak cold island areas. By 2004, thermal patches in the central area of Zhengzhou were concentrated, accounting for only 27.21%, but showing a distinct clustering phenomenon. Thermal patches in the Erqi District transitioned from weak cold island areas to weak heat island areas, accounting for 25%; weak heat island areas in the Guancheng District surpassed normal areas, becoming predominant at the highest proportion; the Jinshui District shifted to predominately weak cold island areas; the Zhongyuan District saw a significant decrease in weak cold island areas. In 2009, thermal patches across Zhengzhou accounted for 30.95% in total, with increases in all districts except the Huiji District. The Guancheng District had the highest proportion at 45%. The increases in these four districts were mainly due to a reduction in differences between weak heat island areas and normal areas. By 2013, the proportion of thermal patches peaked at approximately 33.92%, showing a severe clustering phenomenon. The Huiji District transitioned from weak cold island areas to normal areas, with an increase in weak heat island areas. The Erqi District, Jinshui District, and Zhongyuan District saw significant increases in weak heat island areas, while weak heat island areas in the Guancheng District became predominant. From 2013 to 2020, the proportion of thermal patches decreased to 31.57%, primarily concentrated in the construction areas in the city center. Weak heat island areas decreased in all districts, while weak cold island areas increased in the Huiji District and Zhongyuan District. In summary, the spatiotemporal evolution of the urban heat island effect in Zhengzhou is influenced by urban development and changes in land use, demonstrating a trend of gradual clustering and later stabilization.

According to the spatial distribution maps of heat island patches and urban development gravity (Figure 5) from 2000 to 2020, the heat island patches and urban development gravity in the Zhongyuan District collectively shifted northwestward. From 2000 to 2004, the gravity of heat island patches moved towards the city center, while the gravity of urban development shifted minimally, indicating limited urban expansion. From 2004 to 2020, their gravities overlapped, indicating that urban expansion was a key factor exacerbating the heat island effect. In the Erqi District during the study period, the gravity of heat island patches moved towards the city center, while the gravity of urban development shifted southwestward. From 2000 to 2004, the gravity of heat island patches moved northeastward, primarily influenced by changes in agricultural land vegetation. From 2004 to 2013, their gravities coincided, with urban development continuing to expand southwestward, while heat island patches moved towards the city center, indicating the effect of multiple factors on the heat island effect. The Huji and Jinshui Districts showed similar trends in the migration of heat island patches and urban development gravity to the Erqi District, both moving towards the city center, but with irregular directions for urban development migration. The Huji District’s urban development moved northwestward, while the Jinshui District expanded outward. The urban development gravity of the Guancheng District shifted southeastward, while the gravity of heat island patches initially moved southeastward and later shifted northeastward. The results of the above studies show that there is a close relationship between the heat island effect and urban development and land use change.

These different migration patterns and distances reflect the unique characteristics of each region in the urbanization process, as well as the specific planning decisions or geographical constraints they may face. Furthermore, these observations underscore the important role of urban planning in mitigating the urban heat island effect and promoting sustainable development. Effective planning of urban spatial layout and land use can efficiently control the urban heat island effect, promoting urban ecological balance and environmental quality.

3.3. Analysis of Factors Affecting Urban Heat Island Intensity

Before conducting the analysis, we first used the Fishnet tool in ArcGIS 10.6 to create a 300 m × 300 m grid, generating a regular grid and grid points covering the study area. Given that the independent variables in the geographical detector are categorical, we reclassified the 10 factors mentioned using the Jenks classification method into six classes, consistent with land use categories [36,37]. This method is widely used in discretizing factors affecting the urban heat island effect. Subsequently, we extracted the surface temperature and attribute values of each influencing factor into the same grid for further analysis.

The q-values for each influencing factor are shown in

Table 4. Influence level of each factor.

Factors Influencing	NDBI	NDVI	LULC	MNDWI	NL	SAVI	RD	PD	ELE	AOD
Levels of Influencing Factors (q-value)	0.427	0.413	0.383	0.312	0.301	0.295	0.211	0.163	0.080	0.030

. All 10 factors significantly impact the spatial heterogeneity of surface temperature (p < 0.05). In terms of their explanatory power over surface temperature, the factors rank as follows: NDBI (0.427) > NDVI (0.413) > LULC (0.383) > MNDWI (0.312) > SAVI (0.295) > NL (301) > RD (0.211) > PD (0.163) > ELEVATION (0.080) > AOD (0.030). This indicates that buildings and vegetation have the greatest overall spatial impact on surface temperature distribution, with explanatory powers exceeding 41%. Land-use types follow, while elevation and aerosols have relatively minor explanatory power on surface temperature. Overall, changes in the spatial heterogeneity pattern of urban heat island effects in the city area are largely influenced by buildings and green vegetation.

4. Discussion

4.1. Impact of Urbanization on Urban Heat Islands

The urban heat island effect is closely related to the radiation and thermal characteristics of infrastructure, such as built-up urban land and impervious surfaces. There is a high correlation between solar radiation and surface temperature, and the urban heat island effect can be effectively quantified using surface temperatures from satellite inversion. From 2000 to 2020, the average surface temperature in Zhengzhou City shows a wave-like fluctuation trend, and the magnitude of the temperature change gradually increased. Over time, the areas with lower temperatures were mainly concentrated along the Yellow River and in the reservoir area, while the areas with higher temperatures gradually shifted from the barren rocky and sandy areas in the early period to the urbanized areas. The results of the 2000 study show that the areas with high values of temperature were concentrated in the periphery of the city, while the temperatures in the central urban area were relatively low. This is because at that time, the old urban areas of Zhengzhou were mainly concentrated in the Erqi and Guancheng Districts, while the unused land around other new districts, such as rocky and sandy areas, resulted in a temperature distribution characterized by high temperatures at the edges and low ones in the center. With the continuous development of the city, these rocky and sandy lands were gradually developed and utilized, and the center of gravity of urban construction shifted to the central area, resulting in a distribution pattern of high temperatures in the middle and low temperatures at the edges.

In cities, built-up land and impermeable surfaces gradually replaced vegetation, agricultural land, and waters, and these changes resulted in much higher temperatures in cities than in rural areas. Due to differences in land cover type, surface characteristics, and intensity of human activities, urban and rural areas show significant differences in surface energy balance, which leads to differences in surface temperature between them. Studies have shown that the intensity of the urban heat island effect is closely related to the size of the city. The expansion of the urban area of Zhengzhou City plays an important role in the enhancement of the urban heat island effect, and the building land becomes the main factor driving the increase in surface temperature.

However, as urban development gradually stabilizes, the government’s efforts to mitigate the urban heat island effect become more apparent. The gradual mitigation of the heat island effect through the implementation of environmental management measures is reflected in the shift from a centralized hotspot to a decentralized pattern in strong heat island areas. Subject to specific planning decisions and geographical constraints, the pattern and distance of heat island center migration varies from region to region, highlighting the unique characteristics of each region in the process of urbanization.

4.2. Analysis of Contributing Factors to UHI Intensity

Factors affecting the intensity of the urban heat island effect include the type of land cover, surface materials, and the intensity of human activities. In Zhengzhou City, the natural landscape has gradually been replaced by built-up land and impervious surfaces, which significantly changes the balance of surface energy, leading to an increase in surface heat storage and a rise in temperature. In addition, the density and height of buildings, the area of green space, and the distribution of water bodies are also key factors influencing the intensity of the urban heat island. Solar radiation plays a crucial role in this, as areas under direct sunlight usually exhibit a more significant heat island effect due to its influence.

The interaction between surface temperature and solar radiation reveals a complex relationship, with urban areas being able to absorb and re-radiate heat more efficiently than rural areas due to their higher surface heat capacity, thus exacerbating the heat island effect. In addition, socio-economic factors such as population density, industrial activity, and expansion of transportation networks further amplify the heat island effect by increasing anthropogenic heat emissions and altering natural wind patterns. Understanding these contributing factors is essential for developing effective strategies to mitigate the urban heat island effect and promote sustainable urban development.

5. Conclusions

Zhengzhou, as a national-level central city, has undergone rapid development in recent decades. The urban built-up area of Zhengzhou has significantly expanded, leading to notable thermal environment issues associated with urbanization expansion. This study utilized Landsat satellite imagery from 2000, 2004, 2009, 2013, and 2020, along with GEE and various publicly available datasets, to compute LST and land-use types in the study area. Based on these results, the study analyzed the thermal environment of the region and drew the following conclusions: (1) From 2000 to 2020, Zhengzhou’s land surface temperature exhibited a fluctuating upward trend, primarily influenced by the expansion of built-up areas and urban planning. Spatially, cooler areas were mainly located around water bodies, while hotter areas were predominantly distributed within the built-up areas, displaying significant urban heat island effects. However, as urban development stabilized over time, government-enacted environmental measures have somewhat alleviated the overall urban heat island effect. (2) Further analysis revealed that buildings are the primary factor influencing the spatial distribution of LSTs, explaining up to 42.7% of temperature variations. Vegetation follows as the second most significant factor, with land-use types playing a slightly lesser role. Factors such as elevation and aerosols have comparatively minor effects on the LST. Therefore, to mitigate the urban heat island effect, there should be a focus on urban planning for built-up areas and the conservation and expansion of green vegetation to improve urban environmental quality.