Exploring the Industrial Heat Island Effects and Key Influencing Factors in the Guangzhou–Foshan Metropolitan Area

Abstract

Industrial parks are key contributors to localized urban heat intensification, forming sub-industrial heat islands (IHIs) that influence the urban thermal environment. This study investigates the industrial heat island effect (IHIE) in the Guangzhou–Foshan metropolitan area (GFMA), with a generalized additive model (GAM) to explore the influence of park spatial patterns and land cover characteristics, using indicators such as industrial heat island intensity (IHII), industry warming area (IWA), industry warming efficiency (IWE), and industry warming gradient (IWG). The results show that (1) industrial land significantly contributes to industrial heat islands (IHIs), with heat extending up to 200 m into surrounding areas. (2) IHIE intensity varies notably across park types, with each dominated by different factors: manufacturing parks by landscape shape index (LSI); comprehensive parks by impervious surfaces (IWS) and internal building land (IB); and special parks primarily by IB. (3) In most industrial parks, park area (S), IWS, and LSI are the key factors affecting IHIE. As IWS increases, IHIE strengthens, though this trend can be mitigated by expanding park area. Conversely, a higher LSI weakens IHIE. (4) Several variables, including arable land (AL) and water body (WB), exhibited nonlinear or threshold effects, suggesting that IHIE is shaped by complex mechanisms. These findings offer valuable insights for optimizing land use in urban and industrial planning to reduce IHIE and promote sustainable urban development.

1. Introduction

The urban heat island (UHI) effect, defined as a human-induced urban climate phenomenon characterized by higher temperatures in urban areas than in their rural surroundings [1], has intensified under rapid urbanization and industrialization since the beginning of the 21st century [2,3,4]. Recently, the increase in urban heat island intensity (UHII) and extreme high temperature events worldwide have not only altered local climates and air quality but also adversely affected public health, increasing both morbidity and mortality rates [5,6,7,8]. Consequently, understanding the driving factors behind the UHI effect and developing effective mitigation strategies are critical for sustainable urban development.

Existing research has demonstrated that UHI is closely related to surface biophysical composition and configuration [9,10,11], local climate background [12], and socio-economic driving factors [13,14]. These studies have significantly improved our understanding of the physical mechanism behind UHI formation and evolution, particularly the close relationship between UHII and the distribution of land cover types, a connection that is well documented and widely recognized in the literature. However, as cities expand, the intensity distribution of UHI becomes more heterogeneous, particularly in urban agglomerations where individual heat islands merge to form regional heat clusters [15,16,17]. While these studies have improved our understanding of UHI dynamics and mitigation, they often overlook the contribution of industrial zones, despite their potential to create localized “industrial heat islands” [18,19]. In contrast to the well-studied “cold islands” such as green spaces and water bodies [20,21,22], these high-temperature zones have received limited attention, despite their significances in understanding UHI influencing factors and formulating mitigation measures.

Anthropogenic heat emissions from industry, transportation, and construction are key contributors to UHI [23,24,25]. It is also well known that industrial areas, especially those with high industrial density and complex land use patterns, can intensify local warming, adversely affecting labor productivity, public health, and economic stability [26,27]. Some research has addressed the presence of IHIs and examined factors such as industrial type and park size [8,26,28,29], but these studies have mainly focused on industrial or resource-based cities and have largely neglected urban agglomerations. As a result, the role of industrial land use in shaping the spatial and temporal evolution of UHIs at a finer scale remains poorly understood.

The Guangzhou–Foshan region, located in the core of the Pearl River Delta (PRD) in Guangdong Province, China, provides an ideal setting to address this knowledge gap. This area, with its subtropical climate, is particularly vulnerable to heatwaves intensified by global warming and urbanization [30]. At the same time, Guangdong is a major industrial hub with a large number of industrial parks. Since the 2000s, industrial relocation and upgrading policies have led to a redistribution of industrial activities from central urban areas to peripheral zones [31,32]. These shifts significantly influence urban thermal landscapes, resulting in more pronounced internal differentiation of UHIs and the emergence of extremely high-temperature zones. Although previous studies have examined land use change, urban expansion, and economic development in the PRD’s UHI patterns, they often treated UHIs at a macro-scale, neglecting the micro-scale patterns of industrial land-induced thermal anomalies [16,33,34,35,36]. This omission hampers a detailed understanding of urban internal thermal heterogeneity and constrains the formulation of targeted UHI mitigation measures.

To address these gaps, this study focuses on Guangzhou–Foshan region and investigates the formation and influencing factors of IHIs. Specifically, we examine how industrial park configurations (via landscape metrics), land cover characteristics inside and around these parks, and different industrial park types affect the intensity and spatial distribution of IHIs. By elucidating these relationships, our research aims to enhance the understanding of how industrial land use influences urban thermal environments, thereby providing valuable insights for industrial land planning, environmental management, and the development of effective UHI mitigation strategies in rapidly urbanizing regions.

2. Materials and Methods

2.1. Study Area

The Pearl River Delta Metropolitan Region (PRDMR) and its adjacent areas are located in the coastal region of South China, which exhibits a subtropical monsoon climate. It is warm and humid all year round, with an average annual temperature of 20–24 °C (http://ncc-cma.net/cn). With its comfortable and pleasant climate and unique geographical location, since the “Reform and Opening-up” policy in 1978, PRDMR has played a leading role in China’s economic and social development and developed into one of the most economically advanced regions in China and the most densely populated area globally [16,17,37]. In 2015, the China State Council in its 13th Five-Year Plan for Economic and Social Development strategically initiated the Guangdong–Hong Kong–Macau Greater Bay Area, with emphasis on strengthening its role in economic development [38]. PRDMR has rapid economic growth, and is a new processing industry base in China, with a large number of electronic manufacturing and assembly factories. The region is expected to evolve into a more developed urban agglomeration, possibly leading to more complex IHIE characteristics. Our study area includes Guangzhou and Foshan, which is the core area of the PRD (Figure 1).

Guangzhou is located on the northern edge of the PRD and near the estuary of the lower Pearl River Basin. As the capital of Guangdong Province, Guangzhou has long been a central hub for politics, economics, culture, and education in South China. Foshan is a typical transitional manufacturing city once known as the “world factory” in the developed southern coastal region of China, with an administrative area of 3797 km² and a population of approximately 9 million [39].

The Guangzhou–Foshan metropolitan area (GFMA) is a core development circle of the PRDMR. The two regions are geographically close, share common roots, and have strong economic ties, making the development of the GFMA closely related to the overall growth of the PRDMR [40]. With the rapid development of industry and urbanization, some environmental problems have emerged in PRDMR, and the UHI effect is one of them [33,41]. Therefore, this study selected GFMA as a typical representative of PRDMR to analyze the influencing factors and mechanisms of IHIE.

2.2. Data Source

Four main datasets were used in this study: land surface temperature (LST) data, land use and land cover data, building rooftop area data, and industrial park boundary data.

We obtained 30 m resolution Landsat Level 2 LST data from the United States Geological Survey (USGS) Earth Explorer platform (https://earthexplorer.usgs.gov/ (accessed on 16 January 2022)). The data were acquired on 22 October 2020 (path 122, row 44), covering the Guangdong Province region. Landsat-based LST is considered high-quality and is widely used for urban heat studies [17,42].

We utilized the annual China Land Cover Dataset (CLCD) created by [43] (https://doi.org/10.5281/zenodo.4417810 (accessed on 17 January 2022)), which has a 30 m spatial resolution and provides land cover information from 1990 to 2020. CLCD was generated using multi-temporal training samples, Landsat-based temporal metrics, and a random forest classifier. Validation against independent test sets indicated an accuracy of 79.31%, outperforming other similar products (MCD12Q1, ESACCI_LC, FROM_GLC, and GlobeLand30). After comparing the CLCD dataset with high-resolution Google Maps imagery to verify data accuracy, we classified the land use and cover (LUC) in the study area into five categories: farmland, vegetation (including forest and grass), water, impervious surface, and bare land.

Data on building rooftop areas were obtained from the China Building Roof Area Dataset (CBRA), a 2.5 m resolution dataset produced by [44] using Sentinel-2 images from 2016 to 2021. CBRA is the first full-coverage, multi-annual BRA dataset for China. By employing advanced training-sample-generation algorithms and spatiotemporally aware learning strategies, CBRA achieved an F1 score of 62.55% (improving by 10.61% compared to previous BRA data) based on 250,000 urban samples and a recall of 78.94% based on 30,000 rural samples.

Industrial park boundaries were manually delineated from Google Maps satellite imagery, supplemented by expert knowledge. To ensure reliability and representativeness, we selected industrial parks meeting the following criteria: (1) an area > 60,000 m²; (2) internal green space proportion ≤ 50%; and (3) buildings coverage ≥ 20%. After screening, 94 industrial parks were retained (Figure A1). These parks represent various types, including heavy industrial, light industrial, innovative industrial parks, and so on. Considering that industrial parks differ widely in industrial structure, environmental impact, and management modes [19], we applied a more nuanced classification scheme: a function-oriented classification reflecting building type and park function, categorizing parks into Manufacturing Industrial Parks, Logistics and Trade Industrial Parks, Comprehensive Industrial Parks, and Specialized Industrial Parks (

Table A1. Classification of manufacturing industrial parks, logistics and trade industrial parks, comprehensive industrial parks, and special industrial parks.

Park Type	Park Name	Main Business
Manufacturing	BSC	Hardware (aluminum), electrical appliances, plastic, textile, more comprehensive
	LT	Plastic-based, electronic, and hardware
	JY	Hardware, mechanical machinery factory, hardware innovation center
	CJX	Main hardware, secondary plastic
	LH	Hardware, electronics
	XYKJ	Main hardware, plastic
	SC	Main hardware, electrical appliances
	HGB	Hardware (aluminum), rubber and plastic
	HLF	Hardware (aluminum), logistics
	CCGL	Hardware big factory, have food, electrical appliances
	JZJXL	Hardware factory, small plastic factory
	XB2	Large hardware
	NL	Hardware and plastic appliances
	gyy4	Hardware, electrical appliances, food, more comprehensive
	RT	Main hardware, electrical appliances
	PNPZ	Hardware mainly, there is a shipyard
	MC	Hardware (main steel, with aluminum)
	gyy10	Power grid, a water purification company, you have a little hardware
	XJ	Hardware and plastic appliances
Logistics trade	DDGC	Metal logistics city, trading city
	DAYUAN	Logistics company, advertising company, metal material recycling company
	LC	Sell furniture, send logistics
	gyy8	Sell and make ceramic building
	HH	Refrigeration equipment big factory, horticulture trading city
	XNY	Caohe Industrial zone (main hardware), farmers’ market, automobile trading company
	LYJJ	Sell furniture, hardware, hair logistics
	JCSY	Fuhua Machinery City, Xinbao electrical appliance factory, logistics city
	XX	Plastic factory, logistics city
	HHLX	Main car maintenance (logistics), stationery plastic
Special	ANWL	Agricultural and sideline products trading market
	gyy5	more integrated
	NG	Plastic electromechanical daily necessities
	gyy9	Ceramic building materials
	FXG	Tinfoil
	HM	Warehouse
	GZJY	Guangzhou cigarette factory
	ZC	Electrical factory
	GZGQ	Auto parts, stand in daylight
	HS	Textile, food (with electrical appliances nearby)
	gyy7	Pure hardware factory
	gyy3	Plastic, electronic
	JC	Pure hardware is available near the airport
	MY	Wood trade (gathering)
	DC	Electrical, metal products, printing, integrated type
	NH	Hardware-based, plastic
	LB	Hardware and plastic electrical appliances, there is a food factory
	SX	Hardware, plastic, selling lamps (gathering)
	QY	Hardware, electrical appliances, plastic, paint
Comprehensive	DS	Pure hardware factory
	YB	Pure hardware factory
	CH	Electrical appliances + chemical industry, with electroplating plant (small)
	WSB	Electrical factory
	WLY	Auto parts (small), accessories
	gyy6	Concrete, the brick and tile company
	HGWJ	machine tool
	MAG	Aluminum factory
	DF	Hardware & plastics 1:1
	QC	Automobile general factory maintenance, involving parts assembly and manufacturing
	LZS	Hardware (aluminum), it has a toy factory
	HG	Main printing and dyeing clothing, with ceramics, power generation
	HGA	Rubber, building materials
	SCHD	Hardware industrial park
	XH	Electronic-mechanical robot
	MDX	electrical equipment
	CYY	Pure hardware factory
	MG	Coating small factory
	PTDL	Aluminum factory
	SYQ	Appliances, printing, and locks
	LL	Main electrical appliances, secondary plastic
	LTXL	Hardware, electrical appliances
	XHZY	Hardware, electrical appliances
	DY	Hardware and furniture manufacturing 1:1
	YUX	Aluminum factory
	YX	Pure hardware factory
	XY2	Hardware, there is a steel chemical plant
	TC	Precision machinery, with a ceramic foaming factory

).

2.3. Methods

The integrated datasets in Section 2.2 form the foundation for analyzing industrial heat islands and understanding how land cover composition, building density, and park typologies contribute to urban thermal patterns. After calculating the IHIE indices (IHII, IWA, IWE, and IWG), park pattern indices, and different land cover areas and proportions, we used GAM to develop a framework to examine the mathematical interactions among key drivers (Figure 2). These methods provide a solid foundation for understanding the factors that influence IHIE. The following section outlines the methodologies used in this study.

2.3.1. Classification of Urban Surface Thermal Environment

To characterize anomalous thermal conditions in the city, we applied a classification approach following [45], incorporating two statistical metrics: mean surface temperature and temperature standard deviation [9,46]. Based on these metrics, LST was classified into five categories: L1 (extremely low), L2 (low), L3 (normal), L4 (high), and L5 (extremely high) (

Table 1. LST classification standard.

LST Level	Level Name	Level Range
L1	Extremely Low	∆T ≤ −1.5σ
L2	Low	−1.5σ ≤ ∆T < −0.5σ
L3	Normal	−0.5σ ≤ ∆T ≤ 0.5σ
L4	High	0.5σ < ∆T ≤ 1.5σ
L5	Extremely High	1.5σ ≤ ∆T

), enabling intuitive identification of urban hotspots and their overlap with industrial parks.

2.3.2. Quantifying the Industrial Heat Islands (IHIs) Effect

The selection of appropriate indicators is crucial for quantifying the IHIE of industrial parks. To achieve this, a 3600 m wide buffer zone was established around each industrial park and subdivided into 60 segments, each 60 m wide. The average LST of each buffer zone was calculated to construct LST profiles, enabling a quantitative assessment of the IHIE. As shown in Figure 3, LST decreases with increasing distance from the industrial park, with a turning point beyond which the temperature gradually stabilizes.

Based on the observed LST profiles and previous studies [25,47,48], we identified four key indicators to quantify the IHIE: Industrial Heat Island Intensity (IHII) was defined as the difference in mean LST between the industrial park and the first turning point [25]; Industrial Warming Area (IWA) referred to the maximum area experiencing warming when the first turning point appears within the buffer zone surrounding the industrial park; Industrial Warming Efficiency (IWE) was defined as the ratio of the maximum warming area to the industrial park area, reflecting the spatial extent of the IHIE; and Industrial Warming Gradient (IWG) referred to the ratio of LST reduction to the warming distance, representing the accumulated temperature increase per unit distance. The specific calculation formulas are as follows:

$$\mathrm{I}\mathrm{H}\mathrm{I}\mathrm{I}={\mathrm{T}}_{\mathrm{I}}-{\mathrm{T}}_{\mathrm{D}},$$

(1)

$$\mathrm{I}\mathrm{W}\mathrm{A}={\mathrm{S}}_{\mathrm{b}\mathrm{u}\mathrm{f}\mathrm{f}\mathrm{e}\mathrm{r}},$$

(2)

$$\mathrm{I}\mathrm{W}\mathrm{E}={\displaystyle \frac{{\mathrm{S}}_{\mathrm{b}\mathrm{u}\mathrm{f}\mathrm{f}\mathrm{e}\mathrm{r}}}{{\mathrm{S}}_{\mathrm{i}\mathrm{n}\mathrm{d}\mathrm{u}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{y}}}},$$

(3)

$$\mathrm{I}\mathrm{W}\mathrm{G}={\displaystyle \frac{\mathrm{D}\times {\mathrm{T}}_{\mathrm{d}}-{\int }_0^{\mathrm{D}}{\mathrm{T}}_{\mathrm{d}}\mathrm{d}\mathrm{l}}{\mathrm{D}}},$$

(4)

where D is the distance between the first turning point and the park; ${\mathrm{T}}_{\mathrm{I}}$ is the average surface temperature inside the industrial park; and ${\mathrm{T}}_{\mathrm{D}}$ is the temperature at the first turning point. ${\mathrm{S}}_{\mathrm{b}\mathrm{u}\mathrm{f}\mathrm{f}\mathrm{e}\mathrm{r}}$ is the total buffer zone area up to the first turning point. ${\mathrm{S}}_{\mathrm{i}\mathrm{n}\mathrm{d}\mathrm{u}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{y}}$ is the area of industry park. L is the distance from the edge of the park to the first turning point.

2.3.3. Potential Explanatory Factors

Many studies have shown that landscape pattern indices and land use types have an impact on urban LST [49,50,51]. Therefore, park pattern index and land use type were used as research factors in this study.

Two commonly used landscape pattern indices, namely S and LSI, were used to quantitatively analyze the indices associated with the IHIE. Among them, the LSI was calculated as follows:

$$\mathrm{L}\mathrm{S}\mathrm{I}={\displaystyle \frac{P}{2\sqrt{\pi \times S}}},$$

(5)

where P and S represent the total perimeter and area of the urban park, respectively. The larger the value of LSI, the greater the irregularity of landscape shape [52].

Land use type is also considered to be an important factor affecting the UHI [53,54]. This study calculated the area percentage of the following five land use types within each park: IWS, IB, AL, WB, and forest, shrub, and grass land (FSG), as previous studies have shown that they significantly impact the UHII [17,28,55].

2.3.4. Generalized Additive Model (GAM) Application

To investigate the relationship between IHIE indices and various industrial park characteristics, we employed the Generalized Additive Model (GAM), which is well suited for capturing both linear and nonlinear relationships. GAM has been widely applied in environmental studies due to its flexibility in modeling complex interactions [56,57,58]. In this study, four IHIE indices (IHII, IWA, IWE, and IWG) were used as dependent variables. The independent variables included park layout indices, internal land use types, and their interactions. The GAM formula used in our analysis can be expressed as Equation (6):

$$\begin{gathered} \mathrm{I}\mathrm{H}\mathrm{I}\mathrm{E}\mathrm{i}\mathrm{n}\mathrm{d}\mathrm{i}\mathrm{c}\mathrm{e}\mathrm{s}={\mathsf{\beta }}_0+\mathrm{s}\left(\mathrm{S}\right)+\mathrm{s}\left(\mathrm{L}\mathrm{S}\mathrm{I}\right)+\mathrm{s}\left(\mathrm{I}\mathrm{W}\mathrm{S}\right)+\mathrm{s}\left(\mathrm{I}\mathrm{B}\right)+\mathrm{s}\left(\mathrm{A}\mathrm{L}\right)+\mathrm{s}\left(\mathrm{W}\mathrm{B}\right)+\mathrm{s}\left(\mathrm{F}\mathrm{S}\mathrm{G}\right)+\mathrm{t}\mathrm{i}(\mathrm{I}\mathrm{W}\mathrm{S},\mathrm{S})+\mathrm{t}\mathrm{i}(\mathrm{I}\mathrm{W}\mathrm{S},\mathrm{I}\mathrm{B}) \\ +\mathrm{t}\mathrm{i}(\mathrm{I}\mathrm{W}\mathrm{S},\mathrm{W}\mathrm{B})+\cdots +\mathsf{ϵ} \end{gathered}$$

(6)

where ${\mathsf{\beta }}_0$ is the intercept; s(x) denotes a smooth spline function applied to variable x, capturing its nonlinear effect; ti (${\mathrm{x}}_1$, ${\mathrm{x}}_2$) represents the interaction term between variables ${\mathrm{x}}_1$ and ${\mathrm{x}}_2$; and $\mathsf{ϵ}$ is the random error term. This approach allows us to capture nonlinear relationships between IHIE indicators and explanatory variables while accounting for potential spatial heterogeneity among industrial parks.

3. Results

3.1. Impact of Industrial Parks on the Urban Thermal Environment

The results reveal a strong spatial overlap between thermal hotspots and industrial park boundaries in the GFMA (Figure 4A), with 66% of extreme high-temperature zones and 31% of high-temperature zones located within park areas (Figure 4B). Additionally, the average temperature of industrial parks exceeds that of surrounding urban areas (Figure 4C), which indicates that industrial parks are islands with abnormal high temperature inside the UHI. These sub-industrial heat islands are likely driven by intensive energy use, impervious surfaces, and internal heat emissions and become a sub-industrial heat island in the UHI.

To verify whether industrial parks form new industrial heat islands inside the UHI, we analyzed the LST profiles of all industrial parks. Our findings revealed that 75 out of 94 industrial parks exhibited a significant temperature decline from the industrial park’s center to its surroundings, accounting for over 75% of the cases; our follow-up study was conducted for these 75 parks.

To illustrate typical warming patterns, we selected 10 representative parks of various types (Figure A2). We analyzed the LST differences between the industrial parks, their buffer zones, and the city’s natural surface (Figure 5). The results indicate that industrial parks generally exhibit higher temperatures than various land types. Additionally, the buffer zones surrounding industrial parks also show elevated temperatures compared to other urban areas, highlighting the warming effect of industrial activities on the local thermal climate.

3.2. Industrial Heat Island Effects

To quantify IHIE intensity and its spatial extent, we analyzed the average LST profiles of 75 identified IHI parks (Figure 6a) and calculated four IHIE indices (Figure 6b–e). The results show that the LST profiles of industrial parks consistently exhibit a downward trend from the park center to the surrounding areas across all temperature indicators, with the average warming influence extending approximately 200 m outward.

We conducted Shapiro–Wilk tests to assess the normality of the four IHIE indicators across the 75 industrial parks. The results showed that IHII (W = 0.993, p = 0.952) and IWG (W = 0.974, p = 0.119) followed a normal distribution. In contrast, IWA (W = 0.516, p < 0.001) and IWE (W = 0.371, p < 0.001) significantly deviated from normality, indicating non-normal distribution patterns (Figure 6b–e). The variations in the distributions of these indices may be attributed to the diverse types of industrial parks analyzed.

Given the variability in industrial park functions and spatial configurations, this study analyzes IHIE across park types separately. Significant differences were observed in IHIE indices (Figure 7) and internal characteristics (Figure A3). While IHII values were relatively consistent, manufacturing parks exhibited lower IWA, suggesting more concentrated thermal output. Special parks showed higher IWG, likely due to greener buffers and lower structural density. Comprehensive parks had the highest IWE, indicating stronger spatial spread of heat. These differences underscore the importance of type-based classification when analyzing IHIE drivers.

3.3. Drivers of the Industrial Heat Island Effects

In order to explore what factors (S, LSI, IWS, IB, AL, WB, FSG, and their interaction terms) affect IHIE in different types of parks (manufacturing parks, logistics trade parks, comprehensive parks, and characteristic parks), we calculated the mean p-value of all IHIE indicators versus each impact factor for the different types of parks in the model. Only a few factors were selectively retained to avoid multicollinearity (variance inflation factor < 5).

The models were evaluated using 10-fold cross-validation, yielding average prediction errors around 0.7, which are significantly lower than the corresponding index means (typically between 4 and 6). The relatively small standard deviations indicate consistent model performance across different data partitions. These findings demonstrate that the category-based modeling approach provides robust and generalizable results in analyzing IHIE.

The results (Figure 8) showed that, manufacturing parks are most influenced by LSI, highlighting the role of spatial configuration (Figure 8a). For logistics trade parks, the correlation of all factors is close, except IWS and ti (FSG and IWS) (Figure 8b), which can be attributed to the different planning and construction standards of logistics parks. In comprehensive parks, IWS (p = 0.0344) and IB (p = 0.0395) are highly correlated with the IHIE (Figure 8c). Additionally, in special parks, the IB is strongly correlated with the IHIE (Figure 8d), emphasizing the role of built-up areas in exacerbating the IHIE. The variation in p-values among different park types emphasizes the critical role that green infrastructure, landscape configuration, and impervious surface distribution play in influencing the IHIE. These findings highlight the importance of further investigation into how these factors influence IHIE.

In this study, the factors and IHIE indices of each p-value < 0.05 in the model were analyzed (Figure A4). First, there was a positive correlation between IHII and IWS in different types of parks, which was most obvious in manufacturing parks (p = 0.04) (Figure 9a). Second, LSI was strongly correlated with each index, we found that LSI was negatively correlated with all indices (with one exception of IWG in the manufacturing parks) (Figure 9b). The interaction terms of ti (S, IWS) showed good correlation with multiple indices and influence forms are similar (Figure 9c,d). Higher IWS increased IHIE, while larger park area mitigated it—possibly due to the cooling effects of open space and green infrastructure.

Additionally, although some variables did not reach statistical significance (p > 0.05), they exhibited distinct and potentially meaningful nonlinear patterns (Figure A4). For instance, LSI influenced IWE in a threshold-dependent manner, WB showed a U-shaped effect, and IB demonstrated a saturation curve, peaking at moderate values. Ti (AL and S) showed a saddle-shaped surface where IWE peaked at moderate park area and industrial land area (Figure A4E), suggesting a potential trade-off effect between park size and functional land intensity. These patterns deviate from linear assumptions commonly seen in prior studies and point to complex underlying mechanisms.

4. Discussion

4.1. Dominant Driving Factors of IHIs Vary by Park Type

Although previous studies have shown that some large industrial parks have a sub-heat island effect [18,19,59], our study found that more than 75% of the industrial parks, spanning various types and sizes, exhibited a significant IHIE. This widespread occurrence suggests that, beyond general anthropogenic heat emissions, certain structural or spatial characteristics of industrial parks may play a key role in amplifying surface thermal radiation [18,25,60,61]. In addition to this, we found that the average warming distance of the industrial park was about 200 m. It can provide effective guidance for the future planning of the park scope.

Referring to previous studies, the park pattern and land use type in the park were considered as possible influencing factors [29,34,62]. In addition, different composition types and proportional relationships can produce different thermal environments [59]. Our results revealed clear variation in IHIE drivers across park types. In manufacturing parks (the main industries in the park are hardware and steel mills), IHIE was strongly linked to LSI, consistent with studies on industrial morphology [59]. Logistics parks usually contain large areas of roads, warehouse roofs, parking lots, and other hard surfaces, with more IWS and strong heat absorption capacity [63]. As a result, for logistics trade parks, ti (S and IWS) has a strong correlation with IHIE indices. Comprehensive parks usually contain buildings with high density [64]. The height, material, and density of buildings will affect the thermal environment, which may be the cause mainly affected by IWS and IB. Due to the particularity of the building form and layout of the characteristic park, special parks are mainly affected by IB, while the role of IWS is relatively weak.

In addition, this study found that IHII was positively correlated with IWS in different types of parks, which was consistent with the results of previous studies [10,65]. LSI is negatively correlated with most indices, which is consistent with the conclusions of some studies focusing on urban form and industrial type [55,66,67]. However, we found that LSI of manufacturing parks was positively correlated with IWG, which is different from the previous studies. The planning of manufacturing parks often focuses on functional intensification and space utilization to improve land use efficiency. This compact layout can increase the complexity of the landscape pattern, which is manifested by high LSI values. The more irregular the park is, the more obvious the spillover effect of the IHI [59].

Interaction and nonlinear effects further revealed complex mechanisms. We found that an increase in the proportion of IWS aggravated IHIE, but the IHIE was alleviated to some extent as the area of the campus increased. Previous studies only found a nonlinear relationship between IWS and UHI [29,68] but did not investigate the internal factors, which were well explained in this study. Although not all variables reached statistical significance, LSI, WB, and IB displayed threshold, U-shaped, and saturation patterns, respectively. These patterns go beyond linear assumptions and highlight the need to consider interactive and nonlinear processes in future IHIE studies.

4.2. Implication for the Urban Management and Planning

Industrial heat emission is an important factor for the formation of UHI. In our study, the IHI does exist in urban agglomeration areas, which is consistent with previous studies [17]. Our research results emphasize the contribution of industrial parks to the UHI in urban agglomerations and analyze the influencing factors of IHIE from the perspective of land use type and park pattern. It is important for better planning and management of urban parks to mitigate UHI.

First of all, the formation of urban agglomerations often involves the adjustment of industrial layout and structure [69], which promotes land use change and changes the spatio-temporal dynamics of heat island [70]. For rapidly urbanizing areas, incorporating IHIE considerations into regional urban and industrial planning, including the internal structural design of different types of parks, is critical to achieving the sustainable development goals [71]. To address this, urban planners should integrate IHIE risk maps into spatial development frameworks and set IHIE control targets for new or redeveloped industrial parks.

Second, park size and land composition must be considered jointly. While higher IWS is associated with increased heat intensity, this effect can be mitigated by larger park area and sufficient green coverage. However, we acknowledge that expanding land area or allocating large green spaces may not always be feasible due to rigid land-use constraints in industrial zones. In such cases, alternative strategies such as vertical greening, green roofs, internal green courtyards, and buffer green strips can serve as effective substitutes. These space-efficient interventions can be integrated within existing layouts and help reduce localized heat buildup without compromising industrial functionality.

Third, differentiated management of industrial park types is essential. Our results show that comprehensive parks tend to exhibit the highest IWE, likely due to mixed functions and high human activity levels. In this context, urban policy should promote functional zoning within parks, limit the density of energy-intensive industries, and regulate internal building layout to enhance ventilation and thermal dispersion.

Lastly, planning guidelines should consider interaction effects, not just individual land-use indicators. For example, the compounded impact of IWS and S suggests that heat mitigation cannot rely solely on one factor. Thus, multi-factor assessment tools should be developed to evaluate IHIE risks under different park configurations. These tools can inform adaptive land-use controls and climate-responsive park design strategies.

4.3. Study Limitations and Ideas for Future Research

This study has several limitations that should be acknowledged. First, the data used in this study were confined to LST and CLCD2020 data obtained exclusively from Guangdong Province. The situation may vary in other cities. Increasing the sample size and expanding the study area can be considered in the future.

Second, the spatial resolution of the LST and CLCD data (30 m) may introduce uncertainty in capturing fine-scale spatial variation within industrial parks. Small features such as narrow green belts or roads may be missed or misclassified, leading to potential inaccuracies in land cover assignment and IHIE calculations. In particular, for parks with irregular shapes, boundary mismatches between raster and vector data could affect the precision of LSI extraction. Future studies could explore the use of higher-resolution data or downscaling methods to improve spatial accuracy.

Third, IHI and its drivers were analyzed using data from 22 October 2020 only. Future research should incorporate multi-temporal or seasonal observations to better understand the temporal variability and long-term trends of industrial heat island effects.

Furthermore, the factors that affect the IHIE are extremely complex and diverse, but we only considered land use types for areas and park pattern indices with abnormally high temperatures, without considering other factors that may affect the IHIE within the city. For instance, the impacts of diurnal temperature range and local background climates should be examined, and these can also be investigated in subsequent studies.

4.4. Limitations of Using LST as a Proxy for Surface Heat Intensity

In this study, land surface temperature (LST) derived from remote sensing imagery was used as a proxy to assess surface heat conditions and calculate the IHIE indices. While LST is widely adopted in UHI studies due to its spatial coverage and availability, it has inherent limitations that may affect the accuracy and interpretation of results.

First, LST represents the skin temperature of the Earth’s surface rather than the near-surface air temperature, which means it may not fully capture the heat stress experienced by human populations or reflect the thermal environment at pedestrian level [72,73].

Second, satellite-derived LST can be influenced by surface emissivity, atmospheric conditions, and sensor angle, which introduces uncertainties, especially in heterogeneous urban landscapes like industrial parks [74].

Despite these limitations, LST remains a practical and effective tool for comparative spatial analysis across large areas. However, future studies should consider integrating air temperature data from meteorological stations or sensor networks or applying LST correction methods to enhance the accuracy and representativeness of IHIE indices [75].

5. Conclusions

Using a novel generalized additive model (GAM), this study investigates the impact of industrial heat islands (IHIs) caused by industrial parks on the spatial distribution of urban hot patches in the Guangzhou–Foshan metropolitan area (GFMA) and identifies key influencing factors. Results show that industrial parks exhibit significantly higher land surface temperature (LST), averaging 8.6 °C higher than urban impervious surfaces and 10.45 °C higher than natural landscapes. The warming effect extends approximately 200 m beyond park boundaries, highlighting the significant thermal influence of industrial activities. IHIE intensity varied across different industrial park types due to differences in spatial configurations and land cover. The IHIE of manufacturing parks were mainly affected by landscape shape index (LSI); the comprehensive park was affected by impervious water surface (IWS) and internal building land (IB); special parks are mainly affected by IB. Notably, comprehensive parks exhibited the highest warming intensity, linked to their multifunctional land use and substantial IWS, while special parks displayed relatively moderate effects due to their unique designs that limit heat accumulation. Spatial configuration and land cover emerged as critical determinants of IHIE. Across all park types, IHII showed a consistent positive relationship with IWS, while LSI was generally negatively correlated with IHIE indices, except for IWG in manufacturing parks. Correlation analyses revealed that S and IWS had the most significant influence, with interaction effects indicating that larger parks benefit from increased green space, which mitigates the thermal impact of industrial activities. Furthermore, the study identified several nonlinear relationships and threshold effects, such as saturation patterns for IB and U-shaped trends for water bodies.

This study provides the first large-scale, high-resolution analysis of IHIE across diverse industrial park types. The results emphasize the need for green infrastructure integration, optimized land use planning, and targeted zoning strategies to mitigate IHIs. Future research should assess these patterns in other metropolitan regions, incorporate seasonal variations, and explore landscape-based interventions to reduce IHIE under diverse climatic and socio-economic conditions.