Dynamics of Heat Island Intensity in a Rapidly Urbanizing Area and the Cooling Effect of Ecological Land: A Case Study in Suzhou, Yangtze River Delta

Abstract

Ecological land could play an important function in climate regulation to mitigate urban heat islands (UHIs) and enhance the quality of the living environment. In this work, we chose Suzhou as our case study for urban agglomeration in the Yangtze River Delta (YRD), eastern China. In this city, we analyzed the dynamics of heat island intensity from 2000 to 2021 by retrieving land surface temperature (LST). Subsequently, we examined the relationship between the urban thermal environment pattern and land use change, and finally, we explored the cooling effect provided by ecological land. The results indicated that, in 2000, the city’s UHI effect primarily centered around the central urban region as a singular patch; however, since 2014, the patch UHI effect in the central urban region has been mitigated, and the original small hotspots have converged into a large, contiguous expanse spreading outward. As the shift has occurred from low- to high-temperature zones, the proportion of conversion between ecological land has been decreasing, while the opposite trend has been seen for the proportions of ecological land transferred out and for unchanged artificial surfaces. The normalized difference built-up index was found to be the main contributor to the UHI effect, followed by the normalized difference vegetation index. These findings provide novel insights into the regulation of ecosystem services during urban expansion and offer a reference for improving the function of the cooling effect through urban renewal activities and the optimization of spatial planning.

1. Introduction

The Intergovernmental Panel on Climate Change (IPCC) Sixth Assessment Report shows that climate change directly impacts the functioning of urban systems, significantly modifying their effects, such as the urban heat island (UHI) effect [1,2], which influences urban planning and living quality [3]. The exacerbation of UHIs not only has a very negative impact on residents’ health and the urban environment but also brings a heavy financial burden to urban life [4,5]. Continuous high temperatures cause a city’s production and residential electricity usage, water consumption, and other forms of energy consumption to increase sharply [6]; they also affect the physiological activities of the human body, causing various diseases [7,8,9,10]. At the same time, they reduce the resilience of urban ecosystems, contributing to localized natural disasters and a range of associated problems [11]. To date, numerous researchers at home and abroad have extensively explored the causes [12,13], spatio-temporal distribution [14,15], impacts, and mitigation strategies [16,17] of the heat island effect from diverse perspectives and by applying a range of methodologies. Consequently, the UHI effect has emerged as a prominent and extensively discussed subject in the field of urban climatology. The land surface temperature (LST) significantly influences the energy exchange between the ground and the atmosphere, serving as the primary indicator of the UHI effect, which is generated by material and energy interactions resulting from ground–air interactions. As a single-channel algorithm for retrieving the LST on the basis of thermal infrared remote sensing images, the Qin Zhihao mono-window algorithm [18] stands out thanks to its simplicity and the ease with which atmospheric simulations can be performed without the need for precise real-time atmospheric profile data. Moreover, it exhibits superior accuracy in LST retrieval compared to the generalized single-channel algorithm [19]. Therefore, the mono-window algorithm has been widely adopted in studying UHIs [20,21,22].

As one of the regions with the highest levels of modernization, development, and urbanization in China [23], the urban agglomeration in the Yangtze River Delta (YRD) has traditionally maintained a synergistic balance between a high-quality ecological environment and robust economic development. However, intensive economic development and rapid urban expansion have led to the conversion of increasing areas of ecological land, including cropland, forest, grassland, land for water areas and water facilities, and other such land (including public administration and public service land, as well as transportation land [24]), which is similar to what is known internationally as “blue-green space”, into impermeable surfaces [25]. This transformation has severely fragmented the space structure of ecological land within the YRD urban agglomeration, posing a significant challenge in balancing ecological preservation with economic development and meeting the escalating ecological demands of residents [26]. Simultaneously, the encroachment of impermeable surfaces has altered the subsurface structure of urban areas, significantly impacting local hydrological and thermal conditions. At present, there is no agreement in the academic community regarding the delineation of ecological land, with many scholars having elaborated on or defined ecological land and its connotations from different perspectives [27,28,29]. In this article, we define ecological land as land that can directly or indirectly provide ecosystem services, except for artificial hardened surfaces, and that has a certain capacity for self-regulation, restoration, maintenance, and development [24,30]. Ecological land, as a vital component of the urban ecosystem, is of great importance in urban temperature regulation. Since we are facing a growing problem concerning the thermal environment, understanding the spatio-temporal dynamics of urban ecological land is crucial for further analyzing the driving forces of the thermal environment. This understanding will be pivotal in constructing ecologically balanced cities, enhancing the safety of the urban ecosystem, ensuring urban residents’ health, and, ultimately, achieving a harmonious coexistence between humans and nature.

Suzhou is one of the significant central cities in the YRD, boasting strengths in industry, the economy, and technological advancement. Nevertheless, the rapid urbanization pace, coupled with urban construction expansion and population growth, has raised concerns regarding the UHI issue. When studying the UHI effect, spatial distribution patterns and temporal change characteristics are the two essential components [31]. Previous studies on the UHI effect in Suzhou [32,33] have predominantly focused on analyzing the spatial distribution of heat islands and their relationship with seasonal and climatic variations. However, these studies have been limited by relatively short and discontinuous time periods, primarily focusing on Suzhou’s central urban region, which is insufficient to capture recent shifts in the UHI effect. In addition, land use, as a crucial factor affecting the UHI effect, has also garnered extensive attention from numerous scholars in the academic community. Xu et al. [34] analyzed the spatio-temporal changes in heat islands and land use alterations during Suzhou’s rapid urban expansion since 1986. Their findings revealed that land cover changes have significantly impacted the spatial pattern of the UHI, indicating a spatial correlation between the two factors. Zhu et al. [35] examined the correlation between landscape pattern evolution traits and the UHI effect in Suzhou, utilizing Landsat remote sensing images spanning from 1986 to 2006. Their results revealed an increase in landscape fragmentation and patch complexity over the 20-year period in Suzhou. Additionally, they highlighted the significant influence of sub-bedding pattern evolution on shaping the UHI transformation. Feng et al. [36] used generalized additive models to analyze the spatial factors affecting the LST pattern in Suzhou from 1996 to 2016. They found a consistent upward trend in Suzhou’s UHI intensity during this period, with variations in both spatial distribution and seasonality. Furthermore, they identified the normalized difference built-up index (NDBI) as the dominant factor affecting the LST pattern, followed by the normalized difference water index (NDWI) and normalized difference vegetation index (NDVI). However, studies on LST retrieval in Suzhou using Landsat-8 remote sensing data in recent years are relatively scarce. Moreover, while analyzing the heat island effect and its driving factors, most studies have primarily relied on statistical analysis to evaluate the collective impact of land use transitions on the UHI, often overlooking other natural and anthropogenic factors affecting the intensity of the UHI. Considering these factors, this study aimed to utilize Landsat-5, -7, and -8 remote sensing images for LST retrieval, by using the mono-window algorithm to analyze the spatio-temporal patterns of the UHI in Suzhou from 2000 to 2021 and identify its key driving factors. Furthermore, we explored the cooling service effect of ecological land changes on the intensity and spatial arrangement of the heat island effect using GlobeLand30 data, to provide scientific support for Suzhou’s transformation toward green and sustainable development.

2. Study Area and Dataset

2.1. Study Area

Suzhou is located at coordinates 30°47′–32°02′ N and 119°55′–121°20′ E and comprises five districts and four counties, which have the same administrative level, though the former have a higher level of urbanization (Table 1), encompassing an approximate area of 8657.32 km² (Figure 1). As of the end of 2022, the city had an estimated resident population of approximately 12.91 million, with an urbanization rate of approximately 82.12% [37]. Owing to its favorable geographical location and climatic conditions, Suzhou underwent industrialization and urbanization earlier than many other cities in China. From 2000 to 2021, Suzhou showed significant progress in both its economy and urban development [38]. The built-up area expanded from 86.5 km² to 770.0 km², while land used for green and squares has grown from 14 km² to 98.79 km², approximately seven times its original size. Notably, Suzhou’s cultivated land area decreased from 3020 km² to 1317.2 km² during this period.

Table 1. Basic situation of districts and counties of Suzhou in 2021.

Name	Code	Population (Persons)	Built-Up Area (km2)
Wujiang	P1	895,024	109.53
Taicang	P2	525,897	52.23
Changshu	P3	1,060,994	99.76
Kunshan	P4	1,143,315	72.00
Zhangjiagang	P5	929,339	64.68
Wuzhong	P6	752,114	371.80
Gaoxin and Huqiu	P7	467,703
Industrial park	P8	610,917
Xiangcheng	P9	485,135
Gusu	P10	750,661

Table 1. Basic situation of districts and counties of Suzhou in 2021.

Name	Code	Population (Persons)	Built-Up Area (km2)
Wujiang	P1	895,024	109.53
Taicang	P2	525,897	52.23
Changshu	P3	1,060,994	99.76
Kunshan	P4	1,143,315	72.00
Zhangjiagang	P5	929,339	64.68
Wuzhong	P6	752,114	371.80
Gaoxin and Huqiu	P7	467,703
Industrial park	P8	610,917
Xiangcheng	P9	485,135
Gusu	P10	750,661

Figure 1. A location and elevation map of Suzhou in the Yangtze River Economic Belt.

Figure 1. A location and elevation map of Suzhou in the Yangtze River Economic Belt.

2.2. Data Sources

(1) Remote sensing image data: The Landsat series data, sourced from the Geospatial Data Cloud, were primarily used for retrieving the LST. Considering the typically higher average heat island intensity during the fall and winter seasons compared to summer [39,40], the remote sensing images for autumn and winter were selectively chosen, ensuring a cloud coverage of less than 10%. Detailed information regarding the specific satellite parameters used can be obtained from

Table 2. Landsat data parameters.

Satellite	Sensor	Data Level	Imaging Data	Path and Row	Cloudiness (%)
Landsat-5	TM	L1T	18 September 2006	119/38	0.03
Landsat-7	ETM+	L1T	11 October 2000	119/38	8.93
Landsat-8	OLI-TIRS	L1TP	26 October 2014 14 November 2021	119/38	0.09 2.43

.

(2) Land classification data: The land use data obtained from GlobeLand30 [41] for the years 2000, 2010, and 2020, delineated within the boundaries of Suzhou, segmented the study area into eight categories: cultivated land (CL), forest (FT), grassland (GL), shrubland (SL), wetland (WL), water bodies (WB), artificial surfaces (AS), and bareland (BL), achieving a total classification precision of 83.5%. Based on the current status of the ecological environment and previous studies, all land use categories within the scope of this study, except for AS, were considered ecological land. The main ecosystem service function studied in this paper was the cooling effect attributed to ecological land.

(3) Driver data: Data on the total population at year end and the areas of administrative districts and counties in Suzhou were obtained from the Suzhou Statistical Yearbook (2001–2021) published by the Suzhou Municipal Bureau of Statistics. The building distribution in various years was expressed as the NDBI, while the degree of urban surface vegetation cover was expressed as the NDVI. These indices were calculated from the Landsat images of the same periods and were mainly used to analyze the drivers of UHI changes.

3. Methodology

3.1. LST Retrieval in Accordance with the Mono-Window Algorithm

This study eliminated the impact of atmospheric molecules and aerosol scattering through radiometric calibration and FLAASH atmospheric correction [42,43] to obtain more accurate parameters for the real physical model. Subsequently, the four-phase remote sensing images, obtained after image cropping, were used for LST retrieval. The mono-window algorithm proposed by Qin et al. [44] is built upon the thermal radiation conduction equation [45], which directly integrates the atmospheric and surface effects into the calculation Formula (1):

$${\mathrm{T}}_{\mathrm{LST}}=\frac{1}{{\mathrm{V}}_{\mathrm{atr}}}\{{\mathrm{c}}_1\left(1-{\mathrm{V}}_{\mathrm{atr}}-{\mathrm{V}}_{\mathrm{lse}}\right)+\left[{\mathrm{c}}_2\left(1-{\mathrm{V}}_{\mathrm{atr}}-{\mathrm{V}}_{\mathrm{lse}}\right)+{\mathrm{V}}_{\mathrm{atr}}+{\mathrm{V}}_{\mathrm{lse}}\right]\times {\mathrm{T}}_{6,10}-{\mathrm{V}}_{\mathrm{lse}}\times {\mathrm{T}}_{\mathrm{a}}-273.15$$

(1)

where $T_{LST}$ represents the LST in °C; $c_1$ and $c_2$ represent coefficients fitted to the thermal radiation intensity, with $c_1$ set as −67.355351 and $c_2$ as 0.458606 when LST falls within the range of 0 to 70 °C; and $V_{atr}$ and $V_{lse}$ represent intermediate variables derived from the atmospheric transmittance rate and land surface emissivity, respectively. Specific formulas and parameters for these variables are available in the referenced literature [18] and are not reiterated here.

$T_{6,10}$ represents the radiant brightness temperature value. It is convertible from the spectral radiance value to the brightness temperature value at the satellite altitude through Planck’s formula, as shown in (2):

$${\mathrm{T}}_{6,10}=\frac{\mathrm{K}2}{\ln (\mathrm{K}1/{\mathrm{L}}_{\mathsf{\lambda }}+1)}$$

(2)

where $L_{\lambda }$ represents the pixel gray value of the thermal infrared band. K1 and K2 denote the pre-launch calibration constants, which are accessible by viewing the header file.

$T_a$ represents the mean atmospheric temperature, estimated to approximate the near-LST based on atmospheric conditions, with the specific formula as follows:

$${\mathrm{T}}_{\mathrm{a}}=19.2704+0.91118{\mathrm{T}}_0$$

(3)

where $T_0$ is the near-surface air temperature (at a height of approximately 2 m) measured in K.

3.2. Mean and Standard Deviation

After verifying Suzhou’s historical temperature data, it was observed that the LST on the days of remote sensing data collection for this study consistently exceeded 0 °C. Concurrently, anomalies were detected at the image edges, deemed inconsequential in the overall spatial extent analysis. Thus, anomalies below 0 °C retrieved from the data were eliminated. To delineate the LST distribution, the results were categorized into six classes using the equal interval method, demarcating divisions at 8.5, 11, 13.5, 16, and 18.5 °C to create the LST retrieval map. However, owing to differing temperature intervals acquired by various sensors across fixed periods, a uniform comparison of the LSTs retrieved was challenging [46]. Therefore, the mean and standard deviation method was adopted for segmentation, which involves classifying maximum and minimum temperatures derived from LST retrieval by combining mean LST values (μ) and various standard deviation multiples ($\mathrm{StdDev}$). Employing this approach, the study area was categorized into five surface thermal field classes: low-, sub-low-, medium-, sub-high-, and high-temperature zones (TZs), which reflected LST changes across different periods. The specific grading can be found in

Table 3. Mean and standard deviation method of classifying temperature classes.

Temperature Classification	Interval
Low temperature zone (TZ)	${\mathrm{T}}_{\mathrm{LST}}<\mathsf{\mu }-\mathrm{StdDev}$
Sub-low TZ	$\mathsf{\mu }-\mathrm{StdDev}\le {\mathrm{T}}_{\mathrm{LST}}<\mathsf{\mu }-0.5 \mathrm{StdDev}$
Medium TZ	$\mathsf{\mu }-0.5 \mathrm{StdDev}\le {\mathrm{T}}_{\mathrm{LST}}\le \mathsf{\mu }+0.5 \mathrm{StdDev}$
Sub-high TZ	$\mathsf{\mu }+0.5 \mathrm{StdDev}<{\mathrm{T}}_{\mathrm{LST}}\le \mathsf{\mu }+\mathrm{StdDev}$
High TZ	${\mathrm{T}}_{\mathrm{LST}}>\mathsf{\mu }+\mathrm{StdDev}$

[47].

3.3. Land Use Dynamics Degree and Transfer Matrix

The land use dynamic degree (LUDD) model often represents changes in land use types over a specified period. It can be categorized into single and comprehensive LUDDs [48]. The single LUDD ($D_S$) denotes the magnitude of transition in a particular land use type within a defined spatio-temporal context. A positive or negative value of $D_S$ indicates an increase or decrease in the land use type, with a larger absolute value indicating a faster annual rate of change, as shown in (4) [49]:

$${\mathrm{D}}_{\mathrm{S}}=\frac{{\mathrm{S}}_{\mathrm{j}}-{\mathrm{S}}_{\mathrm{i}}}{{\mathrm{S}}_{\mathrm{i}}}\times \frac{1}{\mathrm{t}}\times 100$$

(4)

where $S_i$ and $S_j$ are the areas of the designated land use types at the initial and final time points, respectively, measured in km², and t denotes the duration covered by the study, measured in years.

The comprehensive LUDD ($D_C$) provides an overall measure of landscape activity for all land use types within a specific region. Greater values of $D_C$ indicate a faster combined annual pace of transformation in land use types within the region, as shown in the following formula:

$${\mathrm{D}}_{\mathrm{C}}=\left\{\frac{{\sum }_{\mathrm{j}=1}^{\mathrm{n}}{\Delta \mathrm{S}}_{\mathrm{j}}}{{\sum }_{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{S}}_{\mathrm{j}}}\right\}\times \left(\frac{1}{\mathrm{t}}\right)\times 100\%$$

(5)

where $S_j$ represents the area of land use type j at the end of monitoring within the region, quantified in km². $\mathsf{\Delta }{S}_j$ denotes the absolute value of the areas converted from land use type j to another type j within the region throughout the study duration, quantified in km². t is the duration covered by the study, measured in years.

The land use transfer matrix depicts the initial and final land use structures, indicating transformations between different land types throughout the study duration. This helps characterize the direction and quantitative structural changes within land use types to determine the origin of each land use type within the region during the study period [50]. This matrix provides insights into transformed land use types and their respective sources of change over the study period. The specific formula can be found in the literature [51].

3.4. GeoDetector Method

GeoDetector, an innovative statistical methodology, was developed by Wang’s [52] research team at the Chinese Academy of Sciences to assess spatial heterogeneity and its driving factors. In this study, we employed factor detection to analyze the social and natural factors affecting the spatial variation in LST, investigating the laws and mechanisms underlying the impact of these factors on LST changes and distribution. The q-value, denoting the degree of impact of these factors on LST, is defined as follows:

$$\left\{\begin{array}{c}\mathrm{q}=1-\frac{{{\displaystyle \sum }}_{\mathrm{h}=1}^{\mathrm{L}}{\mathrm{N}}_{\mathrm{h}}{\mathsf{\sigma }}_{\mathrm{h}}^2}{{\mathrm{N}\mathsf{\sigma }}^2}=1-\frac{\mathrm{SSW}}{\mathrm{SST}}\hfill \\ \mathrm{S}\mathrm{S}\mathrm{W}={\displaystyle {\displaystyle \sum }_{\mathrm{h}=1}^{\mathrm{L}}}{\mathrm{N}}_{\mathrm{h}}{\mathsf{\sigma }}_{\mathrm{h}}^2\hfill \\ \mathrm{S}\mathrm{S}\mathrm{T}=\mathrm{N}{\mathsf{\sigma }}^2\hfill \end{array}\right.$$

(6)

where $SSW$ represents the summation of variance within strata, $SST$ denotes the overall variance across the entire region, while the range of values for q falls between 0 and 1. The greater the value of q, the greater the explanatory power of the assessment factor on the dependent variable.

The UHI effect in Suzhou is influenced by several factors. Based on previous studies and data accessibility, five distinct factors, namely, NDBI, NDVI, land use type change, population density, and proximity to water bodies (Taihu Lake and Yangtze River), were considered to analyze the spatio-temporal variations in surface temperatures and the formation mechanisms. The GeoDetector analysis principle underscores the significance of different forms of spatial discretization and interval combinations on the q-value [53]. Therefore, in this study, sampling was conducted using 2 km grid points and utilizing the natural discontinuous point method to divide and process the five types of driving factors. The q-value with the smallest significance (p) was chosen as the optimal parameter for GeoDetector.

4. Results

4.1. Spatio-Temporal Patterns of LST

The LST in Suzhou exhibited a distinct spatial gradient effect from 2000 to 2021, as shown in Figure 2. The percentages of areas with temperatures exceeding 13.5 °C increased progressively over this period, reaching 9.8%, 39.3%, 69.3%, and 66.8% in 2000, 2006, 2014, and 2021, respectively (

Table 4. Changes in the land surface temperature (LST) classes in Suzhou, 2000–2021. (unit of area: km²; unit of percentage [PCT]: %).

Temperature Class	2000		2006		2014		2021
	Area	PCT	Area	PCT	Area	PCT	Area	PCT
<8.5 °C	2335.34	27.61	1600.83	19.00	456.38	5.31	804.74	9.36
8.5–11 °C	3097.87	36.62	932.32	11.06	1578.79	18.38	1163.20	13.53
11–13.5 °C	2194.74	25.95	2579.12	30.61	605.32	7.05	884.77	10.29
13.5–16 °C	615.36	7.28	2038.05	24.19	1547.52	18.02	3226.96	37.54
16–18.5 °C	176.48	2.09	930.28	11.04	2715.30	31.62	2222.13	25.85
>18.5 °C	38.61	0.46	346.26	4.11	1684.68	19.62	294.62	3.43

). In 2000, approximately 90.2% of Suzhou’s area registered an LST below 13.5 °C, with water bodies such as the Yangtze River and Taihu Lake measuring <8.5 °C. The LST of non-water-body land was confined to the range of 8.5–13.5 °C, while temperatures exceeding 13.5 °C were mainly concentrated in Gusu District. By 2006, Suzhou’s LST was centrally distributed in the ranges of 11–13.5 °C, 13.5–16 °C, and <8.5 °C, constituting 30.6%, 24.2%, and 19.0% of the overall zone, respectively. By 2014, Suzhou’s LST distribution exhibited a shift, concentrating prominently in the ranges of 16–18.5 °C, >18.5 °C, and 8.5–11 °C, making up 31.6%, 19.6%, and 18.4% of the overall zone, respectively. Fast-forward to 2021, and Suzhou’s LST primarily concentrated in the ranges of 13.5–16 °C, 16–18.5 °C, and 8.5–11 °C, representing 37.5%, 25.9%, and 13.5% of the overall zone, respectively. Notably, the LST in the central urban region, along with the center of each district and county, showed markedly higher values than in other regions. The southern segment of the central urban region initially experienced a relatively lower LST than the central city, but this temperature gap gradually diminished over time.

From 2000 to 2021, approximately 97.05% of Suzhou experienced an increasing trend in LST (Slope > 0), with the city of Kunshan recording the highest rate of change occurring, at 8.03 (Figure 3). In accordance with the class division, 81.76% of the area showed a change rate ranging from 0.5 to 2.5, where the percentages of the area distributed in the ranges of 1–1.5, 1.5–2, and 0.5–1 were 23.8%, 21.8%, and 19.9%, respectively (

Table 5. Statistical table of LST slope data for Suzhou, 2000–2021.

Slope Value	≤0	0–0.5	0.5–1	1–1.5	1.5–2	2–2.5	2.5–3	3–3.5	≥3.5
Area (km2)	246.22	531.68	1666.33	1992.26	1824.91	1349.07	521.55	139.8	85.23
PCT (%)	2.95	6.36	19.94	23.84	21.84	16.14	6.24	1.67	1.02

). Across administrative areas, all districts and counties in Suzhou experienced an increasing LST. The city of Taicang recorded the steepest average temperature change (1.87), indicating a significant increase in temperature. In contrast, Gusu District exhibited the smallest average change slope, at 0.38, indicating a comparatively modest temperature change (

Table 6. Statistical table of LST slope data for Suzhou districts and counties, 2000–2021.

Administrative District	MIN	MAX	RANGE	MEAN	STD
Wujiang	−6.20	7.02	13.22	1.35	0.83
Industrial park	−4.11	7.61	11.72	1.59	0.97
Xiangcheng	−3.52	7.03	10.55	1.65	0.71
Taicang	−4.11	6.41	10.52	1.87	0.64
Changshu	−4.86	8.01	12.87	1.81	0.70
Gusu	−3.95	4.18	8.12	0.38	0.79
Kunshan	−5.93	8.03	13.96	1.64	0.76
Zhangjiagang	−5.48	7.42	12.9	1.84	0.77
Wuzhong	−9.41	5.94	15.35	1.15	0.79
Gaoxin and Huqiu	−4.44	7.17	11.61	1.20	0.91

).

4.2. Dynamics of UHI Intensity

The area of the sub-high TZ in Suzhou progressively expanded, accounting for 15.3%, 17.0%, 24.1%, and 30.2% of the total region by 2000, 2006, 2014, and 2021, respectively (

Table 7. Statistical tables of information on Suzhou temperature-rating districts, 2000–2021. (Unit of area: km²; unit of PCT: %).

Temperature Class	2000		2006		2014		2021
	Area	PCT	Area	PCT	Area	PCT	Area	PCT
Low TZ	1684.57	19.92	1764.42	20.94	2065.79	24.05	1897.02	22.07
Sub-low TZ	636.17	7.52	637.79	7.57	462.52	5.39	396.83	4.61
Medium TZ	3610.04	42.68	3307.97	39.26	2711.94	31.58	2757.62	32.08
Sub-high TZ	1296.90	15.33	1432.12	16.99	2071.32	24.12	2599.24	30.24
High TZ	1230.73	14.55	1284.57	15.24	1276.39	14.86	945.69	11.00

). In 2000, the LST intensity distribution in Suzhou was led by the medium TZ (accounting for 42.7% of the area), followed by the low TZ (19.9%), sub-high TZ (15.3%), high TZ (14.6%), and sub-low TZ (7.5%). In 2006, the distribution of UHI intensity across Suzhou was consistent with that in 2000, concentrating on the medium TZ (39.3%), low TZ (20.9%), and sub-high TZ (17.0%). By 2014, the sub-high TZ overtook the low TZ in the heat island intensity distribution, shifting the dominant classes from the medium and low TZs to the medium and sub-high TZs, with intensity ratios in the order of medium TZ (31.6%) > sub-high TZ (24.1%) > low TZ (24.1%) > high TZ (14.9%) > sub-low TZ (5.4%). By 2021, the distribution remained consistent with that in 2014 and concentrated mainly in the medium TZ (32.1%), sub-high TZ (30.2%), and low TZ (22.1%).

The spatial distribution of UHI intensity in Suzhou (Figure 4) as well as the area share of different temperature class zones in each district and county (Figure 5) and specific temperature interval values (

Table 8. Temperature zoning table for each period. (Unit of temperature: °C).

	Low TZ	Sub-Low TZ	Medium TZ	Sub-High TZ	High TZ
2000	0.08–7.00	7.00–8.47	8.47–11.40	11.40–12.86	12.86–43.26
2006	0.02–8.98	8.98–10.70	10.70–14.13	14.13–15.84	15.84–37.87
2014	2.46–11.12	11.12–13.10	13.10–17.06	17.06–19.04	19.04–36.04
2021	0.00–10.76	10.76–12.36	12.36–15.56	15.56–17.16	17.16–34.61

) revealed significant changes in the overall pattern of the UHI effect in Suzhou over the past two decades. At the macroscale, both Taihu Lake and the Yangtze River exhibit significant negative heat island characteristics, categorized as low-TZ and sub-low-TZ areas, respectively. In contrast, built-up and BL areas generally experienced higher temperatures and were thus classified as high-TZ areas. Between 2000 and 2006, the heat island areas with a high TZ were scattered throughout the central urban region and district and county centers. This dispersed pattern gradually consolidated into a large contiguous patch between 2006 and 2014, expanding further outward. However, from 2014 to 2021, while the high TZ extended across most of Suzhou, the overall heat island effect began to alleviate.

At the microscale, in 2000, Suzhou’s UHI effect was localized, concentrating primarily within the central urban regions, such as Gusu, Gaoxin, and Huqiu Districts and the northern part of Wuzhong District. The surrounding districts and counties displayed minimal heat island effects. By 2006, the UHI effect extended beyond the central urban region, reaching industrial parks and Xiangcheng District. Intense heat island effects were observed in Kunshan to the east and Zhangjiagang to the north. By 2014, the heat island influence expanded further across the central urban region and in four cities within Suzhou’s jurisdiction (Kunshan, Zhangjiagang, Taicang, and Changshu). In 2021, the heat island effect covered the entire administrative region, although it was relatively moderate overall, with a significant intensification only in Wujiang District.

4.3. Spatio-Temporal Patterns of Ecological Land

In 2000, 2010, and 2020, Suzhou’s ecological land areas measured 7770.6, 7258.3, and 6407.1 km², accounting for 89.9%, 84.9%, and 74.3% of the total administrative area, respectively. Over the 20-year period, this reflected a cumulative decrease of 1363.5 km², primarily focused on the central urban regions. In 2000, the ecological land structure of Suzhou followed the order of CL, WB, FT, GL, WL, and SL (

Table 9. Area and changes in various land types in Suzhou, 2000–2020. (Unit of area: km²).

	Type	Ecological Land								Artificial Surfaces
Year		CL	FT	GL	SL	WL	WB	BL	Sum
2000		4226.29	204.84	12.95	0.86	2.84	3322.79	0	7770.57	872.85
2010		3743.88	102.38	106.05	1.51	30.77	3273.70	0	7258.29	1385.09
2020		2936.17	168.61	102.65	0.92	53.32	3144.72	0.69	6407.08	2221.17
2000–2010		−482.41	−102.46	93.09	0.64	27.92	−49.09	0	-	512.24
2010–2020		−807.71	66.24	−3.39	−0.59	22.55	−128.98	0.69	-	836.08
2000–2020		−1290.12	−36.23	89.70	0.06	50.48	−178.06	0.69	-	1348.32

). By 2020, significant shifts occurred, with WB surpassing CL to become the first ecological land type, covering 36.5% of the total area, and BL also becoming a new category of ecological land (<1%).

The shift in Suzhou’s land use from 2000 to 2020 mainly involved increased AS expansion and reduced CL coverage, with the AS area proportion increasing from 10.1% to 25.7% and the CL area proportion decreasing from 48.9% to 34.0%. In 2000, the built-up areas in Suzhou were clustered in blocks within the central areas of the main urban region, Wujiang, and four subordinate cities surrounded by extensive CL, making up approximately half of the overall study zone. River-dominated central and southern areas, with small and medium-sized towns and townships along the water network and independent of each other, held significant WB and WL portions, accounting for approximately 40% of the area. By 2020, Suzhou’s built-up area had expanded approximately sixfold, extending into surrounding areas and leading to contiguous district distributions. A large amount of CL was replaced by AS, leading to the further fragmentation of ecological spatial patterns. Additionally, factors such as siltation, lake filling, and land reclamation projects reduced WB areas by 5.36% (Figure 6).

The total areas of land use conversions in Suzhou spanned 1466.81 km² (2000–2010) and 1699.99 km² (2010–2020) (Figure 7), with conversions between ecological and non-ecological lands (AS) accounting for 50.22% and 62.88%, respectively. Over the 2000–2020 period, the expansion of AS was mainly from CL (85.93%) and WB (12.15%), with the area increasing period by period. Most of the reduced CL (76.11%) was used for urban development, with a small portion used toward replenishing WB (19.32%) and FT (2.44%). In addition, there were some conversions from WB and AS to CL, as well as reclamation of AS as CL and WB.

The comprehensive LUDDs of Suzhou were 1.47% and 2.16% for the periods 2000–2010 and 2010–2020 (

Table 10. Changes in the land use dynamic degree (LUDD) in Suzhou, 2000–2020. (Unit of LUDD: %).

	Time Period	Single LUDD
Land Use Type		2000–2010	2010–2020	2000–2020
Ecological land	CL	−1.14	−2.16	−1.53
	FT	−5.00	6.47	−0.88
	GL	71.87	−0.32	34.62
	SL	7.41	−3.89	0.32
	WL	98.21	7.33	88.77
	WB	−0.15	−0.39	−0.27
	BL	−	−	−
Non-ecological land	AS	5.87	6.04	7.72
Comprehensive LUDD		1.47	2.16	1.73

), respectively, indicating an increasing intensity of land type alteration and speed in Suzhou. In the single LUDDs, non-ecological land showed a positive dynamic degree, with a continuously increasing trend. However, the dynamic degrees of CL and WB showed negative values, and the absolute value was increasing. This indicated that the area of non-ecological land had been expanding since 2000, while the areas of CL and WB had been shrinking, with the area of change growing. The largest dynamic degrees in ecological land from 2000 to 2010 were 98.21% and 71.87% for WL and GL, respectively, which mainly manifested as an increase in area. Then, the largest LUDDs in ecological land from 2010 to 2020 were 7.33% and 6.47% for WL and FT, respectively, which also manifested as an increase in area.

From 2000 to 2010, Suzhou exhibited a significant single LUDD but a relatively smaller comprehensive LUDD. However, between 2010 and 2020, the single LUDD remained small, while the comprehensive LUDD increased significantly. Analysis of the change area revealed that even a slight increase in certain land categories with a small initial base could lead to a substantial value for a single LUDD. For instance, the WL area experienced the most significant change from 2000 to 2010, with a dynamic degree of 98.21%, but in 2000, it was only 2.84 km², with an actual increase of only 50.48 km².

4.4. Driving Forces

Based on the factor detector model, the mechanism behind the spatial pattern formation of the UHI in Suzhou was investigated. The findings indicated that the surface temperature in Suzhou was affected by a blend of various natural and social factors, and the explanatory power (q-value) of individual factors on the surface temperature varied significantly (Figure 8). The average magnitude of the q-value for each driving factor in the spatial differentiation of surface temperature from 2000 to 2021 was as follows: NDBI (0.648) > NDVI (0.533) > land use type change (0.466) > population density (0.051) > shortest distance from water bodies (0.044). Among them, the driver with the most significant impact on the spatial variability of the UHI was the NDBI, and the explanatory power was the greatest in the datasets of 2000, 2014, and 2021, at 0.661, 0.703, and 0.581, respectively. The effect of NDVI was slightly weaker than that of NDBI, but the disparity between the q-values diminished over time to 0.194, 0.104, and 0.047, respectively. Both factors maintained a strong correlation with the UHI distribution. Conversely, the explanatory power of land use type change for the spatial differentiation of surface temperature remained consistently high, with q-values exceeding 0.45 throughout all three periods, signifying its pivotal role as a driving factor of the spatial differentiation of surface temperature. Population density and the shortest distance to water bodies, on the other hand, had relatively weak explanatory power for surface temperature in the three periods when compared with the other drivers and showed an overall decreasing trend.

5. Discussion

The urban agglomeration in the YRD is highly susceptible to the adverse impacts of climate change, which may result in vast economic losses [54,55]. In 2021, the Ministry of Ecology and Environment released “China’s Policies and Actions to Address Climate Change”, emphasizing the urgent need to “mitigate the UHI effect and related climate risks through the layout of urban clusters and the construction of urban green environments such as green corridors, greenways, and parks, to enhance the adaptation ability to climate change”. Our study revealed a significant enhancement in Suzhou’s heat island effect, marked by an evolving spatial picture and grading expansion. From the perspective of spatial distribution, owing to rapid economic development, Suzhou experienced a steady increase in building and population densities in its central urban area, which continually expanded over time. In 2000, the city’s UHI effect primarily centered around the central urban region as a singular patch, with small hotspots scattered outside, associated with the four surrounding cities. However, from 2014, Suzhou’s heat island effect spread widely across the region, with the patch UHI effect in the central urban region mitigated. This may be due to the fact that Suzhou became the second batch of national pilot low-carbon cities at the end of 2012, and during the 13th Five-Year Plan period, Suzhou actively promoted green transformation, energy saving, and emission reduction, which resulted in a reduction in carbon emission intensity by more than 22%, thus mitigating the heat island intensity. Meanwhile, the original small hotspots converged into a large, contiguous expanse spreading outward. This was in line with what other authors have suggested for surface temperature studies in Suzhou. For example, Ji et al. [56] found that the UHI of the city of Suzhou had a slowly increasing trend with the development of urbanization during this period, which they ascertained by inverting satellite data from 1986 to 2010; they noted that the heat island was radially distributed around the city with the urban area as the center. Feng et al. [36] found that the UHI in Suzhou was predominantly located in the city center in 1996, expanding to the suburbs in 2004 and 2016. From the perspective of grade change, over the years, the medium-TZ area gradually decreased, while the sub-high-TZ area expanded, with a structural shift primarily led by these two categories; however, the areas occupied by low and sub-low TZs remained relatively unchanged.

In terms of the surface temperature and land use types distribution maps for all three periods, we found a certain degree of overlap between the two. During the period from 2000 to 2021 (

Table 11. The proportion of and change in ecological land change area for each TZ. (Unit of proportion: %).

	Land Use Type Change	The Proportion of Conversion between Ecological Land	The Proportion of Ecological Land Transfer Out	The Proportion of Ecological Land Transfer	The Proportion of Unchanged AS
Change in TZ
Unchanged	Low TZ	100.00	0.00	0.00	0.00
	Sub-low TZ	99.11	0.60	0.30	0.00
	Medium TZ	85.18	12.76	1.09	0.97
	Sub-high TZ	61.67	25.13	2.99	10.22
	High TZ	9.10	25.72	1.95	63.23
Changed	Cooling	67.19	12.36	2.30	18.15
	Warming	66.57	29.13	0.72	3.58

), according to the invariant TZ classification, the low TZ was solely a transition between ecological lands, with a share of 100 percent. For the high TZ, 63.23% of the AS remain unchanged and mostly concentrated in the center of the town. Moreover, 25.72% of the ecological lands, primarily consisting of CL and WB, were transformed into AS, with the area increasing period by period, and the surrounding ecological land dropped significantly. As the shift occurred from low to high TZs, the proportion of conversion between ecological land was decreasing, and the opposite trend was occurring for the proportion of ecological land transferred out and unchanged AS. For the TZs that had changed, we categorized them into cooling and warming. We found that, under the premise that the proportion of internal transformation of ecological land was essentially the same in both, the proportion of ecological land converted to AS in the warming zone was 29.13%, which was more than twice as much as that in the cooling zone, whereas the transfer of ecological land was not even 1%, which can be seen as indicating that the introduction of AS significantly contributed to the generation of heat islands in these areas. This aligns with the discovery made by Mo et al. [46], who noted that the heat class enhancement in Suzhou was accompanied by a high-intensity transformation of vegetation cover and WB to AS during 1986–2004. Therefore, urbanization, if carefully planned, can become a catalyst for enhancing equity and well-being by leveraging the mutual benefits and synergies among climate change adaptation, equitable urban development, and so on [3]. Moreover, leveraging the ecological functions of natural ecosystems for climate regulation and biodiversity preservation is crucial for strengthening Suzhou’s resilience against climate-related disasters [57]. The percentage of ecological land transfer in the cooling zone was 2.3%, more than three times that of the warming zone, which was also attributed to Suzhou’s ecological environmental protection efforts, which included the preservation of water bodies, such as Taihu Lake, Yangcheng Lake, and the Yangtze River, the construction of healthy wetland urban parks, and the expansion of green spaces within residential areas. These water bodies and green spaces have since played crucial roles in moderating the region’s temperature.

To better explore the impact of ecological land on the UHI, we used GeoDetector to explore the cooling service effect of ecological land. The results indicated the NDBI was the primary significant driving factor affecting the spatial variability of Suzhou’s LST, which is consistent with the discovery by Feng et al. [36]. Between 2000 and 2021, the extent of construction land in Suzhou experienced a significant increase, aligned with the direction of urban expansion and amplification of the UHI effect. Given that Suzhou’s terrain is dominated by plains, it is convenient for infrastructural construction and industrial and urban land expansion, resulting in the occupation of a large amount of CL and a rise in building density during the urbanization process. Concurrently, high-NDBI areas expanded from the city center to its fringe, thus strengthening the heat island effect [58]. Therefore, to suppress the negative impacts of the UHI on the urban ecosystem, optimize the urban climate, and improve the urban ecological and living conditions, we should first start with high-temperature buildings. We can use new outdoor construction materials that can cool and conserve energy, along with permeable ground pavement materials. Furthermore, we need to initiate ecologically reasonable energy planning and urban development models, while also adjusting the urban industrial structure [59]. As a secondary driving factor, the explanatory power of NDVI for the spatial differentiation in surface temperature first increased and then decreased. From 2000 to 2014, vegetation reduced surface temperature through transpiration and leaf occlusion, giving full play to the cooling service effect of green spaces. Between 2014 and 2021, vegetation’s influence on the distribution of heat islands weakened slightly. In contrast, with the fading of deciduous plants, the cooling effect of vegetation reached its lowest [60], resulting in a decrease in the influence of NDVI on surface temperature. The data also indicated that the average NDVI value of Suzhou was decreasing, indicating that vegetation growth was worsening, and the canopy coverage area was decreasing. However, the remote sensing images selected in the experiment were taken in autumn and winter, at different times. Since the correlation between NDVI and LST fluctuates seasonally [61], and the NDVI was calculated from the corresponding Landsat data for each period, this may have impacted the results. This also shows that consideration should be given to increasing the amount of greenery within green space by improving the growth state of vegetation, engaging in dense artificial planting and other such activities without changing the green space area [62,63], or considering covering the vertical height and launching roof greening. This facilitates three-dimensional greening of the city, so as to realize a ‘green island’, which will help create an effective urban green ecological space and improve the cooling effect [59]. Meanwhile, rivers and lakes in Suzhou are widely distributed, with the Yangtze River being a prominent natural element affecting the urbanization of Suzhou and numerous other cities along its banks [64]. It absorbs heat through the hydrothermal cycle, evaporates water containing that heat, and dissipates it to reduce temperature. Furthermore, the existence of lakes and lake wind circulation can play a cooling role to a certain extent that lowers the temperature of the urban subsurface of Suzhou, slowing the rate of development of the UHI, while also significantly reducing its area [65]. However, unlike the findings of a previous study [36], our results showed that the shortest distance from water bodies had a weak impact on the surface temperature. This may have been because many water bodies were separated by buildings to satisfy the needs of urban development, which made them unable to remove heat effectively through water flow, weakening the impact of the blue space composed of water bodies [66]. Furthermore, the influence of population density on the UHI effect in Suzhou was very small, aligning with the conclusions reached in the literature [67]. This may have been because the statistics of population density were based on the statistics of the household-registered population at the end of the year, and so most migrant workers were not included in the statistics; therefore, the experiment did not accurately determine the impact of population density on the UHI effect. Meanwhile, land use patterns play a vital role in shaping urban thermal environmental changes, and their interactions with environmental factors are intricate, particularly in rapidly urbanizing regions [68]. In Suzhou, rapid economic development and a rise in population density have resulted in the continuous encroachment of construction land on ecological land, which has forced the spatio-temporal distribution of land use to change significantly [69], thus affecting the surface temperature. However, due to the susceptibility of optical sensors to weather and the limited number of images available to us, the temporal consistency of the data from multiple sources could not be guaranteed, and some seasonal information was inevitably overlooked. In the future, we will utilize longer time-series data to delve deeper into the role of urban ecological land use in mitigating urban heat stress problems.

Generally speaking, ecological land is essential for maintaining the safety of the urban ecosystem, as its cooling effect not only can effectively reduce urban surface temperatures but also regulate the climate at the regional scale. However, in the face of the rapid development of urbanization processes, many conversions have resulted in a decrease in ecological land and an increase in its fragmentation. To remedy this situation, it is necessary to rationally allocate ecological land and enhance its spatial layout through urban renewal activities, etc., in the future. At the same time, strengthening protection and restoration measures for ecological land is crucial to maximize its cooling effects on the basis of the existing foundation, and also to realize sustainable development.

6. Conclusions

In this paper, we utilized long-time-series remote sensing image datasets to invert the spatio-temporal pattern of the UHI effect and its dynamic changes in Suzhou, based on which we explored the cooling effect of ecological land. The experimental findings generated from this study carry practical significance in reconciling urban ecological protection with economic development. The principal conclusions can be summarized as follows: (1) In 2000–2021, the high-temperature UHI of Suzhou showed an expansion trend from point to surface, with the area spreading from the central urban region to various districts. Additionally, the medium-TZ area gradually decreased, while the sub-high-TZ area expanded, shaping a structural shift dominated by these two categories. (2) The enhancement of the UHI effect exhibited a strong correlation with the alterations in land use types. Specifically, the proportion of the area in AS grew from 10.1% in 2000 to 25.7% in 2020, which was the main factor contributing to the increase in heat island intensity. Conversely, ecological land can maintain a lower heat island intensity, and this tended to increase as the proportion of ecological land transferred out increased. (3) Land use type change plays a crucial role in influencing the spatio-temporal differentiation of UHI intensity. Among our findings, the key points to note are that NDBI was positively correlated with surface temperature, with a three-period average q-value of 0.65, and NDVI was negatively correlated with surface temperature, with a three-period average q-value of 0.53. This means that the transfer in and out of ecological land had a significant impact on the spatio-temporal pattern of the UHI, and the cooling service effect of ecological land in the city should be emphasized.