Exploring the Coupling Relationship Between Urbanization and Ecological Quality Based on Remote Sensing Data in Shenzhen, China

Abstract

As a flagship city of China’s reform and opening-up policy and the core engine of the Guangdong–Hong Kong–Macao Greater Bay Area, Shenzhen faces dual challenges of economic development and ecological conservation during its rapid urbanization. This study systematically investigates the relationship between urbanization and ecological quality in this high-density megacity over the past three decades (1990–2020) using multi-temporal Landsat imagery, incorporating an enhanced Remote Sensing Ecological Index (RSEI), impervious surface extraction techniques, and a Coupling Coordination Degree (CCD) model. Key findings include: (1) Impervious surfaces expanded from 458.15 km² to 709.23 km², showing a tri-phase pattern of rapid expansion, steady infill, and slight contraction, with an annual growth rate of 1.47%; (2) Ecological quality exhibited a “decline-recovery” trajectory, with RSEI values decreasing from 0.477 (1990) to 0.429 (2000) before rebounding to 0.491 (2020), demonstrating phased ecological fluctuations and restoration; (3) The CCD between urbanization and ecological environment improved significantly from “marginal coordination” (0.548) to “primary coordination” (0.636), forming a distinct “west-high-east-low” spatial pattern with significant clustering effects. This study reveals a novel three-dimensional synergistic pathway (“industrial upgrading-spatial optimization-ecological restoration”) for sustainable development in megacities, establishing the “Shenzhen Paradigm” for ecological governance in rapidly urbanizing regions worldwide.

1. Introduction

The global urbanization process has accelerated significantly over the past century [1]. According to United Nations data, the proportion of the global urban population was approximately 30% in 1950, rising to over 55% by 2018, and is projected to exceed 68% by 2050 [1,2]. China’s urbanization has been particularly rapid. Since the launch of reform and opening-up in 1978, the country’s urbanization rate has surged from 17.92% to 65.22% by 2022 [3,4]. This transformation has driven rapid economic growth, industrial upgrading, and large-scale population migration, highlighting the increasingly pivotal role of cities in national development. Massive urban construction has promoted the modernization of infrastructure and significantly improved living standards. Meanwhile, population agglomeration has fostered the formation of urban agglomerations and metropolitan areas, further enhancing regional economic coordination and integration [5,6].

While rapid urbanization has significantly contributed to regional economic growth, it has also led to serious ecological and environmental issues [7,8]. The concentration of large populations in urban areas has caused substantial changes in land use patterns and large-scale alterations of natural ecosystems [9]. Urban expansion is often accompanied by the reduction of green spaces and wetlands [10], the destruction of habitats [11,12], and a decline in biodiversity [13,14]. In addition, environmental problems such as air pollution, water pollution, and noise pollution [15,16,17]—byproducts of the urbanization process—have placed tremendous pressure on the stability and functioning of urban ecosystems [18,19].

With the acceleration of urbanization, the relationship between urban development and the ecological environment has become increasingly complex, characterized by mutual influence and constraint [20,21]. Achieving coordinated development between urbanization and ecological sustainability is not only a strategic priority for global sustainable development but also an urgent need for environmental protection. In recent years, research methods and models addressing the complex relationship between urbanization and the ecological environment have continued to evolve, leading to the development of diverse analytical frameworks and theoretical models. This area of study has become a global strategic issue and a research hotspot in Earth system science and sustainability science [22,23,24,25,26,27,28,29,30]. The Environmental Kuznets Curve (EKC) [31,32] and the Urbanization–Environmental Kuznets Curve (UEKC) [33] are among the earliest models used to describe this relationship. These models suggest that environmental quality declines during the early stages of urbanization but improves after a certain level of development is reached. However, such models have limitations in applicability and fail to fully capture the dynamic interactions between urbanization and the ecological environment [34,35]. To address these shortcomings, the coupling coordination degree (CCD) model has been widely adopted to quantify the level of coordination between the two systems [36,37,38,39]. This model reveals their mutual interactions and is applicable across different regions and development stages. Furthermore, the Coupled Human and Natural Cube (CHNC) model offers a more multidimensional analytical perspective, enhancing the systematization of such studies [40].

With the advancement of remote sensing technology, its advantages—such as rapid data acquisition, high precision, broad spatial coverage, and rich datasets—have provided strong data support for assessing ecological environmental quality and urbanization levels. Traditional ecological assessment methods often rely on single indicators such as Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), and land surface temperature (LST), but their limitations hinder a comprehensive evaluation of ecological conditions. To overcome the shortcomings of single-indicator approaches, Xu [41,42] developed the Remote Sensing Ecological Index (RSEI) based on Principal Component Analysis (PCA). RSEI integrates four key ecological factors: greenness, wetness, heat, and dryness, enabling an objective and efficient assessment of regional ecological quality. Due to its strong applicability, RSEI has been widely used in ecological monitoring, modeling, and prediction across various regions, yielding promising results. In recent years, RSEI-based studies have expanded significantly and have been successfully applied to multiple areas, supporting the analysis, modeling, and prediction of urban ecological characteristics [43,44,45,46]. In addition, remote sensing technology has made significant progress in urbanization monitoring through its capabilities for wide-area coverage, long-term time series observation, and high-resolution monitoring. In contrast, traditional statistical data are more difficult to obtain and are often subject to strong subjectivity, potentially leading to discrepancies between research findings and actual conditions [47]. Remote sensing enables the use of key indicators—such as land use/land cover change (LUCC) [24], nighttime light index [25], and impervious surface ratio [48]—to analyze urban expansion, economic activity, and population distribution, thereby indirectly characterizing human activity intensity. This provides essential support for refined monitoring of urbanization processes and informed decision-making.

As an economic special zone under China’s reform and opening-up policy and the core city of the Guangdong–Hong Kong–Macao Greater Bay Area, Shenzhen was the first city in China to achieve complete administrative urbanization. Over the past few decades, accompanied by intensive economic development and urban expansion, Shenzhen’s natural ecosystems have been subject to varying degrees of degradation in terms of components, structure, and function, presenting significant challenges to ecological civilization construction. In the process of urbanization, balancing ecological environment protection with regional sustainable development has become a major challenge for Shenzhen. Although existing studies have conducted dynamic assessments and evolution analyses of the ecological conditions and urbanization process in Shenzhen [49,50,51,52,53], revealing the spatiotemporal characteristics of ecological quality and the speed and patterns of urban expansion. These studies still suffer from the following limitations: (1) In terms of research perspective, most studies focus on unilateral analyses of either urbanization or ecological conditions, with insufficient quantitative research on their coupling coordination relationship, particularly the long-term dynamic coupling mechanisms lack systematic exploration. (2) Methodologically, traditional studies often rely on single indicators or static models, making it difficult to comprehensively reflect the dynamic interactions between urbanization and the ecological environment. Additionally, there are data comparability issues in long-term ecological monitoring in subtropical regions with frequent cloud cover and rainfall.

To address the aforementioned limitations, this study proposes an indicator normalization method based on invariant regions and a multi-temporal fusion principal component analysis (PCA) technique to construct a Remote Sensing Ecological Index (RSEI), effectively resolving data comparability challenges in long-term ecological monitoring in cloudy and rainy regions. Furthermore, a coupling coordination degree (CCD) model was developed to systematically integrate the Impervious Surface Index and an improved RSEI, enabling dynamic quantitative assessment of the interactions between urbanization and the ecological environment. Within this methodological framework, this study focuses on examining the coupling relationship between urban development and the ecological environment in Shenzhen from 1990 to 2020. The specific research objectives include: (1) analyzing the spatiotemporal evolution characteristics of urbanization and ecological quality, and (2) evaluating the dynamic changes and spatial differentiation patterns of their coupling coordination degree. These advancements not only provide a novel analytical perspective for understanding human–environment interactions in high-density megacities but also offer a scientific basis for regional sustainable development decision-making.

2. Materials and Methods

2.1. Study Area

Shenzhen is located in the southern coastal region of China, south of the Tropic of Cancer, and adjacent to Hong Kong, with geographic coordinates between 22°27′ N–22°52′ N and 113°46′ E–114°33′ E (Figure 1). The total area of Shenzhen is approximately 1997.47 km². It has a subtropical marine monsoon climate, characterized by hot, rainy summers and mild weather during other seasons, with an annual average temperature of 22.4 °C and an average annual precipitation of 1935.8 mm.

Shenzhen governs 10 districts: Guangming, Bao’an, Nanshan, Futian, Luohu, Yantian, Longhua, Longgang, Pingshan, and Dapeng District (Figure 1). The city’s topography is primarily flat plains and plateaus, covering about 78% of the total area. The terrain is higher in the southeast and lower in the northwest, with low hills and gentle plateaus, while the coastal areas in the west are characterized by coastal plains.

As China’s first Special Economic Zone, Shenzhen is a window for reform and opening-up, rapidly evolving from a small fishing village to one of the four first-tier cities in China and becoming the core engine of the Guangdong–Hong Kong–Macao Greater Bay Area (GBA). It is one of the most economically dynamic cities in mainland China and an important global financial and port center. Shenzhen was the first city in China to achieve 100% urbanization, and by the end of 2023, its permanent population reached 17.79 million.

2.2. Data Sources and Preprocessing

This study uses Landsat series remote sensing imagery as the base data. The data for 1990, 2000, and 2010 come from Landsat-5 satellite TM images, while the 2020 data are from Landsat-8 satellite OLI/TIRS images. The study area covers Shenzhen, with corresponding Landsat path/row numbers 122/44 and 121/44. The Landsat data were provided by the United States Geological Survey (USGS) and integrated into the Google Earth Engine (GEE) platform. This study selects surface reflectance products (Surface Reflectance, SR) that have undergone radiometric and atmospheric correction.

Shenzhen experiences cloudy and rainy weather year-round, which significantly affects the availability of valid Landsat images. The imagery from different years is influenced by seasonal and weather differences, leading to significant variations in surface reflectance. To enhance the consistency and comparability of the long-term ecological quality assessment and impervious surface extraction results, this study changes the previous approach of selecting a single cloud-free or low-cloud image from a specific monitoring year. Instead, it selects images from the vegetation growing season (September–October) and non-growing season (December–January) within three years before and after the monitoring year. Cloud statistics show that these two seasons have fewer cloudy and rainy days in Shenzhen, making them more likely to yield valid images. Additionally, to further reduce cloud interference, image data with cloud cover less than 30% were selected.

The preprocessing of the Landsat SR product includes cloud/shadow removal, remote sensing index calculation, and median compositing. Cloud and shadow removal is performed using the quality assurance (QA) band in the Landsat SR product to mark clouds and shadows. By marking these interference factors and creating masks, cleaner images are generated. Remote sensing images are applied to RSEI calculation and impervious surface extraction, as detailed in Section 3.1 and Section 3.2. The remote sensing indices are subjected to median compositing for both the vegetation growing season and non-growing season, and the results from both periods are averaged. This method effectively reduces the impact of weather, precipitation, solar angle, and other factors, minimizing interference from extreme observations, and allowing the indicators to more accurately reflect the average levels of each period, avoiding the impact of extreme observations on the overall assessment.

For water body masking, the permanent water bodies are identified using the wetland primary category from the ChinaLand Cover [54] product for the years 1990–2020, and a corresponding water body mask is constructed to effectively mask the water bodies.

2.3. Methods

2.3.1. Remote Sensing Ecological Index (RSEI) Model

The RSEI model integrates four key indicators that are most closely related to human daily life: greenness, wetness, heat, and dryness [41,42].

The greenness factor is represented by the Normalized Difference Vegetation Index (NDVI), as shown in Equation (1).

$$NDVI=\left(B_{nir}-B_{red}\right)/\left(B_{nir}+B_{red}\right)$$

(1)

The humidity factor is represented by the wetness component of the Tassel-Cap Transform, as shown in Equations (2) and (3).

$$\begin{array}{cc}\hfill {WET}_{TM}=& 0.0315B_{blue}+0.2021B_{green}+0.3012B_{red}+0.1594B_{nir}\hfill \\ & \hspace{1em}\hspace{1em} -0.6806B_{swir1}-0.6109B_{swir2}\hfill \end{array}$$

(2)

$$\begin{array}{cc}\hfill {WET}_{OLI}=& 0.1511B_{blue}+0.1973B_{green}+0.3283B_{red}+0.3407B_{nir}\hfill \\ & \hspace{1em}\hspace{1em} -0.7117B_{swir1}-0.4559B_{swir2}\hfill \end{array}$$

(3)

where ${WET}_{TM}$ and ${WET}_{OLI}$ correspond to the calculation methods of the wetness component derived from the Landsat-5 TM sensor and the Landsat-8 OLI sensor, respectively.

The dryness factor is represented by the Normalized Difference Built-up Soil Index (NDBSI), which is composed of the Soil Index (SI) (Equation (4)) and the Index-based Built-up Index (IBI) (Equation (5)), as shown in Equation (6).

$$\mathrm{S}\mathrm{I}={\displaystyle \frac{\left[\left(B_{swir1}+B_{red}\right)-\left(B_{nir}+B_{blue}\right)\right]}{\left[\left(B_{swir1}+B_{red}\right)+\left(B_{nir}+B_{blue}\right)\right]}}$$

(4)

$$\mathrm{I}\mathrm{B}\mathrm{I}={\displaystyle \frac{2B_{swir1}/\left(B_{swir1}+B_{nir}\right)-\left[B_{nir}/\left(B_{nir}+B_{red}\right)+B_{green}/\left(B_{green}+B_{swir1}\right)\right]}{2B_{swir1}/\left(B_{swir1}+B_{nir}\right)+\left[B_{nir}/\left(B_{nir}+B_{red}\right)+B_{green}/\left(B_{green}+B_{swir1}\right)\right]}}$$

(5)

$$\mathrm{N}\mathrm{D}\mathrm{B}\mathrm{S}\mathrm{I}={\displaystyle \frac{\left(IBI+SI\right)}{2}}$$

(6)

In Equations (1)–(6), $B_{blue}$, $B_{green}$, $B_{red}$, $B_{nir}$, $B_{swir1}$ and $B_{swir2}$ represent the surface reflectance values of the blue, green, red, near-infrared, shortwave infrared 1, and shortwave infrared 2 bands of the Landsat imagery, respectively.

The heat factor is represented by land surface temperature (LST). In this study, the LST indicator is derived from the SR products of the Landsat series imagery, with the land surface temperature data selected as the representative metric. These temperature data are obtained from the thermal infrared bands of Landsat, and the land surface temperature information is retrieved through measurements of surface radiation, followed by calibration and inversion processes.

Shenzhen experiences frequent cloud cover and rainfall throughout the year, and its ecological conditions have undergone significant changes over the past 30 years. To enhance the consistency and comparability of long-term ecological quality assessments, this study introduces two improvements to the original RSEI: (1) In terms of index normalization, a normalization method based on invariant areas is proposed. Using land cover data of Shenzhen from 1990 to 2020, a layer of regions with no significant land cover changes over the 30-year period is constructed. Based on this invariant area layer, the four indicator factors from different monitoring periods are masked. The maximum and minimum values of the masked indicator factors are then used to normalize the original (unmasked) indicator factors. (2) In terms of principal component analysis (PCA), the ecological indicator factors from all four monitoring periods are integrated and subjected to a unified PCA. The resulting principal components are consistently applied across all four monitoring periods to generate a unified RSEI (URSEI), thereby ensuring the comparability of ecological quality changes over different time periods. The formulas are as follows:

$$\left[{\lambda }_1, {\lambda }_2, {\lambda }_3,{\lambda }_4\right]=PC1\left(NDVI,WET,NDBSI,LST\right)$$

(7)

$$RSEI={\lambda }_{1}NDVI+{\lambda }_{2}WET+{\lambda }_{3}NDBSI+{\lambda }_{4}LST$$

(8)

where ${\lambda }_i$ represents the eigenvector of the first principal component (PC1), satisfying $\sum {\lambda }_i^2=1$.

2.3.2. Impervious Surface Extraction

One of the key ways urbanization alters ecosystems is by increasing the coverage of impervious surfaces. Impervious surfaces refer to ground types that prevent surface water from infiltrating into the soil. Urbanization is inevitably accompanied by the transformation of natural vegetation cover into impervious surfaces, such as paved roads, building rooftops, and parking lots [55]. In recent years, impervious surface area (ISA) has increasingly been used as a critical environmental indicator [56,57,58].

Various methods have been developed to extract ISA from remote sensing imagery and assess its dynamics. These methods primarily include machine learning techniques, spectral unmixing approaches, and spectral index (SI) methods [58]. Compared to machine learning and Spectral Mixture Analysis (SMA) techniques, index-based methods require less preprocessing and are therefore simpler to implement in practice and offer higher computational efficiency [59,60].

In recent years, researchers have developed a variety of spectral indices (SIs), including the Impervious Surface Area Index [18], the Normalized Difference Impervious Surface Index (NDISI) [61], the Biophysical Composition Index (BCI) [59,60], an improved NDISI [62], and the Combined Built-up Index (CBI) [63]. This study adopts the BCI index, which is calculated as follows:

$$BCI={\displaystyle \frac{\frac{H+L}{2}-V}{\frac{H+L}{2}+V}}$$

(9)

where $H$ represents the normalized first principal component (TC1), indicating high albedo; $V$ represents the normalized second principal component (TC2), indicating vegetation; and $L$ represents the normalized third principal component (TC3), indicating low albedo. These three factors can be derived using the following formulas:

$$H={\displaystyle \frac{TC1-{TC1}_{min}}{{TC1}_{max}-{TC1}_{min}}}, V={\displaystyle \frac{TC2-{TC2}_{min}}{{TC2}_{max}-{TC2}_{min}}}, L={\displaystyle \frac{TC3-{TC3}_{min}}{{TC3}_{max}-{TC3}_{min}}}$$

(10)

where $TCi (i=1, 2, 3)$ (i = 1, 2, 3) represents the first three Tasseled Cap (TC) components, ${TCi}_{max}$ and ${TCi}_{min}$ denote the maximum and minimum values of the TC component, respectively.

In this study, we first performed median compositing on imagery from the growing and non-growing seasons for the years 1990, 2000, 2010, and 2020. Then, the BCI index was calculated, and impervious surfaces were extracted for each year. Finally, further validation was conducted on the extracted impervious surfaces. Taking the year 1990 as an example, for each pixel, it was classified as impervious only if it was identified as such in both the growing and non-growing seasons of that year.

2.3.3. Impervious Surface Proportion and Expansion Indicators

The impervious surface area percentage refers to the proportion of built-up land area in a given year relative to the total area of the study region in that year. It is used to assist in analyzing urban expansion, and the formula is:

$$P=A_{build}/A_{total}$$

(11)

where $A_{build}$ is the impervious surface area for a specific year, and $A_{total}$ is the total area of the study region for that year. When the impervious surface percentage increases, it indicates the expansion of impervious surfaces, thereby demonstrating the occurrence of urban expansion.

The impervious surface expansion rate $V$ represents the average annual increase in impervious surface area during the study period (Equation (12)). The impervious surface expansion intensity $R$ refers to the growth rate of impervious surfaces over a specific period (Equation (13)), reflecting the intensity of impervious surface changes within the study area.

$$V=\left(A_j-A_i\right)/T$$

(12)

$$R=\left(A_j-A_i\right)/\left(A_{i}T\right)\times 100\%$$

(13)

where $A_i$ and $A_j$ represent the impervious surface areas at the beginning and end of the study period, respectively; $T$ is the time interval between the years.

2.3.4. Coupling Coordination Model

The coupling coordination model originates from physics and refers to the phenomenon where two or more systems interact and influence each other to achieve coordinated development [36,37,38,39]. Therefore, in this study, the coupling model is used to measure the interaction between urbanization and the ecological environment. The formula is as follows:

$$C=2\times \sqrt{\left(U\times E\right)/{\left(U+E\right)}^2}$$

(14)

To avoid the situation of “false coordination” where the coupling degree value is high due to both systems having relatively low evaluation indices, a comprehensive coordination index for the two systems is introduced. Based on this, a coupling coordination model is established to more comprehensively assess the coordination level between urbanization and ecological environmental quality.

$$T=\alpha U+\beta E$$

(15)

$$D=\sqrt{C\times T}$$

(16)

where $U$ and $E$ represent the evaluation indices for urbanization and ecological environmental quality, respectively; $C$ is the coupling degree; $T$ is the comprehensive coordination index for the two systems; $\alpha$ and $\beta$ are the relative importance coefficients for the two subsystems. Existing research indicates that the relative values of $\alpha$ and $\beta$ do not significantly affect the overall trend of the coupling coordination degree. Based on the relevant literature and the urbanization development guidelines and functional positioning followed by the study area during the research period, $\alpha$ and $\beta$ are assigned values of 0.65 and 0.35, respectively. $D$ represents the coupling coordination degree, with a value range of [0, 1]. A larger $D$ value indicates a high-level mutual promotion relationship between the two systems, and the entire system tends to be more stable.

Furthermore, to better analyze the coupling coordination between urbanization and the ecological environment in Shenzhen, the coupling coordination degree is classified into levels using an equal interval method. Considering the small differences in coordination degrees among some cities, and referring to the research [64,65,66], the coupling status of urbanization and the ecological environment is divided into 10 categories (

Table 1. Classification table of coupling coordination levels.

Major Categories	Subcategories	Range
Uncoordinated Development	Extreme imbalance	0 ≤ D ≤ 0.1
	Severe imbalance	0.1 < D ≤ 0.2
	Moderate imbalance	0.2 < D ≤ 0.3
	Mild imbalance	0.3 < D ≤ 0.4
Transformational Development	Near imbalance	0.4 < D ≤ 0.5
	Marginal coordination	0.5 < D ≤ 0.6
Coordinated Development	Primary coordination	0.6 < D ≤ 0.7
	Intermediate coordination	0.7 < D ≤ 0.8
	Good coordination	0.8 < D ≤ 0.9
	High-quality coordination	0.9 < D ≤ 1

).

2.3.5. Spatial Autocorrelation Analysis

Spatial autocorrelation analysis is used to examine whether the attribute values of spatial elements at specific locations are related to the attribute values at neighboring spatial points. It includes two aspects: global spatial autocorrelation and local spatial autocorrelation. By calculating the global Moran’s I index, the overall spatial association and spatial variation within the region at different time periods can be analyzed [67,68]. The formula is as follows:

$$I_g={\displaystyle \frac{n\sum _{i=1}^n\sum _{j=1}^n\left(X_i-\overline{X}\right)\left(X_j-\overline{X}\right)}{\left(\sum _{i=1}^n\sum _{j=1}^n{W}_{ij}\right)\sum _{i=1}^n{\left(X_i-\overline{X}\right)}^2}}$$

(17)

where $n$ is the sample size, $X_i$ and $X_j$ are the observed values of attribute $X$ at spatial locations $i$ and $j$, $\overline{X}$ is the average value of attribute $X_i$, and $W_{ij}$ is the spatial weight matrix. The range of Moran’s I index is (−1, 1). When $I_g$ > 0, it indicates a positive correlation between spatial units; when $I_g$ < 0, it indicates a negative correlation between spatial units; and when $I_g$ = 0, it indicates no correlation, representing a random distribution.

The local Moran’s I index can be used to assess the correlation and statistical significance between a region and its neighboring regions, visualized through Local Indicators of Spatial Association (LISA). The formula for calculating the local Moran’s I is as follows:

$$I_l={\displaystyle \frac{n\left(X_i-\overline{X}\right)\sum _{j=1}^n{W}_{ij}\left(X_j-\overline{X}\right)}{\sum _{i=1}^n{\left(X_i-\overline{X}\right)}^2}}$$

(18)

where the meanings of the parameters are the same as those in the previous Equation (17).

3. Results

3.1. Spatiotemporal Evolution of Impervious Surfaces in Shenzhen

The spatiotemporal evolution of impervious surfaces in Shenzhen is shown in Figure 2. Over the past 30 years, the spatial distribution of impervious surfaces in Shenzhen has exhibited significant phased changes. From 1990 to 2000, during the rapid expansion phase, the area of impervious surfaces increased sharply, primarily radiating outward from the original urban clusters (Nanshan, Futian, Luohu), and extending toward regions like Bao’an, Longhua, and Longgang. This change was closely related to the large-scale urban construction in the early stages of Shenzhen. From 2000 to 2010, the pace of urban expansion in Shenzhen slowed, and the growth of impervious surfaces stabilized. The spatial pattern showed a filling-type development, with a triangular high-density cluster formed in the southern parts of Nanshan and Bao’an Districts, a clear axial filling in the central areas of Futian and Longhua, and a more compact block-like distribution in the Eastern Longgang District. During the period from 2010 to 2020, a downward trend in impervious surfaces was observed in areas like Nanshan and Futian, which was closely related to Shenzhen’s strict ecological protection policies and urban renewal strategies.

To further understand the distribution patterns of impervious surfaces in the study area, the area, proportion, expansion rate, and expansion intensity of impervious surfaces in Shenzhen and its districts were calculated. The results are shown in

Table 2. Statistics of impervious surfaces at district level.

District	1990		2000		2010		2020
	Area/km2	Percent/%	Area/km2	Percent/%	Area/km2	Percent/%	Area/km2	Percent/%
Luohu	27.99	35.14	29.26	36.74	25.84	32.45	23.96	30.08
Futian	45.00	61.45	48.22	65.86	38.89	53.12	26.98	36.85
Nanshan	55.49	32.17	72.60	42.09	78.03	45.23	59.77	34.65
Bao’an	103.50	27.67	166.69	44.56	202.36	54.10	193.96	51.85
Yantian	7.32	10.46	14.16	20.25	16.58	23.70	13.99	20.01
Longhua	53.94	30.67	87.99	50.03	98.40	55.95	89.16	50.70
Pingshan	20.82	12.42	42.13	25.13	49.56	29.56	49.15	29.32
Guangming	21.73	13.94	46.84	30.05	60.29	38.68	58.46	37.50
Longgang	109.43	28.22	168.42	43.43	193.37	49.86	173.27	44.68
Dapeng	12.93	4.47	23.93	8.27	22.20	7.67	20.53	7.09
Shenzhen	458.15	23.54	700.26	35.98	785.53	40.36	709.23	36.44

and

Table 3. Expansion rate and intensity of impervious surfaces at district level.

District	1990–2000		2000–2010		2010–2020
	Expansion Speed/km2·a−1	Expansion Intensity/%	Expansion Speed/km2·a−1	Expansion Intensity/%	Expansion Speed/km2·a−1	Expansion Intensity/%
Luohu	0.13	0.45	−0.34	−1.17	−0.19	−0.73
Futian	0.32	0.72	−0.93	−1.93	−1.19	−3.06
Nanshan	1.71	3.08	0.54	0.75	−1.83	−2.34
Bao’an	6.32	6.11	3.57	2.14	−0.84	−0.42
Yantian	0.68	9.34	0.24	1.71	−0.26	−1.56
Longhua	3.41	6.31	1.04	1.18	−0.92	−0.94
Pingshan	2.13	10.24	0.74	1.76	−0.04	−0.08
Guangming	2.51	11.56	1.35	2.87	−0.18	−0.30
Longgang	5.90	5.39	2.50	1.48	−2.01	−1.04
Dapeng	1.10	8.51	−0.17	−0.72	−0.16	−0.75
Shenzhen	24.21	5.28	8.53	1.22	−7.63	−0.97

. From 1990 to 2020, the area of impervious surfaces in Shenzhen grew significantly, increasing from 458.15 km² in 1990 to 709.23 km² in 2020, with an expansion of 251.08 km² over 30 years, at an average annual growth rate of 1.47%. The proportion of impervious surfaces increased from 23.54% to 36.44%. At the district level, in 1990, Futian District had the highest proportion of impervious surfaces, at 61.45%, followed by Luohu District at 35.14%, and Dapeng District had the lowest at only 4.47%. By 2020, Bao’an District had the largest proportion, at 51.85%, while Dapeng District remained the lowest, at 7.09%.

In terms of changes in the proportion of impervious surfaces, from 1990 to 2020, the proportion of impervious surfaces in Luohu and Futian Districts showed a downward trend. Notably, Futian District saw the largest decline, from 61.45% to 36.85%, a drop of over 20 percentage points. Luohu District decreased from 35.14% to 30.08%. On the other hand, the proportion of impervious surfaces in other districts increased, with Bao’an District showing the largest increase, from 27.67% to 51.85%, followed by Guangming District, where the proportion rose from 13.94% to 37.50%. The increases in Nanshan District and Dapeng District were smaller, with changes of 2.48% and 2.62%, respectively. Regarding the expansion rate, from 1990 to 2010, Shenzhen experienced a period of concentrated expansion of impervious surfaces. The expansion rate and intensity were the highest during 1990–2000, at 24.21 km²·a⁻¹ and 5.28%, respectively. From 2000 to 2010, the expansion rate and intensity decreased to 8.53 km²·a⁻¹ and 1.22%. Between 2010 and 2020, both the expansion rate and intensity continued to decline, with negative growth in the area of impervious surfaces. Since 1990, the expansion rate and intensity of impervious surfaces in Futian, Nanshan, Bao’an, Yantian, Longhua, Pingshan, Guangming, and Longgang Districts have continuously decreased. Luohu District showed a trend of first decreasing and then increasing in both expansion rate and intensity, while Dapeng District experienced a decrease in expansion rate and a continuous decline in expansion intensity.

3.2. Ecological Quality Analysis

Table 4. The results of PCA for all phases and unified model.

Year	PC1		The Eigenvector Corresponding to Each Indicator
	Eigenvalue	Eigen Percent	NDVI	WET	NDBSI	LST
1990	0.1338	75.33%	0.2766	0.5919	0.5926	0.4712
2000	0.1413	75.80%	0.3417	0.5692	0.5313	0.5263
2010	0.1859	80.96%	0.3470	0.5096	0.5298	0.5831
2020	0.1942	85.25%	0.3956	0.5341	0.5491	0.5067
Unified	0.1463	81.59%	0.3620	0.5235	0.5288	0.5615

presents the results of PCA for four monitoring phases from 1990 to 2020, along with the unified model. It includes the eigenvalue, contribution rate, and eigenvector for PC1. The contribution rate of PC1 exceeds 70% in all cases, and the contribution rate for the unified result reaches 81.59%, indicating that PC1 integrates most of the features of the four indicators and can be used to create a comprehensive ecological index. In the unified eigenvector, the PC1 weights for the four indicator factors are 0.362, 0.524, 0.529, and 0.562, respectively. Since NDBSI and LST have been inversely normalized, their eigenvectors are positive.

The average RSEI values of Shenzhen in 1990, 2000, 2010, and 2020 were 0.477, 0.429, 0.431, and 0.491, respectively, showing a trend of first decreasing and then increasing, with an overall improvement in ecological quality. The overall RSEI of Shenzhen was below 0.5 in all years, indicating that the city’s overall ecological quality was relatively low, with the lowest value in 2000 (0.429) and the highest in 2020 (0.491). The RSEI increased from 0.477 in 1990 to 0.491 in 2020, representing a growth of 2.94%. Although this growth appears slow, it reflects the trend of gradual restoration and improvement of Shenzhen’s ecological environment during the urbanization process.

To provide a more intuitive and quantitative visual analysis of the RSEI, this study employs an equal interval classification method, dividing the RSEI values into five levels with an interval of 0.2: [0, 0.2), [0.2, 0.4), [0.4, 0.6), [0.6, 0.8), and [0.8, 1], representing poor, relatively poor, moderate, good, and excellent ecological quality, respectively. The classification map of RSEI for Shenzhen from 1990 to 2020 is shown in Figure 3. The areas and proportions of each grade are shown in

Table 5. Area and percentage of Shenzhen’s RSEI grades from 1990 to 2020.

RSEI Grades	1990		2000		2010		2020
	Area (km2)	Percentage	Area (km2)	Percentage	Area (km2)	Percentage	Area (km2)	Percentage
[0, 0.2)	398.43	20.9%	509.69	26.8%	614.72	32.3%	437.96	23.0%
[0.2, 0.4)	417.68	21.9%	482.12	25.3%	364.88	19.2%	418.28	22.0%
[0.4, 0.6)	422.30	22.2%	361.82	19.0%	310.01	16.3%	281.81	14.8%
[0.6, 0.8)	382.56	20.1%	302.41	15.9%	309.43	16.3%	356.37	18.7%
[0.8, 1)	282.71	14.9%	248.61	13.1%	301.99	15.9%	410.64	21.6%

.

The “poor” areas are concentrated in Bao’an, Guangming, Longhua, Longgang, as well as the Futian and Nanshan Districts. These urban areas have experienced rapid urbanization and industrialization, with significant increases in built-up land and a decrease in vegetation coverage. In contrast, the “excellent” areas are mainly located in Dapeng, Yantian, the eastern part of Luohu District, the southern and western parts of Pingshan District, and the urban fringe, where mountainous and hilly terrain predominates, development density is lower, and vegetation coverage is high. This spatial distribution difference reflects the ecological changes in Shenzhen’s various regions during the urbanization process.

At the district level (

Table 6. The RSEI value and ranking at district level.

District	RSEI
	1990	Ranking	2000	Ranking	2010	Ranking	2020	Ranking
Luohu	0.51	4	0.49	3	0.54	3	0.59	3
Futian	0.29	10	0.29	10	0.40	6	0.46	6
Nanshan	0.43	6	0.38	6	0.41	5	0.49	5
Bao’an	0.42	7	0.37	7	0.31	9	0.37	9
Yantian	0.69	2	0.65	2	0.67	2	0.72	2
Longhua	0.34	9	0.30	9	0.28	10	0.36	10
Pingshan	0.52	3	0.46	4	0.46	4	0.52	4
Guangming	0.44	5	0.39	5	0.34	7	0.39	8
Longgang	0.39	8	0.33	8	0.32	8	0.41	7
Dapeng	0.76	1	0.72	1	0.79	1	0.80	1
Shenzhen	0.477	-	0.429	-	0.431	-	0.491	-

), the highest values each year were consistently in Dapeng District, which is closely linked to the region’s strong ecological protection policies implemented since its establishment in 2011. Yantian District, due to its limited industrial development, has maintained good ecological quality, with its RSEI value consistently ranking second. Futian District initially had poor ecological quality due to urban construction and industrial development. However, in recent years, with Shenzhen placing more emphasis on ecological protection, Futian has implemented a series of greening projects and environmental restoration measures, leading to a gradual improvement in its RSEI value. In contrast, Bao’an and Guangming saw a decline in ecological quality, which is closely related to their ongoing urban development and industrialization. Particularly, as Shenzhen’s urban space expanded, Bao’an and Guangming Districts, as key hubs of industrialization, have been under increasing development pressure. Longhua District had the worst ecological quality and consistently ranked last. As an emerging industrial development zone, rapid urbanization and industrial expansion are the main reasons for its long-term low ecological quality.

3.3. Coupling Coordination Analysis

To systematically assess the coupling coordination relationship between urbanization and ecological quality in Shenzhen, this study employs the impervious surface proportion as an indicator of urbanization level, and the RSEI to represent ecological quality. Based on the coupling coordination degree model (Formulas (14)–(16)), the coupling coordination degrees of Shenzhen and its ten administrative districts from 1990 to 2020 were calculated (

Table 7. Coupling coordination degree at district level from 1990 to 2020.

District	CCD	Grade	CCD	Grade	CCD	Grade	CCD	Grade
	1990		2000		2010		2020
Luohu	0.634	Primary coordination	0.638	Primary coordination	0.622	Primary coordination	0.616	Primary coordination
Futian	0.682	Primary coordination	0.701	Intermediate coordination	0.695	Primary coordination	0.631	Primary coordination
Nanshan	0.597	Marginal coordination	0.636	Primary coordination	0.662	Primary coordination	0.625	Primary coordination
Bao’an	0.564	Marginal coordination	0.644	Primary coordination	0.665	Primary coordination	0.679	Primary coordination
Yantian	0.457	Near imbalance	0.553	Marginal coordination	0.585	Marginal coordination	0.562	Marginal coordination
Longhua	0.564	Marginal coordination	0.643	Primary coordination	0.660	Primary coordination	0.674	Primary coordination
Pingshan	0.456	Near imbalance	0.556	Marginal coordination	0.586	Marginal coordination	0.601	Primary coordination
Guangming	0.457	Near imbalance	0.573	Marginal coordination	0.607	Primary coordination	0.617	Primary coordination
Longgang	0.563	Marginal coordination	0.628	Primary coordination	0.654	Primary coordination	0.658	Primary coordination
Dapeng	0.367	Mild imbalance	0.431	Near imbalance	0.430	Near imbalance	0.422	Near imbalance
Shenzhen	0.548	Marginal coordination	0.618	Primary coordination	0.642	Primary coordination	0.636	Primary coordination

). In addition, a 1 km × 1 km grid was generated to cover the entire Shenzhen area, and the coupling coordination degree for each grid cell was computed. Using thematic mapping techniques, the spatial distribution of different levels of coupling coordination was visually represented, as shown in Figure 4. The areas and proportions of each grade are shown in

Table 8. Area and percentage of Shenzhen’s CCD grades from 1990 to 2020.

CCD Grades	1990		2000		2010		2020
	Area (km2)	Percentage	Area (km2)	Percentage	Area (km2)	Percentage	Area (km2)	Percentage
[0, 0.1)	244.10	12.5%	184.94	9.5%	219.12	11.2%	238.56	12.2%
[0.1, 0.2)	135.50	7%	87.74	4.5%	86.67	4.4%	84.01	4.3%
[0.2, 0.3)	227.84	11.7%	142.11	7.3%	145.86	7.5%	127.43	6.5%
[0.3, 0.4)	272.10	14.0%	177.78	9.1%	143.47	7.4%	142.33	7.3%
[0.4, 0.5)	288.10	14.8%	220.87	11.3%	178.46	9.2%	171.43	8.8%
[0.5, 0.6)	319.20	16.4%	349.63	18.0%	296.36	15.2%	248.57	12.8%
[0.6, 0.7)	371.92	19.1%	563.66	29.0%	628.40	32.3%	610.77	31.4%
[0.7, 0.8)	84.18	4.3%	210.79	10.8%	225.76	11.6%	305.14	15.7%
[0.8, 0.9)	3.89	0.2%	9.30	0.5%	22.71	1.2%	18.57	1%

.

In 1990, the coupling coordination degree across Shenzhen’s districts ranged from [0.367, 0.682]. Among them, only the Dapeng District had a value below 0.4, falling into the grade of mild imbalance. Luohu and Futian Districts recorded values above 0.6, indicating a relatively high level of coordination and falling into the grade of primary coordination. By 2020, the coupling coordination degrees ranged from [0.422, 0.679]. Except for Luohu and Futian, where coordination levels slightly declined, all other districts experienced significant improvement. The number of districts reaching a coupling coordination degree above 0.6 (primary coordination) increased to eight. Meanwhile, Dapeng and Yantian Districts were categorized as near imbalance and marginal coordination, respectively.

From 1990 to 2020, the coupling coordination degree between urbanization and ecological quality in Shenzhen showed a significant upward trend (Figure 4). The coupling coordination index increased from 0.548 in 1990 to 0.636 in 2020, a growth of 16.1%, indicating that Shenzhen has made substantial progress in balancing urban development with ecological protection. The period from 1990 to 2010 was a phase of rapid improvement, with the coupling coordination degree growing at an average annual rate of 0.79%, while from 2010 to 2020, there was a slight fluctuation of 1%. It is noteworthy that over the past 30 years, Shenzhen’s coupling coordination level rose from the “marginal coordination” level to the “primary coordination” level. This transformation not only confirms the feasibility of coordinated urbanization and ecological development in megacities but also provides important references for the sustainable development of similar cities.

As shown in Figure 4, the coupling coordination degree between urbanization and ecological quality in Shenzhen exhibits a clear spatial differentiation pattern, characterized by higher values in the western and central regions and lower values in the southeastern region. Specifically, in the Southeastern Dapeng District, land cover is predominantly forest, and the ecological quality has remained in good condition over the long term. However, due to topographic constraints and strict ecological protection policies, development intensity in this area has remained low, and the urbanization process has lagged behind, resulting in a consistently low coupling coordination degree throughout the study period. In contrast, the western and central regions serve as Shenzhen’s core urban development zones, where both RSEI values and impervious surface coverage remain at moderate to high levels. These areas exhibit better coupling coordination performance, indicating that they have achieved a relatively balanced relationship between urban development and ecological protection during the rapid urbanization process.

3.4. Change Analysis of the Coupling Coordination Degree

Based on the spatiotemporal distribution of the coupling coordination degree at the 1 km grid scale shown in Figure 4, this study further analyzes the changes in coupling coordination degree across different periods from 1990 to 2020, as illustrated in Figure 5. The changes (Δ) are classified into eight levels: severe degradation (−1.0 < Δ ≤ −0.3), rapid degradation (−0.3 < Δ ≤ −0.2), moderate degradation (−0.2 < Δ ≤ −0.1), slight degradation (−0.1 < Δ ≤ 0), slight improvement (0 < Δ ≤ 0.1), moderate improvement (0.1 < Δ ≤ 0.2), rapid improvement (0.2 < Δ ≤ 0.3), and high-speed improvement (0.3 < Δ ≤ 1.0).

From 1990 to 2000, the overall coupling coordination degree in Shenzhen exhibited an upward trend, predominantly characterized by slight improvement. Notable areas of moderate improvement were observed in the western part of Bao’an District, the northwestern part of Longhua District, and the northern part of Longgang District. Between 2000 and 2010, the coupling coordination degree generally showed slight degradation, with moderate degradation clusters emerging in the southeastern part of Bao’an District and the northern part of Longgang District. In contrast, emerging districts such as Guangming and Pingshan maintained a slight improvement trend. From 2010 to 2020, core urban districts such as Nanshan and Futian experienced widespread slight degradation, and even Dapeng District—designated as an ecological conservation zone—saw large areas of slight degradation. In contrast, peripheral areas such as Longhua and Guangming continued to exhibit a trend of improvement. Overall, during the period from 1990 to 2020, degradation areas were mainly concentrated in the southeastern part of Bao’an District, as well as in Futian, Luohu, and Dapeng District, while the coupling coordination degree in most other regions was dominated by slight improvement or higher.

3.5. Spatial Autocorrelation Analysis of the Coupling Coordination Degree

To further investigate the spatial agglomeration characteristics of the coupling coordination degree between urbanization and ecological environment quality in Shenzhen, this study utilized the ArcGIS 10.2 platform and generated a 1 km × 1 km regular grid covering the entire city using the fishnet tool, resulting in a total of 1937 sample units. Each grid cell was taken as the basic analytical unit for calculating its coupling coordination degree value, and spatial correlation analysis was conducted using GeoDa software 1.14.0.

Global spatial autocorrelation results (

Table 9. Global Moran’s I index of coupling coordination degree in Shenzhen.

Year	Moran’s I	z	p
1990	0.641	56.248	0.001
2000	0.636	56.058	0.001
2010	0.620	54.393	0.001
2020	0.631	54.435	0.001

) show that the Moran’s I values for all four periods in Shenzhen are positive, with the significance test p-values less than 0.001. This indicates that the coupling coordination degree between urbanization and ecological environment quality in Shenzhen exhibits a significant positive spatial autocorrelation, meaning that neighboring areas tend to have similar levels of coupling coordination, clustering together spatially. Notably, the Moran’s I index reached its highest value of 0.641 in 1990, indicating the strongest spatial clustering of the coupling coordination degree during that period. In the subsequent study periods, the Moran’s I index showed a slight decline followed by a minor rebound, but the overall fluctuations were minimal, reflecting that the coupling coordination relationship between urbanization and the ecological environment in Shenzhen maintained a relatively stable spatial clustering pattern.

Using local spatial autocorrelation analysis, LISA cluster maps were generated to further reveal the spatial distribution pattern of the coupling coordination degree (Figure 6). In the figure, high-high (H-H) clusters indicate areas of high coupling coordination surrounded by other high-value areas, low-low (L-L) clusters represent low coupling coordination areas surrounded by other low-value areas, high-low (H-L) clusters denote high-value areas surrounded by low-value areas, and low-high (L-H) clusters show low-value areas surrounded by high-value areas. The study found that from 1990 to 2020, the spatial distribution pattern of the coupling coordination degree in Shenzhen remained generally stable. High–high clusters were mainly concentrated in highly urbanized areas such as Bao’an District, Nanshan District, Futian District, Longhua District, Longgang District, and the western part of Luohu District. This suggests that these regions, while undergoing rapid urbanization, achieved coordinated development between urban growth and ecological environment through scientific land use planning and ecological management strategies. In contrast, low–low clusters were predominantly located in Dapeng New District, the southern part of Pingshan District, and the peripheral areas on the western urban edge. These zones are mostly ecological protection areas or fringe urban zones, with relatively low levels of urbanization but high ecological quality, resulting in spatial agglomerations characterized by low coordination degrees. Additionally, high–low and low–high clusters were fewer in number and more spatially scattered, further confirming that Shenzhen’s coupling coordination pattern is predominantly shaped by homogenous spatial clustering.

4. Discussion

4.1. The Stage-Specific Response Mechanism of Ecological Quality to the Urbanization Process

The spatiotemporal evolution of ecological quality and its dynamic relationship with urbanization is key to balancing the development of megacities with ecological protection. Based on a long-term time series analysis from 1990 to 2020, the interaction between urbanization and ecological quality in Shenzhen shows distinct stage-based characteristics.

From 1990 to 2000, Shenzhen experienced rapid urbanization, with its impervious surface area surging from 458.15 km² to 700.26 km². The annual urbanization growth rate reached 4.3%, and the expansion intensity of impervious surfaces stood at 5.28%. This rapid urban expansion imposed significant pressure on the ecological system, with the average RSEI dropping from 0.477 to 0.429, a decrease of 10.1%. Spatially, the ecological quality deterioration was particularly notable in emerging districts such as Bao’an and Guangming in the western part of the city, where RSEI declined by more than 11%. This deterioration is closely related to the rapid expansion of industrial land and the reduction of green space. This phenomenon aligns with the Environmental Kuznets Curve (EKC) hypothesis regarding initial ecological pressures during early-stage development [69]. However, compared to peer cities during the same period, Shenzhen exhibited more pronounced ecological quality deterioration, likely attributable to its ultra-rapid urban expansion pattern [70].

Between 2000 and 2010, with the implementation of policies such as the “Shenzhen Ecological Control Line Management Regulations,” the pace of urban expansion slowed significantly. The impervious surface expansion intensity decreased to 1.22%, and ecological quality entered a relatively stable period, with the average RSEI remaining around 0.43. Notably, core urban districts such as Futian and Luohu experienced significant improvements in ecological quality through urban renewal and green space optimization initiatives. The RSEI in Futian increased by 37.9%, and in Luohu by 10.2%. However, in districts such as Guangming and Bao’an, which continued to bear the main burden of urbanization, ecological quality showed limited improvement, indicating the persistent impact of overdevelopment and high-density urban construction. This stage featured a dual trend of “development alongside governance,” with initial signs that policy regulation was positively affecting ecological quality.

From 2010 to 2020, driven by the national ecological civilization strategy, Shenzhen’s ecological quality improved significantly. The average RSEI rose to 0.491, a 14.2% increase. The observed ecological improvement was closely associated with proactive impervious surface regulation. In core urban districts such as Futian and Nanshan, the reduction of impervious surfaces was achieved through the demolition of inefficient buildings, conversion of paved surfaces, and implementation of vertical greening systems, while ecological functions were simultaneously enhanced via sponge city initiatives. This “reduction-and-upgrading” model demonstrates that structural optimization of impervious surfaces served as a key driver of ecological quality enhancement. The outcomes of this phase indicate a fundamental shift in Shenzhen’s ecological governance model—from “end-of-pipe control” to “source prevention,” and from “passive response” to “proactive restoration.” Consequently, the urban–ecological interaction has entered a new phase characterized by synergistic “spatial restructuring and ecological rehabilitation”, marking a virtuous cycle of coordinated development. Similar policy-driven ecological restoration models have demonstrated successful implementation in Hangzhou’s urban renewal practices [71]. However, Shenzhen’s approach of legally demarcating ecological control lines and enforcing rigid regulatory measures has achieved significantly higher efficiency and systematicity in ecological restoration compared to conventional administrative interventions [72].

It is particularly noteworthy that while certain districts exhibited similar trends in impervious surface expansion (Table 2), their ecological responses diverged significantly. Futian and Nanshan demonstrated marked improvements in RSEI values through the systematic implementation of ecological restoration projects. As a designated ecological conservation area, Dapeng District maintained characteristically low impervious surface coverage, a direct consequence of its stringent development restriction policies. Conversely, Longhua District exhibited relatively slower ecological recovery, attributable to persistent urbanization pressures stemming from the rapid agglomeration of emerging industries. Of particular significance is Futian District’s early-stage industrial upgrading, which facilitated substantial pollution reduction and consequently accelerated ecological recovery, while peripheral regions such as Bao’an District displayed delayed ecological improvements due to later industrial transitions. These observed variations principally derive from fundamental inter-district differences in functional orientation, policy frameworks, and industrial composition, collectively emphasizing the critical influence of region-specific development strategies on urban–ecological coordination dynamics.

4.2. Regional Heterogeneity in the Coordinated Development of Urbanization and Ecological Conditions

The coupling coordination degree between urbanization and ecological quality in Shenzhen exhibits significant spatial differentiation, reflected not only in the static spatial distribution but also in the dynamic evolution from 1990 to 2020 (Figure 4 and Figure 5). Overall, the coupling coordination degree follows a “high in the west, low in the east” pattern, closely aligned with the spatial structure characterized by dense development in the west and sparse development in the east. However, influenced by varying urbanization trajectories, ecological policies, and industrial transformation, different regions exhibit diverse trends in coordination, revealing a complex pattern of dynamic heterogeneity.

Futian, Nanshan, Bao’an, Longhua, and Longgang are representative districts with relatively high coupling coordination levels. Among them, Futian District, as a core urban center, reached its peak coordination degree in 2000 (0.701), followed by a slight decline to 0.631 by 2020, still within the “primary coordination” level. This trend may reflect a development threshold in highly urbanized core areas, where further improvements in coordination become increasingly difficult. Nevertheless, Futian maintained a relatively high level of coordination through innovative ecological restoration measures such as vertical greening and sponge city construction. Both Nanshan and Bao’an Districts showed steady improvement in coordination degree from 1990 to 2020, rising from “marginal coordination” to “primary coordination,” indicating the positive impact of industrial upgrading and intensive land use on sustainable development. Longhua and Longgang experienced an initial rise followed by stabilization. While rapid urbanization between 2000 and 2010 slowed the rate of coordination improvement, ecological restoration and urban renewal after 2010 helped stabilize their coordination degree above 0.65, demonstrating the potential of newly emerging suburban districts to achieve balanced development under scientific planning. Notably, Southeastern Bao’an and Northern Longgang saw moderate ecological degradation between 2000 and 2010, likely due to rapid industrial land expansion outpacing ecological restoration efforts. Dapeng District consistently had the lowest coupling coordination level in Shenzhen, remaining at the “near imbalance” level throughout the study period. As an important ecological barrier of the city, Dapeng has maintained the highest RSEI values citywide. However, due to strict ecological protection policies and complex topography, urban development has been limited, resulting in slow coordination improvement. This case highlights that prioritizing ecological protection alone—while neglecting appropriate development—may hinder healthy interactions between urbanization and ecology. In the future, Dapeng should explore a path of eco-prioritized urbanization guided by green development principles to promote harmonious development between urban growth and ecological quality. Guangming and Pingshan Districts significantly improved their coordination levels from “near imbalance” in 1990 to “primary coordination” in 2020 through strict ecological regulation and industrial restructuring, making them the most notable examples of coordinated development optimization. Although Yantian District showed slight improvements, it remains at the “marginal coordinated” level due to limitations posed by mountainous terrain and the ecological impacts of port-related industries. Luohu District, having undergone early saturation in development, has consistently remained at the “primary coordination” level.

Further analysis using local spatial autocorrelation (Figure 6) reveals that “high–high” clusters are concentrated in the western core areas of the city, closely overlapping with Shenzhen’s high-tech industrial corridor. This suggests that economically concentrated zones can achieve high-level coordination through technological innovation and policy regulation. In contrast, “low–low” clusters are concentrated in eastern ecological conservation areas, corresponding to the boundaries of Shenzhen’s ecological control lines, reflecting the constraints imposed by eco-priority policies.

In summary, the relationship between urbanization and ecological quality in Shenzhen demonstrates clear regional heterogeneity and dynamic evolution. The sustained improvement in coordination levels in the western region illustrates that ecological balance is still achievable under high-intensity development through technological innovation. Meanwhile, the long-term stagnation in the eastern region highlights the need to explore synergetic paths that integrate ecological protection with moderate development. These findings provide scientific support for differentiated and region-specific sustainable development strategies in megacities.

4.3. Limitations and Future Work

This study still has several areas that require further improvement. First, in terms of data, due to limitations from cloud cover and imaging conditions, the analysis of urban expansion and ecological quality was only conducted for four discrete years, rather than continuously on an annual basis. This interval-based approach may have missed important changes in key years, potentially leading to an incomplete understanding of the processes of urban expansion and ecological evolution in Shenzhen. Future research should aim to integrate multi-source remote sensing data to enhance the temporal continuity of the time series, thereby improving the accuracy and depth of the analysis. Second, in terms of methodology, the coupling coordination degree model adopted a fixed weight (α = 0.65) for the urbanization subsystem, primarily based on conventional urban development models. However, this fixed weighting may not fully capture the characteristics of Shenzhen as an innovation-driven city. Shenzhen’s urbanization is not solely reliant on physical expansion but is also deeply influenced by high-tech industries and innovation-driven economic growth. Therefore, future studies should consider applying dynamic weighting approaches, such as the entropy weight method or the coefficient of variation method, to better reflect the evolving importance of each subsystem in the coupling model. Regarding indicator representation, although the use of impervious surface area effectively captures the extent of urban spatial expansion, it fails to represent other dimensions of urbanization, such as population density and economic intensity. Future research will explore the integration of emerging data sources, such as nighttime light data and point-of-interest (POI) density, to construct a more comprehensive urbanization evaluation system. Incorporating socio-economic factors and the spatial distribution of public services can further enhance the multidimensional understanding of urbanization impacts. Finally, while this study primarily assessed ecological environment quality using the Remote Sensing Ecological Index (RSEI), it should be noted that air pollution [73], water pollution [74], and soil pollution [75] are also critical components of urban ecological quality. Due to the inherent limitations of remote sensing data, these factors were not directly incorporated into our evaluation framework. Future research could integrate ground-based monitoring data or novel remote sensing indicators to provide a more comprehensive assessment of urban ecological quality.

Despite these limitations, the core findings of this study are still reliable. More importantly, these limitations point to future directions for refinement. By integrating multi-source data, optimizing dynamic weights, and adopting a multi-indicator framework, future research can establish a more precise and refined urban–ecological coupling assessment system. Such a framework would provide stronger decision-making support for the sustainable development of high-density cities like Shenzhen and offer a theoretical foundation for coordinated planning between urban growth and ecological protection.

5. Conclusions

Based on multi-temporal remote sensing data from 1990 to 2020, this study comprehensively applies the remote sensing ecological index, impervious surface area index, and coupling coordination models to systematically examine the relationship between urbanization and ecological quality in Shenzhen. The main conclusions are as follows:

(1)

The expansion of impervious surfaces in Shenzhen exhibited distinct phased characteristics with marked spatial heterogeneity. From 1990 to 2020, the impervious surface area increased from 458.15 km² to 709.23 km², with an average annual growth rate of 1.47%. Spatially, rapid expansion dominated in western regions, infill development occurred in central areas, while the urban core showed slight contraction in later stages. This spatial–temporal pattern was closely associated with regional functional zoning and policy interventions in Shenzhen’s urban development.

(2)

Ecological quality demonstrated a “decline-recovery” trajectory. The mean RSEI decreased from 0.477 (1990) to 0.429 (2000), then recovered to 0.491 (2020). Spatially, eastern ecological conservation areas like Dapeng District maintained optimal quality, whereas central–western built-up areas initially suffered from intensive development but later improved significantly through green space restoration. This transition highlights the critical role of policy regulation in ecological recovery.

(3)

The coupling coordination degree (CCD) between urbanization and ecological environment improved substantially, rising from “marginal coordination” (0.548) to “primary coordination” (0.636). A clear “west-high, east-low” spatial pattern emerged: western and central regions achieved higher CCD through balanced urbanization and ecological management, while eastern areas showed limited CCD improvement due to strict conservation policies constraining urban development. This spatial differentiation underscores how regional functional positioning fundamentally influences coordinated development.