**1. Introduction**

Urban ecological quality (UEQ) evaluation is an important field of urban ecology research and the basis of urban planning and ecological management. With the continuous expansion of urbanization, China's cities have achieved medium-high quality development. However, social problems, such as resource exhaustion, an imbalance of economic structure and environmental pollution, do appear frequently. It is urgent to improve the capacity to implement urban sustainable development. In 2015, United Nations (UN) member states unanimously committed to achieving the Sustainable Development Goals (SDGs) by

**Citation:** Li, J.; Gong, J.; Guldmann, J.-M.; Yang, J. Assessment of Urban Ecological Quality and Spatial Heterogeneity Based on Remote Sensing: A Case Study of the Rapid Urbanization of Wuhan City. *Remote Sens.* **2021**, *13*, 4440. https://doi.org/ 10.3390/rs13214440

Academic Editor: Ronald C. Estoque

Received: 25 August 2021 Accepted: 2 November 2021 Published: 4 November 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

2030 [1]. Although the urbanization process has improved people's living standards, promoted the sustainable development of productive forces and provided economic benefits, it has also broken the balance between human society and the natural environment, and has brought great challenges to UEQ [2,3]. According to the 2018 Revision of World Urbanization Prospects [4,5], the urban population will account for 68% of the global population by 2050, which is an increase of 13% from 2018, and China's urban population will increase by 255 million people. Cities cover less than 2% of the Earth's surface, but consume 78% of the energy generated and produce 60% of greenhouse gas emissions [5]. Additionally, urban land consumption outpaces population growth by approximately 50% [6].

Such changes affect human survival and the sustainable development of the social economy [7–10]. Using UEQ measures to determine the status of the ecological environment could promote the sustainable development of regional economies [8,11–14]. Therefore, the quantitative description and assessment of the spatiotemporal dynamics of urban ecological environments are emerging as leading research topics [11,15].

Numerous studies have been conducted on such an assessment from different perspectives, and several evaluation methods have been suggested. The pressure–state–response model and fuzzy evaluation methods are commonly used in ecological quality assessment. In recent years, geographic information system (GIS) and remote sensing (RS) technologies have provided efficient monitoring and analysis methods for ecological quality research and sustainable development. Progress in satellite-based Earth observation systems facilitates assessing the state of an ecosystem from local to global scales. The scale and scope of this research are expanding constantly. Index systems have been constructed using GIS to conduct strategic environmental assessment for regional and land-use planning [16,17]. In China, research on the ecological environment is based on the Technical Specifications for Ecological Environmental Assessment, promulgated by the National Environmental Protection Agency in 2006 [18]. According to these specifications, the ecological environment index (EI) should encompass biological richness, air pollution, water network density, vegetation cover, land degradation and related factors. The EI is the main tool used to evaluate the quality of the ecological environment [19]. However, as climatic and geological conditions differ across regions, the weight of each index must be adjusted accordingly. Currently, researchers mostly use manual processing, as weight allocation is not strictly required and evaluation criteria vary, making it extremely difficult to accurately compare urban ecological conditions. Therefore, a scientific and logical ecological quality assessment method is required.

The acceleration of urbanization has led to a series of ecological and environmental effects, such as reduced surface water transpiration and water quality. It is generally difficult to monitor these natural processes with on-site instruments. However, remote sensing technologies can provide quantitative physical data with high spatial and temporal resolutions to facilitate the quantitative monitoring and analysis of environmental effects. Among all of the environmental effects of urbanization, the thermal environment has received more attention. The urban thermal environment is an important representative indicator of the urban environment. It is influenced by the physical properties of the urban surface and human social and economic activities, and is a comprehensive summary and embodiment of urban ecosystems. Vegetation is another important component of urban ecosystems. Urban vegetation can selectively absorb and reflect solar radiation energy, adjust the latent and sensible heat exchange, regulate urban air, reduce pollution and other processes that affect the city's natural environment and is another highly comprehensive index of urban ecological evaluation. The spatial distribution and richness of vegetation in cities have always been considered to have important effects on the evolution of the urban ecological environment.

The remote sensing ecological index (RSEI) combines humidity, greenness, heat and dryness indices obtained from RS, and facilitates the monitoring and evaluation of the UEQ. The RSEI, which was first proposed by Hu and Xu [18], could aid in visualizing spatial and temporal analyses and predictions of change in the regional environment, thereby compensating for the deficiencies of the EI. This paper uses existing research from a new perspective to more accurately study urban socio-economic activity intensity and its relationship with the regional ecological environment. Using the RSEI will help in studying the interactions between human activities and natural ecology, and the resulting knowledge of theory, concepts and methods is expected to benefit local governments [20]. In recent years, the RSEI has been applied in ecological quality monitoring in 35 cities of China [19,21,22], Eurasia [23] and America [21,23]. The RSEI and the results of principal component analysis (PCA) have been combined to develop an ecological index [19,24]. However, using the PCA results in insufficient information utilization, as the adaptive nature of PCA algorithms inevitably limits the full use of the available information. For example, the RSEI obtained in two studies using only the first component for normalization ranged from 60% to 90%, which cannot guarantee adequate contribution rates.

Accordingly, the aim of the current study is to improve the RSEI calculation method by proposing an improved-comprehensive remote sensing ecological index (IRSEI) constructed by employing PCA and equal weights (EW). Our study overcomes the shortcomings of previous studies, which only considered the application of PCA in ecological quality assessment, and the resolved knowledge gaps are reflected in the comprehensive consideration of EW and the PCA method to determine the UEQ. The contribution rates of the eigenvalues of PCA and EW are taken as the weights. This method enables the full use of the available data and ensures that the value of the calculated IRSEI is ecologically optimal. In addition, more indicators could be integrated and the IRSEI reduces noise interference and makes optimal use of practical image information. These factors facilitate the reliable and quantitative monitoring of the regional ecological environment.

A comparison and evaluation of the differences in quality in large cities can improve the cognitive ability of the internal mechanism of the reciprocal feed-back relationship between the construction of megacities and regional ecological balance, and can provide a scientific reference for controlling the scale of urban sustainable development and ecological planning and regulation. Wuhan is one of the fastest growing cities in central China, but few studies have been conducted on quantitative UEQ monitoring based on remote sensing data. Therefore, we used a series of parameters obtained from remote sensing imagery to construct the IRSEI for the evaluation of the UEQ of Wuhan city from 1995 to 2020. In addition to the UEQ, we determined the temporal and spatial changes in the city. We present a discussion of the ecological changes caused by economic and social developments and natural conditions. Finally, we provide theoretical guidance and a scientific basis for ecological construction in Wuhan city.

The objectives of this study are to:


#### **2. Materials and Methods**

#### *2.1. Study Area and Data Preprocessing*

We select the rapidly urbanizing city of Wuhan as study area for ecological monitoring and assessment. Wuhan is the capital city of Hubei Province. Its geographical location is 29◦58 –31◦22 N and 113◦41 –115◦05 E (Figure 1). From the perspective of Wuhan's geographical location and the location of its basin, the development of Wuhan has had great impact on the environment of the whole Yangtze River basin, and even the whole country. Therefore, ecological assessment and policy-based restoration and protection in Wuhan are vital for the ecological restoration of the Yangtze River basin. The city has jurisdiction over six central urban areas and seven distant urban areas. The land area comprises 8494.41 km2. The permanent population was 10.91 million in 2018. The Yangtze and Han rivers meet there, forming a geographical pattern referred to as "two rivers and three towns". Wuhan has a subtropical humid monsoon climate, with abundant rainfall and sufficient heat throughout the year. The average annual temperature is 15.8 to 17.5 ◦C. The area is rich in ecological resources, with nearly 40% green coverage and more than 10 m<sup>2</sup> of green space per capita. These ecological resources are crucial for Wuhan to build an ecological civilization city and, therefore, are critical factors in the protection of the ecological environment.

**Figure 1.** Location of Wuhan city.

In order to consider the quality of the remote sensing data, such as cloud cover and vegetation condition, we use data from Landsat 5 TM in 1995, Landsat 5 ETM in 2005 and Landsat 8 OLI in 2015 and 2020 as the main remote sensing data. RS data are particularly useful because they can be used for temporal and spatial monitoring [25]. Details on the satellite images used in this study are provided in Table 1. The source dates of the images are relatively close; therefore, differences caused by different seasons and vegetation growth states can be ignored. Owing to topographic differences in images at different times and the influence of illumination and atmospheric factors on surface reflectance, the selected images required preprocessing with radiometric calibration and atmospheric and geometric correction prior to the calculation of the IRSEI. The corrections were performed using the Environment for Visualizing Images (ENVI) software. The Fast Line-of-sight Atmospheric Analysis of Hypercubes (FLAASH) model was used for atmospheric correction to eliminate the radiation error caused by atmospheric absorption and scattering. The accuracy of radiation calibration was more than 95%, and that of atmospheric correction exceeded 85%. Further, the error in geometric correction was controlled to less than 1 pixel. The quadratic polynomial and the nearest neighbor methods were used to correct the geometry of the images and the preprocessed images of the study area were clipped using the vector data of the administrative districts of Wuhan. Other data sources included the administrative zoning map of Wuhan, digital elevation model data of Wuhan from the geospatial data cloud (http://www.gscloud.cn/sources/accessdata/310?pid=302 (accessed on 15 May 2021)) and the cloud platform of geographical national condition monitoring of China

(http://www.dsac.cn/DataProduct/Search?&cateID=2010&areaID=18 (accessed on 15 May 2021)). Nighttime light data were obtained from the national geophysical data center (https://www.ngdc.noaa.gov/eog/dmsp/downloadV4com-posites.html (accessed on 15 May 2021)).

**Table 1.** Data used and their source.


#### *2.2. Methodology*

#### 2.2.1. Modeling Framework

We combine principal component analysis (PCA) and the entropy value method to design synthetic indicators that facilitate quick and quantitative assessment of UEQ, based on humidity, greenness, dryness and the heat index. Using this method enables prioritizing the natural factors of the ecological evaluation system. The overall framework of IRSEI modeling, as shown in Figure 2, includes four main steps. First, we obtain Landsat Enhanced Thematic Mapper Plus (ETM+)/OLI/TIRS images and perform preprocessing, including atmospheric correction, radiometric calibration and image mosaic (see Section 2.1). Second, we derive four remote sensing indicators: humidity, greenness, dryness and heat. Third, we calculate the PCA components, obtain PC1 and use the entropy method to calculate the results that are used for the construction of the IRSEI. Finally, the characteristics of the spatial and temporal changes in the ecological quality of Wuhan over the past 25 years are determined, and the spatial heterogeneity of the city is analyzed.

**Figure 2.** Overall methodological framework.


The Kauth–Thomas transform (also called the tasseled hat transform) is a linear transformation method based on multispectral imaging [26,27]. This method is widely used in ecological monitoring for data compression and removal of redundancy. The moisture component obtained by this transform reflects moisture information in the soil and vegetation. A low humidity value indicates severe land degradation, low vegetation cover and a poor ecological environment. A high humidity value indicates sufficient soil moisture, rich surface vegetation cover and a good ecological environment.

In this study, *Iwet* was chosen as the humidity index [28], which is expressed as land surface moisture and is generated from Landsat TM, ETM+ and OLI image reflectance using Equations (1)–(3) [10,23,29]:

$$I\_{\text{wrTM}} = 0.0315\rho \mathbf{1} + 0.2021\rho \mathbf{2} + 0.3102\rho \mathbf{3} + 0.1594\rho \mathbf{4} - 0.6806\rho \mathbf{5} - 0.6109\rho \mathbf{7} \tag{1}$$

$$I\_{\text{wetETM}+} = 0.2626\rho 1 + 0.2141\rho 2 + 0.0926\rho 3 + 0.0656\rho 4 - 0.7629\rho 5 - 0.5388\rho 7 \tag{2}$$

$$I\_{\text{wetOLI}} = 0.1511\rho 1 + 0.1973\rho 2 + 0.3283\rho 3 + 0.3407\rho 4 - 0.7117\rho 5 - 0.4559\rho 7 \tag{3}$$

where *ρ*1, *ρ*2, *ρ*3, *ρ*4, *ρ*5 and *ρ*7 represent reflectance in bands 1, 2, 3, 4, 5 and 7 of Landsat TM/ETM+ images and reflectance in bands 2, 3, 4, 5, 6 and 7 of Landsat OLI data, respectively.

• Greenness index (*Indvi*)

The normalized difference vegetation index (NDVI) is often used to monitor vegetation growth [30] and directly reflects the quality of the regional ecological environment. This index is used in the classification of regional land cover, environmental change and vegetation. The NDVI greenness index is computed as follows [31]:

$$I\_{ndvi} = (\rho 4 - \rho 3) / (\rho 4 + \rho 3) \tag{4}$$

where *ρ*4 represents the reflectance of the near-infrared band and *ρ*3 represents the reflectance of the red band.

• Heat index (*Iheat*)

Land surface temperature (LST) refers to heat, which is related closely to vegetation growth, crop yield, surface water circulation, urbanization, other natural phenomena and processes and human activities [32]. LST can be used as a heat index to reflect the surface ecological environment. Several algorithms use thermal infrared technology to retrieve LST, including the atmospheric correction, single-window and single-channel algorithms. Comparison between LST retrieval results obtained using the atmospheric correction method and the actual measurement of LST indicates that the error is within 1 ◦C, thereby meeting research accuracy requirements. LST is generated using Equations (5)–(9) [33,34]:

$$L = \text{gain} \times DN + \text{bias} \tag{5}$$

$$Tb = K2/\ln(K1/L + 1)\tag{6}$$

$$LST = Tb/\left\{1 + \left[\left(\lambda Tb\right)/\rho\right] \ln \varepsilon\right\} - 273.15 \tag{7}$$

where *DN* is the pixel gray value, gain and bias are thermal infrared band excursions and *L* is the radiation brightness value.

Equation (7) is a simplified form of the inverse function of Planck's formula, with *K*1 and *K*2 being the calibration parameters. All of the parameter values are available from the metadata file (MTL) of the satellite data. *ε* is the specific infrared emissivity and is calculated with the method proposed by Min [35]. *λ* is the central wavelength of the thermal infrared band and *ρ* = s 1.438 10−<sup>2</sup> mK.

$$\begin{cases} \in \text{watter} = 0.995 \text{ (NDVI } \le 0\text{)}\\ \in \text{building} = 0.9589 + 0.086 \times F\_{\text{reg}} - 0.0671 \times F\_{\text{reg}}^2 \ (0 < NDVI < 0.7) \\ \in \text{natural} = \ 0.9625 + 0.0614 \times F\_{\text{reg}} - 0.0461 \times F\_{\text{reg}}^2 \ (NDVI \ge 0.7) \end{cases} \tag{8}$$

Vegetation coverage (*Fveg*) refers to the ratio (%) of the vertical projection area of vegetation on the ground to the total statistical area. *Pveg* is based on Landsat NDVI and adopts the dichotomy model of mixed pixels [36]. The calculation formula is as follows [37]:

$$P\_{\text{reg}} = \frac{NDVI - NDVI\_{\text{soil}}}{NDVI\_{\text{reg}} - NDVI\_{\text{soil}}} \tag{9}$$

where NDVI is the normalized vegetation index, *NDV Isoil* is the normalized vegetation index value of bare land and *NDV Iveg* is the normalized vegetation index value of complete vegetation coverage. *NDV Isoil* and *NDV Iveg* were selected as *NDV Imax* and *NDV Imin* with a confidence level of more than 95%.
