Spatiotemporal Detection of Ecological Environment Quality Changes in the Lijiang River Basin Using a New Dual Model

Abstract

Detecting spatiotemporal changes in ecological environment quality (EEQ) is of great importance for maintaining regional ecological security and supporting sustainable economic and social development. However, research on EEQ detection from a remote sensing perspective is insufficient, especially at the basin scale. Based on two indices, namely, the Ecological Index (EI) and the Remote Sensing Ecological Index (RSEI), we established a dual model, combining the remote sensing ecological comprehensive index (RSECI) and its differential change model, to study the spatiotemporal evolutionary characteristics of EEQ in the Lijiang River Basin (LRB) from 2000 to 2020. The RSECI combines the following five indicators: greenness, wetness, heat, dryness, and aerosol optical depth. The results of this study show that the area of good and excellent EEQ in the LRB decreased from 3676.22 km² in 2000 to 2083.89 km² in 2020, while the area of poor and fair EEQ increased from 80.81 km² in 2000 to 1375.91 km² in 2020. From 2000 to 2020, the change curve of the EEQ difference in the LRB first rose, fell, and then rose again. The wetness and greenness indicators had positive effects on promoting EEQ, while the heat, aerosol optical depth, and dryness indicators had restraining effects. The results of stepwise regression analysis showed that, among the selected indicators, wetness and greenness were the key factors for improving the EEQ in the LRB during the study period. The RSECI approach and the difference change model proposed in this study can be used to quantitatively evaluate the EEQ and facilitate the analysis of the spatial and temporal dynamic changes and difference changes in EEQ.

1. Introduction

Climate change and the increasing intensity of human activities have growing impacts on the ecological environment [1], and the demand for effective technologies with which to quantify and assess changes in ecological environment quality (EEQ) is on the rise [2]. The Ecological Index (EI), which serves as a comprehensive assessment criterion for EEQ, consists of the six following indices: the water network denseness, pollution load, biological richness, environmental restriction, land stress, and vegetation coverage indices [3]. The EI approach has been extensively applied in ecological environment assessment at regional levels across China, but there are certain issues with this approach; for instance, it lacks the capacity for spatial visualization and possesses the drawback of artificial index weight determination [4]. Xu [5] proposed the Remote Sensing Ecological Index (RSEI), which was constructed from four remotely sensed indices, namely greenness, wetness, heat, and dryness indices; such an approach eases the challenges associated with ecological indicator extraction and mitigates subjectivity in practical uses, allowing for the quantitative appraisal of the regional ecological environment and offering visual assessment outcomes, thus solving the problems of the absence of spatial visualization in EEQ evaluation and the artificial determination of index weights [6,7,8].

However, compared with the EI, the RSEI lacks the air environmental quality index of pollution load. Aerosol optical depth (AOD), which represents one of the fundamental optical characteristics of atmospheric aerosols, serves as a parameter for depicting the quantity of aerosols present in the atmosphere [9,10]; to a certain degree, it can mirror the level of regional atmospheric pollution [11]. Aerosol studies are of great significance in the fields of climate change and ecological environment surveillance [12,13]. Aerosol particulate matter in the atmosphere exerts considerable adverse effects on the environment, ranking among the most severe environmental pollution issues at present [14,15]; therefore, it is particularly necessary to introduce aerosols as an indicator for ecological environment assessment and monitoring. In this article, we propose a more comprehensive remote sensing ecological comprehensive index (RSECI) based on the EI and RSEI. The RSECI, constructed using principal component analysis, comprises the five following indices, all of which are derived from remote sensing data: greenness, wetness, heat, dryness, and AOD. The RSECI not only addresses the lack of capacity for spatial visualization in EEQ assessment and the artificial setting of index weights but also fully integrates more comprehensive EEQ indicators for more representative results in order to more effectively analyze the rate and tendency of alterations in EEQ, we constructed the RSECI differential change model, which principally focuses on the annual EEQ differential change pattern, aiming to track the directions of EEQ shifts.

The Lijiang River Basin (LRB) is an important ecological functional area with a fragile ecological environment. In the LRB, the implementation of projects such as ecological and farmland protection and restoration, as well as rocky desertification control, has moderately promoted the restoration of the ecological environment in the basin; however, threats to its EEQ remain severe. Therefore, it is urgent to analyze the EEQ spatiotemporal patterns and change trends within the LRB.

The objectives of this research were as follows: (1) construct a remote sensing ecological comprehensive index (RSECI) and its corresponding difference variation model; (2) quantitatively assess the EEQ of the LRB; (3) analyze the temporal and spatial changes, as well as the differential changes, in EEQ in the LRB from 2000 to 2020; and (4) develop additional models and forecasts for the LRB. In this study, we provide a methodological reference for objective, quantitative, timely, and accurate comprehensive monitoring, as well as the evaluation of regional ecological environment quality; it can also help to improve the efficiency of environmental management and promote sustainable development.

2. Materials

2.1. Study Area

The research region selected for this study is the Lijiang River Basin (LRB), located in Guilin City, within the Guangxi Zhuang Autonomous Region of China, as depicted in Figure 1. The LRB is a typical karst landform scenic area [16] within the subtropical humid monsoon climate region in South and Central Asia, with a vegetation coverage rate of approximately 62.0%, encompassing all vegetation types found within the central subtropical regions of China [17]. As global issues, the restoration and management of the fragile karst environment have continually been topics focused upon by relevant scholars [18]. The ecological environment has a lasting influence on the development of urban tourism [19], as the LRB is the essence of Guilin’s tourism resources. Therefore, taking the LRB as an example, the monitoring and assessment of its EEQ hold crucial practical implications for the sustainable growth of both its ecological and tourism sectors.

2.2. Data Sources and Pre-Processing

Remote sensing image data: We downloaded Landsat TM/OLI remote sensing image data from 2000, 2005, 2010, 2015, and 2020 from the United States Geological Survey (USGS) (data portal: http://glovis.usgs.gov/ (accessed on 10 October 2020)), with 4 images per issue, for a total of 20 images. The path numbers of these data are 124 and 125, and the row numbers are 42 and 43. According to the requirements of interpretation, we conducted a series of pre-processing operations on the obtained images, including radiometric calibration, geometric correction, atmospheric correction, and band combination.

AOD data: For this study, we used MODIS MCD19A2 aerosol data (data portal: https://ladsweb.modaps.eosdis.nasa.gov/search/ (accessed on 15 January 2021)). First, we downloaded the daily aerosol data of the LRB from 2000, 2005, 2010, 2015, and 2020. Secondly, according to the data pre-processing specifications, we implemented a series of meticulous pre-processing steps on the remote sensing image data collected during the five aforementioned periods to ensure data quality and consistency. Image stitching was accomplished using the Mosaic tool in ENVI5.3 software. During the stitching process, an overlap area handling strategy was adopted. A weighted average algorithm was applied, in which the weight assigned to each pixel within the overlap region was calculated based on its distance from the overlap boundary; this approach minimized the potential artifacts and discontinuities caused by the stitching operation, ensuring a seamless integration of multiple images. Based on ArcGIS10.4.1 software, the unified coordinate system for reprojection is WGS_1984_UTM_Zone_49N coordinates. For ease of calculation, all data were unified at a resolution of 30 m.

3. Methods

3.1. RSECI

We propose a multi-factor RSECI, based on the EI and RSEI, encompassing five indicators, namely greenness, wetness, heat, dryness, and AOD; among them, greenness, wetness, heat, and dryness are calculated using the vegetation index, bare soil index, humidity component, and surface temperature, respectively. In contrast to the EI, with the RSECI, the greenness vegetation index correlates with the vegetation coverage and biological richness index; the wetness index corresponds to the water network denseness index; the bare soil index, signifying dryness, is related to the land stress index; and the AOD is associated with the pollution load index. Because the environmental restriction index of the EI is a restrictive index, it can only be used when ecological damage and environmental pollution seriously affect the safety of human settlements. There is no significant ecological damage or environmental pollution event in the experimental area, so this index was not considered in this study. In order to further analyze the changes in ecological environment quality, the percentages of poor, fair, average, good, and excellent results were counted, and the change trend was studied. The proposed RSECI is expressed as a function with 5 indicators as independent variables, as follows:

$$RSECI=f(\mathrm{Greenness},\mathrm{Wetness},\mathrm{Heat},\mathrm{Dryness},\mathrm{Aerosol}\mathrm{Optical}\mathrm{Depth})$$

(1)

3.2. Calculation of Each Indicator

3.2.1. Greenness Index

The Normalized Difference Vegetation Index (NDVI) is a commonly utilized index for vegetation [20]; it significantly correlates with plant biomass, leaf area index, and vegetation coverage [21,22,23]. Thus, we chose the NDVI instead of the greenness index for this study, the expression for which is as follows [24]:

$$NDVI=\left(NIR-R\right)/\left(NIR+R\right)$$

(2)

where NIR and R represent the reflectance of the near-infrared and red bands in the TM and OLI sensors, respectively.

3.2.2. Wetness Index

The wetness index, which is expounded upon using the Kauth–Thomas Transformation (K-T Transformation), pertains to the moisture contents of surface water, soil, and vegetation, aspects that have close-knit connections with the quality of the ecological environment [25,26]. The wetness can be calculated as follows [27]:

$$Wet=S_1{\rho }_{blue}+S_2{\rho }_{green}+S_3{\rho }_{red}+S_4{\rho }_{NIR}-S_5{\rho }_{SWIR1}-S_6{\rho }_{SWIR2}$$

(3)

where $S_i(i=1,\dots ,6)$ are parameters; and ${\rho }_{blue}$, ${\rho }_{green}$, ${\rho }_{red}$, ${\rho }_{NIR}$, ${\rho }_{SWIR1}$, and ${\rho }_{SWIR2}$ are the reflectance of the blue, green, red, near-infrared, short-wavelength infrared 1 and 2 bands in TM and OLI sensors, respectively.

3.2.3. Heat Index

Land surface temperature (LST) represents a crucial element in the water–heat equilibrium of the Earth system. Acquiring the land surface temperature constitutes a significant aspect of dynamic resource and environmental surveillance [28,29]. For the heat index in our research, we used the LST retrieved from the thermal bands of Landsat images, bands 6 and 10 of Landsat TM and Landsat OLI, respectively. The calculation process of the LST is as follows [30]:

For TM sensors, the expressions are Equations (4)–(6), as follows:

$$L=gain\times DN+bias$$

(4)

$$T=K_2/\ln (K_1/L_6+1)$$

(5)

$$\begin{array}{l}LST=T/\left[1+\left(\lambda T/\rho \right)ln{\epsilon }_6\right]\hfill \end{array}$$

(6)

For OLI sensors, the expressions are Equations (7)–(9), as follows:

$$L_{10}={\tau }_{10}\left[{\epsilon }_{10}{B}_{10}(T_S)+(1-{\epsilon }_{10})I_{10}^{\downarrow }\right]+I_{10}^{\uparrow }$$

(7)

$$B_{10}(T_S)=\left[L_{10}-I_{10}^{\uparrow }-{\tau }_{10}(1-{\epsilon }_{10})I_{10}^{\downarrow }\right]/{\tau }_{10}{\epsilon }_{10}$$

(8)

$$LST=K_2/\ln \left[K_1/B_{10}(T_S)+1\right]$$

(9)

where $LST$ signifies the surface temperature; $L$ represents the radiance at the TM sensor; $T$ represents the temperature value at the sensor; $DN$ corresponds to the pixel gray value; $gain$ and $bias$ indicates the transmittance of the atmosphere within the thermal infrared band; $K_1$ and $K_2$ represent the scaling coefficients acquired from the metadata of the image; $\lambda$ is the central wavelength; $\rho =1.438\times {10}^{-2}mk$; ${\tau }_{10}$ is the transmittance of the atmosphere in the thermal infrared band; $I_{10}^{\downarrow }$ and $I_{10}^{\uparrow }$ are the downward and upward radiance of the atmosphere; $B_{10}(T_S)$ is the thermal radiation brightness of a blackbody at temperature $T_S$; the NDVI threshold method was used to obtain the specific emissivity ${\epsilon }_6$ and ${\epsilon }_{10}$ of ground objects [31].

3.2.4. Dryness Index

In the regional environment, in addition to the bare soil index (SI), there is a considerable area of construction land, which also causes the “drying” of the ground surface; consequently, the normalized difference built-up index (NDBSI) can be synthesized through the combination of the SI and the IBI [32,33], as follows:

$$NDSI=\left(SI+IBI\right)/2$$

(10)

$$SI=\left[\left(SWIR1+R\right)-\left(NIR+B\right)\right]/\left[\left(SWIR1+R\right)+\left(NIR+B\right)\right]$$

(11)

$$IBI={\displaystyle \frac{2\ast SWIR1/\left(SWIR1+NIR\right)-NIR/\left(NIR+R\right)+G/\left(G+SWIR1\right)}{2\ast SWIR1/\left(SWIR1+NIR\right)+NIR/\left(NIR+R\right)+G/\left(G+SWIR1\right)}}$$

(12)

where $B$, $G$, $R$, $NIR$, and $SWIR1$ represent the reflectance of the blue, green, red, NIR, and SWIR1 bands in the TM and OLI sensors, respectively.

3.2.5. Aerosol Optical Depth Index

Aerosol optical depth (AOD) refers to the integral of the aerosol extinction coefficient throughout the entire vertical column in the atmosphere, serving as a proxy for the level of air pollution [34]. After obtaining the daily AOD data from MODIS MCD19A2, we used ARCGIS10 to calculate the annual average AOD value in the LRB. A calculation example is shown in Figure 2.

3.3. Integration of the Indices

In this work, principal component analysis (PCA) was employed to combine the five measures, namely the NDVI, WET, LST, NDSI, and AOD, so as to construct the RSECI. In order to avoid affecting the load distribution of PCA, the modified normalized difference water index (MNDWI) [35] was used to mask the water before PCA because PC1 explicates more than 80% of the total variation in the dataset (Figure 3), we selected the first component of PCA (PC1) to represent the RSECI.

However, since the dimensions of the calculated evaluation indicators vary, it is necessary to rescale all of the factors to fall within the range from 0 to 1 prior to conducting PCA. The normalized values of all the factors were calculated as follows:

$$NIi=\left(Indicatori-Indicato{r}_{min}\right)/\left(Indicato{r}_{max}-Indicato{r}_{min}\right)$$

(13)

where $NIi$ represents a specific normalized index value. $Indicatori$ denotes the value of a pixel, while $Indicato{r}_{max}$ indicates the maximum value of a pixel and $Indicato{r}_{min}$ represents the minimum value of a pixel; the five normalized indicators are then applicable for computing the principal components. To ensure that a larger value of PC1 corresponds to favorable ecological conditions, the calculated PC1 can be diminished by 1, thereby allowing us to obtain the initial ecological index RSECI₀, as follows:

$$\begin{array}{l}RSEC{I}_0=1-\{PC1[f(NDVI,Wet,LST,NDSI,AOD)]\}\hfill \end{array}$$

(14)

To ensure ease of measurement and comparison, the indicators RSECI₀ can also be normalized [36] as follows:

$$\begin{array}{l}RSECI=\left(RSEC{I}_0-RSE{I}_{0-min}\right)/\left(RSE{I}_{0-max}-RSE{I}_{0-min}\right)\hfill \end{array}$$

(15)

RSECI represents the constructed remote sensing ecological composite index, with its value ranging from 0 to 1. The nearer the RSECI value to 1, the more favorable the ecological condition.

3.4. The Comprehensive Representativeness Test of RSECI

The comprehensive representativeness of the RSECI constructed on the basis of PC1 can also be examined through the correlations between it and each index. Correlation serves as an indicator that reveals the degree of closeness in the relationship between two entities, frequently denoted as the correlation coefficient. The larger the correlation is, the more powerful the explanatory capacity of the RSECI for each index will be, and it can represent the characteristic information of each index in a more comprehensive manner.

The calculation method is as follows: The correlation analysis of the principal component analysis results was carried out using remote sensing software, and the correlation coefficient matrix between each index and the RSECI was established, after which the average correlation degree of each index was calculated. The closer the correlation coefficient value is to 1, the greater the correlation between the RSECI and an individual index. The average correlation degree is expressed as the average value of the absolute value of the correlation coefficient between a certain index and other indices within the same period. The calculation formula is as follows:

$${\overline{AC}}_P={\displaystyle \frac{\left|A{C}_D\right|+\left|A{C}_Q\right|+\cdots +\left|A{C}_R\right|}{n-1}}$$

(16)

where ${\overline{AC}}_P$ is the average correlation degree of the index $P$. $A{C}_D$, $A{C}_Q$, and $A{C}_R$ represent the correlation coefficients of indices $D$, $Q$, and $R$, respectively, in the same period, and $n$ is the total number of indices.

The correlation coefficient matrix among each of the five evaluation indices and the RSECI is shown in Figure 4, where it can be seen that the highest mean correlation over five years (MC₅) among the five indicators and the RSECI, is with the NDSI, with an average correlation coefficient of 0.665, and the lowest MC₅ in the five years is with AOD, with an average correlation coefficient of 0.430. The mean correlation (MC) coefficient between the RSECI and the five indicators is greater than 0.8 every year, and the mean correlation coefficient over five years is 0.826, 0.07 higher than the average with the AOD, the lowest single indicator, results which indicate that the RSECI is capable of integrating the information of each index effectively; it has a higher level of representativeness compared to any individual index and can reflect the EEQ of the LRB more accurately.

3.5. The Calculation of the Difference Change in RSECI

The specific calculation steps of the difference in the RSECI are as follows: (1) Based on the RSECI data from the LRB in 2000, 2005, 2015, and 2020, the RSECI data of other years from 2000 to 2020 were obtained using cubic polynomial interpolation based on a python platform. Cubic polynomial interpolation is a commonly used numerical analysis technique, utilizing a polynomial function of degree three to approximate the relationship between data points; in this case, it takes advantage of the known RSECI data from specific years to estimate the values for the intervening years, a method which shows relative accuracy in capturing the trend and variation patterns in data within a certain range, and which can be used to effectively fill in the gaps from missing data and provide a more continuous and comprehensive dataset for further analysis. (2) Calculation of the difference, as follows:

$$\Delta RSECI=RSEC{I}_t-RSEC{I}_{t-1}$$

(17)

where $\Delta RSECI$ represents the difference in EEQ, $RSEC{I}_t$ represents the EEQ value of year t, and $RSEC{I}_{t-1}$ represents the EEQ value of year $t-1$.

4. Results and Discussion

4.1. Classification of RSECI

With reference to the previous standards, when normalizing the EEQ value between 0 and 1, grading with a level interval of 0.2 can better express the EEQ status [37,38]; therefore, in order to conduct a more in-depth analysis of the representativeness of the RSECI, we divided the results of the RSECI calculation of EEQ into “poor, fair, average, good, and excellent”, with 0.2 as the grade interval. Simultaneously, the area and proportion of each ecological class of the RSECI in the five years of 2000, 2005, 2010, 2015, and 2020 were statistically analyzed, the results of which are shown in

Table 1. Scale and proportion of ecological environment level in Lijiang River Basin from 2000 to 2020.

RSECI Level	2000		2005		2010		2015		2020
	Area (km2)	%	Area (km2)	%	Area (km2)	%	Area (km2)	%	Area (km2)	%
1 (0~0.2) (poor)	5.10	0.09	30.70	0.52	114.15	1.95	146.83	2.52	147.06	2.52
2 (0.2~0.4) (fair)	75.71	1.30	648.17	11.11	1062.65	18.21	1243.17	21.30	1228.85	21.06
3 (0.4~0.6) (average)	2078.58	35.62	2334.55	40.01	2266.93	38.85	2434.94	41.73	2375.81	40.71
4 (0.6~0.8) (good)	3337.43	57.19	1972.48	33.80	1217.17	20.86	1119.87	19.19	1678.05	28.76
5 (0.8~1.0) (excellent)	338.79	5.80	849.71	14.56	1174.71	20.13	890.80	15.26	405.84	6.95
Total	5835.61	100	5835.61	100	5835.61	100	5835.61	100	5835.61	100

. According to the proportion of each grade, the area of the good (0.6~0.8) RSECI level in 2000 was 3337.43 km², with the largest percentage of 57.19%. In 2005, 2010, 2015, and 2020, the percentages of the average (0.4~0.6) RSECI grade were the highest, at 40.01%, 38.85%, 41.73%, and 40.71%, respectively. The area percentage of excellent (0.8~1.0) RSECI grades increased from 5.80% in 2000 to 20.13% in 2010 and then decreased to 6.95% in 2020, showing an unstable trend of change. The proportion of poor (0~0.2) grades increased, from 0.09% in 2000 to 2.52% in 2020, showing a gradually increasing trend. The proportion of fair (0.2~0.4) grades increased from 1.30% in 2000 to 21.06% in 2020, an increase of 19.76%, with a gradually increasing trend.

4.2. Analysis of the Change in RSECI Grade

The following results are illustrated in Figure 5. In 2000, the area of the LRB where the EEQ was rated good or excellent measured 3676.22 km², taking up 62.99% of the total area; in 2005, the corresponding area was 2822.19 km², with a proportion of 48.36%; in 2010, it was 2391.88 km², accounting for 40.99%; in 2015, it was 2010.67 km², making up 34.45%; in 2020, the area of the LRB in which the EEQ rated was good or excellent was 2083.89 km², making up 35.71% of the total area. During the period from 2000 to 2020, the combined area of good and excellent regions witnessed a reduction of 1592.33 km², corresponding to a percentage decline of 27.28%. In 2000, the combined area of the LRB with poor and fair EEQ ratings was 80.81 km², representing 1.39% of the overall area; by 2005, this area had increased to 678.87 km², accounting for 11.63% of the total; in 2010, it further rose to 1176.80 km², constituting 20.16% of the total area; in 2015, the area reached 1390.00 km², making up 23.82% of the total; in 2020, it was 1375.91 km², equivalent to 23.58% of the total area. Over the 20-year span, the area of poor and fair regions expanded by 1295.10 km², with a percentage increase of 22.19%. Thus, from 2000 to 2020, the proportion of good and excellent EEQ in the LRB exhibited a downward trajectory, while that of poor and fair EEQ showed an upward inclination, whereas the proportion of the average ecological grade remained relatively stable.

4.3. Analysis of Temporal and Spatial Variations in RSECI

Figure 6 shows the spatial and temporal distributions of the RSECI with different quality levels on the map. In terms of evolutionary spatial and temporal distributions, remarkable alterations took place in the spatial distribution of each grade between 2000 and 2020; specifically, the areas of poor and fair grades exhibited a progressively increasing tendency year by year. The ecological grades of good and excellent are predominantly located in the northern area of the LRB, as the Maoer Mountain, in the upper reaches of the LRB, is a national nature reserve located in the north; thus, this result corresponds to the reality of the area. The poor and fair areas are distributed in the central and eastern parts of the basin, i.e., the center of Guilin City, and gradually spread southward, indicating that the construction area is gradually increasing and spreading, which is also consistent with the actual situation. The areas of poor and fair ecological environment quality in the center of Xingan County in the upper reaches; the Qixing District, Xiangshan District, and Lingui District of Guilin City in the central reaches; and Yangshuo and Pingle Counties in the lower reaches; share a tendency to expand. The excellent and good EEQ ratings are mainly distributed in the undulating mountains at higher elevations, and the poor and fair areas are mainly distributed in the low-altitude alluvial plain with lower elevations. In order to better reveal the dynamic change characteristics of the ecological environment in the LRB from 2000 to 2020, the ecological environment of the basin was expressed and analyzed in numerical values based on the difference principle. The nine values, from high to low, are +4, +3, +2, +1, 0, −1, −2, −3, and −4; when the grade difference symbol is positive, the EEQ becomes better; when the grade difference is negative, the EEQ becomes worse; when the grade difference is 0, the EEQ remains unchanged. The specific situations are outlined in

Table 2. Ecological environment quality level changes in the Lijiang River Basin from 2000 to 2020.

Class	Level	Level Area (km2)	Class Area (km2)	Class Proportion (%)
Degraded	−4	0.01	2999.16	51.39
	−3	11.23
	−2	452.87
	−1	2535.05
No change	0	2352.39	2352.39	40.31
Improved	+1	466.20	484.06	8.30
	+2	11.93
	+3	5.91
	+4	0.02

, and the dynamic change in spatial information is shown in Figure 6. The classification of the RSECI can be used not only to distinguish different levels of environmental quality in detail but also to compare the changing trends for EEQ over different periods and regions, thus allowing us to intuitively present the degree of superiority or inferiority as well as clarify the spatiotemporal changes at a glance, through the quantification of grades.

From the vantage point of spatial transformation, during the period from 2000 to 2010, a substantial area within the LRB witnessed a decline in EEQ; its distribution was rather concentrated, predominantly in the central and south-central regions of the basin. In contrast, the areas that either improved or remained unchanged were mainly located in the northern part of the basin. From 2010 to 2020, the distribution of regions with deteriorating, unaltered, and improved EEQ values was comparatively dispersed. Over the entire span from 2000 to 2020, the areas with ecological environment degradation accounted for the lion’s share of the basin. In terms of temporal variation, between 2000 and 2020, the areas with deteriorated ecological grades amounted to 2999.16 km², constituting 51.39% of the total area. The areas with unchanged ecological grades totaled 2352.39 km², representing 40.31% of the total area. Meanwhile, the areas with improved ecological grades amounted to 484.06 km², accounting for 8.30% of the total area. Figure 7 reveals that the areas where EEQ increased between 2000 and 2020 are principally situated in the northern section of the basin; in contrast, the regions with degrading ecological environment ratings are mainly located in the central and southern areas of the basin.

4.4. Modeling and 3D Inspection Analysis

4.4.1. RSECI Regression Model

In order to further verify the scientificity of the RSECI, modeling was carried out based on the calculated RSECI data and those of various indicators; the constructed model is applicable for simulating and forecasting the trend of alterations in regional EEQ, as well. The specific steps are as follows: First of all, using ArcGIS10.4 software, the LRB fishing net was constructed, and the NDVI, WET, LST, NDSI, AOD, and RSECI thematic images of the LRB in 2000, 2005, 2010, 2015, and 2020 were sampled. Secondly, after removing outliers, 1542 sample points from the NDVI, WET, LST, NDSI, AOD, and RSECI were collected for each year. Thirdly, the NDVI, WET, LST, NDSI, and AOD were used as independent variables, while the RSECI was used as the dependent variable for stepwise regression analysis. The following regression models were thereby established:

$$\begin{array}{l}RSEC{I}_{2000}=0.415NDVI+0.161WET-0.856LST-0.124NDSI-0.229AOD+0.181\\ \left(R^2=0.999\right)\end{array}$$

$$\begin{array}{l}RSEC{I}_{2005}=0.510NDVI+0.101WET-0.050LST-0.107NDSI-0.824AOD+1.144\\ \left(R^2=0.999\right)\end{array}$$

$$\begin{array}{l}RSEC{I}_{2010}=0.310NDVI+0.041WET-0.241LST-0.091NDSI-0.917AOD+0.493\\ \left(R^2=0.999\right)\end{array}$$

$$\begin{array}{l}RSEC{I}_{2015}=0.394NDVI+0.031WET-0.229LST-0.107NDSI-0.883AOD+0.354\\ \left(R^2=0.999\right)\end{array}$$

$$\begin{array}{l}RSEC{I}_{2020}=0.318NDVI+0.017WET-0.215LST-0.067NDSI-0.586AOD+0.770\\ \left(R^2=0.999\right)\end{array}$$

In the regression analysis, no variables were excluded, and the p-values were less than 0.01 (p < 0.01), indicating that all independent variables were significant to the dependent variable, and the results were extremely statistically significant. As shown in the derived model, during the stepwise regression for each year, none of the five indicators was excluded, implying that all of the five selected indicators are crucial to EEQ and rendering the evaluation results more dependable. Based on the coefficients of each index, it was observed that the coefficients of NDVI and WET were positive, implying that these two particular indices have promoting effects on the EEQ. The negative values of the LST, AOD, and NDSI indicate that these factors have negative relationships with the RSECI and thus have negative effects on changes in EEQ; this outcome is consistent with the contribution of each index to the first principal component in the results of our principal component analysis, as well as the actual scientific facts, indicating that the construction and application of the RSECI are scientific and reasonable.

4.4.2. RSECI 3D Inspection Analysis

With the intent of examining the relationship of each index with RSECI in three-dimensional projection, the year 2020, being the most recent one, was taken as an instance, and all of the indicators, as well as the RSECI, were then projected into a 3D space, illustrated in Figure 8, where, among the five indicators—the LST, NDSI, AOD, WET, and NDVI—the LST, NDSI, and AOD have negative impacts on the RSECI, while the WET and NDVI have positive impacts on the RSECI. The upper part of the scatter plot indicates a favorable ecological environment, while the lower part signifies a poor one. Generally, the bottom of the scatter plot corresponds to regions with relatively concentrated construction land, high population density, and a thick aerosol layer, thus representing areas of inferior ecological and environmental quality. As can be observed in Figure 8, the scatter plot is tightly clustered, indicating that each factor is highly correlated with other factors. Figure 8a,b shows that the maximum values of the LST, NDSI, and AOD correspond to the minimum values of the RSECI, and the slope of Figure 8b is greater than that of Figure 8a, indicating that the negative combined effects of the NDSI and AOD on the RSECI are greater than the negative combined effects of the LST and NDSI on the RSECI, which is the same as the coefficient of the corresponding 2020 RSECI regression model. The tilt direction shared by the scatter plots in Figure 8a,b is opposite to that in Figure 8c, reflecting their opposite impacts on the RSECI.

4.5. Analysis of RSECI Difference Change

Figure 9 shows that the curve of EEQ change first ascends, then descends, and subsequently ascends again, with two evident inflection points in the variation in EEQ difference. The first inflection point appeared between 2007 and 2008, while the second inflection point appeared between 2012 and 2013. From 2000 to 2008, the difference change curve was rising, indicating that the trend of deterioration in EEQ was slowing down during this period; from 2008 to 2013, the difference change curve was declining, indicating that the speed of deterioration in EEQ was accelerating during this period; from 2013 to 2020, the trend of deterioration in EEQ was slowing down. The absolute value of the difference changed from 0.84 to 0.12, and the speed of deterioration in EEQ slowed down by 7 times. Of course, our above conclusions are only interim findings based on the existing data and methods, while more factors need to be considered in actual decision-making; at the same time, we should endeavor to understand the dynamic characteristics of the ecosystem itself when interpreting the trends in ecological environment quality changes.

5. Conclusions

(1)

The temporal and spatial evolutionary processes and dynamic change characteristics of ecological and environmental quality in the LRB during 2000–2020 were studied by constructing the dual model of the RSECI and its difference change. Regarding temporal alterations, the area of good and excellent EEQ ratings in the LRB declined from 3676.22 km² in 2000 to 2083.89 km² in 2020; conversely, the area of poor and fair EEQ ratings rose from 80.81 km² in 2000 to 1375.91 km² in 2020. Concerning spatial modifications, the region of deteriorating EEQ expanded annually, progressively diffusing to the periphery and central–southern areas, with the central–western regions at its core. The areas where the EEQ improved were mainly distributed within the Maoer Mountain National Nature Reserve, located in the northern part of the basin; the areas with unchanged EEQ mainly appeared in the upper reaches of the LRB.

(2)

1542 sample points were taken from NDVI, WET, LST, NDSI, AOD, and RSECI respectively every year. Subsequently, with the NDVI, WET, LST, NDSI, and AOD set as independent variables and the RSECI as the dependent variable, stepwise regression analysis was carried out in order to establish a regression model, in which no variables were excluded, and p-values were all less than 0.01 (p < 0.01), indicating that all of the independent variables were significant to the dependent variable. In the regression models obtained for each of the 5 years, none of the NDVI, WET, LST, NDSI, or AOD indicators were ever removed, indicating that the five selected indicators are key ecological indicators, and the evaluation results are reliable, further confirming the scientific validity of RSECI modeling.

(3)

A difference model was established in order to examine the variation in the RSECI difference within the LRB, the findings of which demonstrated that the curve of EEQ difference in the LRB from 2000 to 2020 initially rose, subsequently declined, and then ascended once more. Throughout the research period, two distinct inflection points were present in the difference change curve, the first of which appeared in 2008, while the second appeared in 2013. An analysis of the RSECI differentiation curve leads to the conclusion that, within the research period, although the EEQ of the LRB was in a state of deterioration, the rate of deterioration was progressively decreasing, thus implying the effectiveness of diverse ecological environmental protection measures. Therefore, in future development, the EEQ of the LRB is expected to gradually improve.

(4)

In this work, we constructed the RSECI and difference change model in order to study the spatial and temporal evolutionary characteristics of EEQ in the LRB from 2000 to 2020. From the perspective of spatial distribution changes, urban expansion has placed significant pressure on the ecological environment. Through regression analysis of the RSECI and each relevant index, the scientific nature of RSECI modeling was further verified, thus confirming that the RSECI can be applied to other regions. The difference analysis unveiled the inflection points and trends in EEQ alteration within the study area. Through an examination of the coupling relationships between the inflection points of EEQ change in the LRB and the principal policies of the study area, it is implied that the shift in EEQ is tightly correlated with policies.

The dual model of the RSECI model and its difference change model, established in the study, can not only be used to study the temporal and spatial evolutionary characteristics of EEQ but also to scientifically reveal the evolutionary trend in EEQ, as well as to analyze the scenario of urban ecological environment sustainability perception and evaluation; however, due to space limitations, the influencing factors of EEQ spatiotemporal evolution have not been taken into account, and further improvements are needed for subsequent research. Regarding data interpolation, future research could combine machine learning and deep learning algorithms to dynamically track complex changes, break through the limitations of traditional interpolation, and obtain more accurate long-term series data.