**1. Introduction**

Climate change and human activities have an overall impact on global ecology [1,2]: on the one hand, human activities such as urban expansion and deforestation affect ecosystem degradation in most parts of the world [3,4]; on the other hand, human activities such as ecological restoration can improve ecosystems to a certain degree [5,6]. Ecological environment quality (EEQ) is the degree of suitability of the ecological environment for human survival and sustainable social–economic development within a certain space–time range [7]. Scientific monitoring and the evaluation of the impact of human activities on EEQ and temporal and spatial changes have shown important theoretical and practical significance to coordinate the relationship between human activities and the ecological environment, and to promote the sustainable development of society.

The United Nations identified poverty eradication as the primary goal of sustainable development, having invested 600 billion CNY in PAR from 2016 to 2020, involving 10 million extremely poor farmers; this is one of the flagship projects to eliminate poverty in

**Citation:** Zhou, Z.; Feng, Q.; Zhu, C.; Luo, W.; Wang, L.; Zhao, X.; Zhang, L. The Spatial and Temporal Evolution of Ecological Environment Quality in Karst Ecologically Fragile Areas Driven by Poverty Alleviation Resettlement. *Land* **2022**, *11*, 1150. https://doi.org/10.3390/ land11081150

Academic Editor: Xiaoyong Bai

Received: 1 June 2022 Accepted: 22 July 2022 Published: 26 July 2022

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

**Copyright:** © 2022 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/).

China [8]. The area of PAR is typical of the vicious circle of poverty and ecological environment deterioration, with a high overlap of ecological fragility and extreme poverty [9,10]. At the same time, the PAR areas also belong to the spatial poverty trap [11], and the task of eradicating poverty in situ is extremely difficult. The PAR is a better development opportunity for farmers who live below the poverty line in ecologically fragile areas, by moving them to cities and towns.

The Chinese government believes that China's PAR has positive significance for eradicating poverty and improving the ecological environment [12]. Should we consider PAR as an effective human activity to improve ecology? However, the existing research on PAR has basically focused on the social effectiveness of its mechanisms of participation and poverty reduction [13,14], with few studies being related to ecological restoration. Similar to the ecological resettlement policy, most scholars are skeptical of the ecological protection effect of the large-scale implementation of ecological resettlement in ecologically fragile areas, despite the effectiveness of ecological restoration or the welfare of relocated farmers [15,16]. However, different from ecologically fragile areas such as Tibet and Inner Mongolia, the climate conditions in the southwest karst mountainous area where PAR is mainly implemented are more suitable for plant growth (the annual average rainfall is approximately 1100–1300 mm, and the average annual temperature is approximately 16 ◦C), so the dominant biophysical limiting factor is not the climate, but the soil resources [17], and human interference is the main factor influencing ecological restoration [18,19]. According to peasants–land coordination theory, in regions with limited resources, human constraints and natural interference are the best choices to coordinate the relationship between peasants and land [20]. PAR reduces human disturbance to natural resources (soil resources), resulting in ecological improvement and the rapid shrinkage of inefficient agricultural production space in the relocated area, which is in line with the Environmental Kuznets Curve theory of ecological economics [21,22]. This paper aims to explore whether the PAR really contributes to the improvement of EEQ as a human activity and actually improves the living environment for human beings. If the EEQ improves the study area, can we consider it to be caused by PAR? We need to quantify the effectiveness of PAR-driven eco-environmental improvements, among which we must distinguish the influence of human activities from natural factors, which will be the focus of this paper.

Due to remote-sensing data being timely and effective, covering a wide area, and being objective and sustainable, the application of remote-sensing technology in ecological environment assessment has increasingly attracted attention from scholars [23]. Using the normalized difference vegetation index (NDVI) to assess ecological environment is the most common method [24], and most scholars use land surface temperature (LST) in evaluating the effect of the urban heat island [25]. Similarly, the wet components form tasseHed captransform (WET), and the normalized difference impervious surface index (NDISI) indicators are the most important indicators for the intuitive human perception of ecological conditions [26,27]. Compared with a single indicator, the ecological status reflected by the comprehensive indicator is more complex and diverse. RSEI, which is based on the Ecological Index (EI) from the Ministry of Ecology and Environment of China, reflects the Technical Criterion for Ecosystem Status Evaluation (HJ 192e2015). EI is authoritative and extensive in regional EEQ assessment in China [28]. Many scholars have verified that RSEI and EI are highly comparable in the ecological sense [29]. The RSEI (Remote Sensing Ecological Index) model integrates intuitive and key influence factors, including greenness, wetness, dryness, and heat. It has the advantages of real and effective evaluation data sources, objective and fair evaluation conclusions, and intuitive and visual evaluation results [30,31]. Additionally, many scholars have evaluated the EEQ improvement effect by using RSEI as a technical means in projects such as Northwest Beijing Ecological Containment Area [32], the Northern Sand-Prevention Belt [33], and the Three-North Shelter Forest Program [34]. Existing research results show that the model of the trend of normalized residuals enables distinguishing between climatic factors and ecological effects caused by human activities [35]. It is necessary to clarify the turning point where PAR causes obvious changes in RSEI; a regression model was constructed using weather factors and human-activity factors during the time before the turning point to predict the RSEI trends during the period of PAR implementation, which was not affected by PAR, and the residual between the observed RSEI and the predicted RSEI, which was thought to be caused by PAR. Because the conclusion of the regression model is predictive, no PAR data are involved, so a correlation analysis model needs to be constructed with PAR implementation data and period RSEI variables to further explain the EEQ effect of PAR. The overall objective of this study is to evaluate the effectiveness of PAR on long-term EEQ dynamics in the study area, by the following means: (1) the analysis of the spatial and temporal evolution trend of long-series RSEI; (2) the elimination of the impact of climate factors and ecological restoration projects such as the Karst Rocky Desertification Restoration Project, to analyze the ecological contribution of the PAR; and, (3) establishing a coupling model between RSEI changes and the village-level PAR population to determine the association.

Southwest Guizhou Autonomous Prefecture, located in the Yunnan, Guizhou, Guangxi, concentrated, contiguous special-hardship area, is one of the most ecologically fragile regions in China, with few resources, a low environmental carrying capacity, a fragile ecological environment and human–land conflict [36]. From 2016 to 2019, 74,600 households with 338,600 people were relocated for the purposes of poverty alleviation, accounting for 3.38% of the total relocated population in China. This paper takes Southwest Guizhou Autonomous Prefecture as the study area, and first applies the Google Earth Engine (GEE) processing platform and cloud computing to obtain and reconstruct remote sensing data from 2000 to 2020, then applies the RSEI model to quantitatively evaluate the spatial and temporal evolution of EEQ in the study area. The ecological contribution of the PAR to the study area is quantified using the model of the trend of normalized residuals, and the association between the PAR and RSEI changes is further determined with a coupled model. Furthermore, we quantitatively reveal the changes in EEQ spatial distribution and the trend of ecological environment improvement caused by PAR, and provide theoretical support for coordinating ecological environment protection and social and economic development in ecologically fragile regions to achieve harmonious development between man and nature.

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

#### *2.1. Study Area*

The Guangxi, Yunnan, and Guizhou areas are located in the karst mountains of southwest China, with a combined population of 220 million people. They span across 0.54 million km2 of carbonate rock area, which is one of the most ecologically fragile and densely populated areas in the world [37]. The Southwest Guizhou Autonomous Prefecture is located in the southwestern part of Guizhou Province between 104◦35 –106◦32 E and 24◦38 –26◦11 N. It has eight counties under its jurisdiction and a land area of 16,800 km2. The area belongs to the subtropical humid monsoon climate and has the most widely distributed carbonate rock layer containing magnesium in the Triassic marine. The karst area in the region is spread across 10,200 km2, accounting for 60.28% of the total land area. Dominant ecological problems in Southwest Guizhou Autonomous Prefecture are stone desertification and soil erosion. The potential stone desertification is spread across 0.21 million km2, where the known stone desertification area is 0.50 million km2, accounting for 42.51% of the land (Figure 1), making it one of the most severely ecologically compromised areas of China [38]. With a rural-poor population of 432,300 in 2015 and a poverty incidence rate of 13.75%, the problem of poverty is relatively prominent. Southwest Guizhou Autonomous Prefecture is a region with high overlap between ecological fragility and extreme poverty.

**Figure 1.** Distribution of degree of stone desertification in the study area.

PAR is the flagship project of the Chinese government's poverty alleviation project. For people living in areas where local resources cannot effectively carry them out of poverty, they relocate to urban areas with better education, medical care, transportation, communication, employment and other improved conditions. Here, they enjoy the favorable development resources of the city, which help them to become free of poverty. The Chinese government concluded that PAR has 9 key achievements, including improved living conditions, broader employment prospects, and the relief of ecological environment pressure [12]. PAR in Southwest Guizhou Autonomous Prefecture involved 1222 villages, 74,600 households and 338,600 people from 2016 to 2019, all of whom were farmers living below the poverty line. The areas from which people were relocated, from low-resource areas to high-resource areas, were Xingyi, Xingren, Anlong, Pu'an, Zhenfeng, Wangmo, Qinglong, and Ceheng, which are concentrated in the north and southeast (Figure 2). The relocated farmers were resettled in 65 resettlement sites in cities and towns, accounting for 99.74% of the total relocated population (the other 0.26% were resettled in centralized rural areas, being relatively scattered), and 26 resettlement sites, with more than 5000 people each, resettled a total of 242,567 people, accounting for 71.66% of the total resettlement. Southwest Guizhou Autonomous Prefecture is dominated by centralized resettlement in cities (Figure 2).

**Figure 2.** Density of relocated household and distribution of 65 resettlement sites of PAR.

*2.2. Data Resources and Pre-Processing*

The remote-sensing data were mainly obtained from Landsat in the GEE platform database, including Landsat 8 (OLI) data for 2013–2020 and Landsat 5 (TM) data for 2000–2012, with a spatial resolution of 30m and a temporal resolution of 16 d. In addition, considering the cloudy climatic attributes of Guizhou Province, effective remote-sensing images of low cloud cover could not be collected in the summer. Therefore, the study screened automatic synthetic Landsat images from April to October for the target years. The GEE official programming algorithm was then used to pre-process data, and to complete geometric correction, radiation correction and atmospheric correction. We achieved cloudremoval processing through the cloud-mask algorithm. In addition, the updated mask was implemented by code to avoid the effect of water area on the load distribution of the principal components. After using the GEE programming calculus to obtain standardized remote-sensing data on the study area from April to October, vector data on the administrative regions at all levels in Southwest Guizhou Autonomous Prefecture State were obtained from the Resource and Environment Science Data Center of the Chinese Academy of Sciences (http://www.resdc.cn, 15 December 2021), the Grain to Green Project, the Karst Rocky Desertification Restoration Project from the Master Plan of National Forestry and Grassland Administration (http://www.forestry.gov.cn, 12 March 2022), and the National Development and Reform Commission (http://www.rdrc.gov.cn, 12 March 2022). The meteorological data were obtained from the China Meteorological Administration Network (http://www.cma.gov.cn, 12 March 2022), and population data for the PAR were obtained from the Ecological Migration Bureau of the Guizhou Province (10 September 2021).

#### *2.3. Methodology*

The RSEI model first proposed by Xu et al., was related to four indicators—greenness, wetness, heat, and dryness—which can be visually determined and are widely used to understand the quality of the ecological environment.

The entire data calculation process was based on the GEE online platform, ArcGIS software, and ENVI software. GEE is the most critical platform for original data acquisition, preprocessing, and RSEI calculation. The GEE operation process was as follows: determine the scope and timeliness; clarify the image type; function cloud mask; function remove cloud; calibrated radiance; function normalization; use unit scale to normalize the pixel values; calculate the NDWI; calculate the NDVI, WET, LST, and NDISI (formula 1–9); collection merge; visualization; map; function PCA model; eigenvalue, eigenvector; return result; normalize the RSEI; and export image to drive (see attachment for original data). ArcGIS and ENVI further process the original data in order to meet the research needs.

(1) Calculation of component indexes. Among the four indexes, the greenness index reflects the regional vegetation coverage, and the normalized difference vegetation index (NDVI) is closely related to the leaf area index and vegetation coverage. The wetness indicators (WET) reflect the moist conditions of the regional surface objects and are expressed as the wet components of a tasseHed captransform of the surface vegetation, soil, etc. Here, the formulae used to calculate EM and OLI were somewhat different. The dryness index reflected the surface drying condition; the soil index and the index-based built-up index IBI were expressed as the normalized difference impervious surface index (NDISI). Heat indicators reflected the surface temperature conditions and were expressed by the land surface temperature (LST). Based on the existing research results, the four indicators were calculated as follows:

$$NDVI = \frac{B\_{NIR} - B\_{red}}{B\_{NIR} + B\_{red}} \tag{1}$$

$$WET\_{TM} = 0.0315B\_{\text{blue}} + 0.2021B\_{\text{green}} + 0.3102B\_{\text{red}} + 0.1594B\_{\text{NIR}} - 0.6806B\_{\text{SWIR1}} - 0.6109B\_{\text{SWIR2}} \tag{2}$$

$$WET\_{OLI} = 0.1511B\_{\text{blue}} + 0.1972B\_{\text{green}} + 0.3283B\_{\text{red}} + 0.3407B\_{\text{NIR}} - 0.7117B\_{\text{SWIR1}} - 0.4559B\_{\text{SWIR2}} \tag{3}$$

$$NDBSI = \frac{SI + IBI}{2} \tag{4}$$

$$SI = \frac{\left(B\_{SWIR1} + B\_{red}\right) - \left(B\_{NIR} + B\_{blue}\right)}{\left(B\_{SWIR1} + B\_{red}\right) + \left(B\_{NIR} + B\_{blue}\right)}\tag{5}$$

*IBI* = 2 × *BSW IR*1/(*BSW IR*<sup>1</sup> + *BNIR*) − [*BNIR*/(*BNIR* + *Bred*) + *Bgreen*/ (*Bgreen* + *BWIR*1)] /2 × *BSW IR*1/(*BSW IR*<sup>1</sup> + *BNIR*) +[*BNIR*/(*BNIR* <sup>+</sup> *Bred*) + *Bgreen*/(*Bgreen* <sup>+</sup> *BWIR*1)] (6)

$$LST = \frac{K\_2}{\ln(K\_1/B(T\_S) + 1)}\tag{7}$$

$$B(T\_S) = \frac{L\_{10} - L\_{up} - \pi\_{10}(1 - \varepsilon\_{10})L\_{down}}{\pi\_{10}\varepsilon\_{10}}\tag{8}$$

$$L\_{10} = \tau\_{10} [\varepsilon\_{10} B(T\_S) + (1 - \varepsilon\_{10}) L\_{down}] + L\_{up} \tag{9}$$

Here, *NDVI* indicates greenness, *WET* indicates humidity, *NDISI* indicates dryness, and *LST* indicates heat. The specific meaning of each variable in the equation is referred to in reference [29].

(2) Calculation of RSEI. Based on the results of the four indicators, NDVI, WET, NDSI, and LST, they were first normalized by using positive normalization to standardize their values. The initial RSEI values were calculated using NEVI software by principal component analysis of the standardized data from the four indicators for PC1. Indicator weights that were not dictated by humans were considered in the calculation of the initial RSEI value. The final RSEI value was obtained using the forward normalization process, where the value was between 0 and 1. The higher the numerical value, the better the EEQ.

$$NI = \frac{I - I\_{\min}}{I\_{\max} - I\_{\min}} \tag{10}$$

$$RSEI\_0 = PC1[f(NDVI, WET, NDBSI, LST)]\tag{11}$$

$$RSEI = \frac{RSEI\_0 - RSEI\_{0-min}}{RSEI\_{0-max} - RSEI\_{0-min}} \tag{12}$$

In the above formula, *NI* denotes the standard index value after processing; *I* is the index value; and *Imax* and *Imin* are the maximum and minimum values of the index, respectively. *RSEI* denotes the final remote sensing ecological index; *RSEI*<sup>0</sup> denotes the primary remote sensing ecological index. *RSEI*0-*max* and *RSEI*0-*min* are the maximum and minimum values of the primary remote-sensing ecological index in the current period, respectively. *PC1* denotes the first principal component.

(3) Residual trends method. RSEI changes are influenced by climatic conditions and human activities. We used the residual trends method to calculate the extent to which PAR contributes to RSEI. Based on the mean value of RSEI from 2000 to 2020, the turning point of RSEI was determined according to linear trend analysis. A multiple linear regression analysis model was established with RSEI (dependent variable) and four factors (independent variables). These were the natural factors—mean annual temperature (MAT) and annual total precipitation (ATP)—and the human activity factors, outlined in the Grain to Green Project and the Karst Rocky Desertification Restoration Project, that may affect EEQ changes during the reference period (before the turning point).
