Spatiotemporal Evolution and Drivers of Ecological Quality in the Tengger Desert (2001–2021)

Abstract

Desert ecosystems, particularly in arid regions like the Tengger Desert, are highly sensitive to both anthropogenic activities and climate change, making the monitoring and evaluation of ecological quality critical for sustainable management and restoration efforts. This study analyses the spatiotemporal evolution of ecological quality in the Tengger Desert from 2001 to 2021 using the Remote Sensing Ecological Index (RSEI), incorporating meteorological factors (temperature, precipitation, wind speed), topographical factors (elevation, slope, relief) and anthropogenic indices (land use and land cover). The mean RSEI fluctuated between 0.1542 and 0.2906, indicating poor ecological quality, with a peak in 2008 attributed to national ecological projects. Despite initial improvements, overall ecological quality declined at a rate of 0.0008 a⁻¹ from 2008 to 2021. Spatially, degradation was most pronounced in the central and southern areas. Due to sand-binding engineering in the Tengger Desert in 2008 and the mountain climate suitable for vegetation growth, improvements occurred in the northeast and southwest. Moran’s I and Hurst index analyses revealed significant spatial clustering of ecological quality and persistence of degradation trends, with over 49.53% of the area projected to experience further deterioration. Geodetector analysis identified land use and land use cover as the most influential factors on RSEI, especially in combination with wind speed, temperature, and precipitation, underscoring the role of both human activities and climate. The study highlights the need for sustained ecological management, particularly in areas showing continuous degradation, to prevent further ecological deterioration.

1. Introduction

The socio-economic-ecological integrated system reflects the interaction between human activities and the natural environment, influencing the sustainable development of a region [1,2]. The Tengger Desert, as a fragile ecosystem, is vulnerable to habitat changes driven by both human activities and climate change [3]. Studying the evolution of ecological quality and driving mechanisms in the fragile ecosystem can provide insights into better environmental protection practices and policies.

Ecological quality refers to the suitability, harmonization and sustainability of a whole system or subsystem of components for human survival and overall development under specific time and regional conditions [1]. Ecological Quality Assessment is an essential tool for understanding and managing the natural environment. Current methods for monitoring and evaluating ecological quality include the hierarchical analysis method [4,5], the ecological footprint method [6,7], and the comprehensive ecological index method [8]. However, these methods face challenges in long-term ecological monitoring and evaluation of large regions, particularly in assigning and determining index weights for comprehensive indicators [9,10]. In comparison, satellite remote sensing technology offers convenient data acquisition with wide coverage and high spatial resolution. It integrates multidisciplinary knowledge, making it indispensable for environmental monitoring, ecological assessment, resource evaluation, etc., [11,12,13,14]. Scholars have extensively studied the feasibility of remote sensing indices for monitoring and evaluating environmental quality, including using the Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI) to monitor vegetation growth and cover [15,16,17], using standardised precipitation evapotranspiration index for assessing drought [18,19], applying a surface temperature index for evaluating urban heat island effects [20,21], etc. However, single indicator-based approaches typically focus on one specific aspect of the ecosystem. Given the complexity of natural systems, it is questionable whether regional ecological quality can be comprehensively evaluated using just one indicator.

With the in-depth exploration and utilization of remote sensing technology and mathematical models, ecological quality evaluation methods have been substantially expanded and improved. The Remote Sensing Ecological Index (RSEI) integrates four key indicators: greenness, humidity, dryness, and heat [22,23]. This approach avoids errors in the definition of weights caused by individual characteristics and provides an objective and reliable technique to more accurately visualize the spatial and temporal changes in regional ecological quality. The RSEI model provides a valuable tool for ecological environment assessment in different regions. However, these indicators mainly apply to areas where the natural environment supports human activities, such as cities, suburbs, managed forests, riparian, etc. [24,25,26]. They are less suitable for regions with extreme environmental conditions, e.g., desert ecosystems. To address the limitations of the RSEI in practical applications, scholars have continuously refined and enhanced it from various perspectives [27]. Firozjaei et al. (2021) [28] incorporated the Soil Index (NDSI) into the original framework, creating a remote sensing ecological index for arid regions. This index was used to evaluate and analyse the ecological conditions of several European cities. Chen et al. (2024) [29] introduced the Desertification Index (DI), and Ju et al. (2023) [30] introduced the Desiccation Index (SI), to study the ecological status of dryland desert oasis areas. Xu et al. (2021) [31] introduced the Land Cover Type Abundance (AI) and developed a regional scale remotely-sensed ecological index model. This model was found to be highly consistent with the environmental indices provided by the Ministry of Ecology and Environment (MOE) at the national scale in China. Therefore, it is feasible to monitor and evaluate the ecological environment using the RSEI, and the improved versions of this index can more accurately reflect the ecological quality of different regions.

The Tengger Desert faces enormous ecological challenges due to its arid climate, sparse population, and backward socio-economic conditions, resulting in a lack of human and financial resources to maintain and protect the ecological environment of the Tengger Desert [32]. Several restoration projects carried out by governments, local residents or non-profit environmental conservation organizations provide a unique opportunity to assess the ecological quality of the desert. Despite some relevant research, studies specifically addressing the ecological quality of the Tengger Desert remain limited. For example, Wang et al. [33] explored the interaction between vegetation coverage and climate factors, identifying precipitation as the primary driver of vegetation growth. However, focusing solely on a single index, like vegetation coverage, is insufficient to fully capture ecological quality. Moreover, their study, limited to four sites, does not adequately represent the broader ecological dynamics of the Tengger Desert. Ecological processes are inherently dynamic, open, and cyclical, shaped by global climate change and regional activities, resulting in temporal fluctuations [34]. Thus, research on ecological quality must extend beyond climate change and address deeper, intrinsic drivers of human activity. This paper adopts a more comprehensive approach to understanding the intricate interactions between human activities and the ecological environment in the Tengger Desert.

To address these challenges, this study analyses the spatial and temporal patterns and trends of ecological quality in the Tengger Desert from 2001 to 2021 using RSEI. Specifically, the research aims to (1) assess the spatiotemporal patterns and trends of RSEI through Theil–Sen slope estimation and the Mann-Kendall test; (2) explore the spatial correlation of RSEI using Moran’s I; (3) predict future trends with the Hurst index; and (4) examine the driving effects of natural and anthropogenic factors on RSEI using Geodetector. By adopting this comprehensive approach, the study seeks to lay a foundation for monitoring of regional ecological changes, assessing spatial ecological patterns, and supporting effective ecological management and protection efforts.

2. Materials and Methods

2.1. Study Area

The Tengger Desert is in the southeastern part of Alxa League (37°32′–40°00′ N, 102°15′–105°41′ E), situated in the north of China. It borders the Badain Jaran Desert to the northwest and the Ulanbuh Desert to the northeast, with its southern and southwestern sections extending into Ningxia and Gansu provinces (Figure 1). Covering a total area of 49,700 km², the terrain gradually slopes from west to east, with elevations ranging from 1124 to 1960 m above sea level. The Tengger Desert exhibits distinct continental climate characteristics, with an annual average temperature of 7.0–9.7 °C [35] and annual precipitation ranging from 186 mm [36]. The growing season (May–September) experiences a potential evapotranspiration of 2300 to 2500 mm, indicative of its temperate continental arid climate [37,38]. The desert’s major landforms include sand dunes, lake basins, mountains, residual hills, and flatlands, with sand dunes occupying approximately 71% of the area. These dunes are primarily composed of lattice dunes and crescent-shaped dune chains.

2.2. Data Collection

2.2.1. Ecological Quality Data

The ecological quality data for this study were sourced from the China High-Resolution Ecological and Environmental Quality Dataset (CHEQ) via the National Earth System Science Data Center (http://www.geodata.cn, accessed on 4 June 2024). This dataset, developed by Xu Dong et al. [31], using principal component analysis on five indices—humidity (WET), greenness (NDVI), Land Surface Temperature (LST), Normalized Difference Bare Soil Index (NDBSI), and Abundance Index for land cover type (AI)—was processed using the Google Earth Engine platform and stored as raster data. It provides annual RSEI data at a 500 m resolution for China from 2001 to 2021, based on MOD09A1, MOD11A2, and MCD12Q1 datasets. Validation has shown high consistency with the Environmental Index (EI) from the Ministry of Ecology and Environment of China, demonstrating excellent temporal stability and continuity [39]. Therefore, this study uses the RSEI from the CHEQ dataset as the standard for ecological quality evaluation. To further understand the spatial and temporal changes of ecological quality in the Tengger Desert, the natural breakpoint method is used to divide the RSEI value into five levels [40]: poor (0–0.2), fair (0.2–0.4), moderate (0.4–0.6), good (0.6–0.8) and excellent (0.8–1). The years 2001, 2006, 2011, 2016 and 2021 were selected as representative years to analyse the ecological quality of the Tengger Desert.

2.2.2. Climate Data

This study’s temperature and precipitation data were sourced from the China 1 km resolution monthly precipitation and mean temperature datasets provided by the National Tibetan Plateau Science Data Center (http://data.tpdc.ac.cn, accessed on 5 June 2024). These datasets, created by Peng Shouzhang, are downscaled for China using the Delta spatial downscaling method based on the global 0.5° climate dataset from the Climatic Research Unit (CRU) and the high-resolution climate dataset from WorldClim. Validation was conducted using data from 496 independent climate observation stations, confirming the reliability of the results [41,42,43]. The wind speed data used in this study were sourced from China’s 1 km resolution monthly average wind speed dataset, provided by the National Earth System Science Data Sharing Service platform (http://www.geodata.cn, accessed on 5 June 2024). This dataset was generated by spatial downscaling of reanalysed climate data [44].

2.2.3. Human, Topographic and Vector Data

Land use and land cover (LULC) data were sourced from Yang and Huang (2021) [45] with a spatial resolution of 30 m. These data include Cropland; Forest; Grassland; Water; Snow/Ice; Barren; Impervious; Wetland; Shrub Prairie (9 LULC); the study area covers only the first six categories. Topographic data, represented by slope, undulation, and elevation, were obtained from the GDEMV2 dataset of the Geospatial Data Cloud (https://www.gscloud.cn/, accessed on 3 June 2024), also with a spatial resolution of 30 m. Additional topographic metrics (slope, relief) were calculated using ArcGIS. The vector boundary data for the study area were acquired from Li et al. [46] and processed into shapefile (shp) format using ArcGIS. All data used in this study were reprojected to the WGS 1984 UTM Zone 48N (EPSG:32648) coordinate system, and the spatial scale was standardized to 30 m for consistency in subsequent analyses.

2.3. Data Analysis

2.3.1. Theil–Sen Slope Estimation

The Theil–Sen slope estimation method is a robust, nonparametric analysis technique that does not rely on distributional assumptions, making it practical for handling missing data in time series and mitigating the influence of data distribution on results. Its strong resistance to data errors makes it applicable to various datasets, ensuring more scientifically credible outcomes [47]. Therefore, this study applies the Theil–Sen median slope estimate to analyse the time-varying trends of the RSEI in the Tengger Desert from 2001 to 2021. The slope is calculated with Formula (1):

$$\beta =median{\displaystyle \frac{{RSEI}_j-{RSEI}_i}{j-i}}(2001\le i<j\le 2021)$$

(1)

where ${RSEI}_j \mathrm{a}\mathrm{n}\mathrm{d} {RSEI}_i$ denote the RSEI values in year i and year j, respectively, and Median is the median function. When β > 0, it indicates that RSEI is improving and, vice versa, is degrading.

2.3.2. Mann–Kendall Trend Analysis

The Mann–Kendall test is a non-parametric statistical method [48,49] used to assess the significance of a trend over time. It does not require the data to follow a specific distribution and is resilient to the presence of outliers. The test is widely used for detecting trends in time series data.

$$\mathrm{Z}=\left\{\begin{array}{c}(\mathrm{S}-1)/\sqrt{\mathrm{V}\mathrm{a}\mathrm{r}\left(\mathrm{S}\right)},\mathrm{S}>0\\ 0\hspace{1em}\mathrm{S}=0\\ (\mathrm{S}-1)/\sqrt{\mathrm{V}\mathrm{a}\mathrm{r}\left(\mathrm{S}\right)},\mathrm{S}<0\end{array}\right.$$

(2)

$$S={\sum }_{i=1}^{n-1}{\sum }_{j=i+1}^{n}sgn({RESI}_j-{RESI}_i)$$

(3)

$$sgn({RESI}_j-{RESI}_i)=\left\{\begin{array}{c}1({RESI}_j-{RESI}_i>0)\\ 0({RESI}_j-{RESI}_i=0)\\ -1({RESI}_j-{RESI}_i<0)\end{array}\right.$$

(4)

$$\mathrm{V}\mathrm{a}\mathrm{r}\left(\mathrm{S}\right)=\mathrm{n}(\mathrm{n}-1,2\mathrm{n}+5)/18$$

(5)

where n denotes the length of the time series; sgn is the sign function; S is the statistic to be constructed for the Mann–Kendall test; Var(S) is the variance of S; Z is the standard normal variable, which is tested for significance using a significance level of ɑ = 0.05; when |Z| ≥ 1.96, it indicates that there is a significant change in RSEI, and this step is done through Matlab R2023a.

2.3.3. Spatial Autocorrelation Analysis

Spatial correlation analysis of ecological quality helps describe the spatial heterogeneity within the study area, determining whether ecological quality exhibits spatial clustering based on the overall degree of dispersion. This analysis reveals whether the attribute values of spatial units are correlated with those of neighbouring units [50]. The global Moran’s index (Moran’s I) measures the overall spatial correlation of attribute values across the entire study area [51]. The significance of Moran’s I is assessed using the Z-score and p-value. The formula for Moran’s I is Formula (6).

$$I={\displaystyle \frac{m\times \sum _{i=1}^m{\sum }_{j=1}^m{W}_{ij}(A_i-\overline{A})(A_j-\overline{A})}{\sum _{i=1}^m\sum _{j=1}^m{W}_{ij}{(A_i-\overline{A})}^2}}$$

(6)

where I is the Moran’s I, m is the total number of elements, $A_i$ is the attribute value of the ith element, $\overline{A}$ is the average attribute value of all elements in the study area, and $W_{ij}$ is the spatial weight. The value domain of Moran’s I is located in [−1,1], which when greater than 0 indicates that the spatial unit attribute has a positive correlation, in which the closer the Moran’s I is to 1 the more significant is the aggregation effect of the attribute; less than 0 suggests that the spatial unit attribute has a negative correlation, in which the closer the Moran’s I is to −1 the more significant is the convergence effect of the attribute; if the value of the Moran’s I tends to be close to 0, the attribute is considered to exhibit a randomly distributed property in the space [52].

2.3.4. The Hurst Index

The Hurst method quantitatively measures the information dependence and continuity of long-time series data, indicating the extent to which past information influences future trends. It is a valuable tool for predicting the future trend of data based on the Hurst exponent [53]. The specific formula is as follows:

$${\displaystyle \frac{R}{S}}={\left(cm\right)}^H$$

(7)

where R is the extreme deviation, S is the standard deviation, c is a constant, and the observations are divided into n sub-sequences ${RESI}_i$, where i = 1, 2, …, n; m is any positive integer and 0 < m < n, H is the Hurst exponent, and the formula for the extreme deviation R is as follows:

$$R=maxX(t,m)-minX(t,m)$$

(8)

where X(t) is the cumulative deviation, calculated as follows:

$$X(t)={\sum }_{i=1}^m({RESI}_i-{\displaystyle \frac{1}{m}}{\sum }_{i=1}^m{RESI}_i)$$

(9)

where 1 < t < m; standard deviation S is calculated as follows:

$$S=\sqrt{{\displaystyle \frac{1}{m}}{{\sum }_{i=1}^m({RESI}_i-{\displaystyle \frac{1}{m}}{\sum }_{i=1}^m{RESI}_i)}^2}$$

(10)

The Hurst index (H) ranges from 0 to 1 and reflects the persistence or anti-persistence of the RSEI trends. When 0 < H < 0.5, it indicates anti-persistence, meaning the future trend will likely be opposite to the past; the closer H is to 0, the stronger the anti-persistence. When H = 0.5, the data exhibit randomness with no clear trend. When 0.5 < H < 1, the data are persistent, meaning future trends will align with past trends; the closer H is to 1, the stronger the persistence [54]. For analysis, H was classified into four categories: strong inverse persistence (<0.35), weak inverse persistence (0.35–0.5), weak persistence (0.5–0.65), and strong persistence (>0.65). Overlaying Hurst and change trends through the MATLAB R3023a software can derive future directions for RSEI; see Section 3.3 for the specific scope.

2.3.5. Geodetector

Considering the study area’s conditions and data availability, key factors closely related to the RSEI but not directly used to produce RSEI—land use and land use cover (LULC), precipitation (PRE), temperature (TMP), wind speed (SWS), elevation (DEM), slope (SLOP), and relief (REL)—were selected to explore the driving factors of RSEI spatiotemporal evolution.

Geodetectors reveal the factors behind the phenomenon, describing the explanatory power of the independent variable over the dependent variable [55]. To explore the dominant factors affecting the spatial variation of RSEI, a one-factor detector was used to reveal the spatial heterogeneity of each influencing factor. The formula is as follows:

$$\mathrm{q}=1-{\displaystyle \frac{\sum _{\mathrm{h}=1}^{\mathrm{L}}{\mathrm{N}}_{\mathrm{h}}{{\mathsf{\sigma }}^2}_{\mathrm{h}}}{\mathrm{N}{\mathsf{\sigma }}^2}}$$

(11)

where q is the degree of explanation of RSEI, h = 1, 2, …, L is the stratification of variables, and ${\mathrm{N}}_{\mathrm{h}}$ and N are the class h and the sample size of the whole region, respectively. ${\mathsf{\sigma }}^2$ and ${{\mathsf{\sigma }}^2}_{\mathrm{h}}$ represent the RSEI variance of the whole region and layer h. The value of q is in the range of [0, 1], corresponding to the spatial differentiation degree of the dependent variable. The larger the value, the stronger the explanatory power for the ecological quality of the study area [56].

Interaction detection can assess the interaction relationship between different factors and evaluate whether there is a change in the influence of varying factor combinations on ecological quality in the Tengger Desert. The specific classification is detailed in

Table 1. Interaction types of eco-environment quality in the Tengger desert.

Basis for Judgment	Types of the Interaction
$q(X_1\cap X_2)<Min\left[q\left(X_1\right),q\left(X_2\right)\right]$	Nonlinear attenuation
$M\mathrm{in}\left[q(X_1),q(X_2)\right]<q(X_1\cap X_2)<M\mathrm{ax}\left[q(X_1),q(X_2)\right]$	Unilinear attenuation
$q\left(X_1\right)+q\left(X_2\right)>(X_1\cap X_2)>Max\left[q\left(X_1\right),q\left(X_2\right)\right]$	Two-factor enhancement
$q(X_1\cap X_2)=q\left(X_1\right)+q\left(X_2\right)$	Independent
$q(X_1\cap X_2)>q\left(X_1\right)+q\left(X_2\right)$	nonlinear enhancement

. In this study, firstly, the effect of a single factor on the dependent variable was calculated. Then, the q-value was calculated after the interaction of the two factors to determine the impact on the dependent variable.

The natural breakpoint method was applied to reclassify the independent variables to meet the requirements of the Geodetector tool. A 1000 × 1000 m grid was constructed using ArcGIS 10.7, and the values of the variables at the centre of each grid cell were extracted for analysis.

3. Results

3.1. Dynamic Change of RSEI

The mean RSEI in the Tengger Desert fluctuated between 0.1542 and 0.2906 from 2001 to 2021, indicating poor ecological quality (Figure 2). The RSEI peaked in 2008, likely reflecting the impact of major national ecological projects such as the conversion of farmland to forest and grassland, and efforts to control sandstorms in Beijing and Tianjin. Though there was a slight, insignificant improvement from 2001 to 2008 (0.0148 a⁻¹), the overall trend showed a gradual decline, with an annual decrease of 0.0008 a⁻¹ (p > 0.05) from 2008 to 2021.

3.2. Spatiotemporal Characteristics of Ecological Quality

3.2.1. Temporal Characteristics

Ecological quality analysis (Figure 3) shows improvement from 2001 to 2006, followed by deterioration from 2006 to 2011, mainly due to shifts between poor and fair classes. Deterioration outweighed improvements, with few sharp changes. From 2001 to 2021, only about 15% of the area had above-moderate ecological quality, while 70% had poor quality, initially decreasing but later increasing. Ecological quality improved between 2011 and 2016 but deteriorated again from 2016 to 2021. By 2021, areas with moderate or better quality had decreased compared to 2001, with more pronounced deterioration than in 2006.

Table 2. Area transfer matrix of different area levels in the Tengger Desert.

RSEI Levels		2021 (km2)					Total (km2)
		Poor	Fair	Moderate	Good	Excellent
2001 (km2)	Poor	2143.4875	1981.7252	1979.3879	1979.1584	1979.0736	10,062.8326
	Fair	1993.3964	1990.817	1980.1176	1979.5741	1979.7191	9923.6242
	Moderate	1980.4561	1983.6106	1981.2788	1979.7469	1979.8999	9904.9923
	Good	1978.972	1979.6497	1980.9664	1979.9197	1980.1457	9899.6535
	Excellent	1978.8713	1979.0449	1979.6182	1980.6172	1990.7458	9908.8974
Total (km2)		10,075.1833	9914.8474	9901.3689	9899.0163	9909.5841	49,700

shows a slight decline in average ecological quality in the study area from 2001 to 2021, with a two-way shift toward lower levels, suggesting that areas of lower ecological quality may have shifted locations. The total reduction in the area above the middle ecological grade amounted to 3.5739 km². The increase in area due to ecological deterioration exceeded the decrease from improvement, indicating an overall decline in the ecological quality of the Tengger Desert.

3.2.2. Spatial Characteristics

Figure 4 illustrates the temporal and spatial changes in ecological quality across the Tengger Desert region for the years 2001, 2006, 2011, 2016, 2021, and the period from 2001 to 2021 (Figure 4f), based on the Remote Sensing Ecological Index (RSEI). Spatially, areas with poor ecological quality are mainly concentrated in the central parts of the study area. In contrast, areas with better ecological quality are primarily located along the margins and some central regions, particularly around the periphery of the Qinghai–Tibet Plateau and in the areas surrounding the Qilian and Helan Mountains. Overall, the ecological quality of the study area fluctuated between 2001 and 2021, with significant degradation in certain areas that require restoration measures.

Figure 5 shows the spatial distribution of RSEI variations from 2001 to 2021. Most areas saw slight or insignificant improvements, especially in the northeast and southwest, with a few areas experiencing significant increases from 2001–2006 (Figure 5a). Improvement trends continued, but slight degradation appeared in some southern and central regions from 2006–2011 (Figure 5b). Ecological stability was observed from 2011 to 2016 (Figure 5c), with improvements in the northeast, while slight declines occurred in the southern and central areas. Degradation intensified in the south and centre from 2016–2021 (Figure 5d), although some slight improvements persisted in the northeast. Over the two decades (Figure 5e), significant improvements ($\left|Z\right|\ge 1.96)$ occurred in the northeast and southwest, while widespread degradation dominated the central and southern regions. Overall, recent years indicate a notable ecological decline, especially in central and south areas, underscoring the need for restoration.

Significant decreases ($\left|Z\right|\ge 1.96)$ in ecological quality were most notable during the 2001–2021 period, affecting 10.33% of the area (

Table 3. Changes in the ecological grade of the Tengger Desert from 2001 to 2021.

RSEI Variation Category	2001–2006		2006–2011		2011–2016		2016–2021		2001–2021
	Area (km2)	Ratio (%)	Area (km2)	Ratio (%)	Area (km2)	Ratio (%)	Area (km2)	Ratio (%)	Area (km2)	Ratio (%)
Significant decrease	511.91	1.03	298.20	0.6	109.34	0.22	745.5	1.50	5134.01	10.33
Slight decrease	1103.34	2.22	1804.11	3.63	278.32	0.56	1689.8	3.40	2584.4	5.2
No significant decrease	22,176.14	44.62	31,435.25	63.25	12,812.66	25.78	29,750.42	59.86	22,449.49	45.17
Unchanged	4234.44	8.52	4796.05	9.65	6192.62	12.46	5258.26	10.58	5511.73	11.09
No significant increase	19,914.79	40.07	10,934	22.00	27,856.85	56.05	11,639.74	23.42	11,704.35	23.55
Slight increase	1192.8	2.4	273.35	0.55	1868.72	3.76	367.78	0.74	839.93	1.69
Significant increase	566.58	1.14	159.04	0.32	581.49	1.17	248.5	0.50	1476.09	2.97

). Slight and insignificant decreases dominated all periods, especially from 2006–2011 and 2016–2021, where over 60% of the area experienced some decline. No significant increases ($\left|Z\right|\le 1.645)$ peaked between 2011–2016, covering 56.05% of the area. The proportion of areas with significant increases remained relatively small but stable throughout the years, reaching 2.97% over the entire 20-year span. These data highlight a general trend of ecological decline, with only small areas experiencing significant improvement.

3.2.3. RSEI Spatial Autocorrelation Analysis

The values of Moran’s I range from approximately 0.0309 (in 2011) to 0.0612 (in 2001) (Figure 6). The highest Moran’s I, valued in 2001, suggests a more substantial positive spatial autocorrelation of ecological quality, meaning similar ecological conditions were more clustered in that year. After 2001, Moran’s I values fluctuated, with lower values in 2011 and 2016, indicating reduced clustering and more spatial randomness. A slight increase was observed again in 2021. Z-scores reflect the statistical significance of the spatial autocorrelation. Higher z-scores in 2001 (36.2941) indicate a more significant clustering of similar ecological quality levels in that year. The z-scores drop significantly in 2011 (19.5043), suggesting weaker spatial autocorrelation, before rising again in 2016 (26.4672) and 2021 (30.979), indicating moderate to strong spatial autocorrelation during those years. Overall, these results suggest that the spatial clustering of ecological quality was most pronounced in 2001, with a notable decline in 2011, followed by a moderate recovery in subsequent years.

3.3. RSEI Trend Persistence Analysis

The Hurst index was calculated through pixel-by-pixel analysis of RSEI results from 2001 to 2021 in the Tengger Desert, yielding an average value of 0.5567 (

Table 4. Future change trend of eco-environment in the Tengger Desert.

${\mathit{\beta }}_{\mathit{R}\mathit{S}\mathit{E}\mathit{I}}\times {10}^{-5}$	Hurst	Change Directions	Future Change Trend	Area (km2)	Ratio (%)
<−5	<0.35	From deterioration to improvement	Anti-strength continues to deteriorate	1153.04	2.32
	0.35~0.5		Anti-weak persistent deterioration	8613.01	17.33
	0.5~0.65	Continuous degradation	Weak persistent deterioration	13,394.15	26.95
	>0.65		Strong persistent deterioration	6684.65	13.45
≥5	<0.35	From improvement to deterioration	Anti-strong continuous improvement	521.85	1.05
	0.35~0.5		Anti-weakness continuous improvement	4015.76	8.08
	0.5~0.65	Continuous improvement	Weak persistent improvement	5869.57	11.81
	>0.65		Strong continuous improvement	3300.08	6.64
−5~5	0~1	Basically unchanged	Basically unchanged	6147.89	12.37

). This indicates a strong positive correlation in the RSEI changes across the study area. Areas with a Hurst index less than 0.5, accounting for 31.44% of the total area, exhibit weak continuity, suggesting unstable ecological quality changes. In contrast, 63.50% of the area has a Hurst index greater than 0.5, indicating more persistent RSEI changes and relatively stable ecological quality in these regions.

To further investigate the variation patterns of ecological quality in the study area, the Hurst Index and Sen Index for the Tengger Desert were combined to project the future trend of the RSEI (Figure 7). The results indicate that the ecological quality in the Tengger Desert is expected to mostly deteriorate, with 40.40% of the area showing continuous deterioration and 9.13% transitioning from improvement to deterioration.

The areas likely to experience the most significant ecological deterioration are located along the edges of towns and in the eastern part of the study area. These regions will require sustained ecological management efforts over the long term. Additionally, approximately 12% of the study area has a Hurst Index close to 0.5, making it difficult to predict future changes in ecological quality. However, areas with deteriorating ecological quality account for more than 49.53% of the total area, surpassing those with improving ecological quality (38.1%) and areas with no change (12.37%). Future studies should closely monitor these regions, particularly those undergoing significant ecological changes.

3.4. Factors Driving RSEI

3.4.1. Single-Factor Detection

The results in Figure 8 show that all driving factors, except slope, significantly impact RSEI, passing the 0.05 significance test. LULC consistently had the highest explanatory power, especially in 2021, with a q-value of 0.4968, passing the 1% significance test. This suggests that human activities have strongly influenced ecological changes in the Tengger Desert over the past 21 years. Besides LULC, temperature (TMP) and precipitation (PRE) also had high explanatory power in 2001 and 2021, while wind speed (SWS) and TMP were significant in 2006, 2011, and 2016. In contrast, slope (SLOP), relief (REL), and elevation (DEM) had the least influence. The ranking of PRE varied over time, indicating that land use changes affected local microclimate conditions. Reduced rainfall and expanding desertification negatively impacted local habitats, highlighting the need for stronger ecological protection in the Tengger Desert to prevent further degradation.

3.4.2. Factor Interaction Detection

The interaction detection results, shown in Figure 9, reveal that the explanatory power of factors on RSEI is significantly enhanced when combined, with no cases of mutual independence or weakening. This indicates that the impact of interacting factors on ecological quality changes in the Tengger Desert is greater than that of individual factors alone. Among all combinations, LULC and TMP—both strong single-factor influencers—showed high explanatory power when interacting with other factors. The combination of wind speed (SWS) and LULC had the strongest explanatory power in 2001, 2011, and 2016, reaching a maximum q-value of 0.5847. This suggests that wind speed, interacting with land-use practices, significantly influences RSEI, likely due to wind erosion in desertified areas with low vegetation cover. Overall, the ecological quality of the Tengger Desert is shaped by anthropogenic activities and the synergistic effects of multiple factors. Studying these interactions provides a more comprehensive understanding of their impact on the region’s ecological changes.

4. Discussion

4.1. RSEI Spatial and Temporal Characteristics

This study used CHEQ datasets (NDVI, WET, NDBSI, LST, AI) to evaluate the ecological quality of the Tengger Desert. This approach mitigates the limitations of using single or partial indicators and reduces the subjectivity of human factors. It also simplifies data collection, enabling quantitative analysis and visualisation of ecological quality over a long time series. The RSEI remains a comprehensive and reliable method for assessing environmental quality when constructing ecological indicators.

This study analysed the spatiotemporal evolution of RSEI in the Tengger Desert over 21 years and developed a future change model. From 2001 to 2021, RSEI showed an overall declining trend (Figure 2), with a significant peak in 2008. According to Liu et al. [57], this peak aligns with the launch of the edge-locking project at the desert’s periphery, which temporarily improved ecological quality. However, the ecosystem was in the early recovery phase, and human-induced damage such as farmland reclamation, industrial and mining construction, irrational land use, and overgrazing outweighed the benefits of restoration [58], leading to a noticeable decline in RSEI between 2008 and 2009. Despite initial restoration efforts, the region’s ecological quality continued to degrade, highlighting the need for us to strengthen ecosystem protection and pre-assessment of environmental monitoring after the project is implemented.

The increasing trend of average temperature and decreasing rainfall (Figure 10) has caused an imbalance between evaporation and water recharge, worsening drought conditions and contributing to fluctuations in RSEI in certain years. These findings suggest a strong correlation between mean temperature, precipitation, and RSEI dynamics, as noted by Yang et al. [59]. Rising temperatures have made the mountain climate more favourable for plant growth, resulting in higher NDVI values in areas near towns and mountains (Figure 11). Since NDVI carries significant weight in RSEI calculation [60,61], the RSEI in these marginal mountain areas is high. However, unused land still accounts for 88.51% of the study area, keeping the overall RSEI value low. Additionally, 49.53% of the Tengger Desert is projected to deteriorate in the future, indicating increasing fragility in the natural environment due to changes in human activities [54,62,63,64]. Some areas showing signs of reversing ecological improvement trends require ongoing monitoring and field research [61].

4.2. Factors Driving RSEI

The results of the Geodetector analysis indicate that LULC is a key factor directly influencing changes in RSEI, likely due to the inclusion of AI in the CHEQ dataset’s principal component analysis and LULC’s reflection of human activities. This makes the CHEQ dataset more effective in capturing the relationship between ecological quality changes and LULC [31]. Previous research [65,66,67] supports the conclusion that human factors are the primary drivers of desert changes, aligning with the findings of this study. From 2001 to 2021, land use in the study area shifted primarily from unused land to cultivated land. Liu et al. noted that newly cultivated land tends to have low agricultural stability, which helps explain the overall decline in RSEI during this period [68].

The interaction between factors had a stronger influence on RSEI spatial differentiation than single factors. In 2021, the interaction between LULC and TMP had the largest effect on RSEI, while the interaction between LULC and SWS was most significant in 2001, 2011, and 2016. This suggests that both anthropogenic and climatic factors drive RSEI dynamics [69,70]. Although topographic factors, as noted by Yang et al. [25], can influence RSEI, their explanatory power in this study was lower, likely due to their longer evolutionary period and lower sensitivity on the 21-year timescale [71]. This aligns with findings from other arid desert regions [29].

The findings of this study offer valuable guidance for improving and protecting the ecological quality of the Tengger Desert. Short-term efforts should focus on optimising land use and protecting key ecological areas, while medium- and long-term strategies should consider regional natural conditions to balance ecological protection with sustainable tourism and economic development.

4.3. Limitations and Future Perspectives

Based on the CHEQ dataset, the RSEI index incorporating AI was used to quantitatively analyse the evolution and driving factors of the Tengger Desert’s ecological quality from 2001 to 2021. The RSEI index, enhanced by AI, performed well at the regional spatial scale and showed a high correlation with EI data [31,39]. However, water bodies were masked in the RSEI calculation to prevent interference, meaning the index cannot assess water quality, a key ecosystem component [30,58]. Given the importance of water and atmospheric environments, future studies should consider integrating related indicators to develop a more comprehensive evaluation system for watershed ecological quality [72,73,74].

This study ended in 2021 due to data limitations, and the data’s spatial resolution was standardised to 30 m, which may have caused information loss from lower-resolution datasets. Zhuo Lu also concluded, by comparing the effects of 10 m and 30 m resolution, that higher-resolution data would give better results in assessing habitat quality, that are not captured by analyses at low resolution [75]. Additionally, the explanatory power of impact factors varies across spatial scales, and the results are influenced by data discretisation methods [76,77]. While this study experimented with different discretisation techniques, it could not guarantee that the same indicators passed the 0.01 significance test in all years. These subtle differences may affect the findings. Future research should focus on building higher-resolution, more timely datasets and refining data discretisation methods to improve detection results across different scales and partitions, ensuring more accurate outcomes.

5. Conclusions

In this study, the Tengger Desert was used as the research area, and the RSEI index was used as the evaluation index of ecological quality. The spatial and temporal dynamics of ecological quality from 2001 to 2021 were analysed, and the dominant factors of ecological quality change in the study area were revealed by Geodetector. The results show that:

(1)

From 2001 to 2021, the average RSEI value in the Tengger Desert decreased by 9.33%, reflecting an overall decline in ecological quality. There are periodic changes in the ecological quality in the study area: 2001–2006 is the improvement stage, 2006–2011 is the accelerated deterioration stage, 2011–2016 is the improvement stage, and 2016–2021 is the deterioration stage. Through the analysis of a spatial transfer matrix, it is concluded that the ecological positions of different levels have migrated, mainly between two adjacent levels.

(2)

From 2001 to 2021, the ecological quality of the Tengger Desert was in a deteriorating state, mainly distributed in a large area, with the largest area of mild deterioration. The improvement area accounted for 28.21%, mainly distributed in the edge and centre of the study area. In addition, 49.53% of the study area may be at risk of ecological environment deterioration in the future. It is necessary to pay attention to the ecological situation in the eastern part of the study area and the edges of the towns and take ecological restoration measures to curb the ecological deterioration trend of the Tengger Desert.

(3)

Land use and land use cover emerged as the most significant factors influencing RSEI. The interaction of land use with wind speed had the strongest explanatory power, highlighting the role of anthropogenic factors and natural forces in shaping ecological conditions. Other climate variables, such as temperature and precipitation, also played important roles, though their influence was less pronounced than that of land use.

Overall, this study highlights the spatial and temporal changes in ecological quality in the Tengger Desert, with a general trend of deterioration from 2001 to 2021. Future efforts should focus on sustainable land use practices to mitigate further degradation and protect the region’s fragile environment.