Spatiotemporal Evolution and Driving Factors of Desertification Sensitivity During Urbanization: A Case Study of the Beijing–Tianjin–Hebei Core Region

Abstract

Desertification sensitivity in semi-arid urbanizing regions remains a critical challenge for sustainable land management. This study analyzes the spatiotemporal dynamics (2018–2022) of desertification sensitivity in the Beijing–Tianjin–Hebei core region using the Normalized Difference Vegetation Index (NDVI), soil texture, the Digital Elevation Model (DEM), and nighttime light data. Using a GIS-based model, we found a decline in overall desertification sensitivity, with vegetation degradation (post-2020) emerging as a key factor. Key recommendations include optimizing urban spatial patterns via ecological red lines, prioritizing vegetation restoration in high-sensitivity zones, and establishing dynamic remote sensing-based monitoring systems. These strategies aim to coordinate urban growth with ecological resilience, offering actionable pathways for semi-arid regions facing similar pressures. Future work should integrate socioeconomic drivers to refine adaptive governance frameworks.

1. Introduction

Desertification is one of the most severe ecological and environmental challenges worldwide, particularly in semi-arid agricultural regions, where it poses a significant threat to ecosystem stability, soil fertility, and agricultural productivity [1]. The formation of desertification results from the combined effects of multiple natural and anthropogenic factors, including climate change, soil properties, vegetation degradation, and human disturbances. In recent years, the development of remote sensing and Geographic Information System (GIS) technologies has provided essential technical support for the monitoring, assessment, and prediction of desertification [2].

Traditional ground-based vegetation monitoring methods typically rely on a visual assessment or coverage estimation using photographs and grid-based reference measurements. While these methods are straightforward, they are susceptible to observer subjectivity, resulting in accuracy limitations and challenges in supporting a long-term dynamic analysis. In contrast, remote sensing technology, with its advantages of large-scale coverage, dynamic updating, and cost-effectiveness, has become an essential tool for vegetation monitoring [3]. The vegetation index, as a key parameter for quantifying vegetation characteristics, has been widely used in desertification assessment. For instance, Fensholt et al. integrated MODIS and AVHRR data and found that the vegetation index in semi-arid areas around the world showed an upward trend from 1981 to 2007, attributing precipitation as the primary driving factor [4]. Li et al. revealed the spatiotemporal heterogeneity of vegetation changes based on global vegetation time-series product data. In India, Banerjee et al. demonstrated that vegetation dynamics are jointly dominated by precipitation and human activities [5].

In China, domestic studies have also highlighted regional vegetation dynamics. Guo Zichen’s research team utilized MODIS-based Normalized Difference Vegetation Index (NDVI) products to observe an overall increasing vegetation cover trend in the Mu Us Sandy Land but noted significant spatial and temporal lag effects in response to hydrothermal factors [6]. Zhang Xu et al. demonstrated through a regression analysis that desertification in the Loess Plateau is predominantly concentrated in the northwestern region, exhibiting marked temporal fluctuations, yet showing an overall improving trend [7].

In northern China, semi-arid climatic conditions coupled with intensive human activities have exacerbated land degradation issues. The core area of the Beijing–Tianjin–Hebei (BTH) region, which includes the Wuqing District, the Daxing District, the Tongzhou District, Gu’an County, and Yongqing County, is an important region for studying urbanization-driven ecological transformation. As one of the most rapidly urbanizing regions in China, the BTH region has experienced unprecedented urban expansion, leading to fragmented landscapes and intensified pressure on fragile ecosystems [8]. However, the existing studies predominantly focus on isolated factors (e.g., climate or soil erosion), while the synergistic effects of urban expansion, vegetation degradation, and human activity intensity remain underexplored. This gap limits the ability to formulate holistic strategies for balancing urban development and ecological resilience.

This study addresses this need by integrating multi-dimensional drivers—such as the NDVI [9], soil texture, the Digital Elevation Model (DEM), and nighttime light data—to construct a spatially explicit desertification sensitivity model. Using high-resolution datasets from 2018 to 2022, we quantify the spatiotemporal evolution of sensitivity and disentangle the dominant roles of urbanization and vegetation dynamics. For instance, nighttime light data, a proven proxy for human activity intensity, reveal how infrastructure expansion directly correlates with localized sensitivity spikes [10]. NDVI trends capture the vegetation degradation patterns linked to land-use conversion, a critical yet understudied pathway in semi-arid regions. Through a weighted overlay analysis, transition matrices, and a hotspot analysis, this study quantifies desertification sensitivity changes and explores the impacts of the NDVI and human activities, providing insights for regional land resource management and ecological restoration.

2. Materials and Methods

2.1. Study Area

The study area includes the Wuqing District, the Daxing District, the Tongzhou District, Gu’an County, and Yongqing County, and is located at the junction of Beijing, Tianjin, and Hebei Provinces (Figure 1). This region is situated on the North China Plain, characterized by flat terrain with elevations ranging from −18 to 55 m, predominantly consisting of alluvial plains. The climate is classified as a temperate monsoon climate, with an average annual temperature of 11.5 °C–12.5 °C and an annual precipitation of 450–600 mm, over 70% of which occurs between June and September. The major rivers in the study area include the Yongding River, the North Canal, and the Ziya River. The dominant soil types are fluvo-aquic soil, Shajiang black soil, and aeolian sandy soil, with the latter being widely distributed and highly susceptible to erosion.

2.2. Data Sources

This study utilizes multi-source remote sensing datasets to comprehensively evaluate desertification sensitivity in the study area by integrating key factors such as vegetation coverage, soil properties, topographic characteristics, and the intensity of human activity disturbances. To guarantee the reliability and spatial compatibility of the data, authoritative global or regional datasets were selected. All datasets underwent rigorous preprocessing procedures, including a quality assessment (QA) and consistency correction; all spatial datasets were projected to the same coordinate system, WGS_1984_UTM_Zone_51N, to ensure spatial consistency across layers, and resampled to achieve uniform spatial resolution (30 m). Specifically, NDVI data and soil texture data were resampled from their original coarser resolutions to 30 m resolution using bilinear interpolation, which minimizes accuracy degradation compared to the nearest neighbor methods. The terrain data (DEM) underwent void-filling preprocessing by interpolating neighboring pixel values using the inverse distance weighted (IDW) interpolation method. Nighttime light data were corrected for consistency and continuity across the study period and resampled to 30 m using bilinear interpolation to match the other datasets. Although resampling to a finer resolution does not increase the actual detail or accuracy, employing the bilinear interpolation method reduces artificial precision and minimizes information loss. Details of the datasets and precise preprocessing procedures are elaborated in

Table 1. Data source.

Data Type	Database Source	Resolution	Website	Pretreatment
NDVI	MOD13Q1 (Collection 6), annual average for growth season (May–September, 2018–2022)	250 m	https://search.earthdata.nasa.gov (accessed on 3 February 2025)	QA layer screening to remove clouds and low-quality pixels; resampled to 30 m resolution using bilinear interpolation method.
Soil texture	HWSD v2.0 Global Sand content grid (Sand%, 2018)	1000 m	https://www.fao.org/soils-portal (accessed on 3 February 2025)	Resampled to 30 m resolution using bilinear interpolation.
Terrain data	ALOS World 3D DEM v3.2 (2018)	30 m	https://www.eorc.jaxa.jp (accessed on 3 February 2025)	Void filling performed through inverse distance weighted (IDW) interpolation method, utilizing values from surrounding pixels to estimate missing elevations.
Human activity intensity	NPP-VIIRS Annual Stable Nighttime Light data (2018–2022)	500 m	https://eogdata.mines.edu (accessed on 3 February 2025)	Consistency and continuity corrections applied to reduce inter-annual variability; resampled to 30 m resolution using bilinear interpolation.

.

Soil texture was classified based on sand content into low erodibility (<30%), moderate erodibility (30%–60%), and high erodibility (>60%) classes [11]. Nighttime light intensity was categorized by DN values into low disturbance (<5), moderate disturbance (5–10), and high disturbance (>10) categories [12].

2.3. Methods

This study utilizes the spatial analysis module of the ArcGIS platform (ArcMap 10.8) to explore the evolutionary mechanisms of desertification sensitivity through multi-factor weighted integration and spatiotemporal coupling analysis. Based on a unified spatial reference system and standardized datasets, a hierarchical analytical approach is employed to achieve a quantitative integration of the indicator system.

In accordance with the “Technical Specification for Ecological Quality Assessment in Desertification Areas” (LY/T 3242-2020), weight coefficients are assigned to each factor as follows: vegetation degradation (0.35), topography (0.25), soil (0.25), and human activity (0.15). This assignment reflects the relative importance of each factor in influencing desertification sensitivity, with vegetation degradation being the most significant due to its direct impact on land cover and ecosystem stability. Topography and soil properties are equally weighted, acknowledging their combined role in determining soil erosion potential and water retention capacity. Human activity intensity is assigned a lower weight, recognizing its indirect but notable influence on desertification processes. These weightings are derived directly from the LY/T 3242-2020 standard, which is based on extensive empirical research and expert consensus in the field of ecological assessments in desertification-prone areas. Alternative weighting schemes were not tested in this study as the focus was adhering to the standardized methodology to ensure consistency and comparability with other assessments conducted under the same framework. Utilizing the Weighted Overlay tool in ArcMap 10.8, these weighted factors are integrated to generate a comprehensive desertification sensitivity classification map, facilitating the identification of areas at varying risk levels and informing targeted land management strategies.

To provide a clear overview of the research process, the technical workflow for data processing, sensitivity modeling, and a spatiotemporal analysis is illustrated in Figure 2. This schematic outlines the sequential steps from data acquisition and preprocessing to the desertification sensitivity assessment, hotspot analysis, and correlation evaluation of driving forces.

Surface roughness is calculated based on DEM, representing the impact of topography on soil erosion. The vegetation degradation index is computed using the temporal trend of NDVI changes [13], using the following equation:

$$\begin{array}{c}\mathrm{Vegetation} \mathrm{degradation} \mathrm{index}={\displaystyle \frac{\mathrm{n}\times \sum _{\mathrm{i}=1}^{\mathrm{n}} {\mathrm{x}}_{\mathrm{i}}{\mathrm{y}}_{\mathrm{i}}-\left(\sum _{\mathrm{i}=1}^{\mathrm{n}} {\mathrm{x}}_{\mathrm{i}}\right)\left(\sum _{\mathrm{i}=1}^{\mathrm{n}} {\mathrm{y}}_{\mathrm{i}}\right)}{\mathrm{n}\times \sum _{\mathrm{i}=1}^{\mathrm{n}} {\mathrm{x}}_{\mathrm{i}}^2-(\sum _{\mathrm{i}=1}^{\mathrm{n}} {\mathrm{x}}_{\mathrm{i}}{)}^2}}\end{array}$$

(1)

In the equation, y_i represents the value for year x_i; when the vegetation degradation index > 0, it indicates an increasing trend, whereas when the vegetation degradation index < 0, it signifies a declining trend. NDVI-based vegetation degradation indices effectively reflect vegetation conditions by capturing the long-term vegetation dynamics driven by climatic variations, land-use practices, and ecological management efforts. Specifically, a increasing trend may signify successful ecological restoration or sustainable land management practices, while a declining trend typically indicates stress or degradation due to drought, human disturbances, or unsustainable agricultural activities [14]. However, the index does have limitations, as it simplifies complex vegetation dynamics into linear trends, potentially overlooking short-term fluctuations and nonlinear vegetation responses to disturbances, thus necessitating a cautious interpretation and complementary field-based validation.

The spatiotemporal evolution analysis employs a sensitivity transition matrix to quantify the changes in different sensitivity levels from 2018 to 2022. Additionally, the hotspot analysis (Getis-Ord) is used to identify highly sensitive regions. Further, the Getis-Ord Gi hotspot analysis is applied to pinpoint areas with the most significant sensitivity changes [15], and vegetation change data along with human activity disturbance data are extracted within these hotspot areas to analyze their roles in sensitivity variations.

The driving factor analysis adopts correlation testing, calculating the Pearson correlation coefficient between nighttime light variation, NDVI change trends, and desertification sensitivity changes to quantify the influence of urban expansion and vegetation degradation on desertification sensitivity. This approach also evaluates their impact across different time scales.

By integrating multiple indicators, this study reveals the spatiotemporal variation patterns of desertification sensitivity in the study area and provides a scientific basis for ecological management and land resource planning.

3. Results

3.1. Spatiotemporal Evolution of Sensitivity

In this study, the Natural Breaks (Jenks) classification method was employed to categorize desertification sensitivity into five distinct levels: Level 1 (Extremely Low), Level 2 (Low), Level 3 (Moderate), Level 4 (High), and Level 5 (Extremely High). The Natural Breaks method is a data clustering technique that identifies breakpoints by minimizing the variance within classes and maximizing the variance between classes, effectively grouping similar values and highlighting the differences among classes [16,17]. The rationale for selecting this classification approach lies in its ability to accommodate the inherent distribution patterns of the data, ensuring that each class is both statistically and practically significant. The specific breakpoints were determined based on the distribution of the desertification sensitivity index values, resulting in consistent thresholds across the study period: 1.39, 1.65, 1.90, 2.15, and 2.65. These thresholds were chosen to reflect natural groupings in the data, facilitating a more accurate representation of the varying degrees of desertification sensitivity within the region. By applying the Natural Breaks method, the classification effectively captures the spatial heterogeneity of desertification sensitivity, providing valuable insights for targeted land management and ecological restoration efforts.

Based on the desertification sensitivity classification results from 2018 to 2022, the spatiotemporal dynamics of different sensitivity levels within the study area can be observed. Overall, the region is still dominated by low sensitivity (Levels 1 and 2) and medium sensitivity (Level 3). However, the distribution of the sensitivity levels shows notable variations across different years (Figure 3).

In 2018, the high sensitivity (Level 4) and extremely high sensitivity (Level 5) areas were mainly concentrated in the northern and western parts of the study area. By 2022, these high-sensitivity regions had shrunk, with some areas transitioning to lower sensitivity levels. Meanwhile, the low-sensitivity areas (Levels 1 and 2) gradually expanded year by year, particularly in the southeastern part of the region, where sensitivity significantly declined. This improvement can be primarily attributed to effective vegetation restoration initiatives and soil conservation practices. For example, from 2000 to 2020, approximately 60.9% of the Beijing–Tianjin–Hebei region experienced substantial vegetation recovery, especially in the southern regions, driven by ecological restoration projects and favorable climatic conditions [18]. Moreover, targeted land cover adjustments, such as converting marginal agricultural and desert lands to forests and grasslands, significantly enhanced regional ecological resilience, thereby reducing sensitivity to desertification [19].

To further quantify the spatiotemporal variations in desertification sensitivity within the study area, the area proportions of different sensitivity levels from 2018 to 2022 were statistically analyzed and graphed using percentage stacked bar charts (Figure 4).

From the perspective of area proportion changes across sensitivity levels, in 2018, the study area was primarily dominated by low sensitivity (Level 2) and moderate sensitivity (Level 3) regions, accounting for 45.19% and 28.37%, respectively. Meanwhile, high sensitivity (Level 4) and extremely high sensitivity (Level 5) regions covered 8.97% and 2.44%, respectively (Figure 2). By 2022, the proportion of low sensitivity areas further increased to 48.73%, while that of moderate sensitivity areas decreased to 25.21%. Additionally, the high sensitivity (Level 4) area decreased from 417.79 km² to 282.41 km², and the extremely high sensitivity (Level 5) area shrank from 113.71 km² to 60.68 km², indicating an overall decline in desertification sensitivity.

This transition suggests that some moderately sensitive areas shifted to low sensitivity categories over the five-year period, likely reflecting the positive impact of vegetation restoration and optimized land-use management [20].

A further analysis of the annual variation trends across sensitivity levels (Figure 5) reveals that the low sensitivity areas (Levels 1 and 2) exhibited a steady increase from 2018 to 2022, whereas the high sensitivity areas (Levels 4 and 5) consistently declined, with a particularly sharp decrease occurring between 2021 and 2022. This trend may be attributed to multiple factors, including the implementation of ecological restoration projects, the optimization of land-use practices, and climatic variations.

It is noteworthy that the moderate sensitivity area (Level 3) exhibited complex changes, with fluctuations observed across different years. While some areas experienced an increase in sensitivity, others transitioned to lower sensitivity levels. This variability is likely influenced by dynamic changes in land-use types, such as adjustments in farmland management practices or the progression of urbanization within the region.

Through the analysis of the sensitivity transition matrix (Figure 6), the conversion characteristics among the different sensitivity levels can be further revealed. From 2018 to 2022, the overall trend in the study area indicates a shift toward lower sensitivity levels. Specifically, the area of extremely low sensitivity (Level 1) increased from 700.18 km² to 870.79 km², while the low sensitivity (Level 2) area expanded from 2104.46 km² to 2269.68 km², with their combined proportion rising from 60.22% to 67.43%, suggesting that the desertification risk has been alleviated to some extent.

Meanwhile, the area of moderate sensitivity (Level 3) decreased from 1321.19 km² to 1174.14 km², though it still constitutes a significant proportion of the study area. This indicates that certain regions experienced short-term fluctuations in sensitivity levels, likely influenced by land-use changes or temporary environmental variations. Additionally, the high sensitivity (Level 4) and extremely high sensitivity (Level 5) areas declined from 417.79 km² to 282.41 km² and from 113.71 km² to 60.68 km², respectively, with their combined proportion decreasing from 11.41% to 7.36%. This trend further confirms that high-sensitivity areas have gradually diminished over the past five years, transitioning toward lower sensitivity levels.

Analyzing the spatial transition pathways of different sensitivity levels reveals that high sensitivity (Level 4) and extremely high sensitivity (Level 5) areas primarily shift toward moderate sensitivity (Level 3) or low sensitivity (Level 2) regions. In contrast, the transition patterns of the moderate sensitivity areas (Level 3) are more complex, with some regions shifting to lower sensitivity levels, while others evolve into high sensitivity zones [21]. This bidirectional transformation of sensitivity levels is likely driven by multiple factors, including land-use adjustments, vegetation cover changes, and human activity disturbances.

In urban expansion areas, the reduction in vegetation cover and intensified land development activities have led to shifts in desertification sensitivity levels. Specifically, the regions undergoing rapid urbanization have experienced a decline in vegetation coverage due to land conversion and construction activities, resulting in transitions from moderate sensitivity (Level 3) to high sensitivity (Level 4) or even extremely high sensitivity (Level 5). This is attributed to the loss of green spaces and increased impervious surfaces, which exacerbate soil erosion and reduce land resilience.

Conversely, the areas implementing effective ecological restoration measures and improved agricultural management have witnessed a decrease in desertification sensitivity. For instance, the Beijing–Tianjin Sandstorm Source Control Project has focused on afforestation and grassland restoration, leading to significant vegetation recovery [22]. These interventions have facilitated transitions from moderate sensitivity (Level 3) to low sensitivity (Level 2) or even extremely low sensitivity (Level 1), as enhanced vegetation cover stabilizes the soil and improves moisture retention, thereby mitigating the desertification risks.

The area variation data for different sensitivity levels further support this trend (

Table 2. Proportion of area categorized by sensitivity level from 2018 to 2022.

	2018		2019		2020		2021		2022
Category	Area/km2	Proportion/%	Area/km2	Proportion/%	Area/km2	Proportion/%	Area/km2	Proportion/%	Area/km2	Proportion/%
1	700.18	15.03	636.70	13.67	618.56	13.28	634.97	13.63	870.79	18.70
2	2104.46	45.19	1987.74	42.68	1936.96	41.58	1957.95	42.03	2269.68	48.73
3	1321.191	28.37	1309.33	28.11	1331.21	28.58	1301.69	27.95	1174.14	25.21
4	417.76	8.97	530.23	11.38	560.71	12.04	557.15	11.96	282.41	6.06
5	113.71	2.44	193.60	4.16	210.43	4.52	206.19	4.43	60.68	1.30

). Over the five-year period, the low sensitivity areas (Levels 1 and 2) increased by a total of 336.83 km², while the high sensitivity (Level 4) and extremely high sensitivity (Level 5) areas decreased by a total of 188.48 km². Despite the overall trend of decreasing sensitivity, certain localized areas still exhibited an increase in sensitivity, particularly in the northern and western regions. These areas may remain at high sensitivity levels due to factors such as topographical constraints, vegetation degradation, and human disturbances.

Therefore, in future desertification prevention and land-use planning efforts, it is crucial to monitor the changes in these high-sensitivity regions and implement targeted management strategies to mitigate desertification risks.

From the perspective of spatiotemporal variation characteristics, desertification sensitivity in the study area exhibited an overall declining trend from 2018 to 2022. However, the changes across different regions displayed significant spatial heterogeneity. In the areas where sensitivity has decreased, this trend may be attributed to vegetation restoration, ecological rehabilitation policies, or optimized land management. Conversely, for the regions where the sensitivity levels remained unchanged or even increased, it is essential to further investigate the driving mechanisms to formulate targeted management strategies.

For instance, in the areas with high sensitivity but minimal change, enhanced vegetation protection and restoration measures may be necessary to improve soil erosion resistance. In contrast, for the regions where sensitivity has increased, an integrated analysis of land-use changes should be conducted to optimize urban expansion and agricultural production layouts, thereby reducing the adverse impacts of human activities on soil environments.

In summary, from 2018 to 2022, the study area experienced an overall decline in desertification sensitivity, with the low-sensitivity areas expanding and the high-sensitivity areas decreasing, indicating that desertification risk has been mitigated to some extent. However, sensitivity variations are influenced by multiple factors, leading to notable spatial heterogeneity, particularly in certain regions where sensitivity has increased. A more in-depth driving factor analysis is needed to clarify the specific roles of land-use changes, vegetation dynamics, and human activities in sensitivity evolution. This will provide a scientific basis for regional desertification prevention and land resource management.

3.2. Driving Force Response Analysis

By examining desertification sensitivity changes, nighttime light variation rates, and NDVI trends within the study area, the primary driving factors influencing sensitivity changes can be further explored.

The driving force response analysis aimed to establish a causal relationship between urban expansion and desertification sensitivity changes, while also verifying whether vegetation degradation plays a crucial role in this process. Through the correlation analysis of desertification sensitivity changes (2018–2022), nighttime light variation rates, and NDVI trends, this study identified key driving factors affecting sensitivity variations in the region.

Using the Getis-Ord Gi* hotspot analysis implemented in ArcMap 10.8, the most significant zones of sensitivity change within the study area were identified [23]. A fixed distance band method with a 1 km neighborhood radius was used, and statistical significance was determined using a z-score threshold of ±1.96 (p < 0.05). To ensure result stability, sensitivity tests were conducted using multiple neighborhood radii (500 m, 1 km, 2 km), confirming that the hotspot patterns remained consistent across spatial scales. These hotspot regions are primarily distributed in the northwestern part of the study area and localized sections in the south.

By integrating the nighttime light variation rates and an NDVI trend analysis, it becomes evident that the regions with a significant increase in sensitivity generally correspond to the areas with enhanced nighttime light intensity and declining NDVI values. This suggests that intensified human activities and vegetation degradation may be the key factors driving changes in desertification sensitivity (Figure 7).

A further quantitative analysis was conducted using IBM SPSS Statistics 26, where the Pearson correlation coefficient (R) was calculated to assess the relationship between sensitivity changes and both nighttime light variation and NDVI trends. The analysis employed a confidence interval of 0.01, ensuring high statistical significance. The results showed that between 2018 and 2022, the Pearson correlation coefficient (R) between sensitivity changes and nighttime light variation was 0.63, while the correlation with NDVI changes was −0.54 (

Table 3. Correlation analysis between sensitivity changes, lighting changes, and NDVI changes.

Year	Light vs. Sensitivity (R)	NDVI vs. Sensitivity (R)
2018–2019	0.68	−0.62
2019–2020	0.81	−0.37
2020–2021	0.61	−0.5
2021–2022	0.56	−0.57
2018–2022	0.63	−0.54

). This indicates a positive correlation between increasing nighttime light intensity and rising desertification sensitivity [24], whereas NDVI decline is negatively correlated with sensitivity enhancement.

On an annual basis, from 2018 to 2019, the correlation between nighttime light variation and sensitivity changes was 0.68, while the correlation with NDVI changes was −0.62, suggesting that urban expansion had a significant impact on sensitivity during this period. In 2019–2020, the correlation with nighttime light further increased to 0.81, while the correlation with NDVI decreased to −0.37, indicating that human activities intensified, whereas the influence of vegetation degradation on sensitivity became relatively weaker.

This trend shifted in 2020–2021, as the correlation with nighttime light decreased to 0.61, while the correlation with NDVI increased to −0.5, suggesting that the impact of urban expansion weakened slightly, whereas vegetation changes played a greater role. By 2021–2022, the correlation with nighttime light further declined to 0.56, while the correlation with NDVI rose to −0.57, indicating that during this phase, the effect of urban expansion stabilized, whereas vegetation changes became the dominant factor in desertification sensitivity evolution. This shift may be attributed to the implementation of large-scale ecological restoration projects, such as the Three-North Shelter Forest Program, which have led to significant improvements in vegetation cover and land quality [25]. For instance, the completion of a 3000 km green belt around the Taklamakan Desert in 2024 has contributed to increased forest coverage and reduced desertification in the region [26]. These findings imply that land management strategies focusing on vegetation restoration and sustainable land-use practices have become increasingly effective at mitigating desertification sensitivity during this period.

In the areas with slower urban expansion, as indicated by the hotspot analysis, the average nighttime light variation is lower than 2.014, significantly below that of the high-sensitivity growth regions (

Table 4. Changes in lighting and NDVI within hotspots.

Quick Statistical Verification	Area/km2	Average
Hot spot area (Gi_Bin = 3)	158.427	--
Average rate of light change within the hotspot area	158.427	2.014
Average NDVI trend within the hotspot area	158.427	−0.012

). This suggests that between 2018 and 2022, the regions with less urban development, limited infrastructure expansion, and reduced industrial activity experienced more stable land-use structures, leading to lower sensitivity increases.

During periods of limited urban expansion, the growth of built-up land is slower, often correlating with maintained or even increased vegetation coverage. This trend is reflected in the NDVI variation, where in regions with lower sensitivity growth, the mean NDVI change is above −0.012, indicating less vegetation loss. This relative stability in vegetation cover helps preserve land resilience and mitigates the increase in desertification sensitivity.

From the perspective of temporal variation trends (Figure 8), urban expansion had a more pronounced impact on sensitivity changes in the early stages, whereas NDVI variations played a more critical role in later periods. This shift may be attributed to the transition from large-scale urban expansion to more intensive land-use practices. Additionally, it may also reflect the effectiveness of ecological restoration and vegetation recovery measures, which have to some extent mitigated the impact of urban expansion on desertification sensitivity.

Overall, urban expansion and vegetation degradation are the key driving factors influencing desertification sensitivity changes in the study area, though their impact varies across different periods. In the early stage (2018–2020), urban expansion played a more significant role in sensitivity changes, whereas in the later stage (2020–2022), the influence of vegetation changes became more pronounced.

Future ecological management efforts should focus on land-use regulation in rapidly expanding urban areas, while simultaneously strengthening vegetation restoration and ecological rehabilitation measures to mitigate desertification risks and enhance regional ecosystem stability.

4. Discussion

This study analyzes the spatiotemporal variations in desertification sensitivity within the study area from 2018 to 2022, revealing the impacts of urban expansion and vegetation degradation on sensitivity changes and further exploring their potential driving mechanisms.

The results indicate that while the overall desertification sensitivity in the study area shows a declining trend, some regions still experience increasing sensitivity, which is closely linked to urbanization processes and vegetation changes. This phenomenon suggests that urban expansion may contribute to vegetation degradation, subsequently increasing land desertification sensitivity [27]. To gain a deeper understanding of this process, further research is needed to explore the driving forces of urbanization, the impact of human activities on ecosystems, and potential regulatory measures.

Urbanization is one of the key driving forces affecting terrestrial ecosystems, and its impact on desertification sensitivity can be realized through multiple pathways, with land-use change being a critical factor [28]. Numerous studies have shown that as urban expansion progresses, construction land gradually encroaches upon forests, farmlands, and grasslands, leading to significant shifts in land cover types [29]. During the urbanization process, soil sealing and the reduction of natural vegetation weaken the soil’s water retention capacity and exacerbate surface water erosion and wind erosion. In the study area, the regions with higher nighttime light variation rates generally exhibit increased desertification sensitivity, indicating that urban development and infrastructure expansion are key factors influencing sensitivity changes [30].

The correlation analysis results of this study indicate that between 2018 and 2022, the average correlation coefficient between nighttime light variation and desertification sensitivity changes was 0.63, suggesting that the urbanization process had a significant impact on sensitivity changes. However, this influence may not be solely driven by urban expansion, but rather by the combined effects of land-use transformation and human disturbances [31]. Vegetation degradation is a key ecological process in desertification, and urban expansion is often considered a major contributing factor to vegetation loss [32]. During urban development, forests, farmlands, and wetlands are frequently converted into built-up land, leading to a decline in regional vegetation coverage, a reduction in soil water retention capacity, and consequently, changes in desertification sensitivity. The NDVI variation trends in this study further confirm this relationship. In the regions where desertification sensitivity increased, the mean NDVI value was −0.012, providing additional evidence that vegetation degradation may be a key factor influencing changes in desertification sensitivity.

Further analysis revealed that the correlation between NDVI changes and desertification sensitivity changes remained negative across all years (R = −0.54), indicating a significant inverse relationship between vegetation reduction and increasing sensitivity. However, the driving mechanism behind this relationship is not solely attributed to urban expansion; other factors such as climate change, variations in agricultural activity intensity, and ecological restoration projects may also play a role [33].

In certain areas, vegetation degradation may occur even in the absence of significant urban expansion due to factors such as overgrazing, improper farming practices, or water scarcity. Therefore, future research should further quantify the direct contribution of urban expansion to vegetation degradation to clarify its specific role in the process of sensitivity changes.

The following are targeted recommendations to mitigate desertification sensitivity risks: First, urban expansion must be strategically managed to minimize land sensitivity. In high-sensitivity areas, strict control over construction land expansion is critical to prevent vegetation loss and soil destabilization [34]. Complementary measures such as establishing green infrastructure (e.g., urban forests and wetlands) in urban fringe zones can enhance ecosystem stability by mitigating the edge effects of urbanization. Additionally, adopting low-impact development (LID) practices—including permeable pavements and rainwater harvesting systems—can reduce the hydrological stress on degraded lands while supporting sustainable urban growth.

Second, vegetation protection and restoration should be prioritized to improve land resilience. Areas exhibiting significant NDVI decline require targeted interventions such as enclosure management (restricting human interference for natural regeneration), ecological replanting with native drought-resistant species [35], and bioengineering techniques (e.g., hydroseeding) for re-greening degraded slopes. In regions severely affected by wind and water erosion, integrated soil stabilization measures are essential. For instance, windbreak forests can reduce aeolian sediment transport, while artificial grassland restoration enhances soil organic matter content and stability. These approaches collectively address the dual challenges of anthropogenic pressure and climatic stressors, offering a pathway to balance urban development with ecosystem preservation in semi-arid regions.

There is a need to establish a threshold monitoring system for early warnings and intervention; future ecological governance should focus on multi-dimensional intervention strategies, leveraging remote sensing technology, ecological threshold setting, and real-time monitoring to enhance early warning capabilities in high-sensitivity desertification areas. Desertification is often a cumulative effect, where long-term changes in soil, water, and vegetation conditions may approach or exceed critical thresholds, leading to rapid increases in sensitivity. Therefore, it is essential to use remote sensing imagery and time-series ecological monitoring data to define key ecological thresholds and establish a desertification sensitivity early warning system. Specifically, based on this study’s findings, indicators such as NDVI trend thresholds, where sustained negative trends indicate vegetation degradation; nighttime light intensity increases, which reflect continuous and significant increases in urbanization and human activity; and soil erodibility classes, particularly transitions toward high erodibility (e.g., areas with >60% sand content), should be employed. Establishing precise thresholds for these indicators would facilitate the proactive identification of areas at imminent risk, thus allowing for a timely implementation of targeted ecological interventions and land management practices [36].

This study elucidates the mechanisms by which urban expansion and vegetation changes influence desertification sensitivity and provides a scientific basis for land management. However, several aspects warrant further investigation: First, this study primarily relies on remote sensing data for analysis. Future research could integrate field surveys and long-term meteorological data to further verify the causal relationship between urbanization and desertification sensitivity. Second, this study does not deeply explore the differences in how various urban development models impact desertification. Future studies could incorporate different urban expansion patterns for a more in-depth analysis. Additionally, this study does not explicitly account for the impact of climate change on desertification. Future research could incorporate meteorological data to explore the evolution of desertification sensitivity in the context of global environmental change.

5. Conclusions

This study analyzes the spatiotemporal evolution of desertification sensitivity in an urbanization context from 2018 to 2022 and explores the mechanisms by which urban expansion and vegetation degradation influence sensitivity changes. The results indicate an overall decline in sensitivity, with a significant increase in low-sensitivity areas and a reduction in high-sensitivity areas, suggesting that desertification risks have been alleviated to some extent. However, in the areas experiencing rapid urban expansion, sensitivity continues to rise, implying that urbanization may mediate changes in desertification sensitivity through land-use conversion and ecosystem disturbances.

The spatiotemporal variation analysis reveals that the proportion of low-sensitivity areas (Levels 1 and 2) has continued to increase, while the proportion of high-sensitivity areas (Levels 4 and 5) has decreased, forming a spatial pattern characterized by “overall improvement with localized deterioration”. The areas with increasing sensitivity highly overlap with regions experiencing enhanced nighttime light intensity, indicating that urban expansion may contribute to increasing sensitivity through the expansion of built-up land, vegetation loss, and land resource consumption. Additionally, the areas with declining vegetation cover are more likely to exhibit increased sensitivity, demonstrating that NDVI reduction plays a crucial role in the evolution of desertification sensitivity.

The driving factor analysis reveals a time-lag effect between the roles of urban expansion and vegetation degradation in sensitivity changes. Between 2019 and 2020, the correlation between nighttime light variation and sensitivity change reached its highest value (R = 0.81), suggesting that urban expansion had a strong direct impact during this period. In contrast, from 2021 to 2022, the correlation between NDVI changes and sensitivity was the most significant (R = −0.57), indicating that vegetation degradation played a more dominant role in the later stage. These findings suggest that urban expansion not only directly affects sensitivity through land-use change but may also indirectly exacerbate desertification risks through long-term ecosystem alterations.

Future efforts should focus on optimizing urban expansion patterns, rationally planning construction land, and minimizing damage to natural vegetation, while also strengthening ecological restoration and soil and water conservation measures to mitigate land degradation risks. Additionally, remote sensing technology can be utilized to establish a desertification sensitivity early warning system, using multi-temporal data to define ecological thresholds, enabling dynamic monitoring and targeted governance. Future research should integrate field observations, socioeconomic data, and meteorological factors to further quantify the contributions of different driving forces and explore the complex mechanisms of desertification evolution under urbanization, thereby enhancing the scientific and strategic precision of regional ecological management.