Spatiotemporal Dynamic Monitoring of Desertification in Ordos Section of Yellow River Basin

Abstract

The Ordos section of the Yellow River Basin represents a typical semi-arid zone in northern China. Due to dual pressures from natural drivers and human activities, this region is at the forefront of desertification. Therefore, rapidly and accurately identifying desertification and analyzing its evolutionary trends plays a vital role in desertification control. Using six-phase Landsat imagery (2000–2023) of Ordos City, this study extracted NDVI and Albedo to construct a fitting model, thereby analyzing desertification severity, spatial distribution patterns, and evolutionary dynamics. Through integrated analysis trends in meteorological and anthropogenic data, key driving factors of desertification processes were further investigated. Conclusions: (1) By 2023, the area of extremely severe and severe desertification reduction accounted for 12.67% of the total study area, the proportion of no desertification area increased by 11.27%, and the expansion of desertification was effectively curbed. (2) Desertification intensification cluster near residential zones and grazing lands, while improved areas concentrate in the western and southern of Mu Us Sandy Land vicinity. (3) Spatial autocorrelation analysis revealed statistically significant clustering patterns across the study area, predominantly characterized by distinct low–low and high–high aggregations. (4) Wind speed, temperature, and pastoral activities were major factors contributing to desertification. These research findings provided references for the ecological restoration and sustainable development of semi-arid areas in the Yellow River Basin.

1. Introduction

Desertification refers to the land degradation in arid, semi-arid, and dry sub-humid regions caused by the combined effects of climate change and human activities. This definition was first proposed by the United Nations Convention to Combat Desertification [1]. This concept emphasizes that the formation and development of desertification cannot ignore the impact of human activities. Land degradation caused by natural factors such as climate change cannot be regarded as desertification [2]. At the same time, desertification includes wind erosion, water erosion, freezing and thawing, and salinization in a broad sense [3]. In semi-arid regions, the most prominent manifestation of desertification is wind erosion, attributed to climatic features such as intense wind action, scant precipitation, and high evaporation, as well as anthropogenic factors like excessive reclamation and overgrazing. Research indicates that desertification is intimately linked to one-fifth of the global population, with annual losses of 6 to 7 million hectares of land resulting from land degradation processes, including soil erosion and desertification [4,5]. The semi-arid regions, characterized by scarce and erratic precipitation, possess highly fragile ecosystems and serve as the “frontier zone” for desertification. The hazards associated with this have a chain effect, making it extremely challenging to control [6].

To combat the occurrence and development of desertification in the semi-arid area, it is important to identify the distribution of desertification in the region and carry out driving force analysis according to local conditions quickly and accurately [7]. Traditional desertification monitoring is mainly conducted through field surveys and visual interpretation. However, this method is highly subjective and difficult to form objective evaluation standards. In addition, large-scale field surveys are time-consuming and laborious [8,9]. In 1984, the Food and Agriculture Organization (FAO) and United Nations Environment Programme (UNEP) jointly proposed the “Desertification Assessment and Mapping Methodology”. Since then, remote sensing technology has been widely used in global desertification monitoring research [10,11]. Remote sensing monitoring technology can provide multi-scale and long-term monitoring data [12], and the normalized difference vegetation index (NDVI), net primary productivity (NPP), Albedo, and other monitoring indicators can be extracted by image processing. Since desertification is manifested as the degradation of land productivity, and NDVI is an excellent indicator of the vegetation status and coverage in a region, it is widely used in desertification assessment [13]. However, a single evaluation index such as NDVI is not sufficient to extract desertification information comprehensively and accurately [14]. Many studies have begun to explore the combination of multiple indicators of desertification information extraction. In the process of exploring desertification information extraction based on multi-index fusion, evaluation models such as the Soil-Adjusted Vegetation Index (SAVI)–Fractional Vegetation Cover (FVC)–Enhanced Vegetation Index (EVI), NDVI-NPP, and NDVI–Albedo were developed. SAVI-FVC-EVI is the organic integration of the soil vegetation regulation index and vegetation coverage and vegetation enhancement index [15,16]; this model uses MSAVI to extract severe and extremely severe desertification information in low vegetation coverage and EVI to extract non-desertification and slight desertification information in high vegetation coverage. Therefore, it can make up for the limitation of NDVI in distinguishing high coverage vegetation and low coverage vegetation. The NDVI-NPP model is based on the vegetation index combined with the net primary productivity of vegetation for desertification evaluation [17]. NPP not only reflects the production capacity of vegetation in the natural environment but also is a key parameter of the terrestrial ecosystem and surface carbon cycle. Although representing refinements over NDVI-based approaches, these models exhibit inherent limitations due to exclusive reliance on vegetation-derived metrics for desertification assessment, which fail to capture non-phytogenic manifestations of land degradation, necessitating further advancement.

The Yellow River spans arid, semi-arid, and semi-humid zones. The semi-arid area within the Yellow River Basin encompasses its middle and parts of upstream and downstream in China, specifically the southwestern region of the Inner Mongolia Autonomous Region, Shanxi Province, Shaanxi Province, and the western part of Henan Province [18]. This region experiences scant rainfall and high evaporation rates, predominantly featuring hills and plains. In certain parts, it is interspersed with mountainous areas featuring significant elevation variations. The soil is predominantly of the loess type, with deep layers. However, due to inadequate rainfall, the soil fertility is limited. The semi-arid region within the Yellow River Basin faces unique challenges and presents opportunities in terms of natural resources, economic development, and ecological conservation [19]. Ordos is situated in the upstream section of the Yellow River Basin and nestled within the heart of the “J-shaped” bay of the Yellow River. The climate is classified as a temperate continental type. The vegetation predominantly consists of drought-tolerant and sandy shrubs and grasslands, while the carrying capacity of water and soil resources is relatively low. It stands as a quintessential region within the semi-arid zone of the Yellow River Basin [20]. Water scarcity has emerged as one of the key constraints on regional development, adversely impacting both agricultural production and domestic water supply [21]. While certain-scale pastoralism and abundant mineral resources provide substantial economic development opportunities, this region faces the critical challenge of maintaining ecological equilibrium amidst growing risks of desertification and soil erosion. For the desertification monitoring of the Ordos section in the Yellow River Basin, many scholars used remote sensing to carry out research on various aspects. Since desertification is characterized by the degradation of land productivity, the vegetation index has become the most commonly used indicator for desertification monitoring. Previous studies employed vegetation indices for this region’s desertification assessment [22,23], such as the NDVI, SAVI, Modified Soil-Adjusted Vegetation Index (MSAVI), etc. However, as research advances, vegetation-based metrics alone have become insufficient for comprehensive monitoring. The purpose of this study is to explore the coupling relationship between other indicators and vegetation indicators to monitor desertification.

Albedo is a characteristic parameter reflecting the reflection of surface to solar shortwave radiation, and its change is related to the influence of surface conditions such as vegetation, soil, and snow [24]. Previous studies have found that the vegetation index corresponding to different degrees of land desertification has a significant linear negative correlation with surface albedo [25]. Consequently, NDVI–Albedo can extract desertification information more comprehensively. This study has driven the exploration of multi-parameter integration, where the coupling of NDVI with Albedo establishes a novel paradigm for remote sensing-based desertification surveillance.

In this paper, based on the long-term remote sensing data of Ordos City, a desertification information extraction model by the integration of NDVI and Albedo is constructed, and the spatial–temporal evolution characteristics and spatial cluster characteristics of desertification in 24 years have been analyzed through this model. According to the characteristics of natural factors and human activities in semi-arid areas, the evaluation index system of desertification driving factors was constructed to analyze the driving factors. We hope the conclusion can provide a reference for the prevention and control of desertification in the semi-arid area of the Yellow River Basin. The authors focus on the following issues:

(1) How to construct the NDVI–Albedo evaluation model of the study area to extract the desertification information from 2000 to 2023 quickly and accurately.

(2) What are the spatial–temporal patterns and trends of desertification in Ordos City from 2000 to 2023?

(3) Among the natural factors and human activities in the semi-arid area of the Yellow River Basin, which factors contribute greatly to the occurrence and development of desertification?

2. Materials and Methods

2.1. Study Area

Ordos City is located in the southwest of the Inner Mongolia Autonomous Region, China, which is in the hinterland of the Ordos Plateau (Figure 1). The coordinates of Ordos City are in the range of 106°42′40″~111°27′20″ E, 37°35′24″~40°51′40″ N. It is about 400 km long from east to west, 340 km wide from north to south, with a total area of 86,752 km². It is composed of two municipal districts and seven counties. The terrain of Ordos City is high in the northwest and low in the southeast, and the east, west and north are surrounded by the Yellow River and the south borders on the Loess Plateau. Ordos City is located in a semi-arid area, the climate type is temperate continental climate, the temperature variation between day and night is large, the annual precipitation is about 276.3 mm, and it is concentrated in July–September.

There is the Kubuqi Desert, one of the ten largest deserts in China, and Mu Us, one of the four major Sandy Lands in Ordos City. The wind in sandy areas is strong and frequent, precipitation is scarce and concentrated, and evaporation is very high. Due to historical reasons such as excessive reclamation and grazing, as well as the role of climate factors, the ecosystem in the desert area is fragile, the ecological environment is sensitive, and the environmental carrying capacity is weak. It is particularly difficult to control desertification. Desertification progression inflicts substantial ecological and socioeconomic losses across Ordos. Research confirms its causal linkage to pasture degradation and cropland desertification. Pastoral surveys reveal that in severely affected core grazing zones, the proportion of high-quality forage has diminished, with grazing capacity declining by over 40%. Meanwhile, farmlands along the Mu Us Sandy Land periphery experience 20–30% reductions in maize yields. Additionally, Kangbashi District incurs approximately 15 annual work-stoppage days due to dust storms. Desertification is severely impeding economic development. Under the scientific guidance, relevant departments have worked hard in building a green shield to combat desertification in the northwest, north, and northeast of China. According to relevant reports, at the beginning of the 21st century, the control rate of the Kubuqi Desert was only 4.6% [26], which increased to 32% by 2024, realizing the “green advance and sand retreat” (Figure 2).

2.2. Data and Preprocessing

NDVI and Albedo data are from GEE (https://earthengine.google.com/ (accessed on 19 June 2024)). Six Landsat remote sensing images of Ordos City in 2000, 2005, 2010, 2015, 2020, and 2023 were downloaded and processed, which were extracted with a 5-year research period. The image selection was from June to August, and the data effects met the needs of this study.

Precipitation, wind speed, temperature, and evapotranspiration data are from the annual average data of 11 meteorological stations in Ordos from 2000 to 2023, which are obtained by using the Kriging interpolation method in Arc GIS. The soil wind erosion data were obtained by using the modified soil wind erosion equation (RWEQ) combined with the climatic conditions, vegetation conditions, surface soil roughness, soil erodibility, and soil crusting factors in the study area.

The social and economic data are from the Ordos Statistical Yearbook (http://sj.tjj.ordos.gov.cn/datashow/pubmgr/publishmanage.htm?m=queryPubData&cn=C03 (accessed on 12 August 2024)). For the selection of agricultural indicators, considering that in the statistical yearbook, since 2023, the cultivated land area has been changed by the third national land survey, this data is not comparable with previous years, so the total sown area of crops is selected. In addition, in order to study the driving effect of animal husbandry on desertification development, the statistical data of the livestock head index is the total number of large livestock and sheep (including cattle, horses, donkeys, mules, camels, and sheep), rather than the total number of livestock, excluding pigs (

Table 1. Date source.

Date Type	Spatial Resolution	Format
NDVI data	250 m	Tif
DEM data	30 m	Tif
Albedo data	250 m	Tif
Meteorological data	monitoring stations (11 different)	Txt
Socioeconomic data		Txt

).

2.3. Methods

2.3.1. Desertification Information Extraction

Desertification information includes the DDI (desertification difference index) and the ADI (Aeolian Desertification Index). The DDI is the desertification degree corresponding to a unit pixel in each phase of the image. There is a significant correlation between the vegetation index and surface albedo in the process of desertification. Through the NDVI–Albedo space scatter diagram, it is obvious that the scatter plot presents a typical trapezoidal distribution. Through further research, it is concluded that the NDVI corresponding to different degrees of land desertification has a significant linear negative correlation with the Albedo. These results prove that with more serious desertification, the Albedo gradually increases, while the NDVI gradually decreases [27]. In this paper, according to 2000–2023 in Ordos, six time periods of 2000, 2005, 2010, 2015, 2020, and 2023 are selected, and the maximum point of Albedo corresponding to the NDVI of each time period, namely the dry-edge scatter, was linearly fitted to construct the Albedo–NDVI linear model. Through the linear model, the DDI, the desertification difference index calculation formula is established:

$$Albedo=a\times NDVI+b$$

(1)

$$\mathrm{D}\mathrm{D}\mathrm{I}=-{\displaystyle \frac{1}{a}}\times NDVI-Albedo$$

(2)

where Albedo is the surface albedo of dry-edge scatters, NDVI is the normalized difference vegetation index, a is the coefficient, and b is the parameter.

ADI is a comprehensive quantitative index including desertification degree information and desertification area, namely the regional desertification index [27], which can be used to analyzed the desertification evolution trend by calculating the regional desertification index (ADI) of each period:

$$\mathrm{A}\mathrm{D}\mathrm{I}=(S_l+2S_m{+3S}_s{+4S}_{es})/S_t$$

(3)

where ADI is the regional desertification index, S_l, S_m, Ss, and S_es are the land areas of slight, moderate, severe, and extremely severe desertification, respectively, and S_t is the total land area of the study area. The ADI value range is from 0 to 4; the more serious in regional desertification is, the higher the ADI value is.

2.3.2. Trend and Spatial Cluster Pattern Analysis

Linear regression is a statistical method used to analyze the linear relationship between one variable (independent variable) and another variable (dependent variable) [28]. After the desertification information of Ordos City in 2000, 2005, 2010, 2015, 2020, and 2023 was extracted, the trend of desertification in the study area was analyzed by the linear regression method. The trend of desertification can be expressed as follows:

$$Slope={\displaystyle \frac{n\times {\sum }_{i=1}^{n}i\times {DDI}_i-{\sum }_{i=1}^{n}i{\sum }_{i=1}^n{DDI}_i}{n\times {\sum }_{i=1}^n{i}^2-{\left({\sum }_{i=1}^{n}i\right)}^2}}$$

(4)

where slope is the trend of DDI, i is the annual variable, i = 1, 2, …, n, and n is the year (n = 6); if the slope > 0, it indicates that desertification has a deteriorating trend, otherwise it has an improving trend.

The interdependent relations (including global and local relations) in space are spatial autocorrelation [29]. The method of spatial autocorrelation can be used to analyze whether there is a clustering relationship between variables in the same region. To analyze the global clustering pattern of regional units and adjacent units in a certain region, it can be reflected by calculating the global autocorrelation index. However, the global autocorrelation cannot highlight the unit clustering status of local regions. At this time, the interdependence characteristics of regional units and adjacent units can be analyzed by calculating the local spatial autocorrelation index [30]. In this paper, the spatial autocorrelation index (Moran’s index) of DDI in Ordos City is calculated in pixels. The calculation formula is as follows:

$$\mathrm{I}={\displaystyle \frac{n\sum _{i=1}^n\sum _{j=1}^n{\omega }_{ij}(x_i-\overline{x})(x_j-\overline{x})}{\sum _{i=1}^n\sum _{j=1}^n{\omega }_{ij}\sum _{i=1}^n{(x_i-\overline{x})}^2}}={\displaystyle \frac{n\sum _{i=1}^n\sum _{j\ne 1}^n{\omega }_{ij}(x_i-\overline{x})(x_j-\overline{x})}{S^2\sum _{i=1}^n\sum _{j\ne 1}^n{\omega }_{ij}}}$$

(5)

$${\mathrm{I}}_i={\displaystyle \frac{(x_i-\overline{x})\sum _{j=1}^n{\omega }_{ij}(x_j-\overline{x})}{\frac{1}{n}\sum _{i=1}^n{(x_i-\overline{x})}^2}}={\displaystyle \frac{(x_i-\overline{x})\sum _j{\omega }_{ij}(x_j-\overline{x})}{S^2}}$$

(6)

where I is the global value of spatial autocorrelation, ${\mathrm{I}}_i$ is the local value of spatial autocorrelation, n is the total number of pixels in the study area, $x_i$ and $x_j$ are the observed values of the desertification index (DDI) in the study area, $\overline{x}$ is the mean value of DDI in the study area, and ${\omega }_{ij}$ is the spatial weight matrix.

The global Moran index is used to indicate whether there is an interactive relationship between desertification in the whole study area and the degree of correlation of desertification. The value is [−1, 1], which is divided as follows: [−1, 0), desertification tends to be spatially negatively correlated on the whole; approximately 0, there is no spatial autocorrelation in desertification; [0, 1), which tends to be positively correlated in space as a whole. According to the local Moran index, the spatial clustering patterns of desertification can be divided into five categories: high–high cluster, high–low cluster, low–high cluster, low–low cluster, and not significant, in which high–high clusters and low–low clusters are positively correlated in space, and high–low clusters and low–high clusters are negatively correlated in space.

2.3.3. Grey Relational Analysis

In this paper, the grey correlation model is introduced to explore the main driving factors affecting desertification in the study area. The grey correlation model is simple and easy to calculate and can be used to explore the variation characteristics between data [31], which can be calculated as follows:

$${ϵ}_{ij}={\displaystyle \frac{\underset{i}{\min }\underset{j}{\min }\left|r_{oj}-r_{ij}\right|+\rho \underset{i}{\max }\underset{j}{\max }\left|r_{oj}-r_{ij}\right|}{\left|r_{oj}-r_{ij}\right|+\rho \underset{i}{\max }\underset{j}{\max }\left|r_{oj}-r_{ij}\right|}}$$

(7)

where ${ϵ}_{ij}$ is the correlation coefficient, $r_{oj}$ and $r_{ij}$ are the optimal value and the normalized value of the evaluation index, respectively, $\rho$ is the resolution coefficient, the value is 0.5, $\underset{i}{\max }\underset{j}{\max }\left|r_{oj}-r_{ij}\right|$ is the two extreme maximum differences, and $\underset{i}{\min }\underset{j}{\min }\left|r_{oj}-r_{ij}\right|$ is the two extreme minimum differences.

$${\omega }_i={\displaystyle \frac{{\sum }_{i=1}^m{ϵ}_{ij}}{n}}$$

(8)

where ${\omega }_i$ is the correlation degree, m is the number of evaluations, and n is the number of evaluation years.

3. Results

3.1. NDVI–Albedo Model Construction

In this section, a total of six time periods in 2000, 2005, 2010, 2015, 2020, and 2023 are selected. After processing the remote sensing image, 3000 sample points are randomly selected from the image at each time period. For each sample point, the NDVI value is used as the abscissa, and the Albedo value is used as the ordinate to obtain the maximum and minimum Albedo values at the same point. The scatter plot of the NDVI–Albedo feature space of Ordos City from 2000 to 2023 was constructed. From the scatter plot (Figure 3), the R² of each image is greater than 0.65, and the fitting effect is preferable, indicating that the NDVI obtained from the six images has a strong negative correlation with Albedo.

Through the NDVI–Albedo dry-edge scatter equation, the DDI extraction formula of the six-period image is as follows (

Table 2. DDI extraction formula in Ordos City from 2000 to 2023.

Year	Dry-Edge Scatter Fitting Model	R2	DDI Extraction Formula
2000	A = −0.52361N + 0.7951	0.75033	DDI = 1.9098 × NDVI–Albedo
2005	A = −0.33576N + 0.7941	0.70316	DDI = 2.9783 × NDVI–Albedo
2010	A = −0.53324N + 0.8660	0.76668	DDI = 1.8753 × NDVI–Albedo
2015	A = −0.34918N + 0.7840	0.65527	DDI = 2.8639 × NDVI–Albedo
2020	A = −0.33341N + 0.8206	0.89248	DDI = 2.9993 × NDVI–Albedo
2023	A = −0.29682N + 0.7985	0.73439	DDI = 3.3690 × NDVI–Albedo

):

3.2. Spatial–Temporal Evolution Trend of Desertification

3.2.1. Temporal Evolution Trend of Desertification

Through the natural breakpoint method, this study divides the DDI of Ordos City into five grades: no desertification, slight desertification, moderate desertification, severe desertification, and extremely severe desertification (

Table 3. Desertification area and proportion in Ordos from 2000 to 2023.

	2000		2005		2010
	Area/km2	Percentage/%	Area/km2	Percentage/%	Area/km2	Percentage/%
Extremely severe	12,219.43	14.09%	15,054.34	17.35%	10,877.55	12.55%
Severe	25,948.04	29.91%	28,950.44	33.37%	28,808.80	33.21%
Moderate	20,237.85	23.33%	26,608.24	30.67%	21,564.88	24.86%
Slight	19,809.35	22.83%	13,498.89	15.56%	18,845.33	21.72%
No	8537.32	9.84%	2640.09	3.04%	6645.44	7.66%
	2015		2020		2023
	Area/km2	Percentage/%	Area/km2	Percentage/%	Area/km2	Percentage/%
Extremely severe	10,397.44	11.99%	9425.48	10.86%	9146.38	10.54%
Severe	31,351.40	36.14%	17,758.36	20.47%	18,034.23	20.79%
Moderate	17,716.88	20.42%	28,237.52	32.55%	21,733.04	25.05%
Slight	18,028.63	20.78%	23,471.07	27.06%	19,523.20	22.50%
No	9257.66	10.67%	7859.57	9.06%	18,315.15	21.11%

). It can be seen from Table 2 that about 79% of the land in Ordos has different degrees of desertification. The area of extremely severe and severe desertification has decreased significantly, which has decreased from 38,167.47 km² in 2000 to 27,180.61 km² in 2023, the area of decreased accounting for 12.67% of the city’s area. The ADI calculation results of the study area can be seen in Table 2. The ADI of Ordos City in 2000, 2005, 2010, 2015, 2020, and 2023 is 2.16, 2.46, 2.21, 2.18, 1.97, and 1.77, respectively. In general, the trend of desertification in Ordos has experienced a process of intensification and improvement. The degree of desertification is the lowest in 2023 and the highest in 2005. By superimposing and analyzing the desertification images of two adjacent periods in the study area, a desertification degree transfer matrix can be obtained (Figure 4).

Temporally, the evolution of desertification in Ordos City in the past 24 years can be divided into two stages. The degree of desertification from 2000 to 2005 was generally aggravated, and 2005 to 2023 was the overall improvement stage in the study area. It can be seen from Figure 4 that the area of extremely severe desertification increased first and then decreased over the 24 years, with a decrease of 3073.05 km² compared to 2000, accounting for 3.54% of the total study area. The area of moderate and slight desertification was relatively stable, with a growth of 1.72% and a decrease of −0.33% compared to the initial year. The area of no desertification increased most significantly, with a total increase of 9778.83 km² from 9.84% in 2000 to 21.11%. By 2023, the proportion of severe desertification areas decreased by 9.12%, with a total reduction of 7913.81 km².

From 2000 to 2005, the conversion of no desertification to severe and severe to moderate was the most significant, with conversion areas of 14,863.99 km² and 10,715.31 km² accounting for 17.13% and 12.35% of the total area of the study area, respectively. Due to the rapid increase in the area of conversion to severe and moderate desertification, the desertification trend during this period deteriorated. By 2005, the ADI value was the highest in 24 years.

From 2005 to 2010, the transition from moderate to slight and severe to moderate was the most obvious in the study area, with the transformation areas of 8965.17 km² and 8829.15 km², respectively. During this period, the overall desertification degree was significantly improved.

From 2010 to 2015, the area of moderate desertification transfer was the most significant, among which the largest proportion was moderate to severe and moderate to slight, with an area of 6613.44 km² and 4716.28 km², respectively. Combined with ADI in 2015, it can be seen that compared with ADI in 2010, it only decreased by 0.03, and this period had the weakest desertification improvement in 24 years.

From 2015 to 2020, the sum area of slight and no desertification transferred in this period was second only to 2005–2010, reaching 14,588.84 km². At the end of this period, ADI decreased significantly, and desertification improved obviously.

From 2020 to 2023, the transfer area of low-degree desertification continued to increase on the whole, and the total transfer area of slight and no desertification reached 14,111.42 km². At the end of this period, ADI continued to decline compared with the previous stage.

3.2.2. Spatial Evolution Trend of Desertification

The spatial distribution pattern of desertification in Ordos City is shown in Figure 5. The distribution pattern of desertification in Ordos City from 2000 to 2023 is generally stable. The high degree of desertification areas is mainly distributed in the northwest, central, and southwest of the study area, while the degree in the eastern region is low on the whole. Locally, the distribution pattern of desertification has changed significantly.

From 2000 to 2005, the degree of desertification in Ordos generally intensified. Due to the diffusion of the Kubuqi Desert to the east and the Mu Us Sandy Land to the surrounding areas, the southern and eastern parts of Etorok Banner have extremely severe and severe desertification. Locally, the area of moderate desertification in Zhungeer Banner and Yijinholo Banner increased. The degraded areas were the west and north of Zhungeer Banner and most of Yijinholo Banner.

From 2005 to 2010, the extremely severe and severe desertification areas in the study area significantly shrank. The improved areas were the south of Etorok Banner, the north of Etorokqian Banner, and the central part of Wushen Banner, and this region was the Mu Us Sandy Land. It can be seen that from 2005 to 2010, the control of Mu Us Sandy Land achieved remarkable results, which not only curbed the expansion of Mu Us Sandy Land but also turned a large area of high desertification into a low degree, achieving the “green advance and sand retreat”.

From 2010 to 2015, the distribution of extremely severe desertification in Etorokqian Banner was extended, which occurred sporadically in the west and southwest. However, due to the conversion of slight desertification areas in the central and southern parts of Zhungeer Banner to no desertification, the area of no desertification areas in the study area increased, resulting in an overall improvement in desertification control in 2015 compared to 2010.

From 2015 to 2020, a large area of extremely severe desertification in the northern part of Dalad Banner and the southern part of Etorok Banner was improved to severe and moderate desertification, which has greatly restrained the spread of Mu Us Sandy Land.

From 2020 to 2023, severe and extremely severe desertification spread in the central area of Etorokqian Banner, and the control of Mu Us Sandy Land still had a long way to go. In addition, benefiting from the effective control of the Kubuqi Desert, the area of no desertification and slight desertification in the whole area of Zhungeer Banner and the northern and central parts of Dalat Banner increased. In general, the northwest of Hangjin Banner had a high degree of desertification over the years. This area is the Kubuqi Desert. Combined with field surveys and access to relevant data, the expansion of the Kubuqi Desert was effectively curbed through the hard work of the past generations. However, due to frequent human activities and climatic conditions, it is difficult to control the Mu Us Sandy Land. Although desertification control has achieved great results compared with 2000, this region is still a land desertification-prone area.

3.2.3. Trend Analysis

In order to further explore the evolution trend of desertification in the study area from 2000 to 2023, the trend of desertification in Ordos City can be obtained (Figure 6). It can be seen that the trend of desertification in the west and south of the study area has improved in 24 years, while the northeast has a slight trend of aggravation. Locally, the trend of improvement in the west from north to south is gradually enhancing, which is manifested in the significant improvement of extremely severe and severe desertification in the southwest. The significant improvement areas are mainly Etorokqian Banner, the east of Etorok Banner, and the south of Wushen Banner, while the northwest of Hangjin Banner is slightly improved. Combined with the field investigation, it can be seen that the improvement area is concentrated near the Mu Us Sandy Land and the Kubuqi Desert, reflecting the remarkable achievements of desertification control in Ordos City.

The slight aggravation of the desertification trend is mainly distributed in the northeast of the study area, concentrated in the north of Zhungeer Banner and the north and southwest of Dalat Banner, while Yijinholo Banner, Dongsheng District, and Kangbashi District have sporadic distribution. This area is the leading area of economic development in Ordos City with intensive industries, which is greatly affected by human negative activities on nature.

Overall (

Table 4. Proportion of desertification trend in Ordos City from 2000 to 2023.

Trend	Classification Criteria	Percentage
Significantly Intensified	slope ≤ −0.01	1.62%
Slightly Intensified	−0.01 ≤ slope ≤ −0.005	1.35%
Stable	−0.005 ≤ slope ≤ 0.005	62.21%
Slightly Improved	0.005 ≤ slope ≤ 0.01	29.90%
Significantly Improved	0.01 ≤ Slope	4.92%

), from 2000 to 2023, the area of aggravation trends in desertification was 2573.63 km², accounting for 2.97% of the study area, and the proportion of stable trends was 62.21%. The area of significant improvement and slight improvement of desertification was 30212.76 km², accounting for about 34.82% of the total area of Ordos City.

3.3. Spatial Autocorrelation Characteristics Analysis

Through the global spatial autocorrelation analysis of the desertification index of Ordos City over 24 years, the global spatial autocorrelation parameters of each year were obtained (

Table 5. Global spatial autocorrelation parameters from 2000 to 2023.

Year	Moran’s Index	Z Value	p Value
2000	0.192	138.672	0.000
2005	0.223	133.666	0.000
2010	0.171	184.833	0.000
2015	0.155	154.368	0.000
2020	0.152	147.888	0.000
2023	0.157	151.872	0.000

). It can be seen from Table 5 that the global spatial autocorrelation indexes in 2000, 2005, 2010, 2015, 2020, and 2023 are 0.192, 0.223, 0.171, 0.155, 0.152, and 0.157, respectively. The standardized Z scores are greater than 100, and the p values are less than 0.005. It can be seen that the degree of desertification in the study area shows significant autocorrelation, and its spatial distribution pattern has very strong cluster characteristics; that is, for regions with a high degree of desertification, the degree of desertification around them is generally high, and for regions with a low degree of desertification, the degree of desertification around them is also low. During the 24 years, the global Moran index of Ordos City showed a downward trend. It can be seen that the areas with a similar desertification degree in the study area from 2000 to 2023 showed a decreasing trend of agglomeration in the overall spatial pattern.

In order to further explore the cluster characteristics of the spatial distribution of desertification in Ordos, the local Moran index was calculated and analyzed (Figure 7). It can be seen from Figure 7 that the spatial cluster patterns of DDI in Ordos are mainly low–low clusters and high–high clusters. The proportion of not significant reduced in the early stage but gradually increased in the later stage, and the proportion of low–high clusters and high–low clusters is very small. It can be seen that the spatial pattern of desertification in Ordos has a significant positive correlation.

Temporally, the characteristics of the DDI cluster in Ordos over 24 years gradually changed from insignificant to a low–low cluster and a high–high cluster. The changes over 24 years were mainly in the western part of the study area and from a low–high cluster and a low–low cluster to not significant. The spatial heterogeneity of DDI in the western region showed an upward trend. Spatially, the low–low cluster is mainly distributed in the western part of Ordos, the high–high cluster is mainly distributed in the eastern part of Ordos, and the high–low cluster and low–high cluster are mainly distributed in the transition zone of the two types of regions.

3.4. Driving Factors Analysis

The formation and development of desertification are influenced by different factors. It is very important to analyze the main driving factors of desertification for controlling the development of desertification [32]. The main driving factors of desertification in different regions are different, and the contribution of different driving factors in the same region to the formation and development of desertification is also different [33]. Therefore, in order to diagnose the driving factors of desertification in a certain area, it is necessary to adjust measures to local conditions, fully combine the natural climatic conditions and socioeconomic conditions of the region, and construct an index system with local characteristics for evaluation.

In this study, combined with the climate, social, and economic conditions of Ordos City, 10 indicators including natural factors and human activities were selected. Among them, natural factors include annual average wind speed, annual average precipitation, annual average temperature, soil wind erosion, and evapotranspiration; human activity factors include population density, per capita GDP, total sown area of cropland, number of livestock, and raw coal output. The driving factors of desertification evolution in the study area over 24 years were analyzed by the grey correlation method (

Table 6. The correlation coefficient between each factor and ADI.

Year	Maternal Sequence	Subsequence (Natural Factor)
	ADI	Wind Speed	Precipitation	Temperature	Soil Wind Erosion	Evapotranspiration
		(m/s)	(mm)	(°C)	(t/(km2·a))	(mm)
2000	2.16	1	1	1	1	1
2005	2.46	0.7756	0.7749	0.7712	0.2148	0.7494
2010	2.21	0.7806	0.6544	0.7685	0.5521	0.7360
2015	2.18	0.7280	0.6085	0.6926	0.5315	0.7924
2020	1.97	0.7524	0.6371	0.7568	0.4275	0.6832
2023	1.77	0.7305	0.5522	0.6872	0.5640	0.6975
Year	Maternal Sequence	Subsequence (Human Factor)
	ADI	Population Density	GDP per Capita	Cropland Sown Area	Number of Livestock	Raw Coal Output
		(People/km2)	(Million Yuan)	(Thousand Hectares)	(10,000 pigs)	(10,000 ton)
2000	2.16	1	1	1	1	1
2005	2.46	0.7335	0.6341	0.7098	0.6532	0.5937
2010	2.21	0.7164	0.3427	0.7297	0.7866	0.3532
2015	2.18	0.6874	0.8333	0.6960	0.7434	0.2625
2020	1.97	0.7068	0.4590	0.7427	0.7638	0.4469
2023	1.77	0.6576	0.2586	0.6734	0.7161	0.3158

). In this paper, ADI is selected as the parent sequence, and the other 10 driving factors are used as subsequences. After dimensionless processing of the data, the correlation coefficient between each driving factor and ADI is obtained with 0.5 as the resolution coefficient (Figure 8).

Grey correlation evaluation can objectively reflect the contribution of each index to desertification in Ordos. It can be seen from Figure 8 that the correlation coefficients between the evaluation indexes selected in this paper and desertification are all greater than 0.6, among which the annual average wind speed, annual average temperature, number of livestock, evapotranspiration, total sown area of cropland, population density, and annual average precipitation are closely related to desertification, and the correlation index is higher than 0.8, which is the main influencing factor. In particular, the correlation index of annual average wind speed, annual average temperature, number of livestock, and evapotranspiration is the highest, reaching more than 0.94, and the correlation degree is among the top four.

From the perspective of climate, the study area is located in the western part of Inner Mongolia, which is a typical area of semi-arid climate type. The climate is dry, the spring and autumn seasons are windy, and the precipitation is scarce, concentrated in July–September. Therefore, the natural factors of desertification development in the study area and the correlation index from high to low are annual average wind speed > annual average temperature > evapotranspiration > annual average precipitation > soil wind erosion.

From the analysis of human factors, the study area is economically developed, and animal husbandry and mineral resources are the local pillar industries. Especially in terms of animal husbandry, Ordos City is renowned for cashmere, especially famous for Albas sheep. These are mainly distributed in some areas of Etorok Banner, Hangjin Banner, and Etorokqian Banner in Ordos City, and this region is made up of semi-desert grassland. While the abundant animal husbandry resources bring economic advantages to Ordos City, they also bring the risk of desertification. During disaster years when grassland carrying capacity dramatically decreases, local authorities invest substantial human, material, and financial resources in disaster mitigation and livestock preservation efforts. This practice frequently leads to an imbalance between pasture availability and livestock numbers, exacerbating ecological pressures. Especially, the escalating livestock population has resulted in excessive grazing density across pastures. Intensive hoof compaction disrupts soil crust integrity and creates numerous micro-wind erosion points. Under persistent aeolian forces, these discrete erosion patches coalesce into contiguous sandy expanses—a process of accelerated pasture desertification. This environmental degradation mechanism significantly complicates ecological restoration efforts in the Mu Us Sandy Land, where progressive vegetation degradation and substrate destabilization create self-reinforcing desertification cycles. Therefore, the human factors of desertification development in the study area are as follows: the number of livestock > total sown area of cropland > population density > raw coal production > per capita GDP.

4. Discussion

This study utilized six phases of remote sensing images spanning from 2000 to 2023 in Ordos City to investigate desertification dynamics. Through the extraction of Albedo and NDVI parameters, an Albedo–NDVI-based model for desertification monitoring was developed. Building upon this model framework, a comprehensive analysis was conducted to reveal the spatiotemporal evolution patterns, dynamic trends, and spatial autocorrelation characteristics of desertification in Ordos City over the past 24 years. Through field investigations and systematic analysis of the existing literature, the results of this study are basically consistent with the current situation of desertification distribution in the study area and the conclusions of the existing literature. This section focuses on the discussion of the influence mechanism of natural factors and human activities on the occurrence and development of desertification, aiming to establish a scientific foundation for formulating targeted intervention strategies to combat desertification in the region.

4.1. Natural Drivers for Desertification Evolution

Through the analysis of the driving factors in this paper, it is found that temperature change and wind force are the two most critical natural factors affecting the evolution of desertification. Under the effect of global warming, the warming rate in the semi-arid region is about 20% higher than the global average (for example, the Mongolian Plateau has increased by 2.1 °C in the past 50 years), and the increase in temperature will lead to an increase in evapotranspiration, which will lead to a lack of soil moisture [34]. Studies have shown that for every 1 °C increase in temperature, the potential evapotranspiration increases by 5–10%, which significantly accelerates soil desiccation and leads to land productivity degradation [35]. Concurrently, thermal gradient variations drive accelerated near-surface wind velocities, the spring wind speed in Central Asia has increased by 0.15 m/s in the past 20 years, the wind erosion potential index has increased, the surface roughness has decreased, and the threshold of sand-raising wind speed has decreased, resulting in frequent sandstorms. Furthermore, the dust released by desertification forms a positive feedback mechanism of regional climate deterioration by reducing surface albedo and inhibiting precipitation.

A spatial analysis of desertification severity within the study area reveals distinct geographical gradients: western Ordos exhibits higher degradation intensity than eastern sectors, northern regions surpass southern counterparts in desertification levels, with the northwest being the most serious, and the northeast maintains a stable state of slight desertification. This spatial configuration primarily stems from the influences exerted by the Kubuqi Desert and Mu Us Sandy Land within Ordos’ territory. Kubuqi Desert, one of the eight deserts in China, is located in the northwest of Ordos. This region exhibits distinct climatic zonation, with the eastern periphery classified as a semi-arid zone transitioning to an arid zone in its western reaches. Characterized by xeric conditions and sparse vegetation cover, the area experiences 25–35 annual windy days, resulting in severe aeolian erosion and geomorphic dynamics. In the southwest, one of the four major sandy lands is located, the Mu Us Sandy Land. The annual precipitation in the sandy area is about 300 mm, with large interannual variation. This region is prone to alternating droughts and floods, with frequent torrential rainstorms during the summer seasons that exacerbate severe soil erosion. The fragile sandy ecosystems demonstrate heightened susceptibility to climate change perturbations, thereby compounding the complexity of integrated desertification control strategies within the study region.

4.2. Anthropogenic Activities for Desertification Evolution

Anthropogenic activities are the key driving force for the occurrence and evolution of desertification in semi-arid areas because most of the areas are farming–pastoral ecotones, which are seriously affected by human activities. It is found that human activities can aggravate the desertification process in coordination with natural factors by changing surface cover, interfering with the hydrological cycle, and destroying ecosystem stability [36].

Through the analysis of the driving factors in this paper, it is found that the number of livestock and the sown area of crops are the two key human factors affecting the occurrence and evolution of desertification. Ordos is a typical area of farming–pastoral ecotone in the north. The increase in meat consumption promotes the overload of livestock, and the overloaded grazing leads to a significant decrease in vegetation coverage [37]. The biomass of dominant forages (such as Leymus chinensis and Stipa) decreased by 60–80%, while the soil bareness rate increased to more than 40%. Pasture degradation elevates the aeolian erosion modulus by 3–5 times in semi-arid steppe ecosystems, triggering patchy sandy desertification patterns. In terms of agriculture, cultivated land expansion and tillage methods are the key factors affecting the occurrence and evolution of desertification [38]. The scale of agriculture in semi-arid areas is limited by ecological conditions. For the reason of expanding the scale of production, agricultural activities will expand to ecologically fragile zones (such as sandy grasslands and riparian buffer zones), resulting in the destruction of native vegetation. For tillage methods, traditional tillage will destroy soil aggregates and make the annual loss rate of organic matter reach 1–2% [39]. It is found that the surface soil coarsening rate in the farming–pastoral ecotone of northern China increased by 20–40% per year due to traditional reclamation [40]. In addition, the predatory exploitation of medicinal plants, such as Cistanche deserticola, Cynomorium songaricum, and other psammophytes, leads to the collapse of sand-fixing plant community structure, and the destruction rate of biological crusts is as high as 90%. The impact of agricultural activities on desertification has a significant time lag and spatial diffusion. Agricultural production activities in semi-arid areas need more scientific and systematic management to achieve the coordinated development of food security and ecological security.

4.3. Recommendation for Desertification Control

Desertification control in semi-arid regions necessitates region-specific strategies incorporating local wind dynamics and precipitation patterns. This requires the holistic integration of ecological rehabilitation, land-use optimization, and policy-driven governance frameworks to establish equilibrium between ecosystem restoration and sustainable development objectives. Intense aeolian erosion in semi-arid regions necessitates prioritizing windbreak–sand fixation vegetation systems as a paramount anti-desertification measure. This phytoremediation framework should adhere to native species predominance, utilizing cold-tolerant plants (e.g., Caragana korshinskii and Hippophae rhamnoides) synergistically combined with nitrogen-fixing species like Haloxylon ammodendron, establishing tri-strata vegetation community systems comprising arboreal, shrub, and herbaceous layers. Husbandry management requires a dynamic forage–livestock equilibrium through grazing capacity alert systems and rotational grazing protocols. The precipitation in semi-arid areas is scarce and concentrated in July–September, implementing slope water-harvesting trenches coupled with precision drip irrigation systems, optimizing water utilization efficiency. Agricultural management should prioritize bio-based biodegradable mulching films to mitigate evapotranspiration losses while maintaining soil health. Furthermore, policy is also an important measure to combat desertification [11]. Agricultural governance policy mandates the rigorous implementation of ecological offset mechanisms for grassland and woodland utilization, effectively curbing forest/grassland conversion for cropland. Husbandry governance requires strict adherence to carrying capacity-informed husbandry scaling through determined rational grazing thresholds, achieving the sustainable intensification of livestock production.

The semi-arid area is the key area of desertification control, and its ecological vulnerability requires scientific and systematic control measures [41]. This study reveals a substantial reduction in desertification extent across Ordos’ Kubuqi Desert and Mu Us Sandy Land over a 24-year monitoring period. Statistical records demonstrate that 80% of the Mu Us Sandy Land has been brought under effective control since the 1979 launch of China’s Three-North Shelterbelt Program, achieving a remarkable 94% sand stabilization rate through early 21st-century conservation efforts [42]. Research demonstrates significant progress in desertification control across Ordos since 1979, with overall degradation trends showing marked reversal [43,44]. After 2000, the vegetation coverage of Mu Us Sandy Land increased to 65%, which proved that desertification could be reversed by human intervention, and the Three-North Shelterbelt Project was quite effective.

4.4. Limitations and Prospects

Changes in surface characteristics during desertification are clearly visualized in the Albedo–NDVI feature space. Desertification assessments utilizing the Albedo–NDVI combination demonstrate high accuracy. But this method is limited to the present situation of desertification. However, desertification is not a static process but a dynamic and continuous process. In future research, the desertification extraction model index can be improved and optimized to extract the non-desertification land with a high desertification trend more deeply in order to carry out desertification prevention and control work more effectively.

In the analysis of desertification driving factors, this study only discusses the contribution of natural factors and human activities to desertification and briefly analyzes its influence mechanism. However, the coupling effect of natural factors and human activities has not been analyzed in depth. It is hoped that the coupling driving mechanism of natural factors and human activities can be analyzed in depth in future research.

5. Conclusions

(1) Temporal analysis reveals that desertification evolution in Ordos over the past 24 years comprised two distinct phases. During 2000–2005, the area of extremely severe and severe desertification expanded significantly; from 2005 to 2023, the area of moderate and above desertification progressively decreased, while slight and no desertification areas increased markedly, demonstrating a distinct rehabilitation phase.

(2) Spatial analysis reveals a distinct desertification pattern in the study area: arid west versus vegetated east and barren north contrasting with greener south. The overall assessment indicates that from 2000 to 2023, improved areas exceeded degraded ones. Restoration zones primarily cluster in central and southern sectors; particularly within the southern Mu Us Sandy Land, the degraded area is mainly distributed in the northeast of the study area, which is a densely populated and industrial area.

(3) Spatial clustering analysis demonstrates significant positive autocorrelation across the six-phase imagery in the study area, with global Moran’s I indices consistently exceeding 0, Z scores surpassing 100, and p values remaining below 0.005. Local clustering patterns in Ordos predominantly manifest as low–low (L-L) and high–high (H-H) clusters.

(4) Desertification drivers in the study area primarily involve climatic factors and human activities. Climatic influences are strongly associated with annual average wind speed, mean temperature, and evapotranspiration, while anthropogenic drivers show close links to livestock population and cropland acreage.