Spatio-Temporal Patterns and Driving Forces of Desertification in Otindag Sandy Land, Inner Mongolia, China, in Recent 30 Years

Abstract

Background: Desertification is one of the main obstacles to global sustainable development. Monitoring, evaluating and mastering its driving factors are very important for the prevention and control of desertification. As one of the largest deserts in China, the development of desertification in Otindag Sandy Land (OSL) resulted in the reduction in land productivity and serious ecological/environmental consequences. Although many ecological restoration projects have been carried out, the vegetation restoration of OSL and the impact mechanism of climate and human activities on desertification remain unclear. Methods: Taking OSL as the research area, this paper constructs the desertification index by using the remote sensing images and meteorological and socio-economic data, between 1986 and 2016, and analyzes the spatio-temporal evolution process and driving factors of desertification by using trend analysis and spearman rank correlation. Results: The results showed that: (1) Desertification in the OSL has fluctuated greatly during the past 30 years. Desertification recovered between 1986 and 1990, expanded and increased between 1990 and 2000, reduced between 2000 and 2004, developed rapidly between 2004 and 2007, and recovered again between 2007 and 2016; (2) The desertification of OSL is dominated by a non-significant change trend, accounting for 73.27%. In the significant change trend, the area of desertification rising trend is 20.32%, which is mainly located in the north and east, and the area of declining trend is 6.41%, which is mainly located in the southwest; (3) Desertification is the result of the superposition of climate and human activities. Climate change is the main influencing factor, followed by human activities, and the superposition effects of the two are spatio-temporal differences. Conclusions: These results shed light on the development of desertification in OSL and the relative importance and complex interrelationship between human activities and climate in regulating the process of desertification. Based on this, we suggest continuing to implement the ecological restoration policy and avoid the destruction of vegetation by large-scale animal husbandry in order to improve the situation of desertification.

1. Introduction

Desertification refers to land degradation in arid, semi-arid and dry sub-humid areas caused by various factors, including climate variability and human activities [1,2]. Desertification is a compound disaster under the influence of natural and human factors, which has a wide range and deep degree of harm. As a major global ecological problem, desertification poses a severe challenge to world food security and ecological security [3]. Desertification, known as the “cancer of the earth”, threatens the survival and development of two-thirds of countries and regions, and one-fifth of the population, in the world [4,5].

Around the world, about 33,100 ha of land is degraded into deserts every day, resulting in 1.3 billion USD in economic losses [6]. China is one of the countries most seriously affected by desertification in the world. Since 1995, China has organized the monitoring of desertification and sandification land every five years [7]. According to the fifth monitoring of desertification and sandification in China, the area of desertification land was 26.12 million ha, accounting for 27.20% of the total land area [8]. The development of desertification leads to the deterioration of the ecological environment, the decline of soil quality, sandstorms and other problems, which seriously restricts the sustainable development of the ecosystem [9,10,11].

Since the 1970s, scholars have begun to use satellite images to study desertification, mainly focusing on the causes, spatial pattern, evolution process, prevention and control measures and vulnerability assessment of desertification [12,13,14,15], among which dynamic monitoring of desertification is the basis of other studies. There are three common research methods. One is to study the sandy land changes after 2000 on the basis of continuous interannual middle-low resolution remote sensing data (MODIS data of 250 m, 500 m or 1000 m). For example, using the classification method, linear regression or Mann-Kendall test and other methods, based on the leaf area index (LAI), Net Primary Productivity (NPP) or Normalized Difference Vegetation Index (NDVI) as the basic data, to analyze the change of vegetation characteristics, and then invert the degree of desertification [16,17]. The other is to study the change of sandy land based on the medium resolution remote sensing data (10 m or 30 m) within a period of between five and ten years. For example, the Landsat TM/ETM+/OLI or SPOT data were used for interpretation to distinguish different landscape areas of sandy land and then analyze the dynamic trend of desert [18,19,20]. The last one is to interpret the morphology and area of surface rheumatoid pits by using medium-high resolution remote sensing data, such as Quickbird, WordView and GF series data, to reflect the process and extent of desertification [21]. However, because the development of desertification is a long-term process, it is difficult to fully grasp the change rules of desertification in short and long-time intervals. In addition, due to the large desert area, low resolution image data is difficult to reflect the detailed changes of desertification. Therefore, it is necessary to use medium-high resolution remote sensing data with a more continuous time series to monitor the process of desertification change, in order to comprehensively understand the dynamic spatio-temporal evolution of desertification.

Desertification is the result of the joint action of climate change and human activities. The comprehensive study on the historical climate and stratigraphic profile of Otindag Sandy Land (OSL) showed that climate change was the key factor triggering OSL desertification [22,23]. Vegetation in this region was sensitive to climate change, especially in the early growing season [24]. The frequency of spring drought in the region was up to 60%. Spring drought would affect the grass greening, slow down the growth of pasture and result in reduced pasture production. In addition, livestock trampling on food and spring wind would further accelerate the development of desertification [25]. However, the extent to which climate change and human activities affect desertification in OSL is still worthy of further study and analysis.

This study takes OSL as the research area and aims to: (1) clarify the overall restoration of OSL through the process of ecological protection; (2) analyze the difference and fluctuation of restoration in different desertification areas; (3) discuss the influence of man-made driving factors and climate driving factors on desertification is discussed. This paper has important reference value for clarifying the formation mechanism of sandy desertification and putting forward effective control methods.

2. Methodological Framework and Study Area

2.1. Methodological Framework

The study collected meteorological, topographic, remote sensing, social and economic data. Using the Landsat remote sensing image data, the desertification index was constructed to represent the desertification situation of the sandy land. The spatio-temporal distribution, evolution and trend of the index were analyzed, and the spatio-temporal evolution law of the sandy land in recent 30 years was obtained. The driving force of desertification was analyzed from two aspects: the meteorological factor and the human activity factor (Figure 1).

2.2. Study Area

Otindag Sandy Land (OSL) (111°27’34.2”E-117°10’46.9”E, 41°10’10.5”N-42°58’30.7”N) is located in the Middle East of Inner Mongolia Plateau [26]. It belongs to the sand land of eastern and northern China, with a total area of 49,000 km² [27]. The study area possess a temperate continental monsoon climate, with a mean precipitation of approximately 250 to 400 mm and a mean annual evaporation of approximately 1643 to 2969 mm. The annual precipitation is mainly concentrated between July and September, accounting for approximately 80% of the annual precipitation. The annual mean temperature is approximately 0.5 to 3.5 °C. The mean temperature in January is −18.3 °C. The mean temperature in July is 18.7 °C. The number of strong wind days in the whole year is, approximately, between 50 and 80 days. The wind period is between March and May, and the dominant wind direction is northwest wind (Figure 2).

As one of the top ten largest deserts in China, OSL is located in the south end of Xilingol Grassland in central Inner Mongolia [28]. It is one of the most serious areas of desertification in China and a typical grassland desertification area. The sandy land is located in the arid and semi-arid grassland area. Desertification has increased the sensitivity of the originally fragile ecosystem and poor ecological environment. OSL is a part of the agricultural and pastoral ecotone in northern China [29]. The development of desertification has seriously affected the development of the local social economy. The development of local animal husbandry depends on the good and stable ecological environment; however, it also promotes the process of desertification. In addition, OSL is 180 km away from Beijing, the capital of China. It is an important part of the northern ecological barrier of the “Beijing-Tianjin-Hebei Region” [30]. Grassland desertification destroys its important function as the northern ecological barrier, making it one of the dust source areas of dust storms in Beijing and Tianjin. The vegetation of OSL is dominated by Stipa grandis, Stipa hirylovi, Leymus chinensis, Agropyron cristatum zonal vegetation, elmus pumila, Caragana sinica, Tamarix chinensis, Salix gordejevi, Artemisia halodendron and other sandy vegetation.

3. Data and Methods

3.1. Data Preprocessing and Collection

The Landsat-5 Thematic Mapper images, Landsat-7 Enhanced Thematic Mapper Plus and Landsat-8 Operational Land images were acquired on the few clouds in summer, between 1986 and 2016, for OSL (path 30–31 and row 123–127), with a spatial resolution of 30 m (http://earthexplorer.usgs.gov/, accessed on 7 March 2022). Data preprocessing includes radiometric correction, geometric correction, mosaic, clipping and spatial projection. The meteorological data included the daily data (temperature and precipitation) of 43 meteorological stations in and around OSL (http://cdc.cma.gov.cn, accessed on 7 March 2022). The socio-economic data comes from the Inner Mongolia statistical yearbook between 1985 and 2017 (IMARBS, 1985–2015.), including the year-end population data, gross national product, cropland area and livestock number (livestock quantity, large livestock and sheep).

3.2. Research Methods

3.2.1. Desertification Index

In the grassland desertification area, the distribution of vegetation and desertification are oppositional, which can reflect the desertification indirectly by reflecting the vegetation status. The normalized difference vegetation index (NDVI) was used to calculate the vegetation coverage [31,32], and the desertification index ($D_{des}$) was obtained to characterize the degree of desertification in the study area. The formula is as follows:

$$NDVI=\left(R_2-R_1\right)/\left(R_2+R_1\right)$$

(1)

$$V_{veg}=\left(N-N_{soil}\right)/\left(N_{veg}+N_{soil}\right)$$

(2)

$$D_{des}=1-V_{veg}$$

(3)

In Equations (1)–(3), $R_2$ is the reflectance in the near infrared band; $R_1$ is the reflectivity of the red band; $N_{soil}$ is the NDVI value of bare soil or non-vegetation covered pixels; $N_{veg}$ represents the NDVI value of pixels completely covered by vegetation. $V_{veg}$ is vegetation coverage; $D_{des}$ is the desertification index. $D_{des}$ is divided into five grades: I (<0.15), II (0.15–0.45), III (0.45–0.65), III (0.65–0.85) and IV (>0.85), representing non-desertification (ND), mild desertification (MD), moderate desertification (MOD), severe desertification (SD) and extremely severe desertification (ESD), respectively.

3.2.2. Theil-Sen Median Trend Analysis

Theil-sen Median trend analysis is a method to fit lines to sampling points in the plane, steadily, by selecting the Median slope of all of the lines of paired points. Its advantage is that it is not interfered by outliers and does not require samples to follow a certain distribution, so it can better avoid errors and outlier data [33,34]. The formula is as follows:

$$T_{Sen}=Median\left(\frac{D_{desj}-D_{desi}}{j-i}\right), 1985\le i<j\le 2016$$

(4)

In Equation (4), $T_{Sen}$ is the trend in desertification change; $D_{desj}$ is the $D_{des}$ of the jth year; and $D_{desi}$ is the $D_{des}$ of the ith year. If $T_{Sen}$ > 0, $D_{des}$ is increasing, indicating desertification improvement or recovery. If $T_{Sen}$ < 0, the $D_{des}$ is decreasing, indicating desertification degradation.

3.2.3. Mann–Kendall Test

The Mann–Kendall test is a non-parametric test method, also known as the non-distribution test, which is used to judge the significance of trends. It has been widely used to analyze the trend changes of hydrological and meteorological time series and to study the trend changes of vegetation annual series. Its samples have no specific distribution requirements, and its results free from the interference of outliers. [35,36]. The formula is as follows:

Set $D_{desj}$, where j = 1986, 2000, …, 2016. The V statistic can be defined as:

$$V=\left\{\begin{array}{c}\frac{S-1}{\sqrt{s\left(S\right)}}, S>0\\ 0, S=0\\ \frac{S+1}{\sqrt{s\left(S\right)}}, S<0\end{array}\right.$$

(5)

$$S={\displaystyle \sum }_{j=1}^{n-1}{\displaystyle \sum }_{i=j+1}^{n}sgn\left(D_{desj}-D_{desi}\right)$$

(6)

$$s\left(S\right)=\frac{n\left(n-1\right)\left(2n+5\right)}{18}$$

(7)

$$sgn\left(D_{desj}-D_{desi}\right)=\left\{\begin{array}{c}1, D_{desj}-D_{desi}>0\\ 0, D_{desj}-D_{desi}=0\\ -1, D_{desj}-D_{desi}<0\end{array}\right.$$

(8)

In Equations (5)–(8), $D_{desi}$ and $D_{desj}$ represent the $D_{des}$ of pixel i and j years, respectively; n is the length of the time series. The sgn is a sign function. The value of the V statistic is in the range of (−∞, +∞). When |V| is greater than V_1−α/2 at a given significance level α, there is a significant change in the time series at the ⍺ level. In this study, α = 0.05 was used to assess the significance of the pattern in the regional NDVI change between 1986 and 2016 when |V| > 1.96 at the confidence level of 0.05.

3.2.4. Meteorological Data Interpolation

The required meteorological data were spatially interpolated using the multiple linear regression Kriging (MLRK) method. By establishing the multiple linear regression relationship between the precipitation and mean temperature and interpolation auxiliary factors, the residuals of the real value and predicted value were obtained. The kriging spatial interpolation method was used to carry out the spatial interpolation of the residual values, and the residual interpolation results of the regions to be predicted were added to the predicted values of the linear regression to obtain the final interpolation results [37]. The formula is as follows:

$$\widehat{Z}\left(s_0\right)={\displaystyle \sum }_{k=0}^{\rho }{\widehat{\beta }}_k{q}_k\left(s_0\right)+{\displaystyle \sum }_{i=1}^n{\lambda }_{i}e\left(s_i\right)$$

(9)

where $\widehat{Z}\left(s_0\right)$ represents the interpolation results of the predicted position points; ${{\displaystyle \sum }}_{k=0}^{\rho }{\widehat{\beta }}_k{q}_k\left(s_0\right)$ is the deterministic part of the regression fitting; ${{\displaystyle \sum }}_{i=1}^n{\lambda }_{i}e\left(s_i\right)$ is the interpolation result of the regression residual by ordinary Kriging; K represents the position serial number during regression fitting; β represents the total number of spatial positions; ${\widehat{\beta }}_k$ is the coefficient of the regression model, when $k=0$, ${\widehat{\beta }}_0$ is the intercept; i represents the position serial number of the regression residual interpolation, n represents the total number of spatial positions, $q_k\left(s_0\right)$ is the value of the auxiliary variable of the position point prediction, and ${\lambda }_i$ is the weight of ordinary Kriging interpolation, determined by the spatial correlation structure of the regression residual; $e\left(s_i\right)$ is the residual at position $s_i$. In this study, the accuracy discriminant coefficient R² of the meteorological data is over 90% (Figure A1 and Figure A2).

3.2.5. Correlation Analysis

The Spearman rank correlation was used to reveal the relationship between meteorological factors and human activities and desertification [38,39]. The method is a common non-parametric statistical analysis method used to analyze the correlation between two groups of variables. This method does not require the data required for analysis to conform to the assumption of normality and can avoid the interference of outliers on the analysis results. The formula is as follows:

$$r=1-\frac{6{\displaystyle \sum _{i=1}^n{\left(R_i-Q_i\right)}^2}}{n\left(n^2-1\right)}$$

(10)

where r is the correlation coefficient, and its value is between −1 and 1; r < 0 means relevant, r > 0 means positive correlation, r = 0 means no correlation; n is the data length; i is the ordinal time; R_i and Q_i are the numerical order of the two groups of comparison data.

4. Results

4.1. Temporal Variation, Spatial Distribution and Evolution Trend of Desertification Degree

4.1.1. Spatial and Temporal Distribution Pattern of Desertification Degree

In terms of time, the $D_{des}$ increased from 0.58, in 1986, to 0.64 in 2016, indicating that the $D_{des}$ showed an increasing trend during the study period (Figure 3 and

Table 1. Desertification index and mean value of OSL.

	1986	1990	1995	2000	2004	2007	2011	2016	Mean
$D_{des}$	0.58	0.56	0.63	0.64	0.60	0.75	0.61	0.64	0.63

). The $D_{des}$ in 1990 and 2007 were 0.56 and 0.75, respectively, which were the lowest and most serious period. In 2000 and 2007, the desertification degree was relatively high, and the spatial distribution was relatively uniform, and the regional $D_{des}$ of each county was more than 0.5, indicating that the desertification degree difference between the eastern and western OSL was small, and the area of desertification expansion was large. In terms of spatial distribution, the study area spans ten equal longitudes (Figure 4), and the spatial distribution of desertification in each period has an obvious decreasing trend from west to east. The results indicated that the desertification degree is high in the western part and low in the eastern part of the OSL in recent 30 years (Figure 3).

4.1.2. Evolutionary Characteristics of Desertification Degree

From 1986 to 1990, the $D_{des}$ decreased from 0.58 to 0.56, which was lower than the mean value from 1986 to 2016 of 0.63. SD and ESD decreased by 964.02 km² and 1540.32 km², respectively. The area of ND, MD and MOD increased by 427.5 km², 1285.91 km² and 790.92 km², respectively. From 1990 to 2000, the $D_{des}$ increased from 0.56 to 0.64, and the area of ND and MD decreased by 1687.74 km² and 6605.34 km², respectively. MD, SD and ESD increased by 230.87 km², 5016.33 km² and 3045.88 km² respectively (Figure 5 and

Table 2. Area of different levels of desertification in OSL (km²).

Year	ND	MD	MOD	SD	SED
1986	4261.79	12137	11,063.94	12,225.89	10,155.69
1990	4689.29	13,422.91	11,854.86	11,261.87	8615.37
1995	3431.33	8004.65	12,600.81	15,981.46	9826.06
2000	3001.56	6817.57	12,085.73	16,278.2	11,661.25
2004	4225.37	12,058.69	9288.56	12,085.4	12,048.86
2007	2385.3	2940.62	5212.97	20,390.64	18,914.78
2011	4183.16	10,465.8	10,643.24	10,686.85	13,865.26
2016	3466.83	7214.08	11,714.96	16,632.61	10,815.82

Note: ND, MD, MOD, SD, ESD represent non-desertification, mild desertification, moderate desertification, severe desertification and extremely severe desertification, respectively.

).

From 2000 to 2004, the $D_{des}$ decreased to 0.60. The area of ND, MD and ESD increased by 1223.82 km², 5241.12 km² and 387.61 km², respectively. MD and SD decreased by 2797.16 km² and 4192.79 km², respectively, indicating that the desertification degree was decreasing as a whole, but the area of ESD increased slightly, indicating that the desertification degree was still aggravating in some regions. From 2007 to 2016, the $D_{des}$ decreased to 0.64, and the area of ND, MD and MOD increased by 1081.54 km², 4273.46 km² and 6501.99 km², respectively. SD and ESD decreased by 3758.03 km² and 8098.96 km², respectively (Figure 5 and Table 2).

4.1.3. Trend Characteristics of Desertification Degree

The increasing and decreasing trend of desertification in OSL accounted for 70% and 30%, respectively. The areas of non-significant increase, significant increase and extremely significant increase were 23,236.52 km², 5629.07 km² and 3851.14 km², respectively. The areas of non-significant decline, significant decline and extremely significant decline were 10,952.14 km², 1724.09 km² and 1267.26 km², respectively. The types of down trend were dispersed in space, and the only concentrated distribution area was located in the western part of the study area. The distribution range of the rising trend types is wide, and the spatial distribution is relatively uniform, which indicates that the desertification of the sandy land has risen in the last 30 years and the desertification degree is increasing. The desertification of OSL was dominated by a nonsignificant trend (73.27%). The areas with a non-significant decline and rising trend were 10,952.14 km² (23.47%) and 23,236.52 km² (49.80%), respectively, indicating that the ecological environment of OSL has been unstable in the last 30 years (Figure 6 and

Table 3. Area and proportion of trend significant types of desertification index.

Types	Area (km²)	Percentage (%)
Extremely significant decrease	1267.26	2.72
Significant decrease	1724.09	3.69
Nonsignificant decrease	10,952.14	23.47
Nonsignificant increase	23,236.52	49.80
Significant increase	5629.07	12.06
Extremely significant increase	3851.14	8.25

Note: The test results were classified as extremely significant changes (p < 0.01, Z > 2.32), significant changes (p < 0.05, Z > 1.64) and nonsignificant change.

).

4.2. Meteorological Driving Factors of Desertification

4.2.1. Spatial and Temporal Distribution of Meteorological Factors

In the last 30 years, the change rate of precipitation in OSL was −1.78 mm/a, which did not pass the significance test. The mean temperature change rate was 0.068 °C/a (p < 0.05). The correlation analysis showed that the correlation coefficient between precipitation and desertification is −0.652, and that between the mean temperature and desertification was 0.51, both of which fail to pass the significance test. The fluctuation of precipitation and mean temperature was stronger than that of desertification, and the influence of hydrological and thermal conditions on desertification were different. From 1986 to 1990, the overall extent of desertification decreased, and precipitation and mean temperature increased by 14.14 mm and 1.57 °C. From 1990 to 2000, precipitation decreased by 62.23 mm, and mean temperature did not change much. From 2000 to 2004, precipitation and temperature increased by 48.99 mm and 1.20 °C. From 2004 to 2007, desertification developed rapidly, precipitation decreased by 53.07 mm, and mean temperature increased slightly. From 2007 to 2016, the precipitation increased by 100.49 mm, and the mean temperature decreased by 0.38 °C. The effects of meteorological changes on desertification in OSL in recent 30 years were not consistent in different periods (Figure 7 and Figure A3).

4.2.2. Analysis of Spatio-Temporal Law of Meteorological Factors Driving Action

The effects of the mean temperature and annual precipitation on desertification were different in different regions. The area with a positive correlation between mean temperature and desertification was 40,370.13 km², accounting for 87.19% of the study area. The area of positive correlation through the correlation test was 4406.76 km² (p < 0.05), accounting for 9.52% of the study area. The area of negative correlation through the correlation test was 6939.99 km² (p < 0.05), accounting for 14.98%. The area with a positive correlation between annual precipitation and desertification was 12,907.08 km², accounting for 27.79%, and the area with negative correlation was 33,545.7 km², accounting for 72.22%. The area of positive correlation and negative correlation were 1006.56 km² (2.17%) and 5163.57 km² (11.12%), respectively (Figure A4 and

Table 4. Area and proportion of types associated with climate and desertification.

	Types	Area (km²)	Percentage (%)
Annual precipitation	Significant negative correlation	5163.57	11.12
	Non-significant negative correlation	28,382.13	61.10
	Non-significant positive correlation	11,900.52	25.62
	Significant negative correlation	1006.56	2.17
Annual mean temperature	Non-significant negative correlation	5933.43	12.81
	Non-significant positive correlation	35,963.37	77.67
	Significant positive correlation	4406.76	9.52

).

4.3. Human Activity Driving Factors of Desertification

4.3.1. Analysis on Driving Force of Population and Economic Development to Desertification

In the past 30 years, the Gross Domestic Product (GDP) of the study area has grown rapidly, from 876 million yuan in 1988 to 76.956 billion yuan in 2015, nearly 88 times, with rapid economic development. After 2000, the economic development speed greatly exceeded that of before 2000. From 1986 to 2000, the GDP increased by five times, from 876 million yuan to 5.489 billion yuan in 2000. From 2000 to 2015, the GDP increased by 13 times from 5.489 billion yuan to 76.956 billion yuan. In terms of population, the population of the OSL increased from 723,500 in 1986 to 876,800 in 2015, an increase of 153,300 (an increase of 21.1%). The correlation coefficient between GDP and desertification was 0.198, and that between population and desertification was 0.442 (Figure A5).

4.3.2. Analysis on the Driving Force of Agriculture and Animal Husbandry to Desertification

In the past 30 years, the cropland area in OSL increased from 162,300 ha in 1988 to 183,800 ha in 2011, an increase of 13.24%. The change in the cropland area experienced three stages; from 1988 to 1997, the cropland increased by 76,200 ha. From 1997 to 2003, it decreased by 86,300 ha. From 2003 to 2011, it increased again by 31,600 ha. The number of livestock also experienced three stages. From 1986 to 1999, the livestock increased from 90.68 × 10⁴ to 12.21 × 10⁴. From 1999 to 2006, the livestock decreased to 72.27 × 10⁴, and the livestock breeding reached the lowest value in the study period. Between 2006 and 2016, livestock started to recover, increasing by 1.288 million. The correlation coefficients between the area of cropland and the number of livestock and desertification were 0.177 and −0.420, respectively, which did not pass the significant test (Figure A6).

5. Discussion

5.1. Effects of Climate Change on Desertification

Three-quarters of the global expansion of arid and semi-arid areas will take place in developing countries, which will be at risk of further land degradation and increased poverty in these areas [40]. Studies have shown that desertification in the east and west of arid Asia [41,42], El-Dakhla oasis in Egypt [43], Horqin Sandy Land and Hulun Buir Sandy Land in China were mainly affected by climate change [44]. Desertification was mainly affected by human factors in the semi-arid highland of central Mexico [45], Heihe River Basin of China and Gangcha County of Qinghai Province [46]. Desertification in HognoKhaan of Mongolia, Dharmapuri of India and Mu Us of China were affected by climate factors and human activities [47,48,49]. Wang et al. found that the desertification process in the east and west of the arid region of Asia is sensitive to climate change [50]. In our study, although, overall, there was no significant correlation between the climate and human activity indicators and desertification between 1986 and 2016, the desertification process in arid and semi-arid areas of China was affected by natural and human factors. The area with a positive correlation between the average temperature and desertification and the area with a negative correlation between precipitation and desertification account for approximately three-quarters of the total area of sandy land. Natural factors were the main driving force of desertification, which was consistent with the relevant research [51]. Desertification in arid and semi-arid areas of China was sensitive to climate change, which was similar to Wang et al. [50]. Among the climate factors, precipitation had the greatest influence, which was consistent with the relevant studies [52,53]. Our results also showed that the regions with a significant correlation between the average temperature and desertification, precipitation and desertification in the OSL were in the north central part of the study area. In these areas, temperature and precipitation were significantly related to desertification, accounting for about one-tenth of the total area, and the impact of precipitation was slightly greater. This might be due to the high volatility and spatial heterogeneity of desertification in the short term.

5.2. Effects of Human Activities on the Process of Desertification

In order to curb the development of desertification, four major afforestation projects have been carried over the world: The Great Stalin Plan for the Transformation of Nature [54], the American Roosevelt Shelterbelt Project [55], the Green Dam Project of the Five Countries in Northern Africa and Three Norths Shelter Forest Program in China [56]. These four major afforestation projects have had a significant impact on the international community and have triggered enthusiasm for repairing the ecological environment all over the world [57]. In addition, OSL has also carried out the Beijing Tianjin Sand Storm Source Control Project, the Combating of Destruction Program, the Natural Forest Protection Project, and the Grain for Green Project, etc. Studies have shown that in the past 30 years, Chinese ecological restoration projects have played an important and promoting role in the process of greening the world [58].

In addition to the ecological protection projects guided by policies, the rapid economic development of the study area and the environmental development, such as industrial development, urban construction and agricultural water use, have also led to the gradual shrinkage of high-quality grassland around the river, thus increasing the risk of desertification [59]. The rapid but irregular development of tourism leads to tourists trampling on and destroying grassland, and the trend of desertification expands outwards, with roads as the center of tourism routes [60]. This excessive exploitation and utilization of natural resources has led to the rapid development of desertification in OSL. During the development of OSL’s animal husbandry in the past 30 years, the change trend of sheep and big livestock is oppositional, which reflects the adjustment of the animal husbandry industry structure in OSL [61]. According to the relevant industrial development data of the OSL region, in recent years, the research area has carried out the strategy of “reducing sheep and increasing cattle” to promote economic development and protect the ecological environment by reducing the breeding number of sheep and increasing the breeding amount of large cattle [62]. From the perspective of ecological protection, the grazing mode of sheep is “gnawing feeding”, while the grazing mode of cattle is “rolling feeding” [63]. At the initial stage of pasture growth, the growth rate of pasture in cattle grazing areas was higher than that in sheep grazing areas. Therefore, the number of livestock was still an important reason for the reduction of desertification between 2007 and 2016 due to the adjustment of the breeding structure.

5.3. Measures and Suggestions to Prevent or Slow down the Desertification Process

The zonal vegetation of OSL is typical grassland, and animal husbandry is the pillar industry. According to the statistical data, the primary industry accounted for 80% in 1990, but by 2015, the proportion had dropped to 10%, which was still higher than the average level of Inner Mongolia (9.1%) and the country (8.4%) [64]. To maintain the sustainable development of OSL, it is still necessary to explore the reasonable prevention and control of land desertification and the sustainable use mode of land resources. First, natural recovery plays a huge role in the restoration of regional degraded ecosystems. We should follow the path of the intensive development of agriculture and animal husbandry and engage in production in a small area and ecology in a large area [65]. For example, grazing in the enclosure is forbidden for natural recovery. Research shows that after four years of enclosure, the grass layer height inside the enclosure is 410% higher than that outside the enclosure, the coverage is 397% higher, and the aboveground biomass is 126% higher [66,67]. In addition, we can also consider adjusting the industrial structure of OSL. OSL has the largest lignite field in China and the largest alkali mine in Asia. These rich resources need to be developed and utilized [68]. Moreover, a sound policy guarantee system is also the key to desertification control, and a guaranteed system for the paid development, utilization and protection of regional resources should be established [69]. Finally, we should strengthen the dynamic monitoring and early warning of grassland in this area and grasp the dynamic status of grassland resources in a timely manner in order to scientifically verify the livestock carrying capacity of OSL, regulate the relationship between grass and livestock, and promote regional sustainable development.

5.4. Uncertainties in the Process of Desertification and the Limitations of This Paper

Policy and investment in science and technology can also influence desertification. Due to the limitations of the data and regional scale, this study did not quantitatively analyze the effects of these factors on the desertification process [70]. In addition, the response of desertification to driving factors has a spatial difference and lag, which should be strengthened in the follow-up research [71]. Our research cycle is only 30 years, so in order to pursue the spatial resolution accuracy of the remote sensing data, we chose the Landsat series data at the cost of temporal resolution. In fact, if MODIS data are used, it is possible to calculate data with a wider time range and shorter intervals [72,73] Future studies that comprehensively consider temporal resolution, spatial resolution and spectral resolution will reveal the driving mechanism of desertification in the study area more systematically.

6. Conclusions

In the past 30 years, the evolution process of desertification in OSL has fluctuated greatly and has experienced five periods. The degree of desertification in OSL gradually decreased from west to east. The area with a significant desertification trend accounted for one-fifth of the whole study area, the area with a significant improvement of desertification was less than one-tenth, and the non-significant area accounted for approximately three-quarters of the whole area. During the evolution of desertification in OSL, there was a strong correlation between the meteorological changes and desertification evolution, which was the main reason to promote desertification, and the driving effect of precipitation on desertification was stronger than that of temperature. Human activities were the secondary caused of desert evolution, including economic development, population increase and the development of animal husbandry. In order to prevent the further development of desertification and to restore the regional ecological environment, global climate change should be taken seriously. Policies to increase carbon sinks and reduce carbon emissions should be developed. In addition, scientific and reasonable industrial development policies should be formulated to regulate tourism development and the production activities of farmers and herdsmen, supplemented by scientific and effective means of desertification control.