Long-Term Dynamics of Sandy Vegetation and Land in North China

Abstract

Owing to the lack of long-term, continuous, large-scale, and high-resolution monitoring data and methods, we still cannot accurately understand the detailed processes of sand change in northern China. To some extent, this hinders the scientific implementation of sand prevention and control actions. To gain a more accurate and detailed understanding of the process of sandy land change, we conducted an investigation using a reconstructed, long-term, continuous, 250 m-high spatial resolution normalized difference vegetation index (NDVI) and fractional vegetation cover (FVC) data from 1982 to 2018 to examine vegetation changes in sandy land in northern China. This study revealed that vegetation activity (NDVI _slope = 0.011/a, R² = 0.148) and vegetation coverage (FVC _slope = 0.011/a, R² = 0.080) in the northern sandy land (NSL) have slowed the desertification trend. The NSL desertification and reverse areas show decreasing and increasing trends, respectively, indicating an improvement in the degree of desertification from 1982 to 2018. Furthermore, we employed a newly proposed sandy classification method to investigate the area changes in mobile, semi-mobile, semi-fixed, and fixed sandy lands. Over the past 37 years, the total NSL area has shown a significantly weak decreasing trend (slope = −0.0009 million km²/year, r = −0.374, p = 0.023), with relatively small changes in the total area. However, the distribution area of large mobile sandy lands has significantly decreased, whereas the area of fixed sandy lands has significantly increased. Additionally, a survey of changes in the location of sandy lands revealed that 71.86% of the distribution of sandy land remained relatively fixed between 1982 and 2018, with only 28.14% of the distribution remaining in an unstable state. Stable mobile and fixed sandy lands accounted for 85.40% and 82.41% of the total area of mobile and fixed sandy lands, respectively, whereas there were more unstable sandy land distribution areas in the semi-mobile and semi-fixed sandy lands. These results indicate the alleviation of NSL desertification. The new sandy classification and monitoring methods proposed in this study will help improve the remote sensing monitoring of large-scale sand dynamics and offer new ideas for monitoring desertification on a large scale using remote sensing techniques.

1. Introduction

Sandy land is a crucial land type in the arid and semi-arid regions of northern China and serves as a physical foundation for the operation and maintenance of ecosystems in desert areas [1,2]. Changes in sandy ecosystems impact the ecological security of the northern desert region. Moreover, alterations in the physical properties of sandy land surfaces can regulate changes in carbon, water, and heat in desert areas. Dynamic variations in sandy vegetation can control the degradation of sandy land, thereby influencing the ecological quality of the area [1,3]. This, in turn, affects the energy, information, and life characteristics of desert ecosystems. The degradation of sandy land exacerbates desertification in arid and semiarid regions. Desertification is a phenomenon of land degradation caused by climate change and human activity in arid and semi-arid regions that sequentially affects and disrupts the survival and sustainable development of human society [2,4,5,6,7]. Land desertification represents environmental degradation and acts as a positive feedback mechanism for environmental instability. Approximately 6–12 million km² of land is suitable for desertification, and approximately 1–6% of the drylands reside in these areas [8]. Desertification is a complex and multifaceted process influenced by factors such as climate change, land-use practices, and population growth. Although estimating the current global potential for desertification is challenging, studies have suggested that various regions worldwide are vulnerable to this process. Identifying areas sensitive to land desertification and assessing their spatiotemporal patterns and evolutionary trends provide an important basis for planning desertification control and ecological protection measures [7]. Consequently, conducting large-scale dynamic changes in sandy lands is important for guiding ecological protection and construction [4,9,10,11].

When examining existing research reports, it is evident that many studies have primarily focused on dynamic changes in sandy land on a small scale [12,13,14,15]. However, with the advancement of rapid processing technology for remote sensing big data (such as GEE), research on the dynamic monitoring of sandy lands on a larger regional scale is increasing [16,17]. These studies contribute to a deeper understanding of the dynamic changes in the sandy lands in northern China. Yan et al. [10] utilized Landsat MSS and TM/ETM to investigate desertification and dynamic changes in sandy lands in northern China over four periods: 1975, 1990, 2000, and 2010. Zhao et al. [9] further investigated desertification and dynamic changes in sandy land in northern China at three time points, 2010, 2015, and 2020, using FVC data from Landsat TM/OLI. Duan et al. [18] utilized high-resolution temporal continuous MODIS remote sensing data to focus on the dynamic changes in Horqin Sandy Land from 2000 to 2015. Various studies have demonstrated that the remote sensing index serves as an important indicator for monitoring dynamic changes in sandy lands through remote sensing [6,12,13,16,19,20,21,22,23,24,25,26,27], and remote sensing data monitoring methods have become the primary approach for monitoring the characteristics of sandy land changes. However, owing to the influence of NDVI changes on the FVC, which is affected by meteorological conditions, there may be some randomness in the dynamic characteristics of sand changes during isolated years (such as a sudden increase in precipitation in the sand area that year, leading to a contraction in the area of sand monitored by the FVC). This undoubtedly increases the uncertainty of dynamic sand monitoring and does not provide accurate guidance for the ecological protection of sand.

The majority of dynamic sand changes are currently being investigated using time segments, but there is still a lack of detailed high-frequency observational research on long-term continuous time series. The lack of long-term, continuous, and large-scale ground observation data makes it difficult to accurately monitor the evolutionary trends of changes in vegetation and sandy areas in sandy regions. Sandy desertification occurs over a long period of time, and it is challenging to monitor it accurately using remote sensing data over shorter periods. This is because the sparse vegetation distribution in sandy regions makes it difficult for existing limited, discontinuous, and low-spatial-resolution remote sensing monitoring data to accurately depict the dynamic characteristics of sandy vegetation. Moreover, the current methods for monitoring and distinguishing between different types of sand do not effectively capture sand signals in response to global warming. Additionally, there is a lack of objective descriptions for estimating the movement characteristics of sandy land positions over extended periods. With sustained global warming, changes in temperature, precipitation, and wind speed patterns may alter the type of sandy land or exacerbate desertification. In recent decades, as the planet has experienced rapid warming, the distribution of mobile and fixed sand has changed. However, there remains an incomplete understanding of the distribution of stable and easily changing sandy areas.

To address these gaps, this study utilized long-term continuous NDVI and FVC data with a high spatial resolution of 250 m from 1982 to 2018 to investigate the dynamic change characteristics of the sandy lands in northern China. Additionally, this study conducted further classification and investigation of the change characteristics of different types of sandy land to diagnose the degree of desertification and identify the stability of the sandy land distribution positions. These findings provide support for the healthy operation and maintenance of sandy land ecosystems.

2. Materials and Methods

2.1. Study Region

The northern sandy land (NSL) is primarily located in the potentially arid to semi-arid desert areas of northern China, between 35–50° N and 75–125° E. This region experiences a variety of climate types, ranging from extreme drought to semi-humid conditions. It is mainly found in the following eight provinces: Xinjiang, Qinghai, Gansu, Inner Mongolia, Ningxia, Shaanxi, Shanxi, and Hebei. China’s main deserts, including Taklimakan Desert, Gurbantunggut Desert, Badain Jaran Desert, Tengger Desert, and sandy lands, such as Maowusu Sandy Land, Hunshandake Sandy Land, Horqin Sandy Land, and Hulunbeier Sandy Land, are distributed in this area. The annual rainfall in this region sharply decreases from over 400 mm in eastern semi-arid regions to below 50 mm in northwestern arid regions. Additionally, the vegetation transitions from temperate steppes to desert steppes, and eventually to desert. The sandy area covered in this study was 1.223 million km² and was divided into nine sand zones based on China’s natural zoning: Hulunbuir, Nenjiang, Horqin, Hunshandak, Ordos, Alxa and Hexi Corridor, Qinghai, Tarim, and Junggar.

2.2. Data

To accurately describe the dynamic changes in vegetation activity in sandy areas, constructed NDVI annual cumulative sum data (CD NDVI) from 1982 to 2018 with a spatial resolution of 250 m were used. This dataset was obtained from Wang et al. [28] and possesses long-term continuity and a high spatial resolution. It effectively monitors vegetation change characteristics in northern desert areas of China on a large regional scale. Wang et al. [28] demonstrated that NDVI and FVC data can better describe vegetation activities and changes in sandy lands. Furthermore, the high spatial resolution (250 m) and long-term continuous constructed fractional vegetation cover (CD FVC) dataset from Wang et al. [28] covering the period from 1982 to 2018 was used to provide detailed descriptions of the dynamic changes in the northern sandy land. The FVC is an important parameter used to measure surface vegetation cover. The index is considered a suitable criterion for identifying land degradation and specifications in arid and semiarid regions. These methods can also be used to study these processes [14,29]. Additionally, we reconstructed FVC data using VH (blended vegetation health product, hereafter VH NDVI), GIMMS3g, SPOT, and MODIS NDVI annual cumulative sum data to verify the classification results of the sandy areas using CD FVC data recognition. The formula for calculating the FVC is as follows:

$${\mathrm{FVC}}_y=\frac{NDV{I}_{y,i,j}-NDV{I}_{aver,\min }}{NDV{I}_{aver,\max }-NDV{I}_{aver,\min }}$$

(1)

The maximum and minimum NDVI values are represented by multi-year average $NDV{I}_{y,\max }$ and $NDV{I}_{y,\min }$, respectively. To mitigate uncertainty and randomness when determining extreme NDVI values [29], we selected the average NDVI values (NDVI > 0) at the 5% and 95% percentiles for their respective periods, denoted as ${\mathit{NDVI}}_{aver,\min }$ and $NDV{I}_{aver,\max }$. The VH NDVI product (1982–2018) was retrieved from VIRS (available since 2013) and the Advanced Very High-Resolution Radiometer (AVHRR, available from 1982 to 2012) land surface data. This product has a spatial resolution of 4 km and a temporal resolution of 7 days (access from https://www.star.nesdis.noaa.gov/smcd/emb/vci/VH/vh_ftp.php, accessed on 1 October 2023). For the 16 day composite 250 m MODIS NDVI, we used the MOD13Q1 Terra Vegetation Indices available since 2000. The GIMMS3g NDVI dataset is an updated version of the GIMMS NDVI dataset [30] and has a spatial resolution of 8 km and temporal resolution of 16 days. GIMMS3g data were derived from Advanced Very High-Resolution Radiometer (AVHRR) sensors on several NOAA satellites and can be obtained from the NASA Earth Exchange (NEX) at https://daac.ornl.gov/VEGETATION/guides/Global_Veg_Greenness_GIMMS_3G.html, accessed on 1 October 2023. Finally, the SPOT NDVI dataset, derived from SPOT-VGT, had a spatial resolution of 5 km and a temporal resolution of 10 days. Access to this dataset can be obtained from https://docs.terrascope.be/#/DataProducts/SPOT-VGT/Level3/Level3, accessed on 1 October 2023. In summary, the reconstructed FVC data effectively monitored the degradation status of sandy land in the potential desert areas to the north [28].

2.3. Sandy Desertification Monitoring

A sandy desertification index (SDI) was constructed based on FVC data to monitor the dynamics of desertification in sandy land. This index characterizes the desertification dynamics of sandy land and is calculated as follows:

$$\begin{array}{l}SD{I}_{y,i,j}=\frac{FV{C}_{y,i,j}-FV{C}_{aver,i,j}}{FV{C}_{aver,i,j}}\\ FV{C}_{aver,i,j}=\frac{1}{n}{\displaystyle \sum _{y=1}^{n}FV{C}_{y,i,j}}\end{array}$$

(2)

In the above equation, $i$ and $j$ represent the grids in rows $i$ and $j$, respectively. y represents the time period and n represents the total number of years. If $FV{C}_{aver,i,j}<0$, this indicates the intensification of desertification in sandy land. If $FV{C}_{aver,i,j}>0$, this indicates a reversal and slowing of desertification in sandy lands.

2.4. Sandy Classification

Because most current research is primarily based on changes in vegetation coverage, sandy land is classified into four types: mobile (MS), semi-mobile (SMS), semi-fixed (SFS), and fixed sandy (FS). Based on this classification, the spatiotemporal change trend of the sandy land was evaluated. The probability distribution curves of the FVC in the coverage areas of different types of sandy land were depicted using the spatial distribution data of Wang et al.’s [31] four types of sandy land as prior classification information. The 1:100,000 Chinese sandy land classification data [31] were used with a resolution of 250 m. The distribution frequencies of CD FVC corresponding to the four types of sandy land from 1982 to 2018 were calculated to be between 0 and 1.0, and the frequency distribution curves of the four types of sandy land were plotted (Figure 1a). As the four types of sandy land (MS, SMS, SFS, and FS) represent different forms of evolution in northern China, there is an intersection between the types. Therefore, the point at which the two probability distribution curves intersect, that is, the point at which the distribution probability is equal, is used as the boundary point. However, owing to the nonunique and continuous distribution characteristics in the geographical space of the transition, the first intersection point among the continuous occurrence points was used as the boundary point. It is worth noting that the existing intersection points are not unique and fall into two categories: continuous and breakpoint.

2.5. Stability Identification of Sandy Land Distribution Positions

To further investigate the changing characteristics of the location distribution of different types of sandy lands, an index was constructed to describe the stability of the sandy land location. The index calculation formula is as follows:

$$\begin{array}{l}{C}_j=1,j=1,2,3,4\\ Su{m}_j={\displaystyle \sum _{y=1}^{n=37}{C}_{j,y}}\\ S_j=\left\{\begin{array}{l}Su{m}_j\ge 23,Steady\\ 0<Su{m}_j<23,Unsteady\end{array}\right\}\end{array}$$

(3)

In the above equation, $j$ = 1, 2, 3, and 4 refer to MS, SMS, SF, and FS, respectively. $Su{m}_j$ refers to the total number of occurrences of the sandy land type $j$ between 1982 and 2018. $S_j$ indicates whether the location distribution of type $j$ sand was stable. Cluster analysis was used to determine the classification threshold, and sandy land was divided into stable and unstable categories based on the location distribution.

3. Results

3.1. NDVI-Detected Vegetation Activity Dynamics in the NSL over 1982–2018

As an important vegetation index for remote sensing in Earth observations, the NDVI has become a significant indicator for monitoring the dynamics of desertification in potential desert areas. Therefore, long-term continuous NDVI data were used to investigate desertification dynamics in the NSL region. Initially, the spatial distribution of the multi-year average NDVI in the NSL region from 1982 to 2018 was examined (Figure 2a). Figure 2a illustrates the spatial distribution of NDVI in nine sandy areas, displaying an average NDVI value of 1.89 for the entire region. Among these areas, Hulun Bur (S1), Nunkiang (S2), Horqin (S3), and Otindag (S4) had average NDVI values greater than 1.89, whereas regions below 1.89 included Ordos (S5), Ala Shan and Hexi Corridor (S6), Qinghai (S7), Tarim (S8), and Junggar (S9). The Tarim region exhibited the lowest average NDVI value (0.79) among the potential desert areas in northern China, indicating a weaker vegetation activity in the western region than in the eastern region. Compared with 1982–1991 (Figure 2b), the average NDVI in the NSL region increased by 10.8%. The S5–S9 region showed an average increase of 37.6%, whereas the S2 and S3 regions demonstrated an average increase of 3.1%. Only regions S1 and S4 experienced decreases of −10.1% and −4.8%, respectively. The highest percentage increase in NDVI occurred in S8 (56%), whereas the lowest (0.2%) was observed in S2, indicating an overall increasing trend in vegetation activity in the NSL region.

Furthermore, the spatial distribution of the NDVI change rate in the NSL region over the past 37 years was analyzed (Figure 2c). As shown in Figure 2c, approximately 61.7% of the coverage area had an NDVI change rate slope > 0, whereas only 38.1% of the areas exhibited slope values less than zero. The significantly increased and decreased coverage area ratios were 18.4% and 6.2%, respectively, with significant differences at the 90% confidence interval. The average NDVI slope value in the NSL region was 0.006 per year, indicating an increasing trend in NDVI and further suggesting an augmentation in vegetation activity in the region.

Additionally, a detailed investigation and analysis of the annual changes in the NDVI for the nine partitions (S1–S9) were conducted (Figure 3). Figure 3 depicts the different increasing trends in the Nunkiang, Horqin, Ordos, Ala Shan, Hexi Corridor, Qinghai, Tarim, and Junggar regions. Among them, the increasing trend in the Ordos, Tarim, and Junggar regions was significant at the 95% confidence interval, with Ordos exhibiting the highest rate of increase (0.023 per year). Among the nine regions, only Hulun Buir and Otindag displayed insignificant decreasing trends, with the Hulun Buir region demonstrating a significant decrease in NDVI (−0.012 per year). The annual variation in the average NDVI of the NSL region also showed an insignificant increasing trend (slope = 0.006 per year; R² = 0.148).

In summary, the results provided further evidence of an increasing trend in vegetation activity in the NSL region from 1982 to 2018, with the western region contributing significantly to this enhancement, whereas the eastern region exhibited a decreasing trend in vegetation activity. This suggests that strengthened vegetation activity in the western region helps to mitigate the potential desertification process in the NSL region.

3.2. FVC-Dectected Desertification Dynamics in the NSL over 1982–2018

Can the increasing vegetation activity over the past 37 years slow the process of desertification in sandy lands? To investigate this, we analyzed the annual trend of the fraction of FVC changes in the NSL using FVC data from 1982 to 2018 (Figure 4). Figure 4 shows that the average FVC in the overall NSL region exhibited an insignificant increasing trend (slope = 0.011/a; R² = 0.080). However, there were varying increasing trends in the FVC across different regions, such as Nunkiang, Horqin, Ordos, Ala Shan and Hexi Corridor, Qinghai, Tarim, and Junggar, whereas the Hulun Buir and Otingdag regions demonstrated an insignificant decreasing trend. Notably, the increasing trend in the FVC in the Ordos and Junggar regions was significant at the 95% confidence interval, with Ordos experiencing the highest rate of increase (0.054/a). Conversely, the Hulun Buir area exhibited a significant decrease in FVC (−0.043/a).

Furthermore, we assessed the reversal of desertification in the NSL region over different periods using a desertification index (Figure 5 and Figure 6). Figure 5 indicates that, from 1982 to 1989, only the Hulun Buir and Otingdag regions displayed an evident reversal. Between 1990 and 1999, desertification was prevalent in the Nunkiang, Horqin, and Ordos regions, whereas the other regions exhibited reversed phenomena. From 2000 to 2009, the Ordos, Ala Shan, Hexi Corridor, Qinghai, and Junggar regions had a greater distribution of reversal phenomena. Additionally, the distribution area of the reversal phenomenon decreased from 2010 to 2018 compared with that in the previous decade. This suggests that the reversal of desertification in the NSL region varied over time. Moreover, from 1982 to 1989, the reversal phenomenon was mainly observed in the northeast of the NSL and the northern part of the Tarim region, whereas from 1990 to 2018, it predominantly occurred in the northwest of the NSL. This indicates a shift in the desertification reversal phenomenon from the northeast, where vegetation activity is strong, to the northwest, where it is weaker.

In Figure 6, the changes in the coverage area of desertification and reversal phenomena in different periods of sandy land are statistically analyzed. Compared to the period of 1982–1989, the coverage area of desertification reversal increased by 68.07%, 33.31%, and 27.64% during 1990–1999, 2000–2009, and 2010–2018, respectively. Meanwhile, the coverage area affected by sandy desertification decreased by −36.60%, −17.91%, and −14.92% during the respective time periods. Moreover, the proportion of the total NSL area covered by sand desertification reversal during the four periods was 34.96%, 58.77%, 46.61%, and 44.65%, respectively, whereas the proportion of sand desertification coverage area decreased from 65.04% in 1982–1989 to 55.35% in 2010–2018. Figure 6 further indicates that the desertification reversal area in the NSL is expanding, whereas the desertification area is decreasing. These findings suggest a decreasing trend in desertification in the NSL region over the past 37 years.

In summary, the results partially support the increasing trend in vegetation activity within the NSL region from 1982 to 2018. Notably, the western region significantly contributed to this increase, whereas the eastern region exhibited a decline in vegetation activity. This could imply that the augmented vegetation activity in the western region has aided in slowing the potential desertification process in the NSL region.

3.3. Area Changes in Four Sandy Land Types over 1982–2018

To accurately investigate the spatial distribution characteristics of the four types of sandy land in the NSL area, we initially utilized the methods described in Section 2.4 to examine the spatial distribution of MS, SMS, SFS, and FS.

3.3.1. Comparative Verification of the Distribution Area of the Four Types of Sandy Land

Based on the FVC thresholds for sand classification presented in

Table 1. Thresholds for classifying mobile sand (MS), semi-mobile sand (SMS), semi-fixed sand (SFS), and fixed sand (SF) based on FVC values.

Class	Thresholds of FVC
	CD	VH	GIMMS3g	SPOT	MODIS
MS	(0.0, 0.13]	(0.0, 0.1]	(0.0, 0.07]	(0.0, 0.08]	(0.0, 0.04]
SMS	(0.13, 0.2]	(0.1, 0.28]	(0.07, 0.2]	(0.08, 0.21]	(0.04, 0.13]
SFS	(0.2, 0.39]	(0.28, 0.52]	(0.2, 0.37]	(0.21, 0.42]	(0.13, 0.26]
FS	(0.39, 1)	(0.52, 1)	(0.37, 1)	(0.42, 1)	(0.26, 1)

, we investigated the distribution areas of mobile sand, semi-mobile sand, semi-fixed sand, and fixed sand in the potential northern sand areas of China from 2000 to 2013 using different FVC data, as illustrated in

Table 2. Figure 5 shows the mobile sandy (MS), semi-mobile sandy (SMS), semi-fixed sandy (SFS), and fixed sandy (FS) areas indicated by FVC data from 2000 to 2013 (unit: 10,000 km²).

Data	Sandy Land
	MS	SMS	SFS	FS	Total Area
CD	52.90 (43.09%)	15.53 (12.64%)	24.26 (19.76%)	30.09 (24.51%)	122.78
VH	40.15 (32.84%)	31.48 (25.75%)	25.32 (20.71%)	25.29 (20.69%)	122.25
GIMMS3g	39.30 (32.23%)	28.83 (23.65%)	21.57 (17.69%)	32.23 (26.43%)	121.94
SPOT	39.2 1 (32.01%)	29.34 (23.96%)	24.76 (20.22%)	29.17 (23.82%)	122.48
MODIS	41.44 (33.67%)	30.82 (25.04%)	22.06 (17.92%)	28.78 (23.38%)	123.10

Note: The percentages in parentheses represent the percentage of the distribution area of this type of sandy land in the total area of the sandy land.

. The results indicate that the average total area of sand derived from the five sets of FVC data was 1,225,104,500 km². The total area of sand retrieved from the CD FVC was 1,227,800 km², surpassing the average area by 0.2% (2700 km²) and exceeding the total areas monitored by the VH, GIMMS3g, and SPOT FVC by 0.43% (5300 km²), 0.69% (8400 km²), and 0.24% (3000 km²), respectively. However, this total area was less than that monitored by the MODIS FVC by 0.26% (−32,200 km²). These findings indicate that the total area of potential desertification observed with the CD FVC aligns with the results obtained from the other four FVC datasets. Hence, the CD FVC proved to be effective in monitoring dynamic changes in sandy lands within potential desertification areas.

3.3.2. Analysis of Spatiotemporal Variation Characteristics of the Four Types of Sandy Land

Furthermore, we analyzed the spatial distribution of the four types of sandy land from 1982 to 2018 (Figure 7). The figure illustrates that mobile and semi-mobile sandy lands are primarily concentrated in the northwestern region of the potential desert areas, whereas semi-fixed and fixed sandy lands are mainly found in the northeast, with a smaller presence in Qinghai, Xinjiang, and Gansu provinces. Mobile and fixed sandy lands maintained relatively stable distribution positions throughout different time periods, whereas the distribution positions of semi-mobile and semi-fixed sandy lands appeared relatively fixed but exhibited unstable characteristics and easy adaptability. Between 1982 and 2018, sandy land predominantly resided in unstable and changeable regions within Ala Shan, Hexi Corridor, and Ordos. For instance, the Ordos region was predominantly composed of semi-mobile sandy land from 1982 to 1989, which subsequently transitioned to mostly semi-fixed sandy land between 1990 and 1999 and between 2000 and 2009. Furthermore, some semi-fixed sandy lands were transformed back into fixed sandy lands from 2010 to 2018.

Figure 8 provides a further analysis of the annual variation trend in the coverage area of the four types of sandy lands from 1982 to 2018. Figure 8a,b show a decreasing trend in the distribution area of mobile and semi-mobile sandy land during this period, particularly highlighting a significant decrease in the area of mobile sandy land at the 90% confidence level. Conversely, Figure 8c,d indicate an increasing trend in the distribution area of semi-fixed and fixed sandy land, with a significant increase in the area of fixed sandy land at the 95% confidence level. Additionally, the fractional vegetation coverage (FVC) of mobile and semi-fixed sandy land displayed a declining trend, whereas the FVC of semi-mobile and fixed sandy land exhibited an increasing pattern. Figure 8 reveals that the distribution areas of mobile, semi-mobile, semi-fixed, and fixed sandy land from 1982 to 2018 were 56.55 ± 9.53, 14.34 ± 3.10, 22.88 ± 4.55, and 29.00 ± 54,500 km², respectively, accompanied by corresponding FVC values of 0.08 ± 0.03, 0.17 ± 0.02, 0.30 ± 0.05, and 0.62 ± 0.16, respectively. These findings indicate that mobile sandy land, with the largest distribution area, possessed the lowest FVC value, whereas fixed sandy land, with the second-largest distribution, exhibited the highest FVC value. Thus, mobile and semi-fixed sandy lands demonstrated a trend towards lower vegetation coverage, whereas semi-mobile and fixed sandy lands showed a trend towards higher vegetation coverage.

Moreover, the average distribution area of northern sandy land from 1982 to 2018 was 122.36 ± 0.03 million km², and its annual variation displayed a significant linear decreasing trend (linear slope of −0.0009 million km²/year, r = −0.374, p = 0.023). However, the reduction in this area was relatively small. Compared with the total distribution area of sandy land (1.2239 million km²) from 1982 to 1989, the total sandy land area (1.2234 million km²) decreased by approximately 0.03% from 2010 to 2018. Furthermore, when compared to the distribution area of each sandy land type from 1982 to 2018, mobile sandy land decreased by approximately −15.45% from 2010 to 2018, whereas semi-mobile, semi-fixed, and fixed sandy land increased by 4.62%, 16.29%, and 24.66%, respectively. The analysis demonstrated that over the past 37 years, the total distribution area of sandy land in northern China has changed relatively little and has remained relatively constant. However, there were significant changes in the distribution area of the different types of sandy land, indicating the influence of varying distribution locations during different time periods on the distribution area of each sandy land type.

3.3.3. Stability Analysis of Distribution Positions of Different Types of Sandy Land

To investigate the changes in the distribution positions of the different types of sand, we utilized the method for identifying the stability of the sand distribution positions discussed in Section 2.5. Our analysis focused on the four types of sand between 1982 and 2018 (see Figure 9). The results depicted in Figure 9 indicate that stable flowing sandy land is primarily concentrated in the northeast of Junggar, Ala Shan, and the Hexi Corridor and northwest of Qinghai. On the other hand, stable fixed sand distribution was mainly found in the northeast of Hulun Buir, Horqin, and Oringdag. Between 1982 and 2018, approximately 71.86% of the sandy land in the north maintained a relatively fixed distribution, whereas the remaining 28.14% was in an unstable state. Stable flowing and fixed sandy lands accounted for 85.40% and 82.41% of the total area of flowing and fixed sandy lands, respectively. The area of unstable sandy land was estimated to be 877 km² for flowing sandy land and 52,700 km² for fixed sandy land.

However, semi-mobile and semi-fixed sand areas displayed a greater proportion of unstable areas. Specifically, the proportions of unstable semi-mobile and semi-fixed sandy land distribution areas were 91.78% and 51.65%, respectively. These proportionate figures correspond to areas of 841 km² and 119,900 km², respectively. From 1982 to 2018, the stable sandy land area in northern China was 879,600 km², but the proportion of unstable sandy land area was only 28.14%. This unstable area covers 344,400 km², accounting for 57.36% of the total mobile sandy land area, which is 6.36% more than the combined sum of the semi-mobile and semi-fixed sandy land areas.

These findings indicate that, over the past 37 years, the distribution of sandy land in northern China has remained fixed. This indirectly demonstrates that the total area of the NSL distribution experienced relatively minor changes in the short term.

4. Discussion

4.1. Advantages and Limitations

Dynamic changes in sandy land affect the potential desertification process in arid and semi-arid regions of northern China. We investigated the spatiotemporal dynamics of vegetation activity and coverage in the NSL by utilizing high spatial resolution and long-term continuous remote sensing observations of NDVI and FVC data. This allowed for a detailed analysis of the desertification process as well as an in-depth examination of area changes and spatial distribution characteristics of the four types of sandy land. Compared to previous research, our study provided a more comprehensive understanding of vegetation activity in northern sandy land using 250 m high spatial resolution and long-term continuous remote sensing data. In particular, the analysis of long-term vegetation activities, coverage, and changes in the sandy land area offers detailed evidence of the declining trend of desertification in the north. The utilization of this high-resolution data enabled the analysis of annual variation characteristics for each of the four types of sandy land distribution. For instance, the slowdown in desertification in northern China’s sandy land was mainly attributed to the transformation from mobile and semi-mobile sandy land to other types, resulting in an increase in the distribution area of semi-fixed and fixed sandy land and a decrease in the distribution area of mobile sandy land. Additionally, we proposed a novel method for investigating the stability of large-scale sandy land distribution positions, which not only provides a solid foundation for data and methodology but also offers new technical means for precise desertification prevention and control in sandy land.

Furthermore, this study proposes a new method for large-scale investigation of the distribution area of different sandy land types using FVC data based on the concept of sandy land type succession. This method considers the interconnectedness of sand types that undergo a succession process rather than a sudden interruption or transition. This allows for a more objective determination of classification thresholds, particularly for accurately distinguishing highly similar types that coexist in boundary areas. This study contributes to a new method for classifying remote sensing images of desert areas.

The observed increase in vegetation activity and coverage attributed to climate warming and humidification may directly lead to the conversion of mobile and semi-mobile sandy land into semi-fixed and fixed sandy land types. However, based on the succession process of alluvial sand types, the probability of direct transition from mobile or semi-mobile sand types to fixed sand is relatively low. The changes in climatic conditions over a specific period typically follow stable processes.

4.2. Future Applications

Short-term reports indicate a gradual warming and wetting trend in the climate of arid and semi-arid regions in the north, which is closely related to the current slowdown in desertification in the northern sandy land. However, considering the long-term trajectory of climate change, the northern sandy land still faces the risk of drought, which can exacerbate desertification. This study assessed the stability of sandy land distribution positions and analyzed the distribution characteristics of sandy land types throughout historical periods from a probability distribution perspective. This introduces a new technical method for identifying sensitive areas susceptible to future changes in sandy lands and accurately investigating large-scale desertification. Sensitive areas refer to locations where various factors can potentially influence land conditions, such as soil and vegetation characteristics, climate, and human drivers [32,33,34].

5. Conclusions

Owing to the lack of long-term continuous, large-scale, and high-resolution monitoring data and methods for tracking dynamic sand changes in northern China, our understanding of the detailed process of sand changes remains incomplete. This limitation hinders the accurate and in-depth implementation of sand prevention and control measures. To address this issue and gain a more precise understanding of sandy land changes, we used reconstructed long-term continuous normalized difference vegetation index (NDVI) and fractional vegetation cover (FVC) data with a spatial resolution of 250 m from 1982 to 2018. This allowed us to investigate the vegetation activity and vegetation cover changes in sandy land over the past 37 years in northern China. From 1982 to 2018, there was an increase in vegetation activity (NDVI _slope = 0.011/year, R² = 0.148) and an increase in vegetation coverage (FVC _slope = 0.011/year, R² = 0.080) in northern sandy land. These changes have contributed to the decreasing desertification trend in the region to a certain extent. For instance, over the past 37 years, the desertification area in the northern sandy land showed a decreasing trend, whereas the reverse area exhibited an increasing trend. The coverage area ratio of sandy land desertification decreased from 65.04% in 1982 to 55.35% in 2018, indicating an improvement in the degree of desertification in northern sandy land from 1982 to 2018. Furthermore, we employed a newly proposed classification method for sandy land types to examine the characteristics of area changes in mobile, semi-mobile, semi-fixed, and fixed sandy lands. During the past 37 years, the total area of sandy land distribution demonstrated a weak decreasing trend (slope = −0.0009 million km²/year, r = −0.374, p = 0.023), albeit with relatively small changes in the total area. However, the primary distribution area of mobile sandy land has significantly decreased, whereas the fixed sandy land area has significantly increased. Additionally, an analysis of the location changes in different types of sandy lands revealed that 71.86% of the sandy land distribution remained relatively fixed between 1982 and 2018, whereas only 28.14% of the sandy land distribution was in an unstable state. Among the flowing and fixed sandy lands, stable flowing and fixed sandy lands accounted for 85.40% and 82.41% of the total area, respectively, with more unstable sandy land distribution areas observed in the semi-flowing and semi-fixed sandy lands, respectively.

In summary, these findings indicate that desertification in sandy areas of northern China has been mitigated. The sand classification and monitoring methods proposed in this study for tracking sand location changes will contribute to the enhancement of remote sensing monitoring of large-scale sand dynamics and provide novel insights into remote sensing monitoring of desertification on a large scale.