Spatiotemporal dynamic evolution and driving factors of desertification in the Mu Us Sandy Land in 30 years

Desertification, especially wind erosion desertification, is a major environmental and socio-economic issue affecting arid to semi-humid regions globally and is severe in northern China. In China, wind erosion desertification affects about 1.8 million km², notably across the northwest, north, and northeast. The Mu Us Sandy Land, situated in a sensitive agro-pastoral transition zone, is a critical ecological barrier whose desertification trajectory directly impacts regional economy, ecology, and the prevention of dust storms and further spread of desert conditions. Traditional monitoring is labor-intensive and limited, whereas remote sensing offers efficient, objective, and large-scale monitoring capabilities. This study aims to analyze the spatiotemporal evolution of desertification and its driving factors in the Mu Us Sandy Land from 1990 to 2017 using Landsat imagery, quantifying temporal dynamics (annual change rates, ADI), spatial evolution (transfer matrices, gravity-center shifts), and assessing climatic and human drivers to inform management and sustainable land use.

Early concerns (Zhu, late 1950s) identified anthropogenic drivers converting sandy lands into deserts in the Mu Us region. International and national initiatives (FAO/UNEP methodologies, World Atlas of Desertification) and Chinese efforts (Desertification Indicator Handbook) established frameworks for assessment. Remote sensing-based studies have monitored desertification dynamics in China and the Mu Us Sandy Land: Wang and Li used TM data in the 1980s; Wu et al. detected a 1,936 km² decrease in desertified area between 1987 and 1993; Hao et al. linked ongoing development (mid-1980s to late 1990s) to unreasonable land use. More recent works with Landsat TM/OLI (1990–2014) reported increased fixed/semi-fixed sand, decreased mobile sand, rapid expansion of human activities, and limited climatic influence in some analyses; other studies showed continuous downward trends in desertified extents and overall improvement. Numerous studies examined spatial patterns, drivers (climate, land use), and restoration outcomes (e.g., afforestation, policy impacts) in the Mu Us and surrounding areas, highlighting the role of both natural factors (monsoon-driven climate variability, wind regimes) and human activities (overgrazing, cultivation, mining, policy interventions).

Study area: The Mu Us Sandy Land spans ~42,200 km² across southern Ordos (Inner Mongolia), northern Yulin (Shaanxi), and northeastern Yanchi (Ningxia) at 37.45°–39.37° N, 107.67°–110.5° E. It exhibits semi-arid continental climate (mean annual temperature 6.5–10.3 °C; precipitation 250–440 mm), with elevation decreasing west to east (avg 1254 m, up to 1595 m). Soils are nutrient-poor sandy/chestnut/brown calcareous types; vegetation includes Artemisia ordosica, Caragana microphylla, Sabina vulgaris. Population density is higher in the east (e.g., Yuyang: 78.0 people/km²) than west (as low as 4.9 people/km²). Data sources: - Remote sensing: Landsat TM/OLI scenes for 1990, 1995, 2000, 2007, 2010, 2017 (geospatial data cloud). All images were from the growing season (June–September) to ensure comparability. False-color composites used TM 4-3-2 and OLI 5-4-3. Rigorous geometric and atmospheric corrections were applied; post-correction error < 0.5 pixels. - Interpretation and accuracy: Visual interpretation via human–computer interaction on ArcGIS 10.3, aided by field investigation data and 2010 interpretation marks. In 2017, 350 field verification points targeted typical/confusable classes. Overall classification accuracy was 92.6% with Kappa 0.91 via confusion matrix. - Ancillary data: Meteorological variables (annual mean precipitation, temperature, wind speed) from China Meteorological Data Sharing Service. Human factors (population, livestock, cultivated land) from national statistical yearbooks. Classification scheme: Based on national indicators and prior studies, land was categorized into five types: non-desertified; mild, moderate, severe, and extreme desertification. Metrics and models: - Annual change rate (R): Computed to capture interannual dynamics of desertified land extent over intervals. - Transfer matrix: Quantified transitions among desertification classes between time steps. - Aeolian Desertification Index (ADI): ADI = (S_mild + 2 S_moderate + 3 S_severe + 4 S_extreme) / S_total; ranges 0–4, larger values indicate more severe regional desertification. - Gravity-center model: Computed barycentric longitude-latitude for each desertification class per year using area-weighted patch centroids to analyze spatial migration trajectories and distances over time. Software: ENVI 5.3 and ArcGIS 10.3 for preprocessing, classification, mapping, and spatial analysis. Temporal segmentation: Analyses identified three phases (1990–2000, 2000–2010, 2010–2017) for comparative assessment.

- Overall trend (1990–2017): Total desertified area decreased by 1,680 km² (~60 km² per year on average). Over the full period, mild desertification area increased (annual rate ~3.0%), while severe and extreme desertification declined (−2.7% and −1.7% per year, respectively). - Turning point and phases: 2000 marked the turning point. • Development (1990–2000): Mild and moderate desertified areas decreased annually; severe and extreme increased (e.g., mild −4.4%/yr in 1995–2000; severe +3.0%/yr). • Rapid reversal (2000–2010): Severe and extreme classes both shrank; mild desertification increased strongly (+12.8%/yr in 2000–2007). • Steady reversal (2010–2017): Mild increased slowly; moderate, severe, and extreme decreased with smaller annual rates (e.g., moderate −0.8%/yr; severe −2.7%/yr; extreme −1.0%/yr). - Class-specific area milestones: Mild desertification reached a low in 2000 (2,710 km²) and a high in 2017 (11,090 km²). Moderate desertification had a valley in 2000 (11,980 km²) and a peak in 2007 (19,000 km²). Severe and extreme classes increased to 2000 then declined thereafter. - Transitions (selected): 1990–1995: 38.7% of mild became moderate; 26.1% of moderate became severe; only 18.4% of extreme improved to severe. 1995–2000: 58.1% of mild converted to moderate; 37.3% of moderate deteriorated to severe. 2000–2007: Severe improved to moderate (36.8%) and mild (18.1%); extreme improved to severe (21.0%) and moderate (19.1%). 2007–2010: Moderate and severe to mild were 30.1% and 11.5%, respectively. 2010–2017: Predominantly stepwise reversal. - Composition shift: Main types shifted from moderate (42.6%) and severe (35.4%) in 1990 to moderate (45.0%) and mild (32.0%) in 2017. - ADI dynamics: ADI increased during development (avg ~2.5; +0.3 from 1990 to 2000), decreased markedly during rapid reversal (avg ~2.3; −0.6 from 2000 to 2010), and stabilized during steady reversal (avg ~2.0; <0.1 change 2010–2017). ADI peaked in 2000 (~2.7) and reached a minimum in 2017 (~2.0). - Spatial evolution: 1990–2000 dominated by development/expansion, especially in western/central areas (e.g., Otog Qi, Otog Qian Qi, Uxin Qi). Post-2000 saw widespread improvement, strongest in the northwest (2000–2007), though localized rebounds occurred (central/west 2007–2010; eastern Yuyang and Shenmu 2010–2017). - Gravity-center shifts (1990–2017): Extreme desertification center moved ~3 km NW; severe ~8.8 km NW; moderate ~5 km N; mild ~10 km SE. By 2017, overall desertification severity ranked spatially: central area > northwest > southwest > east > south. - Climatic drivers: 1990–2010 warming with declining precipitation (warm-dry) favored desertification; 2010–2017 warm-wet favored vegetation recovery and reversal. Average wind speed declined by ~0.1 m s⁻¹ per year over ~30 years, reducing aeolian transport capacity and aiding restoration. - Human drivers and policies: Population increased by ~2.4×10⁴ persons/year; livestock numbers rose 2000–2010 then declined; cultivated land expanded overall with a dip after 2001 due to “Grain to Green” policy. Post-2000 measures (Three-North Shelterbelt, returning farmland to forest/grass, grazing prohibition/rotation, enclosures, aerial seeding, Desertification Prevention Law) substantially contributed to reversal. From 2000–2013, afforestation totaled ~13,800 km²; in Uxin Qi, woodland increased from 71 to 203 km² (2000–2015; ~7%/yr).

The study demonstrates a clear temporal transition from desertification development (pre-2000) to rapid and then steady reversal (post-2000), aligning with shifts in both climate and human land management. The warm-dry conditions and human pressures (overgrazing, cultivation expansion, woodcutting, mining) prior to 2000 coincided with deterioration and expansion of severe-to-extreme desertification, especially in western/central Mu Us. After 2000, policy-driven ecological restoration (afforestation, grazing bans, enclosures) and structural adjustments in land and livestock management, together with climatic shifts toward warmer-wetter conditions and a sustained decline in wind speeds, collectively promoted widespread reversal, particularly in areas of mild-to-moderate desertification that are more responsive to management. Spatial gravity-center analyses confirm the east–west gradation of severity and the relative stability of extreme desertification cores versus greater mobility of mild-to-moderate zones, underscoring where targeted human interventions can most effectively alter trajectories. The findings validate the efficacy of integrated ecological policies under favorable climatic trends and highlight the need to sustain and refine interventions in hotspots (e.g., Shenmu, Yuyang) where local deterioration persists.

- Temporal evolution: Desertification exhibited three stages with a turning point around 2000. Development (1990–2000) featured growing severe/extreme desertification mainly in the west; rapid reversal (2000–2010) saw severe desertification reverse (~9.7%/yr) and mild increase (>12.8%/yr), especially in the northwest; steady reversal (2010–2017) continued with slower rates, and central/southern areas mostly stable or improving, though local deterioration occurred in parts of Shenmu and eastern Yuyang. - Spatial patterns: Severity gradients from central (most severe) to northwest, southwest, east, and south persisted. Gravity-center analysis showed mild and moderate classes shifting more than severe/extreme, indicating higher responsiveness to management. By 2017, Uxin Qi remained most severely affected; Otog Qi was dominated by moderate–severe; Otog Qian Qi, Shenmu, Yuyang, Yanchi, and Ejin Horo primarily mild–moderate; Dingbian, Jingbian, and Hengshan ranged from non- to mild–moderate. - Drivers: A ~0.1 m s⁻¹ per year decline in wind speed reduced aeolian transport; 1990–2010 warm-dry versus 2010–2017 warm-wet tendencies modulated vegetation and aeolian processes. Policy measures and human interventions since 2000 played crucial roles in reversing desertification in concert with climate. - Implications: Under ongoing warming-wetting conditions, continued rational use of sandy land resources and strengthened, science-based control of desertification drivers are key to maintaining and enhancing reversal. The integrated approach of monitoring, targeted interventions, and policy support should continue, with attention to localized hotspots of deterioration.