Dynamic Changes and Driving Factors of Vegetation Cover in the Oasis-Desert Ecotone: A Case Study of Cele, Xinjiang

CAO Yongxiang¹, MAO Donglei¹, XUE Jie²,³, SU Songling¹, Kaimaierguli Abulaiti¹, CAI Fuyan⁴

¹College of Geography Science and Tourism, Xinjiang Normal University, Xinjiang Laboratory of Lake Environment and Resources in Arid Zone, Urumqi 830054, Xinjiang, China
²State Key Laboratory of Desert and Oasis Ecology, Xinjiang Institute of Ecology and Geography, Chinese Academy of Sciences, Urumqi 830011, Xinjiang, China
³Cele National Station of Observation and Research for Desert Grassland Ecosystem in Xinjiang, Cele 848300, Xinjiang, China
⁴College of Application Engineering, Urumqi Vocational University, Urumqi 830002, Xinjiang, China

Abstract

The oasis-desert ecotone serves as an ecological buffer zone between oasis and desert ecosystems, playing a vital role in maintaining oasis stability and providing windbreak and sand fixation functions. Taking the Cele oasis-desert ecotone as an example, this study utilized Landsat remote sensing imagery data alongside meteorological and socioeconomic data to investigate long-term vegetation dynamics and their driving factors in a small-scale region using the Normalized Difference Vegetation Index (NDVI), pixel dichotomy model, and principal component analysis. The results indicate that: (1) the multi-year average vegetation coverage was 0.23, with vegetation coverage continuously increasing over time; (2) vegetation coverage was predominantly medium and low, accounting for 30.73% and 21.47% of the study area respectively, followed by low, higher, high, and very low coverage vegetation at 21.48%, 20.39%, 20.12%, and 7.26% respectively; (3) areas with significantly improved vegetation coverage were relatively small and concentrated primarily in the central and southeastern regions and around the Cele River basin, while other vegetation cover change classes were more dispersed; and (4) at the interannual timescale, human activities represent the dominant factor influencing vegetation change in the Cele oasis-desert ecotone, with vegetation showing greater sensitivity to precipitation and atmospheric relative humidity than to temperature. Specifically, artificial afforestation area, cultivated land area, population size, and shelterbelt area accounted for 0.850, 0.810, 0.853, and 0.779 respectively in the first principal component, while atmospheric relative humidity, precipitation, and livestock inventory accounted for 0.845, 0.753, and 0.608 in the second principal component, and temperature accounted for 0.883 in the third principal component. This study provides a theoretical basis for further understanding vegetation dynamics, influencing factors, and vegetation conservation and restoration in oasis-desert ecotones.

Keywords: Normalized Difference Vegetation Index; vegetation coverage; oasis-desert ecotone; influencing factors; principal component analysis

1. Introduction

Vegetation constitutes a crucial component of terrestrial ecosystems and plays a significant role in maintaining sustainable development of global and regional ecosystems. Vegetation coverage, as a key indicator for measuring surface vegetation conditions and describing ecosystem environments, serves as a sensitive indicator of ecological environmental changes and holds great significance for soil and water conservation, regional environmental quality assessment, and ecosystem evaluation. Previous studies have demonstrated that precipitation and temperature are two primary factors affecting regional vegetation cover change, and with intensifying human activities, vegetation cover change is increasingly recognized as the combined result of climate change and human activities.

Oases represent typical landscape features in the arid regions of northwestern China, where the stability and healthy development of their ecological environments are critical. The oasis-desert ecotone, composed of oasis and desert ecosystems, is highly sensitive to climate change and human activities, serving as an important natural ecological barrier for maintaining oasis ecological security and preventing wind erosion and sand fixation. Therefore, in-depth investigation of vegetation dynamics and driving factors in this special zone is of important theoretical and practical significance for ensuring the stability of oasis and desert ecosystems.

The Cele oasis-desert ecotone, located on the southern margin of the Taklimakan Desert, has become a hotspot for recent research. Under the dual influence of global warming and human activities, ecological environmental issues in northwestern China's arid regions have become increasingly prominent, with vegetation changes serving as a key entry point for studying regional ecological evolution and climate change responses. With the rapid development of remote sensing technology, using remote sensing data to study regional vegetation changes has become a major trend. However, most studies have focused on large- and medium-scale regions using short-term, low-resolution MODIS NDVI data, while research using higher-resolution Landsat imagery to analyze long-term vegetation changes and their driving factors in small-scale regions remains relatively rare.

This study takes the Cele oasis-desert ecotone in Xinjiang as an example, selecting the northwestern ecotone—most affected by wind-sand activities and intense human disturbance—as the research area. Using multi-year Landsat imagery data, we analyzed vegetation cover dynamics and driving factors in this region to understand recent vegetation changes and their influencing factors, providing a theoretical basis for further understanding vegetation dynamics, influencing factors, and vegetation conservation and restoration in oasis-desert ecotones.

2. Data Sources and Research Methods

2.1 Data Sources and Preprocessing

Landsat remote sensing imagery data were obtained from the Geospatial Data Cloud platform of the Chinese Academy of Sciences' Computer Network Information Center (http://www.gscloud.cn). Due to significant weather impacts on satellite imagery, we selected images with cloud cover below 10% during the vegetation growing season. Using ENVI 5.3 software, we performed spatial cropping, radiometric calibration, and atmospheric correction as standard preprocessing steps. Given that vegetation growth peaks during summer and autumn, and following previous research indicating that July-August remote sensing data best represents vegetation conditions in southern Xinjiang, we selected imagery from these months to calculate NDVI and vegetation coverage.

Specifically, we used Landsat 5-TM images for years up to 2011 and Landsat 8-OLI images for 2014 and 2017. Vegetation change is a gradual process; to clearly demonstrate dynamic changes in vegetation coverage, we performed visual expression and spatiotemporal change analysis on the imagery. Meteorological data (temperature, precipitation, and atmospheric relative humidity) and human activity data (population, cultivated land area, afforestation area, shelterbelt area, and annual livestock inventory) were all provided by the Cele National Station of Observation and Research for Desert Grassland Ecosystem in Xinjiang, Chinese Academy of Sciences.

2.2 NDVI Calculation and Vegetation Coverage Classification

The Normalized Difference Vegetation Index (NDVI), also known as the standardized vegetation index, offers high sensitivity, wide identification range, and effective elimination of terrain and community structure shadow and radiation interference, making it widely applicable in vegetation coverage research. The calculation formula is:

$$NDVI = (NIR - R)/(NIR + R)$$

where NIR represents the near-infrared band (band 4 in Landsat 5-TM, band 5 in Landsat 8-OLI) and R represents the red band (band 3 in Landsat 5-TM, band 4 in Landsat 8-OLI).

NDVI values range between [-1, +1], with positive values indicating vegetation coverage (higher values = greater coverage) and negative values representing water, snow, or bare rock. To classify vegetation coverage levels—an important indicator for assessing desert vegetation growth status—we used the pixel dichotomy model based on calculated NDVI values:

$$FVC = (NDVI - NDVI_{soil})/(NDVI_{veg} - NDVI_{soil})$$

where $NDVI_{soil}$ represents NDVI values for completely bare soil or non-vegetated pixels, and $NDVI_{veg}$ represents NDVI values for fully vegetated pixels. Due to uncertainty in these values under varying temporal, atmospheric, and surface moisture conditions, we adopted approximate values based on cumulative frequency statistics: $NDVI_{soil}$ at 0.5% cumulative frequency and $NDVI_{veg}$ at 99.5% cumulative frequency.

Using ArcGIS software, we reclassified the calculated vegetation coverage into five levels based on national desertification survey standards and previous research: very low coverage (<10%), low coverage (10-30%), medium coverage (30-60%), high coverage (60-80%), and higher coverage (>80%). To analyze vegetation cover change, we performed raster difference operations between consecutive periods ($D = FV_{t2} - FV_{t1}$) and between the final and initial periods to determine overall change trends, where negative values indicate degradation and positive values indicate improvement.

2.3 Principal Component Analysis

Based on previous research, we selected eight factors influencing vegetation dynamics: temperature (°C), precipitation (mm), atmospheric relative humidity (%), afforestation area (hm²), cultivated land area (hm²), population (people), shelterbelt area (hm²), and livestock inventory (heads). Principal component analysis was employed to quantify the contribution rates of these factors to vegetation cover change in the Cele oasis-desert ecotone. The process involved: (1) standardizing raw data to eliminate scale effects; (2) calculating correlation coefficient matrices; (3) computing eigenvalues and eigenvectors; (4) determining principal components based on cumulative contribution rates exceeding 85%; and (5) analyzing factor loadings to identify primary influences.

3. Results

3.1 Spatiotemporal Characteristics of NDVI

Between 1993 and 2017, the overall NDVI of the Cele oasis-desert ecotone showed an increasing trend, with multi-year average values of 0.20, 0.18, 0.19, 0.23, 0.27, and 0.28 for the respective years. The spatial distribution pattern remained generally consistent, with high values concentrated in the southeast near the Cele oasis and low values in the west and northwest near the Taklimakan Desert margin. During 1993-1998, the high-value range expanded slightly but not significantly. From 1998-2008, vegetation coverage increased overall, though remaining predominantly low to medium. Changes between 2008-2011 were minimal, with only very low coverage areas showing some reduction. After 2011, the spatial distribution of high NDVI values expanded noticeably, particularly near the oasis edge where vegetation coverage increased significantly in blocks. After 2014, overall vegetation coverage improved markedly, with very low coverage areas gradually decreasing and vegetation cover exhibiting a spatial pattern of decreasing from the oasis center outward toward the desert.

3.2 Vegetation Coverage Characteristics

3.2.1 Overall Vegetation Coverage Features

The multi-year average vegetation coverage of the Cele oasis-desert ecotone was 0.23. Medium coverage represented the largest area (30.73% of the study area), followed by low (21.48%), higher (20.39%), high (20.12%), and very low (7.26%) coverage. Between 1993-1998, low and medium coverage areas increased by 1.45 km² and 2.20 km² respectively, while very low and higher coverage areas decreased by 0.82 km² and 0.19 km². During 1998-2008, vegetation coverage showed an overall decreasing trend, with very low and low coverage areas increasing by 0.85 km² and 1.06 km² respectively, while medium and higher coverage areas decreased by 0.14 km² and 2.06 km². From 2008-2011, very low, higher, and high coverage areas increased by 0.10 km², 0.19 km², and 0.12 km² respectively, while low and medium coverage areas decreased by 1.65 km² and 0.19 km². During 2011-2014, low and medium coverage areas increased by 3.06 km² and 0.88 km², while very low and higher coverage areas decreased by 0.07 km² and 3.1 km², indicating overall vegetation improvement. In 2014-2017, very low coverage and higher coverage areas increased by 0.1 km² and 1.45 km² respectively, while low and medium coverage areas decreased by 2.06 km² and 0.19 km², with other classes showing minimal change.

3.2.2 Vegetation Coverage Change Classification

Analyzing spatiotemporal change data from 1993-2017, we classified vegetation coverage changes into seven levels. The classification standards and area proportions show that "basically unchanged" dominated all periods, accounting for 60.73%, 59.70%, 68.61%, 62.27%, and 63.72% of the total area respectively. Areas of slight degradation and slight improvement showed minor differences, while areas of significant improvement and significant degradation were relatively small. Throughout the entire study period, the ranking of change levels by area was: basically unchanged > slight degradation > slight improvement > significant improvement > significant degradation, with average proportions of 62.27%, 19.73%, 15.00%, 2.87%, and 1.70% respectively.

Spatially, areas with significantly improved vegetation coverage were concentrated in the central, southeastern, and Cele River basin regions, distributed in patches, while other change classes were more dispersed. The vegetation cover change pattern indicates that while overall vegetation improved, fluctuations were substantial, with very low coverage areas being unstable, low coverage areas expanding, medium and higher coverage areas decreasing, and high coverage areas showing large fluctuations.

3.3 Analysis of Driving Factors

3.3.1 Trends in Climate Change and Human Activities

Meteorological factors showed that mean annual temperature exhibited a fluctuating upward trend from 1993-2017, while precipitation and atmospheric relative humidity showed consistent patterns with large fluctuations. In terms of human activities, both population and cultivated land area increased significantly, with cultivated land area growing slowly before 2000 but increasing dramatically afterward, from 16,790 hm² to 25,500 hm², and continuing to grow slowly thereafter. Population showed an approximately linear growth trend, increasing by nearly 2×10⁴ people every five years. Artificial afforestation and shelterbelt areas increased sharply, particularly after 2000, while livestock inventory increased slowly with large fluctuations. These trends indicate that population growth drove cultivated land expansion, which in turn necessitated increased shelterbelt and afforestation areas to protect against wind erosion.

3.3.2 Factor Selection and Principal Component Analysis

Principal component analysis of standardized data yielded a correlation matrix showing most inter-factor correlation coefficients exceeded 0.3, indicating suitability for PCA. We selected the first three principal components with cumulative variance contribution exceeding 85% to analyze factors influencing vegetation change. The first principal component (contribution rate: 52.38%) was heavily loaded with afforestation area (0.850), cultivated land area (0.810), population (0.853), and shelterbelt area (0.779). The second principal component (contribution rate: 21.77%) showed high loadings for atmospheric relative humidity (0.845), precipitation (0.753), and livestock inventory (0.608). The third principal component (contribution rate: 12.69%) was dominated by temperature (0.883).

These results demonstrate that human activities are the primary driver of vegetation change in the Cele oasis-desert ecotone, with climate factors being secondary. Vegetation shows greater sensitivity to precipitation and atmospheric relative humidity than to temperature, consistent with previous research findings. Although some studies suggest vegetation responds more to temperature than precipitation, analysis shows that while Cele's mean temperature has risen, maximum temperatures haven't changed significantly, whereas precipitation has increased with anomalous years. Temperature rise may provide needed heat for vegetation growth but also accelerates evaporation, making vegetation more sensitive to precipitation and humidity changes.

4. Discussion

Throughout the 1993-2017 study period, vegetation coverage in the oasis-desert ecotone showed an overall increasing trend, though with banded reduction patterns. The spatial distribution exhibited a decreasing gradient from the oasis center outward to the desert, with the ecotone gradually shifting toward the desert and narrowing in width. This aligns with previous research showing vegetation cover and species diversity generally increase from the desert front toward the oasis, though with scattered local distributions. The primary cause is farmland reclamation at the oasis edge, reducing natural vegetation. The northwestern ecotone of Cele lies downwind of prevailing westerly and northwesterly winds, where severe wind-sand hazards damage vegetation through sand cutting and burial, with damage decreasing from desert to oasis. Additionally, population growth, cultivated land expansion, and shelterbelt construction have significantly increased vegetation coverage within the oasis and extended it toward the desert, though patchy distributions in the ecotone likely result from local topography and soil nutrients.

The region's extremely arid climate and abundant sand sources create harsh conditions for vegetation growth, with even lower coverage in the desert-adjacent ecotone. While interannual variations in precipitation, temperature, and vegetation coverage are relatively small, intra-annual changes are more pronounced. At interannual timescales, human activities thus have more significant impacts on vegetation cover than climate factors. Livestock inventory indirectly reflects grazing impacts, showing a negative correlation with vegetation coverage, though its relatively small loading in PCA suggests limited influence, likely due to grazing prohibition and firewood collection bans implemented in recent years. Additionally, improved canal lining rates with oasis expansion have reduced groundwater seepage, causing decline in some desert vegetation dependent on groundwater.

Government policies have also played important regulatory roles. Early rapid population growth and irrational water resource use, deforestation, and over-cutting caused severe vegetation degradation and wind-sand invasion. However, increased attention to wind-sand control, the "Three-North" Shelterbelt Program, sand enclosure for vegetation restoration, and other measures have significantly improved the ecological environment and increased vegetation coverage since the mid-1990s, particularly through afforestation and sand control efforts in the northwestern ecotone since 2000. Besides climate and human activities, factors such as groundwater, topography, soil moisture, temperature, and nutrients directly affect vegetation changes and should be incorporated in future research using multiple data sources and quantitative analysis methods.

5. Conclusions

This study analyzed the Cele oasis-desert ecotone using 1993-2017 Landsat imagery and meteorological and socioeconomic data, employing NDVI, pixel dichotomy model, and principal component analysis to investigate long-term vegetation dynamics and driving factors. The main conclusions are:

1) Vegetation coverage in the Cele oasis-desert ecotone showed large spatiotemporal fluctuations but an overall increasing trend, expanding gradually from southeast to northwest. Temporally, vegetation coverage increased yearly, with multi-year average NDVI of 0.23. Spatially, high NDVI values were concentrated in the southeast near the oasis, while low values occurred in the west and northwest adjacent to the Taklimakan Desert, forming an expansion pattern from the oasis center outward through the ecotone.

2) During 1993-2017, low and medium vegetation coverage dominated, with very low coverage showing significant change. Low coverage areas expanded, medium and higher coverage areas decreased, and high coverage areas fluctuated substantially. Medium coverage represented the largest proportion (30.73%), followed by low (21.48%), higher (20.39%), high (20.12%), and very low (7.26%) coverage. Vegetation change was predominantly classified as "basically unchanged," accounting for 60.73-68.61% of the area across periods, with slight degradation and improvement showing minor differences, and significant changes occupying small areas.

3) Human activities constitute the dominant factor influencing vegetation change, followed by climate factors. Artificial afforestation area, cultivated land area, population, and shelterbelt area had large loadings in the first principal component (0.850, 0.810, 0.853, and 0.779 respectively), while atmospheric relative humidity, precipitation, and livestock inventory dominated the second principal component (0.845, 0.753, and 0.608 respectively). Temperature was the main factor in the third principal component (0.883). Vegetation sensitivity to precipitation and atmospheric relative humidity exceeded that to temperature.

References

[1] Forkel M, Carvalhais N, Rödenbeck C, et al. Enhanced seasonal CO2 exchange caused by amplified plant productivity in Northern ecosystems[J]. Science, 2016, 6274: 696-699.

[2] Peng J, Liu Z H, Liu Y H, et al. Trend analysis of vegetation dynamics in Qinghai Tibet plateau using hurst exponent[J]. Ecological Indicators, 2012, 14 (1): 28-39.

[3] Bao Gang, Qin Zhihao, Bao Yuhai, et al. Spatial temporal changes of vegetation cover in Mongolian Plateau during 1982-2006[J]. Journal of Desert Research, 2013, 33(3): 916-927.

[4] Qi Yan, Wang Xiulan, Feng Zhongke, et al. Study on coverage changes of the vegetation in Beijing City based on RS and GIS[J]. Forest Inventory and Planning, 2009, 34(2): 1-4.

[5] Ma Zhiyong, Shen Tao, Zhang Junhai, et al. Vegetation changes analysis based on vegetation coverage[J]. Bulletin of Surveying and Mapping, 2007(3): 45-48.

[6] Gan Chunying, Wang Xizhi, Li Baosheng, et al. Changes of vegetation coverage during recent 18 years in Lianjiang River watershed[J]. Scientia Geographica Sinica, 2011, 31(8): 1019-1024.

[7] Dong Lu, Zhao Jie, Liu Xuejia, et al. Responses of vegetation growth to temperature during 1982-2015 in Xinjiang, China[J]. Chinese Journal of Applied Ecology, 2019, 30(7): 2165-2170.

[8] Ci Hui, Zhang Qiang. Spatio temporal patterns of NDVI variations and possible relations with climate changes in Xinjiang Province[J]. Journal of Geo-Information Science, 2017, 19(5): 662-671.

[9] Ding Yue, Abudureheman Halike, Chen Xiangyue, et al. Spatial temporal changes in vegetation characteristics and climate in Hotan prefecture[J]. Acta Ecologica Sinica, 2020, 40(4): 1258-1268.

[10] Chen Xiuyan, Fu Bihong, Shi Pilong, et al. Vegetation dynamics in response to climate change in Tianshan, Central Asia from 2000 to 2016[J]. Arid Land Geography, 2019, 42(1): 162-171.

[11] Ma Zice, Yu Hongbo, Cao Congming, et al. Spatio temporal characteristics of fractional vegetation coverage and its influencing factors in China[J]. Resources and Environment in the Yangtze Basin, 2020, 29(6): 1310-1321.

[12] Liu Hai, Liu Feng, Zheng Liang. Effects of climate change and human activities on vegetation cover change in the Yellow River basin[J]. Journal of Soil and Water Conservation, 2021, 25(4): 143-151.

[13] Dong Diwen, Abudureheman Halike, Wang Dawei, et al. Spatio temporal variations in vegetation cover in Hotan oasis from 1994 to 2016[J]. Acta Ecologica Sinica, 2019, 39(10): 3710-3719.

[14] Wand Yuchao, Zhao Chengyi. Study in desert-oasis ecological fragile zone[J]. Arid Land Geography, 2001, 24(2): 182-188.

[15] Abuduwasiti·wulamu. Research on Model of Groundwater Level Distribution in Oasis and Desert Ecotone Using Remote Sensing[D]. Urumqi: Xinjiang University, 2000.

[16] Mao Donglei, Lei Jiaqiang, Wang Cui, et al. Characteristics of sand flown structure and sand transportation particles in Cele desert-oasis ecotone of Xinjiang Wei Autonomous Region[J]. Bulletin of Soil and Water Conservation, 2015, 35(1): 25-33.

[17] Mao Donglei, Lei Jiaqiang, Zeng Fanjiang, et al. Sand erosion and deposition on different underlying land surfaces in the desert-oasis ecotone in Cele, Xinjiang, China[J]. Journal of Desert Research, 2014, 34(4): 961-969.

[18] Zhou Sasa, He Qing, Jin Lili, et al. Radiation characteristics of the oasis-desert transition zone in the northern margin of the Taklimakan Desert: A case study of Xiaotang[J]. Journal of Desert Research, 2020, 40(4): 43-51.

[19] Yang Fan, Wang Xueqin, He Qing, et al. Morphological features and spatial distribution pattern of Tamarix ramosissima Nebkhas in an oasis-desert ecotone[J]. Arid Zone Research, 2014, 31(3): 556-563.

[20] Han Fugui, Xu Xianying, Yu Qiushi, et al. Response of reproductive phenology of typical sand fixing plants to climate change in the oasis-desert transitional zone in Minqin, Gansu, China[J]. Journal of Desert Research, 2015, 35(2): 330-337.

[21] Ren Xiao, Mu Guijin, Xu Lishuai, et al. Characteristics of artificial oasis expansion in south Traim Basin from 2000 to 2013[J]. Arid Land Geography, 2015, 38(5): 1022-1030.

[22] Chang Xueli, Ji Shuxin, Qiao Rongrong, et al. NDVI based identification of oasis-desert transitional zone width: A case study in the central Hexi corridor[J]. Acta Ecologica Sinica, 2020, 40(15): 5327-5336.

[23] Ye Guixiang, Li Weiqing, Tian Yuan. Dynamic changes of vegetation cover in typical oasis of arid areas based on NDVI[J]. Journal of Arid Land Resources and Environment, 2009, 23(9): 128-133.

[24] Mu Xiangting, Bai Jie, Li Guanglu, et al. Comparison of methods based on MODIS for estimating sparse vegetation fraction across desert in Xinjiang[J]. Arid Land Geography, 2013, 36(3): 502-511.

[25] Chen Tao, Li Pingxiang, Zhang Liangpei. Dynamic analysis of vegetation fraction change in Wuhan region from 1988 to 2002[J]. Remote Sensing Technology and Application, 2008, 23(5): 511-516.

[26] Yang Xinghua, He Qing, Cheng Yujing, et al. Near surface horizontal aeolian sand dust flux over oasis-desert ecotone in Qira county, Xinjiang[J]. Arid Zone Research, 2013, 30(6): 1100-1105.

[27] Li Zhen, Yan Fuli, Fan Xiangtao. The variability of NDVI over Northwest China and its relation to temperature and precipitation[J]. National Remote Sensing Bulletin, 2005, 20(3): 308-313.

[28] Tian Jing, Yan Yu, Chen Shengbo. The advances in the application of the remote sensing technique to the estimation of vegetation fractional cover[J]. Remote Sensing for Land & Resources, 2004(1): 1-5.

[29] Zhang Hao, Qu Jianjun, Zhang Kecun. Vegetation cover information extraction technology for Dunhuang oasis based on remote sensing images[J]. Journal of Desert Research, 2015, 35(2): 493-498.

[30] Bai Hongying. The Response of Vegetation to Environmental Change in Qinba Mountains[M]. Beijing: Science Press, 2014.

[31] Guo Ni, Zhu Yanjun, Wang Jiemin, et al. The relationship between NDVI and climate elements for 22 years in different vegetation areas of Northwest China[J]. Journal of Plant Ecology, 2008, 32(2): 319-327.

[32] Wu Chaoqun, Zhang Xubing, Wang Yao, et al. Analysis of vegetation coverage extraction and time-space change in Muli coalfield based on landsat image[J]. Geomatics & Spatial Information Technology, 2020, 43(2): 67-72.

[33] Wang Ziyu, Xu Duanyang, Yang Hua, et al. Impacts of climate change and human activities on vegetation dynamics in Inner Mongolia, 1981-2010[J]. Progress in Geography, 2017, 36(8): 1025-1032.

[34] Zeng Jia, Guo Feng, Zhao Can, et al. Climate change of small oases in the southern margin of Taklimakan Desert in recent 50 years[J]. Arid Land Geography, 2014, 37(5): 948-957.

[35] Duan Zhengrong, Zubaidai Muyibula, Xia Jianxin. Analysis of vegetation coverage dynamic change in typical oasis of arid areas based on NDVI: A case study of Aksu region in Xinjiang[J]. Journal of MUC(Natural Sciences Edition), 2018, 27(2): 5-14.

[36] Xue Wei. SPSS Statistical Analysis Method And Its Application[M]. Beijing: Publishing House of Electronics Industry, 2013: 262-279.

[37] Liu Xiao, Xue Ying, Ji Yupeng, et al. An assessment of water quality in the Yellow River estuary and its adjacent waters based on principal component analysis[J]. China Environmental Science, 2015, 35(10): 3187-3192.

[38] Chen Pei. Principal Component Analysis and Its Application in Feature Extraction[D]. Shaanxi: Shaanxi Normal University, 2014.

[39] Mao Donglei. Research on space heterogeneity of vegetation and soil moisture in oasis-desert ecotone in Cele county[J]. Modern Agricultural Science and Technology, 2013(13): 252-256.

[40] Yue Shengru, Zhou Jiyun, Hu Xuefei, et al. Study on driving factors and spatial-temporal evolution of vegetation coverage in Xinjiang from 2000 to 2018[J]. Journal of Tarim University, 2020, 32(2): 97-105.