Ecological transitions in Xinjiang, China: Unraveling the impact of climate change on vegetation dynamics (1990-2020)

HAO Haichao; YAO Junqiang; CHEN Yaning; XU Jianhua; LI Zhi; DUAN Weili; Sadaf ISMAIL; WANG Guiling

doi:10.1007/s11442-024-2238-7

PDF(14161 KB)

Journal of Geographical Sciences ›› 2024, Vol. 34 ›› Issue (6) : 1039-1064. DOI: 10.1007/s11442-024-2238-7

Research Articles

Ecological transitions in Xinjiang, China: Unraveling the impact of climate change on vegetation dynamics (1990-2020)

Author information +

History +

Abstract

For the past several decades, climate change has been driving vegetation dynamics in arid regions worldwide. This study investigates vegetation dynamics and their links to climate from 1990 to 2020 in Xinjiang, China, using data from 30-m resolution land use and cover change, remote sensing, and climate reanalysis. Our approach encompasses a range of analytical techniques, including transfer matrix analysis, modeling, correlation, regression, and trend analysis. During the study period, there were major vegetation conversions from grassland to forestland in the mountains, and from cropland to grassland in the plains. Climate change emerged as an important trigger of these changes, as evidenced by the increase in net primary productivity in most vegetation types, except for cropland-grassland and grassland-cropland conversions. Precipitation and soil moisture were the most influential climatic factors, contributing 15.1% and 15.2%, respectively, to natural vegetation changes. The study also found that evapotranspiration serves as a key mechanism for moisture dissipation in the hydrological cycle of vegetation dynamics. The interplay between precipitation, soil moisture, and evapotranspiration is a critical pattern of climatic influence that shapes vegetation dynamics across zones of intersection, increase, decrease, and change. These insights are invaluable for informing vegetation conservation and development strategies in Xinjiang and other similar environments facing climate-driven ecological transitions.

Key words

climate change / ecological transitions / LUCC / NPP efficiency / vegetation dynamics / Xinjiang

Cite this article

EndNote

Ris (Procite)

Bibtex

Download Citations

HAO Haichao, YAO Junqiang, CHEN Yaning, XU Jianhua, LI Zhi, DUAN Weili, Sadaf ISMAIL, WANG Guiling. Ecological transitions in Xinjiang, China: Unraveling the impact of climate change on vegetation dynamics (1990-2020)[J]. Journal of Geographical Sciences, 2024, 34(6): 1039-1064 https://doi.org/10.1007/s11442-024-2238-7

1 Introduction

Vegetation is a key component of terrestrial ecosystems and has a profound impact on human society and its economic activities (Chen et al., 2019; Piao et al., 2020). In addition to promoting ecosystem stability and the environmental improvement of soil, vegetation plays a vital role in regulating global temperature and the carbon cycle balance (Kang et al., 2020). Climate change also significantly affects terrestrial ecosystems, mainly by regulating plant respiration, photosynthesis, growth cycles, and soil development processes, thereby altering the spatial distribution patterns of vegetation growth (Huang et al., 2020; Li et al., 2023).

Global warming has resulted in a rising trend in both the frequency and intensity of droughts, severely affecting the stability of land ecosystem functions, such as photosynthesis and carbon sequestration capabilities (Zeng et al., 2023).

In recent decades, researchers have increasingly been focusing on substantial changes in global climate conditions and their impact on vegetation dynamics, especially in terms of changes in vegetation types and productivity (Ge et al., 2021; Yang et al., 2021). Clarifying the mechanisms by which climate affects vegetation changes is of great practical significance for ecosystem services and regional ecological restoration.

Xinjiang, which is located in the hinterland of the Eurasian continent, has a climate typical of arid zones. However, it also has notable geographical heterogeneity and complex climatic processes, making its fragile ecosystem highly sensitive to global climate change (Yao et al., 2022). The main vegetation types in Xinjiang are forests, grasslands, wetlands, scrub, and croplands (Figure 1). Therefore, studying the spatiotemporal pattern changes of vegetation in this region is highly useful not only for understanding the mechanisms of vegetation-climate interactions, but also for revealing the patterns of ecosystem changes in arid areas of the world.

Figure 1 Overview of study area and the study’s targeted climate factor changes, i.e., TMMX, TMMN, PRE, DEF, PDSI, SOIL, SRAD, PET, AET, VSP, VAP, and VPD, from 1990 to 2020, in Xinjiang, China (NO.: GS (2019) 3333)

Full size|PPT slide

Since forests are considered significant carbon sinks under climate change and account for a large portion of the global terrestrial carbon, their carbon dynamics are critically important (Cai et al., 2022). Grasslands, which are mostly comprised of annual and perennial herbs, are extremely sensitive to changes in meteorological elements such as precipitation, soil water, and evapotranspiration (Huang et al., 2018; Bai et al., 2021; Hao et al., 2022). The degradation of grasslands is attributable to climate change and grazing and has become a major issue in Xinjiang. Climate change is also affecting the region’s wetlands, which are predominantly located in sub-alpine moist areas and river and lake basins, such as the Bayanbulak Grasslands and the Bosten Lake basin. Along with climate change, human activities are threatening these fragile ecosystems as well, leading to a reduction in wetland area and overall degradation of ecological functioning (Salimi et al., 2021). Scrub ecosystems are widely distributed in arid and semi-arid areas, with their carbon absorption/release primarily controlled by soil moisture availability (Wang et al., 2021), while croplands primarily consist of irrigated farmlands in the oases (Abulizi et al., 2017).

Vegetation changes triggered by climatic factors generally affect vegetation greenness at both regional and global scales. Previous studies used climate datasets and time series of the Normalized Difference Vegetation Index (NDVI) to analyze how vegetation responds to climate change in China's arid regions (Meng et al., 2023). The key indicator of material cycling and energy flow in terrestrial ecosystems is Net Primary Productivity (NPP), as calculated through vegetation greenness NDVI (Piao et al., 2005). NPP not only reflects the ecosystem's capacity as a carbon source or sink, but also serves as an important measure for evaluating ecosystem regulatory functions and ecological processes (Zhao et al., 2024).

In-depth examination of NPP's spatiotemporal variability and its interactions with climate variables (e.g., temperature, precipitation, solar radiation, etc.) reveals that climate factors impact NPP on several levels, from directly affecting plant photosynthesis and respiration to indirectly influencing the availability of soil moisture and the cycling of nutrients (Tang et al., 2020; Zhao et al., 2024). Therefore, understanding vegetation dynamics and their response to climate change is very important. Studies show that vegetation growth in several arid regions is exhibiting a significant declining trend (Jiao et al., 2021; Hao et al., 2022a), while in other arid regions, vegetation is showing a greening trend (Deng et al., 2020). Thus, the key drivers of vegetation growth are still not well understood, despite significant efforts to elucidate the diverse responses of vegetation changes in China's arid regions to climate fluctuations. How climatic factors over time affect the area and productivity of various vegetation types also remains unknown.

Previous studies were limited to assessing vegetation types and their responses to climate change, without considering how changes affect the vegetation’s climate change response (Hao et al., 2022a; Zhang et al., 2023). Although climate fluctuations have different impacts on different vegetation types, affecting both their distribution and changes, earlier studies only investigated the relationship between vegetation and annual climate factors (Ge et al., 2021; Shi et al., 2021). Vegetation change dynamics were neglected in the research, especially the response of vegetation area changes and type conversions (i.e., intersection, increase, decrease, and change zones) to climate change. An in-depth comprehension of the dynamics of different vegetation types and their relationship to climate change is thus necessary to devise adaptation strategies in Xinjiang to tackle the challenges that climate change and human activities pose to the vegetation ecosystems across the region. To achieve this aim, remote sensing and modeling techniques can be used to elucidate dynamic change patterns.

In studying the impact of land-use changes on vegetation, traditional field methods have been shown to be limited. Remote sensing offers a more effective solution, providing detailed observations over large areas and long time series (Woodcock et al., 2020; Gao et al., 2021) and enabling the monitoring of vegetation location, growth, and seasonal changes. The latest advancements in remote sensing technology offer new insights into how climate change affects vegetation changes (Beamish et al., 2020; Crowley and Cardille, 2020). For example, high-precision land-use data aids can accurately measure vegetation features, track changes, and detect shifts in land types (Yang et al., 2021), all of which is essential for long-term ecological studies and ecosystem modeling (Weisser et al., 2017). Remote sensing data are also better able to explain the complex connections between climate and vegetation, especially with the rapid vegetation changes driven by global warming and land-use changes (He et al., 2019).

To date, the widespread application of remote sensing data has provided a big data foundation for interdisciplinary model borrowing and application. The concept of NPP efficiency draws from the super-efficiency model in economics, emphasizing the fundamental relationship between inputs (environmental climatic factors and vegetation productivities) and outputs (vegetation productivity efficiency) (Li et al., 2013; Zheng et al., 2023). Understanding these technical aspects is crucial for a deep comprehension of the rapid evolution of vegetation dynamics and the climate factors driving these changes, which is vital for maintaining the health and sustainability of ecosystems in arid regions.

Recent studies have shown that, as an important ecological barrier in northwest China, Xinjiang has experienced substantial land-use changes that have primarily been characterized by the expansion of cropland and the reduction of grassland coverage (Liu et al., 2023). Over the past three decades (1990-2020), Xinjiang has undergone a significant shift of land into agricultural use (Wang et al., 2022). The northern Tianshan region in particular has drawn widespread attention for its expansion of cropland and grasslands, as well as for its reduction of forests, water bodies, and unused lands. All of these changes have significantly impacted vegetation productivity (Xu et al., 2021; Xie et al., 2022).

It is therefore both timely and crucial to investigate changes in vegetation types in Xinjiang and determine their underlying causes. This study aims to explore how climatic factors influence the conversion of different vegetation types over time as well as changes in area and productivity. The study also delves deeper into climate trends affecting different vegetation types within the four main dynamic vegetation regions (intersection, increase, decrease, and change zones). The insights derived from this research are anticipated to enhance our understanding of ecological shifts by uncovering patterns of ecosystem changes in Xinjiang, offering scientific guidance for understanding vegetation-climate interactions, and contributing to the development of suitable policies to address global climate change.

2 Materials and methods

2.1 Study area

Xinjiang is located in northwest China and covers an area of 1.66 million km² (34°25°N-48°10°N and 73°40°E- 96°18°E). Known for its arid climate, hot summers, and cold winters, Xinjiang’s unique landform includes three mountain ranges (Altai Mountains, Tianshan Mountains, and Kunlun Mountains), two river basins (Junggar Basin and Tarim Basin), and numerous oases, leading to significant climate variations between the north and the south. The north experiences cooler temperatures (4-8℃) with limited rainfall (100-500 mm), while the south is warmer (10-13℃) with higher evaporation rates (>3400 mm).

Despite the region having annual rainfall amounts of mostly less than 100 mm, its climate is still classified as temperate continental type. From 1990 to 2020, Xinjiang saw increases in average annual minimum temperature (TMMN), maximum temperature (TMMX), precipitation (PRE), climate moisture deficit (DEF), solar radiation (SRAD), potential evapotranspiration (PET), actual evapotranspiration (AET), vapor saturation pressure (VSP), and vapor pressure difference (VPD). These increases indicated a trend towards regional warming and wetting, mainly due to increased mountain rainfall. In contrast, a decreasing trend emerged in soil moisture (SOIL), vapor pressure (VAP), and the Palmer Drought Severity Index (PDSI) (Figures 1 and 7).

Xinjiang’s harsh climate predominantly supports deserts with sparse vegetation, and the area's natural vegetation includes forests (1.17%), grasslands (23.38%), wetlands (0.03%), and scrub (0.0002%). Grasslands are prevalent in the sub-alpine and pre-mountain plains and forests, mainly as desert riparian dominated by Populus, while wetlands are common near Bayinbruck and along the river and lake shorelines of the Tianshan Mountains. Most of the scrub is scattered throughout the desert-oasis transition zones, and cropland (5.38%) is primarily situated in the oases. Overall, this ecologically fragile region exhibits limited quality and resilience. Recent ecological restoration efforts have partially improved oasis environments, but degradation persists in peripheral areas where human activities continue to pose threats to the natural vegetation.

2.2 Climate, vegetation, and topographical data

The Carnegie-Ames-Stanford approach (CASA) net primary productivity (NPP) data from 1990-2015 were obtained from the Global Change Research Data Publishing & Repository (Chen, 2019a; 2019b) and the CASA-NPP data for 2016-2020 were calculated for this study (Table 1). MODIS (Justice et al., 2002) is a key instrument aboard the Terra and Aqua satellites. The MODIS-MOD13A3 Normalized Difference Vegetation Index (NDVI) is computed from atmospherically corrected bi-directional reflectance and MCD15A3H Fraction of Photosynthetically Active Radiation (FPAR), composited from the best four-day MODIS acquisition pixels. The MODIS Land Cover Types (MCD12Q1) Version 6 data product provides global land cover types (2001-2020), derived from supervised classification of MODIS Terra and Aqua reflectance data. The Global Land Data Assimilation System (GLDAS) integrates satellite and ground data using land surface modeling for optimal land states and fluxes (Rodell et al., 2004).

Table 1 Data product type and source

Model/ Chapter	Product	Type	Temporal resolution	Spatial resolution	URL source
CASA	MOD13A3	NDVI	30 d	500 m	https://modis.gsfc.nasa.gov/
	MCD15A3H	FPAR	4 d	500 m	https://modis.gsfc.nasa.gov/
	T3H (GLDAS)	TEM	3 h	0.25°	http:/ldas.gsfc.nasa.gov/
	TerraClimate	PRE	Monthly	1/24°-4 km	https://www.ecmwf.int
	TerraClimate	SOL	Monthly	1/24°-4 km	https://www.ecmwf.int
	MCD12Q1	LUCC	96 d	500 m	https://modis.gsfc.nasa.gov/
Chapter	TerraClimate	AET	Monthly	1/24°-4 km	https://www.ecmwf.int
	TerraClimate	DEF	Monthly	1/24°-4 km	https://www.ecmwf.int
	TerraClimate	PET	Monthly	1/24°-4 km	https://www.ecmwf.int
	TerraClimate	PRE	Monthly	1/24°-4 km	https://www.ecmwf.int
	TerraClimate	SOIL	Monthly	1/24°-4 km	https://www.ecmwf.int
	TerraClimate	SRAD	Monthly	1/24°-4 km	https://www.ecmwf.int
	TerraClimate	TMMX	Monthly	1/24°-4 km	https://www.ecmwf.int
	TerraClimate	TMMN	Monthly	1/24°-4 km	https://www.ecmwf.int
	TerraClimate	VSP	Monthly	1/24°-4 km	https://www.ecmwf.int
	TerraClimate	VAP	Monthly	1/24°-4 km	https://www.ecmwf.int
	TerraClimate	VPD	Monthly	1/24°-4 km	https://www.ecmwf.int
	TerraClimate	PDSI	Monthly	1/24°-4 km	https://www.ecmwf.int
Chapter	CLCD	LUCC	Yearly	30 m	https://zenodo.org/record/5210928
Chapter	SRTM	DEM	—	30 m	https://www.usgs.gov/

Note: Table 1 shows the NDVI, fraction of photosynthetically active radiation (FPAR), LUCC, total solar surface radiation (SOL), as well as the AET, DEF, PET, PRE, SOIL, SRAD, TMMX, TMMN, VSP, VAP, VPD, PDSI, Shuttle Radar Topography Mission (SRTM), and Digital Elevation Model (DEM).

TerraClimate is a global dataset combining WorldClim climate normal with time-varying data from CRU and JRA55 reanalysis (Abatzoglou et al., 2018). Its climate-assisted interpolation provides validates monthly climate and water balance data. TerraClimate enables consistent attribution analysis of vegetation changes. The annual China Land Cover Dataset (CLCD) from 1985-2021 uses 335,709 Landsat images on Google Earth Engine (GEE) (Yang and Huang, 2021). The CLCD captures China’s urbanization and ecological projects, revealing anthropogenic land cover changes under climate change. Percentage of LUCC in Xinjiang by 2020: Forest (1.17%), cropland (5.38%), grassland (23.38%), wetlands (0.03%), scrub (0.0002%), water bodies (0.64%), ice-snow (2.08%), bare land (67.00%), and cities (0.31%). The Shuttle Radar Topography Mission (SRTM) acquires near-global digital elevation models (Farr et al., 2007). This study uses the 1 arc-second (approximately 30 m) SRTM V3 product (SRTM Plus) from NASA JPL.

Remote sensing and reanalysis data used in the model and computational analyses in this study have been unified at raster resolution by the GEE platform and open-source programming, which, despite uncertainties, is already the state of the science achieved with the current data and technological conditions.

2.3 Carnegie-Ames-Stanford approach model: net primary productivity algorithm

For our first model, we explored the applicability of CASA to our study. We then used it to estimate the NPP and performed a validation for the study region (Zhang et al., 2021; Hao et al., 2022b). For the simulation, we chose two variables, namely, the absorbed photosynthetically active radiation (APAR) (MJ/m²) and light energy conversion (ε) (g C/MJ), formulating them as:

N P P = A P A R \times ε

(1)

A P A R = S O L \times F P A R \times 0.5

(2)

where APAR is the product of PAR and the fraction of photosynthetically active radiation (FPAR). PAR (i.e., radiation in the 400 to 700 nm wave band) is the portion of the photosynthetically active radiation received by an ecosystem that is absorbed by green plants. It can be calculated as half the total SOL (MJ/m²) (Sala et al., 2000). FPAR is estimated by two variables, i.e., the FPAR_NDVI and FPAR_SR (Potter et al., 1993; Zhu et al., 2006):

F P A R_{N D V I} = \frac{(N D V I - N D V I_{i, min}) \times (F P A R_{max} - F P A R_{min})}{N D V I_{i, max} - N D V I_{i, min}} + F P A R_{min}

(3)

where FPAR_max (=0.95) and FPAR_min (=0.001) are independent of the vegetation type. Here, NDVI_i_,max is the NDVI value corresponding to 95% of the NDVI population i, while NDVI_i_,min is the NDVI value corresponding to 5% of the NDVI population i. The relation between FPAR and SR can be represented as:

F P A R_{S R} = \frac{(S R - S R_{i, min}) \times (F P A R_{max} - F P A R_{min})}{S R_{i, max} - S R_{i, min}} + F P A R_{min}

(4)

S R = [\frac{1 + N D V I}{1 - N D V I}]

(5)

where SR_i_,max and SR_i_,min correspond to NDVI_i_,max and NDVI_i_,min.

F A P R = α F P A R_{N D V I} + (1 - α) F P A R_{S R}

(6)

with

α

set at 0.5.

The light energy conversion (ε) (g C/MJ) can be calculated as:

ε = T_{1} \times T_{2} \times w_{ε} \times ε_{max}

(7)

where T₁ and T₂ represent the effects of low and high temperature stress, w_ε denotes the effects of water stress, and ε_max is maximum light use efficiency (g C/MJ). The calculation of each stress factor and the value of the maximum light energy utilization rate of each vegetation type was based on existing research results (Zhu et al., 2006).

2.4 Super-SBM model: NPP efficiency algorithm

The highly efficient slacks-based measure (SBM) model (Li et al., 2013) is formulated as follows. Using Xinjiang as our example, let us suppose there are n provinces to be evaluated, labeled as j=1, …, n. Each Xinjiang raster has m input indicators and s output indicators. Let x_ij= (x₁_j, x₂_j, …, x_mj) and y_rj= (y₁_j, y₂_j, …, y_sj), respectively, denoting the m-input vector and s-output vector of Xinjiang raster j. The x elements of this study are NPP, AET, PDSI, PRE, SOIL, SRAD, VPD, TEM, DEM, and SLOPE (calculated using DEM data from GEE). As this study focuses on the intrinsic response of climate change in the water cycle to vegetation changes, it does not incorporate the effects of elements such as carbon dioxide and instantaneous wind speed. In this framework, y is vegetation NPP efficiency and λ= (λ₁, λ₂, …, λ_n) is the intensity vector of n Xinjiang rasters. The efficiency of the kth Xinjiang raster (denoted as θ_k) can be derived by a linear program, as follows:

\begin{aligned} θ_{k} = Min \frac{\frac{1}{m} \sum_{i = 1}^{m} \bar{x_{i}} / x_{i k}}{\frac{1}{m} \sum_{r = 1}^{s} \bar{y_{r}} / y_{r k}} \\ s . t . \bar{x_{i}} \geq \sum_{j = 1}^{n} j \neq k, x_{i j} λ_{j} \\ \bar{y_{r}} \leq \sum_{j = 1}^{n} j \neq k, y_{i j} λ_{j} \\ \bar{x_{i}} \geq x_{i k} \\ y_{r} \leq y_{r k} \\ λ, s^{-}, s^{+} \geq 0 \\ i = 1, 2, . . ., m; r = 1, 2, . . ., s; j = 1, 2, . . ., n (j \neq k) \end{aligned}

(8)

where x_ik and y_rk are the input and output vectors of the evaluated DMU_k, respectively;

\bar{x}

and

\bar{y}

are the input/output matrices excluded (x_k, y_k) from (x, y), respectively; and s⁻and s⁺corresponds to the slacks in inputs and outputs. The possible production set built by other DMU_s besides DMU_k is:

P = {(x, y) | \sum_{j = 1, j \neq k}^{n} λ_{j} x_{i j} \leq x, y \leq \sum_{j = 1, j \neq k}^{n} λ_{j} y_{r j}}

(9)

If the efficiency value of DMU_k (θ_k) is equal to or larger than 1, the DMU is SBM-efficient; otherwise, the DMU is inefficient.

2.5 Random Forest approach

Random Forest (Biau and Scornet, 2016) is a versatile machine-learning method capable of performing both regression and classification tasks. The software also performs dimensional reduction techniques and effectively handles missing values, outlier values, and other crucial data exploration steps. The described approach is an ensemble learning technique that involves the aggregation of multiple weak models to create a robust one. We used this approach to determine the significant values of DEF, PDSI, PET, PRE, SOIL, SRAD, TMMN, TMMX, VAP, VPD, and VSP for changes in the NPP of different vegetation types.

2.6 Partial least squares regression

Partial least squares regression (PLSR) analyzes relationships between two sets of variables, especially when variables outnumber observations and collinearity exists (Geladi and Kowalski, 1986). PLSR overcomes multicollinearity by constructing linear components from predictors. This study uses PLSR to relate area changes of vegetation types to climate factors, as collinearity occurs between the twelve examined climate variables. The optimal number of PLS components was determined by minimizing root mean square error (RMSE) during cross-validation. Using fewer components prevents overfitting while capturing primary variability (Ma et al., 2023).

2.7 Principle of GeoDectector

The geographic detector model (Wang et al., 2022) is a statistical method that detects spatial differentiation and explains its driving force. The model functions without many constraints and overcomes the shortcomings of traditional statistical methods when dealing with variables. Factor detection is used to find the spatial differentiation of Y and to determine to what extent factor X explains the spatial differentiation of attribute Y.

Based on the relationship between data availability and vegetation succession and climate factors in the study area, three key climate factors were selected. These are: TEM (X1), PRE (X2), and SRAD (X3). The value of vegetation succession and climate factors are spatially matched, and the dependent and independent variables at each discrete point are extracted. The calculation formula can be written as:

q = 1 - \frac{\sum_{h = 1}^{L} N_{h} σ_{h}^{2}}{N σ^{2}}

(10)

where q is the explanatory power of climate factors at a range of [01]. In this case, q=0 indicates that the climate factor quality is randomly distributed: the larger the value of q, the stronger the explanatory power of climate factors. N_h and N are the number of sub-level sample units and the whole research units, respectively, and

σ_{h}^{2}

and σ² are the variance of the climate factors of the sub-level and the whole research unit, respectively.

3 Results and discussion

3.1 Spatio-temporal variations of vegetation dynamics

3.1.1 Spatio-temporal variations of vegetation area

From 1990 to 2020, vegetation cover increased by 11,809.232 km² in Xinjiang, with grassland and cropland accounting for 54.05% and 38.01% of the increased area, respectively (Figure 2a). Grassland also saw the largest decline (81,341.544 km²), representing 89.7% of the total decline in vegetation (Figure 2b, Table S1). Overall, natural vegetation showed a net increase in forest, scrub, and wetlands, but a net decrease in grassland. Conversely, artificial vegetation, mainly cropland, experienced a substantial net increase of 29,689.509 km².

Figure 2 Land-use changes for different vegetation types in Xinjiang from 1990 to 2020 (a. Areas with increased vegetation types; b. areas with decreased vegetation types)

Full size|PPT slide

In southern Xinjiang, there was an increase in forests, scrub, and wetlands, while grasslands saw a decline. Cropland gains exceeded natural vegetation losses, reflecting human demand for grain and economic crops (Zhu, 2013). In contrast, in northern Xinjiang, there was an expansion in forests and wetlands but a decline in grasslands and scrub. Unlike the south, the reduction in natural vegetation in the north was greater than the gains in its cropland, pointing to more pronounced disturbances caused by climate change. These regional variations highlight different influences on vegetation dynamics. Specifically, the cropland expansion in the south was due to heightened demand for grain and economic crops, whereas the north experienced the combined effects of climate change on natural vegetation.

Table 2 Before and after transfer matrices for different land-use types from 1990 to 2020

Type	Cropland	Forest	Scrub	Grassland	Water bodies	Ice- Snow	Bare land	Cities	Wetland
1990 Symbols	Cr1	Fo1	Sc1	Gr1	Wa1	So1	Ba1	Ci1	We1
2020 Symbols	—	Fo	Sc	Gr	—	—	—	—	We

3.1.2 Spatio-temporal dynamics of vegetation types

The 1990-2020 transfer matrix shows the expansion of new forested areas (7824.441 km²), with the majority (7486.517 km²) coming from grassland conversion. Other sources include ice-snow (162.648 km²), cropland (110.625 km²), water bodies (48.616 km²), bare land (13.488 km²), wetlands (1.823 km²), and scrub (0.725 km²). New grassland (55,397.059 km²) was converted from bare land (46,873.192 km²), cropland (7576.247 km²), ice-snow (636.371 km²), water bodies (291.816 km²), wetlands (16.523 km²), forest (1.707 km²), urban land (0.669 km²), and scrub (0.535 km²). Scrub increased 2.033 km², converting from grassland (1.452 km²), forest (0.568 km²), and ice-snow (0.013 km²). Wetlands gained 308.102 km² from grassland (305.088 km²), cropland (1.840 km²), bare land (0.605 km²), and water bodies (0.568 km²) (Figure 3a and Table S2).

Figure 3 Transfer matrix of different land-use types to natural vegetation and spatial distribution of dynamics of different vegetation types in Xinjiang in 1990-2020 (a. spatial distribution of dynamics of different vegetation types from 1990 to 2020; b-j are the transfer matrix figures for different land-use types to natural vegetation; b. Xinjiang; c. mountainous area of Xinjiang; d. plains area of Xinjiang; e. northern Xinjiang; f. mountainous area of northern Xinjiang; g. plains area of northern Xinjiang; h. southern Xinjiang; i. mountainous area of southern Xinjiang; j. plains area of southern Xinjiang. Note: In this study, areas <1500 m are defined as plains and areas >1500 m are defined as mountains (Chen, 2014).)

Full size|PPT slide

In comparing natural vegetation in 2020 with land-use types in 1990, we can see that the main vegetation transitions were cropland to grassland and grassland to forest. Natural vegetation gained 63,531.635 km² from 1990 to 2020, primarily from conversions of grassland to forest and bare land/cropland to grassland (Figures 3b-3j and Table S2). Bare land conversion to grassland dominated most regions, while grassland conversion to forest prevailed in the northern mountains (Figure 3f) and cropland conversion to grassland in the plains. The primary contributors to these shifts were bare land, cropland, grassland (1990), and ice-snow, with forests and grasslands (2020) emerging as the main beneficiaries of the ecological transitions (Figures 3b-3j).

Additionally, glacial melting, expanding forests, and new grassland areas were induced by climate warming (Yao et al., 2022; Figure 7), with only a small proportion of grassland converting to scrub and wetlands. At the same time, increased evaporation reduced some water bodies into natural forests, grasslands, and wetlands, while other grasslands along with croplands transitioned into bare land due to climate change and human activities when the original vegetation and soil conditions became unsuitable for farming or plant growth. However, restoration policies such as cropland retirement and grazing exclusion may also make the transition from cropland to grassland possible (Feng et al., 2005).

3.2 Impact of climate change on vegetation dynamics (NPP)

3.2.1 Spatio-temporal patterns of vegetation dynamics (NPP)

In 1990, the net primary productivity values in Xinjiang spanned from 0 to 571 g C·m^-2·yr^-1 (abbreviated heretofore as g C), with an average of 94.37 g C. By 2020, these values had shifted to a range of 0 to 893.61 g C, with a lower average of 57.82 g C. This indicates that the overall mean NPP decreased over time, despite NPP’s broader range in 2020. Moreover, the spatial extent represented by higher NPP values (above 331.40 g C [red area]) in 1990 was considerably more extensive than in 2020 (Figures 4a and 4c).

Interestingly, despite the general decrease in average NPP values over the 30-year study period, the mean NPP values within specific vegetation zones (cropland, forest, scrub, and wetlands) were higher in 2020. In 1990, NPP values across vegetation zones, ranked from highest to lowest, were as follows: wetlands at 208.55 g C, cropland at 164.87 g C, forest at 139.92 g C, grassland at 136.96 g C, and scrub at 11.15 g C. In 2020, the order was forest at 334.18 g C, wetlands at 271.95 g C, cropland at 218.23 g C, grassland at 127.57 g C, and scrub at 37.19 g C (Figure 4b).

Figure 4 Comparison of changes in NPP at the intersection of different vegetation types in Xinjiang from 1990 to 2020 (a. spatial changes in NPP in 1990; b. annual average value changes in NPP at the intersection of different vegetation types from 1990 to 2020; c. spatial changes in NPP in 2020)

Full size|PPT slide

Additionally, the lowest NPP values for scrub occurred in 1990 and 2020, the highest NPP values for wetlands occurred in 1990, and the highest NPP values for forest occurred in 2020, reflecting climate change-induced dynamic shifts among various vegetation types (Figure 4b). All transitions showed an increase in NPP, except for cropland-to-grassland and grassland-to-cropland conversions (Table 3).

Table 3 Comparison of NPP in Xinjiang in 1990 and 2020 for vegetation type change areas in land-use transition (g C)

LUCC-vegetation (1990-2020)	NPP-1990 (g C)	NPP change and trend (1990-2020)	NPP-2020 (g C)	Rise or fall
	170.28		283.34
	165.40		145.62
	216.12		265.99
	161.57		243.20
	56.57		181.65
	224.87		201.07
	159.96		335.91
	61.66		204.83
	162.29		270.98
	91.94		203.17
	138.14		233.43
	203.95		303.93
	153.97		269.05
	166.14		295.74

Our analysis further delved into the impact of climate factors on the variability of NPP across different vegetation types. Our evaluated climate factors (DEF, PDSI, PET, PRE, SOIL, SRAD, TMMN, TMMX, VAP, VPD, and VSP) each showed a discernible effect on vegetation NPP. Our findings reveal that cropland, forest, and scrub NPP were predominantly influenced by maximum temperature and soil moisture. Furthermore, grassland NPP showed a higher sensitivity to potential evapotranspiration and soil moisture, whereas wetlands were most impacted by precipitation and maximum temperature (Figure 5).

Figure 5 Importance of climate factors to NPP changes of different vegetation types in Xinjiang from 1990 to 2020 (a. cropland; b. forest; c. grassland; d. wetlands; e. scrub)

Full size|PPT slide

3.2.2 Climatic interpretation of NPP by vegetation NPP efficiency

NPP efficiency in Xinjiang underwent notable shifts during the study period. In 1990, NPP efficiency values ranged from 0 to 4.73, with an average of 0.25, while in 2020, they ranged from 0 to 0.89, with a mean of 0.16. These values indicate higher efficiency in 1990. Moreover, the spatial extent of NPP efficiency above 0.495 (red area) in 1990 far exceeded that in 2020, although if cropland, scrub, and wetlands were excluded, forests and grasslands showed improved NPP efficiency in 2020. Specifically, NPP efficiency rankings for each vegetation zone in 1990 were cropland (0.51), forest (0.43), grassland (0.35), wetlands (0.30), and scrub (0.07). In 2020, the order shifted to wetlands (0.64), forest (0.45), cropland

(0.36), grassland (0.27), and scrub (0.17). Overall, in 1990, scrub recorded the lowest efficiency and cropland the highest, while in 2020, wetlands recorded the highest.

Natural vegetation types demonstrated uniform increases in both NPP and efficiency from 1990 to 2020. Croplands, on the other hand, showed increased NPP but decreased efficiency.

NPP and efficiency were also highly correlated during the study period. These results demonstrate that climate change plays a dominant role in vegetation NPP changes while also being the most significant factor in vegetation productivity changes. Prior research indicated that climate change accounts for 50.16%-60.06%, while human activities account for 39.94%-49.84% (Ge et al., 2021). The NPP and efficiency trends provide an ecological basis for attributing vegetation changes to climatic factors.

3.3 Climate attribution of changes in vegetation dynamics

3.3.1 Climate contribution to the overall vegetation dynamics in geographic patterns

Precipitation is such a critical climate factor affecting vegetation, that any variations in it significantly influence vegetation changes. Setting aside other variables, PRE, TEM, and SRAD emerge as the top three climatic factors driving vegetation growth and succession, highlighting the pivotal role these natural climatic elements play in the evolution of vegetation patterns. In this study, the GeoDetector model has been selected to investigate the dynamics of natural vegetation succession in Xinjiang and to gauge the influence of natural factors across varying spatial scales. This approach aims to provide a nuanced understanding of how intrinsic climatic factors shape vegetation transitions in the region.

The GeoDetector model’s mean temperature (TEM) was derived from TMMX and TMMN calculations. Subsequently, TEM, PRE, and SRAD were identified as natural factors and their impact on the succession of forests and grasslands was explored. PRE emerged as the dominant factor, accounting for 48.8% of the influence, underscoring its critical role in the succession dynamics of the study region’s natural forests and grasslands. SRAD, contributing 37.94%, was identified as the second most significant factor, reinforcing its importance in the ecological succession of these habitats. TEM, with a 30.10% influence, also plays a crucial role in this ecological process. Notably, PRE consistently stands out as the foremost natural driver of forest and grassland vegetation succession in Xinjiang, a conclusion supported by previous studies (Cao et al., 2011; Liang et al., 2015).

Figure 6 Comparison of changes in NPP efficiency at the intersection of different vegetation types in Xinjiang from 1990 to 2020 (a. spatial changes in NPP efficiency in 1990; b. annual average value changes in NPP efficiency at the intersection of different vegetation types from 1990 to 2020; c. spatial changes in NPP efficiency in 2020)

Full size|PPT slide

Throughout the entire study period, Xinjiang experienced a climate trend of warming and drying (Figure 7). This period also saw increases in TMMX, TMMN, PRE, DEF, SRAD, PET, AET, VSP, and VPD, indicating positive trends, whereas PDSI, SOIL, and VAP all decreased, indicating negative trends. Despite an uptick, PRE was insufficient to offset the rise in PET, leading to a net drying effect. Consequently, the shift in vegetation types and land-use areas is intricately linked to these climatic factor trends. Notably, the warming and drying trend was more pronounced in southern than in northern Xinjiang, and more so in the plains than in the mountains.

Figure 7 Trends in the spatial distribution of climate factors in Xinjiang from 1990 to 2020 (a-l show spatio-temporal trends for TMMX, TMMN, PRE, DEF, PDSI, SOIL, SRAD, PET, AET, VSP, VAP, and VPD, respectively. Note: “+” indicates a positive trend, and “-” indicates a negative trend.)

Full size|PPT slide

From 1990 to 2020, all of Xinjiang and especially its northern region experienced a decrease in natural vegetation coverage, with annual reductions of 438.96 km² and 504.20 km², respectively, while southern Xinjiang witnessed an annual increase of 65.25 km². In terms of vegetation change trends, Xinjiang and its northern district saw forests and wetlands expand and scrub and grasslands decline. Southern Xinjiang, on the other hand, experienced growth across all vegetation types, with forests in the lead, followed by grasslands, wetlands, and scrub. In sum, although forests and wetlands generally expanded across the entire study region, the north led the south in forest area change ranking, but the south led the north in wetlands area change. Furthermore, grasslands and scrub declined in Xinjiang and its northern area but increased in the south. These differences point to distinct regional trends.

Additionally, natural vegetation shows positive correlations with SOIL, SRAD, and VAP, peaking at a correlation coefficient of 0.40 with SOIL and showing a negative correlation with TMMX, TMMN, PRE, DEF, PDSI, PET, AET, VSP, and VPD. Of these factors, VPD presents the largest negative coefficient (-0.29). Forests and wetlands have positive associations with TMMX, TMMN, PRE, DEF, PET, AET, VSP, and VPD but show negative correlations with PDSI, SOIL, and VAP. Meanwhile, SRAD varies between forests and wetlands, indicating a relatively humid climate. Grassland and scrub exhibit negative correlations with TMMN, PRE, DEF, PET, AET, VSP, and VPD while positively correlating with VAP, mirroring their drier conditions.

The strongest correlations identified were TMMN with forests and SOIL with grasslands, wetlands, and scrubs. Overall, vegetation types across Xinjiang present a 47.92% positive and 52.08% negative correlation with land-use and climatic changes (Figure 8a). Specifically, northern Xinjiang shows 33.33% positive and 66.67% negative correlations (Figure 8b), whereas southern Xinjiang reveals 76.67% positive and 23.33% negative correlations (Figure 8c). This pattern indicates a stark contrast in the interaction between vegetation changes and climatic factors between the north and the south (Figure 8).

Figure 8 Correlations between land-use area change of natural vegetation (Na-Veg) and climatic factors in Xinjiang (a), northern Xinjiang (b), and southern Xinjiang (c) (Notes: * indicates significance, p < 0.05, and ** indicates extreme significance, p < 0.01.)

Full size|PPT slide

Partial least squares regression identified the primary climate factors contributing to vegetation area changes. Water scarcity factors, including SOIL, SRAD, VPD, and PRE, were identified as significant forces behind the decline in natural vegetation (Figure 9a). For cropland, PRE, TMMN, VPD, VSP, and PDSI had high contribution rates, reflecting climatic controls. However, cropland is also influenced by irrigation, which skews the data. Therefore, the climatic factor ranking should be interpreted with caution. Nonetheless, atmospheric drought and water deficit still appear to affect croplands (Figure 9b).

Figure 9 Contribution rates of climate factors to area changes of different vegetation types in Xinjiang from 1990 to 2020 (a. natural vegetation; b. cropland; c. forest; d. grassland; e. wetlands; f. scrub)

Full size|PPT slide

Along with croplands, forests are also strongly influenced by PRE and TMMN, while grasslands are strongly influenced by PRE and SOIL (Figures 9b-9d). Changes in forest areas were primarily driven by PRE, TMMN, PET, and VSP, with PRE and TMMN patterns playing a key role. For grasslands, the climate factors impacting them, in order of importance, were PRE, SOIL, TMMN, VPD, and PDSI (Figure 9d). Wetlands responded to climate factors in the order of PRE, TMMN, TMMX, and PET, and scrub changes were influenced by SOIL, SRAD, TMMX, and PET (Figure 9f).

3.3.2 Climatic trend analysis of vegetation dynamic intersection, increase, decrease, and change zones

Climate trends differed between vegetation dynamic intersection, increase, decrease, and change zones in 1990-2020 (Figures 10-13). In intersection zones, croplands faced warming and drying trends, with increased actual evapotranspiration (AET+) and reduced precipitation (PRE-). Irrigation, however, mitigated water shortages, ensuring relative stability. Conversely, forests, grasslands, and wetlands experienced warming and wetting trends (PRE+), with precipitation remaining lower than evapotranspiration in forests and grasslands (PRE+ < AET+) and higher in wetlands (PRE+ > AET+). Scrub showed a greater decrease in AET than in PRE (AET- > PRE-). These patterns of AET and PRE changes, both positive and negative, were consistent across all vegetation types other than for croplands. Notably, the differences in magnitudes between PRE and AET were not significant (PRE≈AET), with stability primarily arising from consistent PRE and AET trends (Figure 10).

Figure 10 Climate factor trends for the intersection of the same vegetation types in Xinjiang from 1990 to 2020 (a-l show trends for TMMX, TMMN, PRE, DEF, PDSI, SOIL, SRAD, PET, AET, VSP, VAP, and VPD, respectively. Note: “+” indicates a positive trend, and “-” indicates a negative trend.)

Full size|PPT slide

In vegetation increase zones, key parameters such as TEM, PRE, DEF, SRAD, PET, VAP, and VPD experienced an upward trend during the study period, while the PDSI, SOIL, and VSP declined, indicating a warming and drying trend similar to intersection zones. Cropland, forest, grassland, wetlands, and scrub exhibited warming and wetting patterns (TMMX+, TMMN+, PRE+). This shift facilitated the expansion of different vegetation types, underscoring the resilience and adaptability of these ecosystems in response to changing climate conditions.

Among these factors, PRE+ was the primary climatic driver behind the expansion of vegetation types. Cropland growth was attributed to both human-mediated irrigation efforts and rising precipitation levels. In contrast, wetlands and scrub were more directly influenced by PRE+. It is worth noting that although low SOIL and high VPD could undermine soil moisture use efficiency in vegetation, leading to plant decline through hydraulic failure and carbon starvation (Hao et al., 2022), optimal levels of SOIL and VPD support vegetation growth (Liu et al., 2020). The concurrent rise in PRE+ and AET+ trends, coupled with a climate characterized by rain and warmth, facilitated the expansion of both forest and scrub areas (Figure 11).

Figure 11 Regional climate factor trends for the increase zone of different vegetation types in Xinjiang from 1990 to 2020 (a-l show trends of different vegetation types in TMMX, TMMN, PRE, DEF, PDSI, SOIL, SRAD, PET, AET, VSP, VAP, and VPD, respectively. Note: “+” indicates a positive trend, and “-” indicates a negative trend.)

Full size|PPT slide

In decreasing vegetation zones, TMMX+, TMMN+, DEF+, SRAD+, PET+, VSP+, and VPD+ increased over the 30-year timeframe, while PDSI and VAP experienced a decline. Specifically, within the context of PRE climate factors, croplands, forest, and scrub demonstrated PRE- trends, whereas grasslands and wetlands exhibited PRE+ trends. Croplands, forests, and scrub were characterized by a warming and drying pattern, evidenced by increases in TMMX+ and TMMN+ alongside PRE-. Conversely, grasslands and wetlands experienced a warming and wetting trend, though this increase in wetness was not sufficient to offset the higher rates of evapotranspiration (AET- > PRE-), indicating that despite the wetting trend, evapotranspiration consumed more moisture than was gained through precipitation.

The combination of higher temperatures and drought conditions, characteristic of a warming and drying climate, precipitated a reduction in natural vegetation cover. Declines in cropland were primarily attributed to anthropogenic activities, while a slight downward trend in PRE- was also observed. Our analysis reveals that the reduction in forest was intimately linked to PRE-, AET+, and SOIL+. Excess soil moisture in forested areas could induce dynamic changes in vegetation, including tree mortality caused by root hypoxia (Shabala et al., 2014). Similarly, the decline in scrub was driven by PRE- and AET+, contributing to drier climatic conditions (Figure 12).

Figure 12 Regional climate factor trends showing the decrease zone of different vegetation types in Xinjiang from 1990 to 2020 (a-l show vegetation trends for TMMX, TMMN, PRE, DEF, PDSI, SOIL, SRAD, PET, AET, VSP, VAP, and VPD, respectively. Note: “+” indicates a positive trend, and “-” indicates a negative trend.)

Full size|PPT slide

The vegetation shifts observed over the past 30 years are closely tied to the magnitude and trends of climate change in Xinjiang (Figure 13). The transitions between cropland (Cr), forest (Fo), grassland (Gr), scrub (Sc), and wetland (We) over the last three decades have been significantly influenced by both human activities and climate change. Specifically, the transitions from Cr to Fo and Cr to Gr were predominantly driven by human interventions, such as reforestation and revegetation, in response to national policies (Feng et al., 2005; Zinda et al., 2017) and against a backdrop of a warming and drying climate. These activities were further intensified by severe climatic changes (Yao et al., 2022), amplifying vegetation dynamics in the region. On the other hand, the transitions from Fo to Cr and Gr to Cr occurred mainly in a warm and wet climate and were driven by human activities aimed at expanding agricultural land to meet the demands for grain and economic crops. These transitions were facilitated by improvements in the ecological environment and the absence of restrictive government policies.

Figure 13 Regional climate factor trends for the change zone of different vegetation types in Xinjiang from 1990 to 2020 (a-l show vegetation trends for TMMX, TMMN, PRE, DEF, PDSI, SOIL, SRAD, PET, AET, VSP, VAP, and VPD, respectively. Note: “+” indicates a positive trend, and “-” indicates a negative trend.)

Full size|PPT slide

Additionally, the transition from Fo to Gr is attributable to a decrease in precipitation and an increase in soil moisture (i.e., PRE-, AET-, SOIL+, VAP+) under constant warming conditions, leading to forest degradation and subsequent transition to grassland. The transition from Fo to Sc resulted from increased precipitation and decreased soil moisture (i.e., PRE+, SOIL-, VAP-), causing the degradation of forest into scrub under unchanged warming conditions. The transformation of Gr to Fo and Gr to Sc (i.e., PRE+, SOIL-, VAP-), as well as the reverse transition from Sc to Gr (i.e., PRE-, SOIL-), were all influenced by changes in precipitation and soil moisture under stable warming conditions. These transitions reflect the adaptability of ecosystems to varying climatic and moisture conditions.

Finally, the transitions from Gr to We, Sc to Fo, We to Cr, We to Gr, and We to Fo occurred under warm and humid climate conditions, driven by shifts in key climate variables such as TMMX+, TMMN+, PRE+, DEF+, PDSI-, SOIL-, SRAD+, PET+, AET+, VSP+, VAP-, and VPD+. The most significant changes involved PRE, SOIL, TEM, and VSP. These transitions highlight the profound impact of climatic factors on vegetation dynamics.

Overall, the complex interplay between human activities, national policies, and climatic changes has profoundly shaped the vegetation patterns across Xinjiang, underscoring the importance of considering both anthropogenic and natural factors in understanding ecological transitions.

4 Conclusions

This study examined vegetation dynamics in Xinjiang using LUCC data, remote sensing simulations, and climate reanalysis data from 1990 to 2020. Through transfer matrix analysis, model simulations, correlation, regression, and trend analysis, we investigated vegetation dynamics and attribution analysis, with the aim of understanding the climate-driven mechanisms. The results of the study can inform efforts to manage and restore degraded vegetation. Our conclusions are as follows:

(1) The vegetation area in Xinjiang showed an overall expansion (+393.64 km²·yr^-1) from 1990 to 2020, with increases in cropland (+989.65 km²·yr^-1) and decreases in natural vegetation (-596.01 km²·yr^-1). The decrease in grassland area was greater than the increase in forest, wetlands, and scrub area. The dynamics of grassland-to-forest conversion in mountainous areas and cropland-to-grassland conversion in plains areas were dominant patterns across the study region.

(2) Climate change is an important trigger of vegetation dynamics, with all types of NPP showing a net increasing trend. The exceptions were cropland-to-grassland and grassland-to-cropland conversions, which revealed a decreasing trend caused by human activities in a warming and drying climate. The proportion of positive and negative correlations between climatic factors and vegetation area highlighted the opposing patterns in northern and southern Xinjiang.

(3) Changing moisture conditions were the main trigger affecting the decrease in natural vegetation area and the increase in cropland area. Furthermore, the pattern of accumulation and dissipation of precipitation and soil moisture was the predominant driving mechanism in vegetation dynamics, with contribution rates of 15.1% and 15.2% for PRE and SOIL, respectively. It is worth noting that evapotranspiration serves as a key mechanism for moisture dissipation in the hydrological cycle of vegetation dynamics.

Overall, our findings provide important insights into change patterns of vegetation dynamics and climate response mechanisms in arid regions, with practical implications for ecological conservation and government decision-making.

References

List( Publishing order | Descend order by publishing year | Descend order by cited within ) Chart analysis

[1]	Abatzoglou J T, Dobrowski S Z, Parks S A et al., 2018. TerraClimate, a high-resolution global dataset of monthly climate and climatic water balance from 1958-2015. Scientific Data, 5: 1-12. Cited in this article [1]

[2]	Abulizi A, Yang Y, Mamat Z et al., 2017. Land-use change and its effects in Charchan oasis, Xinjiang, China. Land Degradation & Development, 28(1): 106-115. Cited in this article [1]

[3]	Bai X, Zhao W, Wang J et al., 2021. Precipitation drives the floristic composition and diversity of temperate grasslands in China. Global Ecology and Conservation, 32: e01933. Cited in this article [1]

[4]	Beamish A, Raynolds M K, Epstein H et al., 2020. Recent trends and remaining challenges for optical remote sensing of Arctic tundra vegetation: A review and outlook. Remote Sensing of Environment, 246: 111872. Cited in this article [1]

[5]	Bennett A E, Classen A E E T, 2020. Climate change influences mycorrhizal fungal-plant interactions, but conclusions are limited by geographical study bias. Ecology, 101(4): e02978.

[6]	Biau G E R, Scornet E, 2016. A random forest guided tour. Test, 25: 197-227. Cited in this article [1]

[7]	Cai W, He N, Li M et al., 2022. Carbon sequestration of Chinese forests from 2010 to 2060: Spatiotemporal dynamics and its regulatory strategies. Science Bulletin, 67(8): 836-843. Cited in this article [1]

[8]	Cao X, Chen X, Bao A et al., 2011. Response of vegetation to temperature and precipitation in Xinjiang during the period of 1998-2009. Journal of Arid Land, 3(2): 94-103. Cited in this article [1]

[9]	Chen C, Park T, Wang X et al., 2019. China and India lead in greening of the world through land-use management. Nature Sustainability, 2(2): 122-129. Cited in this article [1]

[10]	Chen P F, 2019a. Monthly NPP 1 km raster dataset of China’s terrestrial ecosystems (1985-2015). Global Change Research Data Publishing & Repository, 3(1): 34-41. (in Chinese) Cited in this article [1]

[11]	Chen P F, 2019b. Monthly NPP dataset covering China’s terrestrial ecosystems at north of 18°N (1985-2015). Global Change Data & Discovery, 3(1): 34-41. (in Chinese) Cited in this article [1]

[12]	Chen Y, 2014. Water resources research in Northwest China. Springer Science & Business Media, 1-49. Cited in this article [1]

[13]	Crowley M A, Cardille J A, 2020. Remote sensing’s recent and future contributions to landscape ecology. Current Landscape Ecology Reports, 5: 45-57. Cited in this article [1]

[14]	Deng Y, Wang S, Bai X et al., 2020. Vegetation greening intensified soil drying in some semi-arid and arid areas of the world. Agricultural and Forest Meteorology, 292: 108103. Cited in this article [1]

[15]	Farr T G, Rosen P A, Caro E et al., 2007. The shuttle radar topography mission. Reviews of Geophysics, 45(2): 361. Cited in this article [1]

[16]	Feng Z, Yang Y, Zhang Y et al., 2005. Grain-for-Green policy and its impacts on grain supply in West China. Land Use Policy, 22(4): 301-312. Cited in this article [2]

[17]	Gao F, Zhang X, 2021. Mapping crop phenology in near real-time using satellite remote sensing: Challenges and opportunities. Journal of Remote Sensing, (1): 14. Cited in this article [1]

[18]	Garc I A, Alvarez D, Nanu S F, 2022. Land use cover datasets: A review. Land Use Cover Datasets and Validation Tools, 47.

[19]	Ge W, Deng L, Wang F et al., 2021. Quantifying the contributions of human activities and climate change to vegetation net primary productivity dynamics in China from 2001 to 2016. Science of the Total Environment, 773: 145648. Cited in this article [3]

[20]	Geladi P, Kowalski B R, 1986. Partial least-squares regression: A tutorial. Analytica Chimica Acta, 185: 1-17. Cited in this article [1]

[21]	Hao H, Chen Y, Xu J et al., 2022a. Water deficit may cause vegetation browning in central Asia. Remote Sensing, 14(11): 2574. Cited in this article [3]

[22]	Hao H, Li Z, Chen Y et al., 2022b. Recent variations in soil moisture use efficiency (SMUE) and its influence factors in Asian drylands. Journal of Cleaner Production, 373: 133860. Cited in this article [2]

[23]	He X, Liang J, Zeng G et al., 2019. The effects of interaction between climate change and land-use/cover change on biodiversity-related ecosystem services. Global Challenges, 3(9): 1800095. Cited in this article [1]

[24]	Huang N, Wang L, Song X et al., 2020. Spatial and temporal variations in global soil respiration and their relationships with climate and land cover. Science Advances, 6(41): eabb8508. Cited in this article [1]

[25]	Huang X, Luo G, Ye F et al., 2018. Effects of grazing on net primary productivity, evapotranspiration and water use efficiency in the grasslands of Xinjiang, China. Journal of Arid Land, 10: 588-600. Cited in this article [1]

[26]	Jiao W, Wang L, Smith W K et al., 2021. Observed increasing water constraint on vegetation growth over the last three decades. Nature Communications, 12(1): 3777. Cited in this article [1]

[27]	Justice C O, Townshend J R G, Vermote E F et al., 2002. An overview of MODIS land data processing and product status. Remote Sensing of Environment, 83(1/2): 3-15. Cited in this article [1]

[28]	Kang Z, Chen J, Yuan D et al., 2020. Promotion function of forest vegetation on the water & carbon coupling cycle in karst critical zone: Insights from karst groundwater systems in south China. Journal of Hydrology, 590: 125246. Cited in this article [1]

[29]	Li C, Fu B, Wang S et al., 2023. Climate-driven ecological thresholds in China’s drylands modulated by grazing. Nature Sustainability, 6(11): 1363-1372. Cited in this article [1]

[30]	Li H, Fang K, Yang W et al., 2013. Regional environmental efficiency evaluation in China: Analysis based on the Super-SBM model with undesirable outputs. Mathematical and Computer Modelling, 58(5/6): 1018-1031. Cited in this article [2]

[31]	Liang S, Yi Q, Liu J et al., 2015. Vegetation dynamics and responses to recent climate change in Xinjiang using leaf area index as an indicator. Ecological Indicators, 58: 64-76. Cited in this article [1]

[32]	Liu L, Gudmundsson L, Hauser M et al., 2020. Soil moisture dominates dryness stress on ecosystem production globally. Nature Communications, 11(1): 4892. Cited in this article [1]

[33]	Liu Y, Yuan X, Li J et al., 2023. Trade-offs and synergistic relationships of ecosystem services under land use change in Xinjiang from 1990 to 2020: A Bayesian network analysis. Science of the Total Environment, 858: 160015. Cited in this article [1]

[34]	Meng N, Wang N A, Cheng H et al., 2023. Impacts of climate change and anthropogenic activities on the normalized difference vegetation index of desertified areas in northern China. Journal of Geographical Sciences, 33(3): 483-507. Cited in this article [1]

[35]	Piao S, Fang J, Zhou L et al., 2005. Changes in vegetation net primary productivity from 1982 to 1999 in China. Global Ecology and Conservation, 19(2): 1-16. Cited in this article [1]

[36]	Piao S, Wang X, Park T et al., 2020. Characteristics, drivers and feedbacks of global greening. Nature Reviews Earth & Environment, 1(1): 14-27. Cited in this article [1]

[37]	Potter C S, Randerson J T, Field C B et al., 1993. Terrestrial ecosystem production: A process model based on global satellite and surface data. Global Ecology and Conservation, 7(4): 811-841. Cited in this article [1]

[38]	Rodell M, Houser P R, Jambor U et al., 2004. The global land data assimilation system. Bulletin of the American Meteorological Society, 85(3): 381-394. Cited in this article [1]

[39]	Sala O E, Jackson R B, Mooney H A et al., 2000. Introduction methods in ecosystem science:Progress, tradeoffs, and limitations. In: Methods in Ecosystem Science. Springer, 1-3. Cited in this article [1]

[40]	Salimi S, Almuktar S A, Scholz M, 2021. Impact of climate change on wetland ecosystems: A critical review of experimental wetlands. Journal of Environmental Management, 286: 112160. Cited in this article [1]

[41]	Shabala S, Shabala L, Barcelo J et al., 2014. Membrane transporters mediating root signalling and adaptive responses to oxygen deprivation and soil flooding. Plant, Cell & Environment, 37(10): 2216-2233. Cited in this article [1]

[42]	Shi S, Yu J, Wang F et al., 2021. Quantitative contributions of climate change and human activities to vegetation changes over multiple time scales on the loess plateau. Science of the Total Environment, 755: 142419. Cited in this article [1]

[43]	Tang X, Pei X, Lei N et al., 2020. Global patterns of soil autotrophic respiration and its relation to climate, soil and vegetation characteristics. Geoderma, 369: 114339. Cited in this article [1]

[44]	Wang J F, Xu C D, 2017. Geodetector: Principle and prospective. Acta Geographica Sinica, 72(1): 116-134. (in Chinese)

[45]	Wang L, Wang X, Chen L et al., 2021. Trade-off between soil moisture and species diversity in semi-arid steppes in the Loess Plateau of China. Science of the Total Environment, 750: 141646. Cited in this article [1]

[46]	Wang Y, Shataer R, Zhang Z et al., 2022. Evaluation and analysis of influencing factors of ecosystem service value change in Xinjiang under different land use types. Water, 14(9): 1424. Cited in this article [2]

[47]	Weisser W W, Roscher C, Meyer S T et al., 2017. Biodiversity effects on ecosystem functioning in a 15-year grassland experiment: Patterns, mechanisms, and open questions. Basic and Applied Ecology, 23: 1-73. Cited in this article [1]

[48]	Woodcock C E, Loveland T R, Herold M et al., 2020. Transitioning from change detection to monitoring with remote sensing: a paradigm shift. Remote Sensing of Environment, 238: 111558. Cited in this article [1]

[49]	Xie C, Wu S, Zhuang Q et al., 2022. Where anthropogenic activity occurs, anthropogenic activity dominates vegetation net primary productivity change. Remote Sensing, 14(5): 1092. Cited in this article [1]

[50]	Xu Q, Guo P, Jin M et al., 2021. Multi-scenario landscape ecological risk assessment based on Markov-FLUS composite model. Geomatics, Natural Hazards and Risk, 12(1): 1449-1466. Cited in this article [1]

[51]	Yang J, Huang X, 2021. The 30 m annual land cover dataset and its dynamics in China from 1990 to 2019. Earth System Science Data Discussions, 13(8): 3907-3925. Cited in this article [3]

[52]	Yao J, Chen Y, Guan X et al., 2022. Recent climate and hydrological changes in a mountain-basin system in Xinjiang, China. Earth-Science Reviews, 226: 103957. Cited in this article [3]

[53]	Zhang J, Hao X, Hao H et al., 2021. Climate change decreased net ecosystem productivity in the arid region of Central Asia. Remote Sensing, 13(21): 4449. Cited in this article [1]

[54]	Zhang X, Chen Y, Zhang Q et al., 2023. Potential evapotranspiration determines changes in the carbon sequestration capacity of forest and grass ecosystems in Xinjiang, northwest China. Global Ecology and Conservation, 48: e02737. Cited in this article [1]

[55]	Zhao H, Fan J, Gu B et al., 2024. Carbon sink response of terrestrial vegetation ecosystems in the Yangtze River delta and its driving mechanism. Journal of Geographical Sciences, 34(1): 112-130. Cited in this article [2]

[56]	Zheng H, Wu Y, He H et al., 2023. Urban land use eco-efficiency and improvement in the western region of China. Journal of Cleaner Production, 412: 137385. Cited in this article [1]

[57]	Zhu H, 2013. Underlying motivation for land use change: A case study on the variation of agricultural factor productivity in Xinjiang, China. Journal of Geographical Sciences, 23(6): 1041-1051. Cited in this article [1]

[58]	Zhu W, Pan Y, He H et al., 2006. Simulation of maximum light use efficiency for some typical vegetation types in China. Chinese Science Bulletin, 51: 457-463. Cited in this article [2]

[59]	Zinda J A, Trac, C J et al., 2017. Dual-function forests in the returning farmland to forest program and the flexibility of environmental policy in China. Geoforum, 78: 119-132. Cited in this article [1]

Funding

Key Laboratory Opening Foundation of Xinjiang Uygur Autonomous Region(2023D04048)

Shanghai Cooperation Organization (SCO) Science and Technology Partnership and International S&T Cooperation Program(2023E01022)

China Scholarship Council(CSC)

PDF(14161 KB)

184

Accesses

Citation

Detail

Sections

Recommended

Abstract
Key words
Cite this article
1 Introduction
Figure 1 Overview of study area and the study’s targeted climate factor changes, i.e., TMMX, TMMN, PRE, DEF, PDSI, SOIL, SRAD, PET, AET, VSP, VAP, and VPD, from 1990 to 2020, in Xinjiang, China (NO.: GS (2019) 3333)
2 Materials and methods
2.1 Study area
2.2 Climate, vegetation, and topographical data
Table 1 Data product type and source
2.3 Carnegie-Ames-Stanford approach model: net primary productivity algorithm
2.4 Super-SBM model: NPP efficiency algorithm
2.5 Random Forest approach
2.6 Partial least squares regression
2.7 Principle of GeoDectector
3 Results and discussion
3.1 Spatio-temporal variations of vegetation dynamics
3.1.1 Spatio-temporal variations of vegetation area
Figure 2 Land-use changes for different vegetation types in Xinjiang from 1990 to 2020 (a. Areas with increased vegetation types; b. areas with decreased vegetation types)
Table 2 Before and after transfer matrices for different land-use types from 1990 to 2020
3.1.2 Spatio-temporal dynamics of vegetation types
Figure 3 Transfer matrix of different land-use types to natural vegetation and spatial distribution of dynamics of different vegetation types in Xinjiang in 1990-2020 (a. spatial distribution of dynamics of different vegetation types from 1990 to 2020; b-j are the transfer matrix figures for different land-use types to natural vegetation; b. Xinjiang; c. mountainous area of Xinjiang; d. plains area of Xinjiang; e. northern Xinjiang; f. mountainous area of northern Xinjiang; g. plains area of northern Xinjiang; h. southern Xinjiang; i. mountainous area of southern Xinjiang; j. plains area of southern Xinjiang. Note: In this study, areas <1500 m are defined as plains and areas >1500 m are defined as mountains (Chen, 2014).)
3.2 Impact of climate change on vegetation dynamics (NPP)
3.2.1 Spatio-temporal patterns of vegetation dynamics (NPP)
Figure 4 Comparison of changes in NPP at the intersection of different vegetation types in Xinjiang from 1990 to 2020 (a. spatial changes in NPP in 1990; b. annual average value changes in NPP at the intersection of different vegetation types from 1990 to 2020; c. spatial changes in NPP in 2020)
Table 3 Comparison of NPP in Xinjiang in 1990 and 2020 for vegetation type change areas in land-use transition (g C)
Figure 5 Importance of climate factors to NPP changes of different vegetation types in Xinjiang from 1990 to 2020 (a. cropland; b. forest; c. grassland; d. wetlands; e. scrub)
3.2.2 Climatic interpretation of NPP by vegetation NPP efficiency
3.3 Climate attribution of changes in vegetation dynamics
3.3.1 Climate contribution to the overall vegetation dynamics in geographic patterns
Figure 6 Comparison of changes in NPP efficiency at the intersection of different vegetation types in Xinjiang from 1990 to 2020 (a. spatial changes in NPP efficiency in 1990; b. annual average value changes in NPP efficiency at the intersection of different vegetation types from 1990 to 2020; c. spatial changes in NPP efficiency in 2020)
Figure 7 Trends in the spatial distribution of climate factors in Xinjiang from 1990 to 2020 (a-l show spatio-temporal trends for TMMX, TMMN, PRE, DEF, PDSI, SOIL, SRAD, PET, AET, VSP, VAP, and VPD, respectively. Note: “+” indicates a positive trend, and “-” indicates a negative trend.)
Figure 8 Correlations between land-use area change of natural vegetation (Na-Veg) and climatic factors in Xinjiang (a), northern Xinjiang (b), and southern Xinjiang (c) (Notes: * indicates significance, p < 0.05, and ** indicates extreme significance, p < 0.01.)
Figure 9 Contribution rates of climate factors to area changes of different vegetation types in Xinjiang from 1990 to 2020 (a. natural vegetation; b. cropland; c. forest; d. grassland; e. wetlands; f. scrub)
3.3.2 Climatic trend analysis of vegetation dynamic intersection, increase, decrease, and change zones
Figure 10 Climate factor trends for the intersection of the same vegetation types in Xinjiang from 1990 to 2020 (a-l show trends for TMMX, TMMN, PRE, DEF, PDSI, SOIL, SRAD, PET, AET, VSP, VAP, and VPD, respectively. Note: “+” indicates a positive trend, and “-” indicates a negative trend.)
Figure 11 Regional climate factor trends for the increase zone of different vegetation types in Xinjiang from 1990 to 2020 (a-l show trends of different vegetation types in TMMX, TMMN, PRE, DEF, PDSI, SOIL, SRAD, PET, AET, VSP, VAP, and VPD, respectively. Note: “+” indicates a positive trend, and “-” indicates a negative trend.)
Figure 12 Regional climate factor trends showing the decrease zone of different vegetation types in Xinjiang from 1990 to 2020 (a-l show vegetation trends for TMMX, TMMN, PRE, DEF, PDSI, SOIL, SRAD, PET, AET, VSP, VAP, and VPD, respectively. Note: “+” indicates a positive trend, and “-” indicates a negative trend.)
Figure 13 Regional climate factor trends for the change zone of different vegetation types in Xinjiang from 1990 to 2020 (a-l show vegetation trends for TMMX, TMMN, PRE, DEF, PDSI, SOIL, SRAD, PET, AET, VSP, VAP, and VPD, respectively. Note: “+” indicates a positive trend, and “-” indicates a negative trend.)
4 Conclusions
References
Funding

Received	Accepted	Published
2023-09-07	2024-03-04	2024-06-25
Issue Date
2024-09-11

模态框（Modal）标题

Please choose a citation manager

Content to export

Abstract

Key words

Cite this article

1 Introduction

Figure 1 Overview of study area and the study’s targeted climate factor changes, i.e., TMMX, TMMN, PRE, DEF, PDSI, SOIL, SRAD, PET, AET, VSP, VAP, and VPD, from 1990 to 2020, in Xinjiang, China (NO.: GS (2019) 3333)

2 Materials and methods

2.1 Study area

2.2 Climate, vegetation, and topographical data

Table 1 Data product type and source

2.3 Carnegie-Ames-Stanford approach model: net primary productivity algorithm

2.4 Super-SBM model: NPP efficiency algorithm

2.5 Random Forest approach

2.6 Partial least squares regression

2.7 Principle of GeoDectector

3 Results and discussion

3.1 Spatio-temporal variations of vegetation dynamics

3.1.1 Spatio-temporal variations of vegetation area

Figure 2 Land-use changes for different vegetation types in Xinjiang from 1990 to 2020 (a. Areas with increased vegetation types; b. areas with decreased vegetation types)

Table 2 Before and after transfer matrices for different land-use types from 1990 to 2020

3.1.2 Spatio-temporal dynamics of vegetation types

3.2 Impact of climate change on vegetation dynamics (NPP)

3.2.1 Spatio-temporal patterns of vegetation dynamics (NPP)

Figure 4 Comparison of changes in NPP at the intersection of different vegetation types in Xinjiang from 1990 to 2020 (a. spatial changes in NPP in 1990; b. annual average value changes in NPP at the intersection of different vegetation types from 1990 to 2020; c. spatial changes in NPP in 2020)

Table 3 Comparison of NPP in Xinjiang in 1990 and 2020 for vegetation type change areas in land-use transition (g C)

Figure 5 Importance of climate factors to NPP changes of different vegetation types in Xinjiang from 1990 to 2020 (a. cropland; b. forest; c. grassland; d. wetlands; e. scrub)

3.2.2 Climatic interpretation of NPP by vegetation NPP efficiency

3.3 Climate attribution of changes in vegetation dynamics

3.3.1 Climate contribution to the overall vegetation dynamics in geographic patterns

Figure 7 Trends in the spatial distribution of climate factors in Xinjiang from 1990 to 2020 (a-l show spatio-temporal trends for TMMX, TMMN, PRE, DEF, PDSI, SOIL, SRAD, PET, AET, VSP, VAP, and VPD, respectively. Note: “+” indicates a positive trend, and “-” indicates a negative trend.)

Figure 8 Correlations between land-use area change of natural vegetation (Na-Veg) and climatic factors in Xinjiang (a), northern Xinjiang (b), and southern Xinjiang (c) (Notes: * indicates significance, p < 0.05, and ** indicates extreme significance, p < 0.01.)

Figure 9 Contribution rates of climate factors to area changes of different vegetation types in Xinjiang from 1990 to 2020 (a. natural vegetation; b. cropland; c. forest; d. grassland; e. wetlands; f. scrub)

3.3.2 Climatic trend analysis of vegetation dynamic intersection, increase, decrease, and change zones

Figure 10 Climate factor trends for the intersection of the same vegetation types in Xinjiang from 1990 to 2020 (a-l show trends for TMMX, TMMN, PRE, DEF, PDSI, SOIL, SRAD, PET, AET, VSP, VAP, and VPD, respectively. Note: “+” indicates a positive trend, and “-” indicates a negative trend.)

4 Conclusions

{{custom_sec.title}}

{{custom_sec.title}}

References

{{custom_fnGroup.title_en}}

Footnotes

Funding