Altay Prefecture (85°31′36″- 91°04′23″E, 45°00′00″- 49°10′45″N) is located in northern Xinjiang, China, and includes the eight counties and cities of Altay, Jimunai, Fuyun and others, with a total area of 1.18×10⁵ km² (Fu et al., 2017) (Fig. 1). It extends from the northern part of the Gurbantunggut Desert in the south to the southern portion of the Altay Mountains. The center of Altay Prefecture is the Irtysh River Valley, and the terrain is higher in the northern than the southern altitudes. This area has a temperate continental monsoon climate, and is a typical arid and semiarid region with limited precipitation and cold weather. Specifically, the annual average temperature and sunlight hours in the north of Altay Prefecture are lower than those in the south, while on the contrary, the annual precipitation in the north is higher than in the south (Zhang et al., 2019). Affected by hydrothermal conditions, climatic change and human activities, the land cover in 2020 is mainly cropland (1.91%), forest (4.92%), grassland (11.39%), meadow (10.94%) and desert (61.84%).

In this study, we adopted the temperature, precipitation and sunlight hours to reflect the influence of meteorological factors on vegetation variation. As displayed in Fig. 5, these three meteorological factors in Altay Prefecture demonstrated heterogeneous geographical distributions, and their spatial distributions are extremely uneven. The annual average temperature and annual sunlight hours are lower in the mountains than in the river basin (Fig. 5 A1 and B1). However, the distribution of annual precipitation is inverted (Fig. 5 C1). Over the past 20 years, the temperature in most of area presented an increasing trend with fluctuations (91.7%, 0.023 ℃ yr^‒1, Fig. 5 A3 and A4), and the high slopes are mainly in the southeast and northwest (Fig. 5 A2). In addition, the precipitation of most regions (61.2%) experienced a downward trend, while 38.8% of the study regions displayed increasing trends (Fig. 5 B4). On the whole, the precipitation showed a slight increase (0.0984 mm yr^‒1, Fig. 5 B3), and the high positive trend was mainly concentrated in the northwest, while the low negative trend was mainly concentrated in the southeast. Furthermore, compared with the temperature and precipitation, the sunlight hours in all regions showed a decreasing trend (15.489 h yr^‒1, Fig. 5 C3 and C4), and the sharpest declining trends were primarily concentrated in the north (Fig. 5 C2).

Figure 6 shows spatial correlations between NDVI and the meteorological factors of temperature, precipitation and sunlight hours. In order to provide a more visual view of the effects of climatic factors on NDVI in different regions, we used the Maxwell color triangle (RGB) to compound the significant spatial correlations (P < 0.05) for the effects of the three meteorological factors on NDVI (Fig. 7). We found the significant correlations between NDVI and meteorological factors were mainly distributed in the north, however, except for air temperature, the significant correlation almost did not occur at all in the south. The relationships between NDVI and air temperature were mainly positive correlations (accounting for more than 50%), but the area significantly correlated with air temperature was only 3.3%, and among them, significant positive correlations were mainly concentrated in the southern region, and significant negative correlations are mainly concentrated in the northern region (P<0.05, Fig. 6a). Similarly, the relationships between NDVI and precipitation were mainly positive correlations (accounting for more than 60%), with only 6.9% of the regions showing significant correlations, and among them, 6.6% showed significant positive correlations which were chiefly concentrated in the northwest (r > 0, P < 0.05, Fig. 6b). In contrast, negative correlations were the main relationship between NDVI and sunlight hours, which accounted for about 75%, and the area significantly affected by sunlight hours accounted for 20.3%, with 19.97% showing a significant negative correlation and mainly located in the northeast and northwest (r < 0, P < 0.05, Fig. 6c).

Climate change and diverse human activities could be responsible for the vegetation variations. In this study, we used the threshold segmentation method to determine the contributions of climate change and human activities to the observed vegetation variations. As shown in Fig. 8, in Altay Prefecture during the time-frame of 2000-2019, zones of vegetation with significant restoration accounted for 24.08%, whereas 1.24% of the regions showed significant degradation. The vegetation with significant restoration caused by multiple factors accounted for 18.94% of the entire region, and this phenomenon was widely distributed in Altay, especially in the south of Fuhai County. Zones of vegetation with significant restoration triggered by human activity accounted for 3.4% of the entire region, and were predominately distributed in the Irtysh River Valley. The effects of climate change resulted in the vegetation with significant restoration in a relatively small area (1.74%), and mainly occurred in the northwest of Altay. Simultaneously, zones of vegetation with degradation caused by climate change accounted for 1.07%, and were unevenly distributed in sporadic areas of Altay. However, there were very few areas of anomalous degradation (0.17%), which can be explained by human activities.

Overall, the multi-year (2000-2019) average NDVI in Altay Prefecture presented a significant spatial heterogeneity (Fig. 3a), with a gradient of variation which decreased from north to south that was consistent with the regional elevation changes. Meanwhile, the NDVI increased in most regions of the Altay Prefecture, suggesting that vegetation variation in Altay demonstrated an overall increasing tendency from 2000 to 2019 (Slope > 0, Fig. 3). This result is similar with the observations of vegetation turning green from ongoing studies in Eurasia, arid and semiarid regions and other areas (Zhao et al., 2011). Vegetation variation is inseparable from the effects of climatic changes, anthropogenic activities and geography. In recent years, a great deal of surveillance and research have verified that northwestern China is becoming warmer and wetter (Peng and Zhou 2017; Yang et al., 2017), and this result was demonstrated by the trend analysis of climatic factors in this study. The beneficial hydrothermal conditions provided an advantageous foundation for vegetation growth (Ye et al., 2020). Some artificially implemented positive policies in Altay Prefecture could increase the regional NDVI, such as returning the grain plots to forestry (Li et al., 2020b), reclaiming the oasis in a desert (Su et al., 2007) and the “Three North” Shelterbelt Project (Deng et al., 2019). Concurrently, the grazing and construction of water conservancy facilities also profoundly impact the vegetation of the Altay Prefecture. Previous studies suggested that artificial oases created by agricultural activities contributed to the improvement of vegetation cover (Piao et al., 2003), and the obvious vegetation improvement areas in Altay were mainly concentrated in the oases of the central river basin, although they also have sporadic distribution in the mountainous areas. Meanwhile, with the passage of time, an increase in NDVI also occurred in the areas surrounding the oasis. Such vegetation dynamic variations have also appeared in the relevant studies of Xinjiang (Du et al., 2015).

Climate change is one of the important factors affecting vegetation dynamic variations. The hydrothermal condition is the primary limiting factor for vegetation growth (Ye et al., 2020), which is also affected by radiation (Yao et al., 2018). Vegetation variation (Fig. 3a) and its response to climate change (Fig. 7) in Altay Prefecture can be explained from the following aspects. An appropriate temperature increase can promote the enzymatic activities related to photosynthesis (Davidson and Janssens, 2006; Sun et al., 2020), accelerate the decomposition of organic matter (Kim et al., 2012) and extend the growth period of vegetation (Piao et al., 2007), with an evident promotion of photosynthesis. But higher temperatures will significantly affect the enzymatic activities of photosynthesis (Bao et al., 2015) as well as the transpiration of vegetation and the transport of nutrients (Bao et al., 2014), thus inhibiting the growth of vegetation. The increase of precipitation can increase soil moisture content, and soil water is a necessary medium for guaranteeing soil nutrient transportation (Sun et al., 2020), so when light and CO₂ concentrations are sufficient, the soil water content determines the rate of photosynthesis. Therefore, the increase of precipitation played a crucial role in the restoration of vegetation. In addition, the sunlight hours can affect the amount of solar radiation the vegetation receives, and the radiation input would change the anaerobic soil conditions caused by high soil moisture content, thus indirectly affecting the photosynthesis of vegetation. In arid and semi-arid regions, precipitation is the main factor affecting vegetation growth, while temperature has little influence, which can explain why the area of NDVI is more significantly affected by precipitation than temperature in Altay (Fig. 6). The continuous increase in temperature may promote the circulation of the atmosphere, leading to an increase in precipitation (Qin et al., 2005), which leads to the hydrothermal synchronization in the Altay Prefecture (Fig. 5). Taking Jimunai County as an example, due to beneficial hydrothermal conditions, the NDVI of 58.68% of the areas improved, and significant improvement accounts for 19.95%. However, 5.39% of degradation occurred in this region, which may be caused by the fact that although the increase of precipitation improved the drought stress of the vegetation (Stefanie et al., 2017), the continuous increase of temperature also led to more intense evaporation in the arid region, which reduced the soil moisture content and inhibited the growth of vegetation (Shi et al., 2014). For northern Altay in the mountains, the regional NDVI showed overall improvement which is mainly controlled by multiple factors, but since human activity is relatively difficult in the mountains, the NDVI dynamics is dominated by climate change. This may be due to rising temperatures in the region, leading to the permafrost melting, the precipitation increased and sunshine time decreased, the soil moisture was supplemented, the germination period of the vegetation was advanced, and the deciduous period was pushed back, resulting in an increase in the vegetation growth period (Oliva et al., 2018), which was conducive to the vegetation prosperity. In addition, the vegetation type in this region is primarily coniferous forest, which absorbs water mainly through shallow root systems and does not depend on deep groundwater, so the increase of soil water is more conducive to its growth (Jiang et al., 2017). The southern region of Altay is mainly desert with poor vegetation coverage, so in the past 20 years, as the temperature rose and the precipitation decreased, it was difficult to improve the vegetation situation, but it was also difficult for it to undergo extreme degradation relative to its initially poor state.

Although NDVI is closely related to climatic factors, the influence of human activities cannot be ignored. Human activities can change the vegetation cover in a short period of time, and their impact is more significant. The impacts of human beings on vegetation mainly include two types: one is improvement including afforestation, returning the grain plots to forestry and reclaiming oasis in desert, etc.; and the other is destruction including overgrazing, construction of reservoirs and deforestation. Studies have shown that the afforestation projects widely implemented around the world have played a positive role in vegetation restoration, improvement of ecosystem services (Gao and Bian, 2019) and soil erosion control (Wang et al., 2014). In addition, grazing is one of the main ways of grassland utilization, and livestock activities can change the biomass, soil structure and nutrients of the grassland. Therefore, long-term overgrazing will also reduce the resilience of grassland vegetation and increase the risk of land degradation (Hao et al., 2018). Combined with the analysis of the results obtained here, we found that the reclamation of the Irtysh River Valley had a remarkable effect: The large area of cultivated land has been increased and the original desert area has been turned into oasis, so the NDVI is worth improving. According to the land use data, the cultivated land in Altay Prefecture is mainly distributed in this region. In 2000, the total area of arable land in Altay Prefecture was 4864.35 km², which increased to 6201.58 km² in 2020, for an increase of 27.5%. From the perspective of stocking rate, the overall livestock amount in Altay Prefecture showed a decreasing trend from 2000 to 2019 (Fig. 9), while the actual NDVI during this period showed an increasing trend. Therefore, the decline in stocking rate and the improvement of land use practices have exerted profound effects on the vegetation variation.

In order to verify the reliability of driving force separation by the significance threshold segmentation method, especially separating the influences of climate factors and human factors, we compared the NDVI values of the human factors leading to

significant changes in area (Fig. 10). Based on the previous assumption, the disturbance of climate change has no significant effect on vegetation change in a short time, but human activity or sudden disasters will lead to significant changes in the vegetation coverage condition. If the vegetation is significantly improved or degraded by human factors, then the NDVI value of that area will also have a significant change. In Fig. 10, three verification zones were selected to verify the significant changes led by human factors. Regions A and C are the areas with increased cultivated land near the river, and the result of driving force separation is the significant restoration triggered by human activities. In contrast, region B is the man-made reservoir, and the result of driving force separation is abnormal degradation.

As can be seen from Fig. 10, the NDVI values of validation regions A and C have significantly increased in the past 20 years, while those of validation area B have significantly decreased, indicating that the vegetation coverage of validation areas A and C show significant restoration, while the vegetation coverage of validation area B shows significant degradation. Therefore, the method of significance threshold segmentation is reliable for separating the driving forces.

In this study, we analyzed the spatiotemporal distribution characteristics of vegetation in the Altay Prefecture and its responses to climate change and human activities, including quantitatively separating the contributions of climate change and human activities to the observed vegetation changes. Nevertheless, there are some deficiencies in this study. First of all, in addition to the three meteorological factors selected, other factors, such as soil nutrients, soil microorganisms and carbon dioxide concentration (Zhang et al., 2021), will also have a decisive impact on vegetation changes. Therefore, in future studies, more attention should be paid to the influences of soil properties and other factors on vegetation change. Of course, the influences of evaporation (Hoffman et al., 2011), acquisition date, sensors, etc., on the inter-annual NDVI should also be considered in future research. Also, although the NDVI index can reflect the vegetation productivity to a certain extent (Piao et al., 2005), the net primary production (NPP) can better reflect the situation of vegetation photosynthesis and ecosystem vitality. As a result, NPP is more sensitive to climate change and human activities and can more accurately reflect the vegetation change response mechanism (Liu et al., 2019), thus NPP can be used to reflect ecosystem productivity in future studies. In addition, this study only conducted a simple quantitative separation and analysis of the driving factors, and subsequent in-depth research on the influencing mechanisms of each driving factor can be carried out, to further clarify the size and change of their contribution rates.