The Qinling-Daba Mountains, composed of the Qinling Mountains in the north, the Hanzhong Basin-Hanshui Valley in the middle, and the Daba Mountain in the south, is situated in central China (102°–114°E, 30°–36°N), covering a total area of about 30.60 × 10⁴ km² (Qin et al., 2008), and is known as the transitional zone of China (Kou et al., 2020; Yao et al., 2020) (Figure 1). It extends 1,000 km in the east–west direction and 200–300 km in the south–north direction, covering 155 counties, 31 cities and 6 provinces in central China (Figure 1). As the north–south transitional zone in China, steep elevation gradients and complex climate make it a biodiversity hotspot, the vegetation in this region gradually changes from subtropical evergreen broadleaf forest to deciduous broadleaf forest from south to north, and has vertical zonality in the mountains (Liu and Lu, 1990); it is also an important habitat for rare animals, containing many nature reserves and national parks (Yu et al., 2022). Therefore, this area is one of the most important and concerned areas for biodiversity in China, and one of the most sensitive areas to climate change and human activities (Yao and Cui, 2022; Zhang et al., 2019).

The NDVI dataset used in this study was the annual growing season (May to September) NDVI data (30 m resolution) of Landsat 5/ Landsat 7/ Landsat 8 from 1986 to 2019, which were synthesized using the maximum synthesis method on the Google Earth Engine (GEE) platform. Savitzky–Golay (SG) filtering was applied to the annual growing season NDVI data to further reduce the noise (Kou, 2021). Temperature and precipitation data with 500 m resolution were downloaded from the Data Center of Resources and Environment Science, Chinese Academy of Sciences¹ for 1980–2015, which were generated by 2,400 meteorological stations using the spatial interpolation method (Figures 2A,B). The ASTER GDEM (Advanced Spaceborne Thermal Emission and Reflection Radiometer Global Digital Elevation Model, downloaded from https://earthdata.nasa.gov/) with 30 m resolution was mainly used to analyze the pattern and change of NDVI with variable altitude. Land cover data sets (1 km resolution) for 1990, 1995, 2000, 2005, 2010 and 2015 were downloaded from the Data Center of Resources and Environment Science, Chinese Academy of Sciences (see text footnote 1) (Figure 2C). The soil type data (Figure 2D), population density (1990–2015), and GDP data (1995–2015) (all with 1 km resolution) were also downloaded from the Data Center of Resources and Environment Science, Chinese Academy of Sciences (see text footnote 1). Among of them, the soil type data were digitally generated according to the “1: 1,000,000 Soil Map of the People’s Republic of China” compiled and published by the National Soil Census Office in 1995 (Figure 2D); the population density data and GDP data were based on the statistics of population and GDP by county, using the multi-factor weight distribution method to calculate the distribution weights of land use type, night light brightness, residential area density, and other related factors (Xu and Zhang, 2017) (Figures 2E,F). These data were used to analyze the influencing factors on NDVI pattern and change.

Firstly, the spatial patterns and dynamics of NDVI were analyzed. The Sen trend method (Sen, 1968) and Mann-Kendall (MK) significant test were used to analyze the NDVI trend for 1986–2019. The Sen trend method can effectively avoid the influence of time series data loss and data distribution form, and eliminate the interference of time series outliers (Liu et al., 2010), and the MK significant test (Mann, 1945; Kendall, 1975) was conducted to test the significance of the calculated Sen trend. The MK mutation test (Karpouzos et al., 2010) was used to determine the mutated NDVI change period and to discover the dynamic process of NDVI. The spatial patterns of NDVI along latitude, longitude and elevation in the Qinling-Daba Mountains were investigated by profile analysis (along 33.6° N and 107° E) and statistical analysis methods.

Then, the main environmental influencing factors of NDVI pattern and its change were investigated by Geodetector analysis. As we know, climatic factors such as temperature, precipitation, light and seasonal variation are the main factors influencing vegetation distribution, and soil factors such as soil type, texture, pH, and nutrient content have important effects on vegetation growth (Brady and Weil, 2008; Marschner, 2012). Topography regulates the local re-distribution of precipitation, soil moisture, and solar radiation, which in turn affect the distribution of vegetation (Bonan, 2015; Turner et al., 2001). Therefore, four factors, including soil type, DEM (as a combination of topographical factors such as elevation and slope), temperature and precipitation, were selected as the regional environmental influencing factors of NDVI pattern. Previous studies have found that climate change was the main driving factor for NDVI changes, although human activities also had important effects on NDVI changes, but it was weaker than that of these climate factors (Pei et al., 2019; Yang and Han, 2019; Tao et al., 2020; Zhang et al., 2020). Therefore, temperature, precipitation, and three human activity factors including population density, GDP, and land use type (Table 1) were selected to reveal the driving factors of NDVI change in the Qinling-Daba Mountains based on the above mentioned data every five years. As the input data of the Geodetector requires classified data, all the selected factors were classified into the classified data with 9 categories by the natural breakpoint method.

The Geodetector method was constructed on the assumption that when an independent variable has an important effect on a dependent variable, the spatial distribution of the independent variable and the dependent variable should be similar (Wang and Hu, 2012; Wang et al., 2010; Wang and Xu, 2017). It can quantitatively express the spatial stratification heterogeneity of the research object by analyzing the similarities and differences between the intra-layer variance and the inter-layer variance (Hu et al., 2011; Wang et al., 2013). At present, it has been widely used to detect the driving factors in many studies, such as land use (Hu et al., 2011), public health (Wang et al., 2013), regional economy (Ding et al., 2014), regional planning (Liu and Yang, 2012; Yang et al., 2016), meteorology and environment (Du et al., 2016), and vegetation change (Peng et al., 2019; Wang J. et al., 2019; Wang W. et al., 2019). Therefore, this study used the Geodetector method to detect the influencing factors of vegetation pattern and change. The Q-statistic in Geodetector can be used to measure spatial stratified heterogeneity, detect explanatory factors, and analyze the interactive relationship between variables (Wang and Xu, 2017). The range of the Q-statistic is [0, 1], and a larger value of the Q-statistic indicates that the independent variable has the stronger explanatory power on the dependent variable (Wang and Xu, 2017). At the extremes, a Q-statistic value of 1 indicates that the independent variable (X) completely controls the spatial distribution of the dependent variable (Y), and a Q-statistic value of 0 indicates that the independent variable of X has no relationship with the dependent variable of Y. The P-statistic, which corresponds to the Q-statistic of the independent variable (X), can be used to represent the significance of the variable X on the dependent variable Y. For example, a P-statistic of less than 0.05 means that the effect of the variable X on the dependent variable Y is significant, and a P-statistic of less than 0.01 means that the effect of the variable X on the dependent variable Y is highly significant, which can be interpreted as the smaller the P-statistic, the greater the reliability of the inference that a certain type of independent variable X has an effect on the dependent variable Y.

The Q-statistics of the four factors on vegetation cover pattern were ranked as soil type (X2) > topography (X1) > annual average temperature (X3) > annual precipitation (X4) (Table 2), which showed that physical environmental factors (soil and topography) played stronger influencing roles on the NDVI spatial pattern than climate factors in Qinling-Daba Mountains.

Based on the Geodetector analysis of environmental factors on NDVI change, the explanatory power (Q-statistic) of each factor was ranked as follows: land use type (X7) > annual average temperature (X3) > annual precipitation (X4) > population density (X5) > GDP (X6) (Table 3). This result showed that land use and temperature had stronger effects on regional NDVI changes than precipitation, and the effect of land use was stronger than that of temperature. The effects of population density (X6) and GDP (X7) were relatively weak (Q-statistic less than 0.06).

When studying NDVI changes and patterns, climate and anthropogenic factors are often considered, but other environmental factors such as soil and topography are often neglected. It is well known that soil properties such as soil type, texture, pH, and nutrient content have important effects on vegetation growth (Brady and Weil, 2008; Marschner, 2012), and topography redistributes local hydrothermal conditions and soil nutrients, resulting in the changes in temperature, precipitation and vegetation along elevation (Bonan, 2015; Turner et al., 2001; Zhang et al., 2009). In this study, although temperature and precipitation played important roles in the distribution and growth of vegetation, their effects on the spatial pattern of NDVI at the regional scale were weaker than soil and topography. In particular, the effect of precipitation on NDVI pattern was weaker than that of temperature in the study area. Therefore, when studying the influencing factors of NDVI patterns at different scales, we should fully consider the effects of various environmental factors.

A recent study showed that the “Greening Earth” was attributed to human land use practices in China and India (Chen et al., 2019a,b). This study (Table 3) and other related studies on NDVI change in the Qinling-Daba Mountains also showed that land use had a great influence on NDVI change (Yao and Cui, 2022; Cui et al., 2012; Sun et al., 2010). The Qinling-Daba Mountains is not only an ecological functional area for biodiversity conservation in China, but also a water conservation area for the “South to North Water Transfer Project” in China. Many nature reserves (over 30), national forest parks (about 37), national geological parks (11), and scenic spots (over 7) have been established in the study area from the 1960s to the present. The good condition of the vegetation was one of the achievements of these environmental protections. In addition, the rapid increase of NDVI in the areas below 1,500 m was partly contributed by the land use policies (Chen et al., 2019a,b; Yao and Cui, 2022). The Chinese government issued the Grain-for-Green policy in 1999–2000, and local governments formulated strict implementation measures (Chen et al., 2019a,b; Chen et al., 2006; Zhang et al., 2010). One of the achievements of Grain-for-Green was that croplands in mountainous areas with slopes steeper than 25° were required to be returned to forest (or grassland), and those with slopes between 15° and 25° were conditionally returned to forest or grassland (Chen et al., 2006; Zhang et al., 2010). During the field survey of the “Comprehensive Scientific Investigation of the North–South Transitional Zone” project, it was also found that a large number of croplands in the mountainous areas below 1,500 m were returned to forest. Moreover, the croplands also contributed to the increase in NDVI due to the rapid growth of hybrid cultivars, multiple cropping, irrigation, fertilizer use, pest control, improved seed quality, farm mechanization, credit availability, and crop insurance schemes (Chen et al., 2019a,b; Yao and Cui, 2022). Therefore, places below 1,500 m have higher NDVI increases, and the breakthrough increase period was around in 2005 (Figure 3). All these indicate that land use in the Qinling-Daba Mountains has had a positive effect on vegetation dynamics in recent years, and its effect on NDVI change was even stronger than that of climate warming. The temperature of the study area has been warming significantly, while the precipitation has been increasing slightly (Yao and Cui, 2022). That is why the effect of precipitation was weaker than that of land use and temperature on NDVI change in this area.

Due to the lack of high-resolution quantitative data on human activities (Xie and Yao, 2024), most quantitative analyses of human activities have focused on land use, population density, or GDP. However, except for land use data, the resolution and quality of population density and GDP data are not good enough to characterize human activities. As a result, the results obtained are not satisfactory. Additionally, infrastructure construction such as transportation and roads, social and economic development, and urbanization also affect the NDVI change. Therefore, which indicators can better reflect the impact of human activities on NDVI, especially how to objectively evaluate the impact of human activities on NDVI, remains to be further explored. Moreover, human activities have greatly affected every aspect of the Earth and have a profound influence on vegetation change, so there is a need for a more comprehensive indicator that can synthesize the various human activities. Recently, quantitative methods for assessing and analyzing the impact of human activities on the natural environment, as well as data products, have also developed rapidly. Human activity intensity (HAI) has been widely used to assess and quantify the impacts of human activities on landscapes (Goudie, 2018; Shrestha et al., 2021). There are some useful HAI data products such as HAI data (Xie and Yao, 2024), the global human footprint map (Sanderson et al., 2002; Venter et al., 2016; Mu et al., 2022) and the wildness map (Lin et al., 2016; Cao et al., 2019), which can greatly facilitate our analysis of the impact of human activities on NDVI.