2) The RESTREND method
We used the residual trend analysis (RESTREND) method to explore the degree and scope of the effect of anthropogenic and climatic factors on regional vegetation (Evans and Geerken, 2004; Ma et al., 2017).
NPP-A, which is directly observed by remote sensing and selected from a set of remote sensing datasets, monitors the real growth conditions of local vegetation. NPP-A records the combined effects of human activity and climate change.
NPP-P is calculated from the actual annual evapotranspiration measurements, and this index represents the NPP value under ideal conditions with no human activity; that is, the NPP-P value influenced only by climatic factors. The change in NPP-P can be interpreted as reflecting a climatic control on vegetation dynamics. A positive value of the NPP-P change rate indicates that climate change is beneficial to vegetation recovery, whereas a negative value indicates a detrimental effect.
The residuals of (i.e., differences between) the NPP-Aand NPP-P values are defined as the residual NPP value (NPP-RES) and can be interpreted as reflecting the influence of human activity on the value of NPP:
NPP-RES = (NPP-P) – (NPP-A)
Using the RESTREND method, the NPP-p reflects the absence of anthropogenic influence, while the NPP-RES value represents only the extent to which human activity contributes to or damages regional vegetation recovery, and is interpreted in terms of the anthropogenic influence on vegetation dynamics. A positive value of the NPP-RESregression slope indicates that human activity has been conducive to vegetation recovery, whereas a negative value indicates that human activity has caused degradation.
The slope of the linear regression fit of the NDVI, NPP-P, and NPP-RES time series can be either >0 or <0. Taking all permutations of the NDVI, NPP-P, and NPP-RES slope values into account, we can generate eight potential scenarios that reflect the main driving factors controlling vegetation change and these are detailed in Table 1.

3 Results