3.1 Data sources and preprocessing
The precipitation data used in this study were from the Climate Forecast System Reanalysis (CFSR) (https://globalweather.tamu.edu), which provides daily meteorological indicators for specific latitudes and longitudes, selected from 1 January 1979 to 31 December 2013. The soil data were obtained from the Harmonized World Soil Database (HWSD) (http://webarchive.iiasa.ac.at/Research/LUC/), which was in a grid format with an original resolution of approximately 30 arc seconds. The topography data were provided by Geospatial Data Cloud site, Computer Network Information Center, Chinese Academy of Sciences (http://www.gscloud.cn). ASTER GDEM V2 version of the original elevation data was selected, and the original spatial resolution was 30 m. Elevation data can be used to extract slope in ArcGIS 10.2. The land use data were derived from the interpretation of remote sensing images. According to the research needs, Landsat 5 TM images in 2000 and Landsat 8 OLI_TIRS images in 2015 were selected as the base map for interpretation. The source of remote sensing images data was the same as topography data, and the original spatial resolution was 30m. Land use status in 2000 and 2015 were obtained through manual interpretation (Zhang et al., 2018). The enhanced vegetation index (EVI) data from 2000 to 2016 were obtained from synthetic vegetation index product MOD13Q1 V6 with an original resolution of 250m (Didan, 2015), and the average EVI from the 97th to 289th days (plant growth season) of each year is considered the year’s EVI (Xu et al., 2020). The population data were from the Statistical Yearbook of Yan’an (http://data.cnki.net), and the resident population of each county in 2015 were selected as the calculation standard.
It should be noted that, (1) all data were presented in grids with a resolution of 90m × 90m, (2) the area with a slope below 6° was ignored because the GFG was hardly carried out in such an area (Zhang et al., 2019).