*2.1. Study Area*

The Loess Plateau lies in the north of China, covering an area of 62.4 × 10<sup>4</sup> km<sup>2</sup> (Figure 1). The region is dominated by a continental monsoon climate. The annual accumulated temperature ≥10 ◦C (annual AT10) increases from ~50 ◦C·d in the high-elevation western part to ~5800 ◦C·d in the southern part. Additionally, the annual mean precipitation varies from 50 mm in the northwest to 700 mm in the southeast. The terrain of the Loess Plateau varies significantly, and the altitude ranges from 80 m in the southeast to 5200 m in the west. Land use data with a spatial resolution of 500 m from the MODIS land cover product (MCD12Q1) show that there are seven land cover types in the entire study area; these are grasslands (68.5%), croplands (21.2%), forests (5.1%), barren regions (3.3%), urban and built-up areas (1.8%), and water bodies (0.1%).

**Figure 1.** (**a**) Location of the Loess Plateau; (**b**) digital elevation model (DEM) showing the location of the Weihe Plain, Lvliang Mountains, Mu Us Sandy Land, Hetao Plain, and Qilian Mountains; (**c**) the pattern of accumulated temperature ≥ 10 ◦C (AT10); and (**d**) land cover types in 2001.

#### *2.2. Data Resources and Preprocessing*

In this study, we used four types of satellite data: MODIS NDVI with spatial resolutions of 250, 500, and 1000 m and GIMMS3g NDVI with a spatial resolution of ~10 km. These data have been widely applied for vegetation phenology extraction at regional to global scales [43–45]. MODIS data for the period 2001–2020 were acquired from NASA (ftp://ladsweb.nascom.nasa.gov/allData/6/, accessed on 15 April 2021). MODIS NDVI with a temporal resolution of 16 days is a gridded level 3 product. GIMMS3g NDVI was obtained from the NASA Earth Exchange platform for the period 1982–2015 (https://nex.nasa.gov/nex/, accessed on 15 April 2021). The spatial and temporal resolution of the GIMMS3g NDVI was 15 days and 1/12 degree (~10 km), respectively. Both the MODIS and GIMMS3g time-series datasets were used to identify the start of the growing season (SOS) and the end of the growing season (EOS).

The daily mean air temperature with a spatial resolution of 0.25 degrees for the period 1982–2020 was obtained from the Copernicus Climate Change Service Climate Data Store (CDS) (https://cds.climate.copernicus.eu/#!/home, accessed on 19 June 2021). The daily air temperature was used to obtain AT10 in order to analyze the interaction between phenology and temperature.

The annual MODIS land cover type product at a 500 m spatial resolution was used to examine the influence of vegetation type on the LSP. The classification scheme that is used for the product is the International Geosphere–Biosphere Program (IGBP). Additionally, the year used for the latest MODIS land cover type data was 2019. The land cover type product for the period 2001–2019 was obtained from NASA's Land Processes Distributed Active Archive Center (LP DAAC) (https://lpdaac.usgs.gov/products/mcd12q1v006 /, accessed on 19 June 2021). The spatial resolution of the product was resampled to 250 m, 1000 m, and 1/12 degree using nearest-neighbor interpolation. The Shuttle Radar Topography Mission (SRTM) digital elevation model (DEM) dataset was achieved at the

CGIAR (https://srtm.csi.cgiar.org/, accessed on 11 July 2021). The DEM dataset was used to understand the distribution of the topography in the Loess Plateau. The terrain of the central and western parts of the Loess Plateau is heterogeneous, while the terrain of the Weihe Plain is flat [46].

The 271 detailed ground-observed records in the Loess Plateau were obtained from 14 agro-meteorological stations from 2001 to 2013 (a small set of data with low quality was rejected). The distributions and information of the ground-observed sites can be found in Figure 1a and Table 1. The observed records were used to validate the vegetation phenology that was derived from MODIS data and more information could be used as prior knowledge of the research area [47].


**Table 1.** The description of the ground-observed sites.
