**2. Data and Methods**

### *2.1. Study Area*

Southwest China covers a region bounded by 20◦54 –34◦19 N and 97◦21 –112◦04 E (Figure 1). It includes five provinces (municipalities, autonomous regions): Chongqing, Sichuan, Guizhou, Yunnan, and Guangxi. The area has a total coverage of 1.38 × 106 km2, which is about 14.34% of China. The study area is dominated by the subtropical monsoon climate, affected by the southeast and the southwest monsoon, with a mean annual temperature of 14.6 ◦C and an average annual precipitation of 1195 mm, of which 600 mm occurs in the growing season. The terrain is high in the west and low in the east, with a stepped distribution, diverse landform types, and high landscape heterogeneity [48]. Southwest China is the most concentrated area of karst landforms, accounting for 23.14% of China's karst distribution area. The forest has a large area and is widely distributed in the study area. Grassland is mainly distributed in the western Sichuan Plateau, and cropland is mainly distributed in the Sichuan Basin, the middle of the Guangxi hills, and the eastern part of the Yunnan–Guizhou Plateau. The three main LULC types are grassland, forest, and cropland, accounting for more than 98% of the total area of southwest China.

**Figure 1.** Study area location in China and land cover types in 2019.

#### *2.2. Data*

The normalized difference vegetation index (NDVI) data came from the study derived from the MOD13A2 data product, with a spatial resolution of 1 km and a temporal resolution of 16d (https://ladsweb.modaps.eosdis.nasa.gov/ (accessed on 6 August 2021)). These data were preprocessed on the Google Earth engine platform (https: //earthengine.google.com (accessed on 6 August 2021)), and the NDVI data from May to September 2000–2019 were downloaded. The gross primary productivity (GPP) datasets were obtained from the MOD17A2HV6 data product with a spatial resolution of 500 m and a temporal resolution of 8d (https://lpdaac.usgs.gov/products/mod17a2hv006/ (accessed on 28 August 2021)), and we downloaded the contemporaneous datasets from the Google Earth engine platform (https://earthengine.google.com (accessed on 28 August 2021)). The LULC data came from the first Landsat-derived annual China Land Cover Dataset

(CLCD) from 1990 to 2019, with a spatial resolution of 30 m and an overall accuracy rate of 79.31%, better than MCD12Q1, ESACCI\_LC, FROM\_GLC, and GlobaLand30. The LULC types in these data were divided into nine categories: cropland, forest, shrub, grassland, water, snow/ice, barren, impervious, and wetland. The CLCD dataset introduced in this article is freely available at http://doi.org/10.5281/zenodo.4417810 (accessed on 10 August 2021) [49].
