*2.3. Methods*

The normalized difference vegetation index (NDVI) can reflect the growth and coverage of surface vegetation [50,51]. Therefore, the NDVI values of different LULC types were extracted to represent the growth state of this vegetation type. Vegetation has obvious inter-annual and seasonal variation characteristics, and its NDVI value in the season of the most active growth stage can more accurately represent the vegetation growth state in this region [52]. In order to reduce the NDVI error caused by the seasonal changes in vegetation, this paper used NDVI and GPP values in the growing season (May–October) to analyze vegetation activities, which accurately reflected the status of vegetation cover and GPP in southwest China. To minimize the effect of cloud contamination and atmospheric variability, we calculated the annual growing season NDVI by using the maximum value composites (MVC) method [53]. The MODIS NDVI and GPP data in the study area were resampled to the monthly scale using ArcGIS10.4, with a spatial resolution of 1 km. Additionally, by averaging each pixel, the average values from May to October were obtained as the growing season NDVI and GPP for each year from 2000 to 2019, which meant in pixels of 500 × 500, there was, on average, one gC unit of GPP per 1 m2.

Simple linear regression was used to analyze the annual trends of average NDVI and GPP values in the growing season of different LULC in southwest China from 2000– 2019. All data were analyzed using Python software. Additionally, referring to related research [54], the vegetation coverage was divided into lowest coverage (0 < NDVI ≤ 0.35), low coverage (0.35 < NDVI ≤ 0.55), medium coverage (0.55 < NDVI ≤ 0.75), high coverage (0.75 < NDVI ≤ 0.85), and highest coverage (0.85 < NDVI ≤ 1), five grades. The gravity center migration can show the change directions in the region's center of gravity. Therefore, the change directions in the region's NDVI and GPP gravity centers were calculated by gravity center migration. This, in turn, can reflect the import of human activities or climate change on the migration of vegetation zone. The main LULC types in southwest China were cropland, forest, and grassland, and other land types (shrub, water, snow/ice, barren, impervious, and wetland) accounted for only 2% of the total area (Figure 1). Therefore, this paper mainly analyzed the change characteristics of three LULC types of cropland, forest, and grassland, the temporal and spatial distribution of NDVI and GPP in the growing season of different LULC types, as well as the regular change in the center of gravity.

#### **3. Results**

#### *3.1. Land Types Changes*

During the study period, the composition of LULC types in the study area changed significantly, with a high spatial heterogeneity between the LULC types (Figure 1). The conversion of farmland into forests was the major driving force of LULC change. Overall, the forest area showed increasing trends after 2000, but it showed a decreasing trend first and then increased (1.04 year<sup>−</sup>1, *p* < 0.01). Forest area increased by 16,180.0 km2, accounting for 58.7% of the regional area from 2000 to 2019 (Figure 2, Table 1). The growing regions of forest were mainly distributed in northeastern Sichuan and southwestern Yunnan, and the decreasing regions were mainly distributed in southeastern Guizhou, northwestern Guangxi, and northwestern Yunnan. The cropland area first increased and then decreased, showing a significant downward trend overall (−0.707 year−1, *<sup>p</sup>* < 0.01). The area of conversion of farmland into forests was 61,318.48 km2, which was the LULC change type with the greatest land area transfer, mainly concentrated in the Sichuan Basin, northern Guizhou, and central and southern Guangxi (Figure 1). The grassland area showed a

continuously decreasing trend (−0.64 year−1, *<sup>p</sup>* < 0.01), of which 10,948.30 km2 of the grassland area was converted into forest. The change regions were concentrated in the Hengduan Mountains in the northwest of Yunnan, Kunming–Zhaotong regions in the northwest, and the western Sichuan plateau region. A grassland area of 4231.28 km<sup>2</sup> was converted into cropland, the change regions were mainly concentrated in the northeast part of Yunnan, the western part of Guizhou, and there were also sporadic changes in southern Sichuan.

**Figure 2.** Changes in area of different LULC types: (**a**) cropland, (**b**) forest, (**c**) grassland. (Note: the black solid line represents the trend line, and the formula of cropland is y = −0.707x + 1759.78; the formula of forest is y = 1.04x − 1294.51; the formula of grassland is y = −0.64x + 1469.68).


**Table 1.** Transformation matrix of land cover types from 2000 to 2019 (km2).

#### *3.2. The Characteristics of Inter-Annual Variation in NDVI*

From 2000 to 2019, the average NDVI of vegetation in the growing season in the whole study area showed an overall upward trend with fluctuations (slope = 0.0023 year<sup>−</sup>1, *p* < 0.01). The growing season average NDVI values of different land cover types all showed an increasing trend, but there were some differences. Among them, forest had the highest NDVI value, ranging from 0.79 to 0.84; cropland NDVI value ranged from 0.74 to 0.79; grassland vegetation NDVI varied from 0.70 to 0.73 (Figure 3b). The increase in the NDVI of forest was 0.0025 year−1, and the trend of NDVI of cropland was 0.0023 year−1, which was twice the increase in grassland (0.0011 year<sup>−</sup>1).

**Figure 3.** (**a**) Spatial distribution of NDVI in southwest China and (**b**) Changing trends in NDVI values from 2000 to 2019 for different land use types.

The vegetation coverage in southwest China in 2000 was dominated by medium and high coverage, accounting for 87.05% of the regional area. In 2019, the vegetation coverage was dominated by highest and high coverage, accounting for 83.66% of the regional area. In terms of spatial distribution, there was a large heterogeneity in the level of surface cover (Figure 3a). The coverage situation in the southwest region showed that the area of lowest coverage, low coverage, medium coverage, and high coverage decreased, but the area of highest coverage increased. Overall, the degree of the greening of vegetation was strengthened. From 2000 to 2019, the area of highest coverage (0.85~1.0) increased from 10.52% to 34.37%, and the main change region was concentrated on the edge of the Sichuan Basin and the three provinces of Yunnan, Guizhou, and Guangxi, with the land cover being forest. The area of high coverage (0.75~0.85) decreased from 54.73% to 49.30%, mainly concentrated in the Sichuan Basin, the eastern Yunnan-Guizhou Plateau, and the northern hills of Guangxi, with the land cover being mainly cropland and forest. The area of medium coverage (0.55~0.75) decreased the most from 32.32% to 14.12%. The main change region was distributed in the Yunnan–Guizhou Plateau, with the cover area being mainly grassland and cropland. There was little change in the low coverage (0.35~0.55) and lowest coverage (0~0.35), which decreased from 2.70% to 2.21%, with the main type of cover being grassland, distributed in the northwest of Sichuan. The vegetation cover types remained unchanged at 48.66% of the total area, while vegetation cover improved to 45.60%.

From 2000 to 2019, the vegetation cover in southwest China mainly shifted from "high coverage to highest coverage" and "medium coverage to high coverage and highest coverage" (Table A1). The specific transfer amount was as follows: the area from high coverage to highest coverage was 311,870.61 km2, accounting for 44.17% of the total change, and the area from medium coverage to high coverage was 259,673.48 km2, accounting for 36.77% of the total change. From the perspective of the final change, the transfer-out area of high coverage was the largest, which was 357,704.37 km2, 1.26 times the transfer-in area, accounting for 50.66% of the total transfer-out area; the highest coverage transfer-out area was 23,192.33 km2, accounting for 3.28% of the total transfer-out area; the medium coverage transfer-out area was 307,709.17 km2, accounting for 43.58% of the total transfer-out area and 5.36 times the transfer-in area; the low coverage transfer-out area was 13,993.99 km2, accounting for 1.98% of the total transfer-out area; and the transfer-out area of lowest coverage was 3544.33 km2, accounting for 0.5% of the total transfer-out area. The transferin area of highest coverage was the largest, which was 351,279.44 km2, accounting for 49.75% of the total transfer-in area, and 15.15 times the transfer-in area; the high coverage transfer-in area was 283,127.25 km2, accounting for 40.10% of the total transfer-in area; the area of medium coverage transfer-in area was 57,361 km2, accounting for 8.12% of the total transfer-in area; the transfer-in area of low coverage was 12,805.25 km2, accounting for 1.81% of the total transfer-in area; and the transfer-in area of lowest coverage was 1570.49 km2, accounting for the total transfer-in area of 0.22%. In this period, the vegetation coverage mainly showed a trend of gradual improvement, and the vegetation coverage also gradually increased.
