*3.3. GPP Changes*

Since 2000, the overall trend of the annual mean value of GPP in the growing season in the study area increased significantly, with a change rate of 3.65 gC·m<sup>−</sup>2year−<sup>1</sup> (*<sup>p</sup>* < 0.01). From 2000 to 2019, the average GPP in the growing season increased from 553.77 gC·m−<sup>2</sup> to 624.33 gC·m−2, an increase of 12.74% (Figure 4a). Among them, the average value of GPP in the growing season was the lowest at 553.77 gC·m−<sup>2</sup> in 2000, and the average value was the highest at 728.83 gC·m−<sup>2</sup> in 2016. During the study period, the GPP of the three land cover types showed an increasing trend. Cropland and forest showed a significant increasing trend (*p* < 0.01), while grassland showed a slight upward trend (*p* > 0.05). The GPP of different land cover types showed a great difference in growing seasons, and the GPP of forest and cropland had a high coincidence with the overall GPP trend in southwest China. Forest was the land cover type with the highest gross primary productivity, and the highest average GPP in the growing season was 654.7 gC·m−2; the second was cropland, which was 571.63 gC·m−2; and the lowest value of grassland was 420 gC·m−2. Since forest had the highest average GPP, the transfer of forest to other land cover types led to a decrease in GPP, and the conversion of cropland and grassland to forest led to an increase in GPP. In general, the carbon storage of land cover increased, indicating that the ecological environment in southwest China improved.

The average value of the growing season in southwest China from 2000 to 2019 was 594.61 gC·m<sup>−</sup>2, which showed a decreasing trend from southeast to northwest, corresponding to the distribution of cropland, forest, and grassland, with significant spatial differences. The high-value regions of GPP were mainly distributed in northwestern Yunnan, southeastern Guangxi, Sichuan Basin, etc. In addition, the GPP was relatively large in central and southern Guizhou, which was covered by forest and cropland. However, the concentrated distribution region of grassland in northwestern Sichuan, the concentrated distribution region of cropland in the Sichuan Basin, and the GPP value were generally lower than the average value (Figure 4b).

**Figure 4.** (**a**) Interannual GPP changes in different land use types; the dashed line shows the variation trend of GPP in different land use types, where the black is GPP mean, the formula is y = 3.65x − 6742.5; the yellow is cropland's GPP, the formula is y = 4.36x − 8195.02; the cyan is forest's GPP, the formula is y = 3.97x − 7322.35; the light green is grassland's GPP, the formula is y = 0.94x − 1463.16 and (**b**) spatial distribution pattern of GPP values of different land cover types in Southwest China from 2000 to 2019.

#### *3.4. Changes in the Center of Gravity of NDVI and GPP*

From 2000 to 2019, the migration direction of the NDVI center of gravity of different land cover types also showed varying differences. The center of gravity of the NDVI value of cropland moved to the southwest by 0.12◦ in the meridian and 0.12◦ in latitude. Among them, the center of gravity of cropland moved in the southeast direction, moving 0.037◦ in the meridian and 0.12◦ in latitude from 2000 to 2005; the center of gravity moved in the west direction, moving 0.16◦ in the meridian and 0.01◦ in latitude from 2005 to 2019; and the center of gravity of the woodland moved to the northeast, moving 0.03◦ in the meridian and 0.06◦ in latitude. However, the center of gravity moved to the northwest from 2000 to 2005 and continued to move to the northeast from 2005 to 2019. The center of gravity of the grassland continued to move to the northwest from 2000 to 2019 by 0.06◦ in the meridian and 0.16◦ in latitude. Since 2000–2006 was the main implementation period of the ecological project, the composition of land cover changed greatly, and the change direction of the NDVI center of forest, cropland, and grassland changed (Figure 5).

**Figure 5.** Changes in the center of gravity of NDVI and GPP. (Note: (**a**–**c**): Changes in the NDVI center of gravity of cropland, forest and grassland from 2000 to 2019; (**d**–**f**) Change in the GPP center of gravity of cropland, forest and grassland from 2000 to 2019).

The study found that the migration direction of the GPP gravity center of the same type of LULC type had been consistent with the gravity center migration direction of its NDVI (Figure 5). During the study period, the center of GPP of cropland moved to the southwest. The GPP center of gravity of cropland moved to the southeast from 2000 to 2005, by 0.03◦ in the meridian and 0.12◦ in latitude, while the GPP center of gravity of cropland moved to the southwest from 2005 to 2019, and the migration amplitude increased, moving 0.15◦ in the meridian, while there was almost no change in latitude. The GPP center of gravity of the forest moved to the northeast. From 2000 to 2005, it first moved to the

northwest, moving 0.03◦ in the meridian and 0.052◦ in latitude, and then it moved to the northeast, moving 0.063◦ in the meridian and 0.014◦ in latitude. The GPP center of gravity of the grassland continued to move to the northwest in the study period, by 0.05◦ in the meridian and 0.15◦ in latitude.

From 2000 to 2019, the center of gravity of the lowest coverage shifted to the southeast, moving 0.36◦ in the meridian direction and 0.24◦ in the latitudinal direction, with a small but complex change in the inter-annual migration direction. The center of gravity of low coverage also moved to the southeast, 0.77◦ in the longitudinal and 0.24◦ in the latitudinal direction. During 2000–2008, the low coverage center of gravity first shifted back and forth to the northwest, and then folded to the southeast after 2008. The medium coverage center of gravity moved to the northwest, with a larger migration range, moving 1.24◦ in the meridian and 1.14◦ in the latitudinal direction. The center of gravity of the high coverage migrated to the northwest but migrated to the southeast first from 2000 to 2011, moving 0.157◦ in the meridian and 0.01◦ in latitude. After 2011, it migrated to the northwest, moving 0.26◦ in the meridian and 0.08◦ in latitude; the center of gravity of the highest coverage migrated to the southeast as a whole, moving 1.32◦in the meridian and 0.53◦ in latitude (Figure A1).

Under different vegetation coverage levels, the migration direction of the gravity center of GPP was basically the same as that of vegetation NDVI (Figure A1). During the study period, the gravity center of GPP of lowest coverage and low coverage migrated to the southeast direction, with 0.36◦and 0.77◦in longitude, respectively, with a latitude migration of 0.24◦ for both. The center of gravity of the GPP with medium coverage moved to the northwest, moving 1.30◦ in the meridian and 1.19◦ in latitude. The center of gravity of the high-coverage GPP moved to the southeast from 2000 to 2008, moving 0.21◦ in the meridian and 0.22◦ in the latitudinal direction, then turned around and moved to the northwest direction after 2008, moving 0.31◦ in the meridian direction, and 0.30◦ in the latitudinal direction. The center of gravity of the highest-coverage GPP moved to the southeast direction, moving 1.37◦ in the meridian direction and 0.61◦ in the latitudinal direction.

#### **4. Discussion**

#### *4.1. Spatiotemporal Variation in Land Types and NDVI and GPP*

The results revealed that the rate of returning farmland to forest increased significantly in the study period (2000–2019), and cropland was the main land source for forest expansion. The area changes of LULC types were as follows: the area of cropland increased first and then decreased, while the area of forest decreased first and then increased, and the grassland continued to decrease in southwest China (Figure 2). This trend was consistent with the trend of land-use change in China [19]. In the study region, cropland was mainly distributed in the east of Sichuan, the center of Guangxi, the west of Guizhou, and the east of Yunnan (Figure 1). These regions were suitable for agricultural activities due to their flat terrain. The increase in agricultural intensification and productivity has led to the expansion of cropland in these regions, the main source of which was forest and grassland. Due to the development of ecological engineering, the phenomenon of "returning farmland to forest" appeared in the margin of Sichuan Basin, northwest and southwest Yunnan, southeast Guizhou, and south Guangxi, which effectively controlled the expansion of cropland. This indicates that the ecological restoration project in this region was the main driver of LULC change in most regions of southwest China and has achieved some results at this stage [55]. Second, rapid urbanization was also one of the most common causes of farmland loss [56]. At the beginning of the 21st century, the urban area of the Sichuan Basin increased by 66,000 hm2, accounting for 68.31% of the decreased area of cultivated land [57]. The grassland was mainly concentrated in the northwestern part of Sichuan, where the altitude is higher, combined with a cold and dry climate. There are rivers flowing through this region, and the precipitation is relatively abundant. Under the background of climate warming, this is conducive to the growth of vegetation, and with the LULC change, in some regions, grassland has shifted to forest.

The study found that the NDVI and GPP of vegetation in the growing season had shown a significant upward trend as a whole, and the growth trends had been different for different LULC types in southwest China from 2000 to 2019 (Figures 3b and 4a). This is consistent with the conclusion that the main vegetation types in China show a dynamic greening trend, which is the result of the combined effects of climate change, LULC types of distribution, and human activities (such as ecological engineering and agricultural management), among other factors [58–60]. In general, the increase in the NDVI value in southwest China was mainly due to ecological restoration caused by ecological projects such as returning farmland to forest, natural forest protection plans, and closing mountains for afforestation [55,61]. The average NDVI and GPP of vegetation in the growing season in southwest China were high in the southeast and low in the northwest (Figures 3a and 4b), which was mainly related to the spatial distribution of LULC types. In southwest Yunnan, southeast Guangxi, and Sichuan Basin, high mean NDVI and GPP levels in the growing season were mainly distributed in forests, while northwest Sichuan and Sichuan Basin were mainly distributed in grasslands and croplands, which were lower than those of forests on the whole. Compared with 2000, most regions became greener and more productive in 2019. The conversion of a large amount of cropland or grassland to forest and the ecological restoration of most regions (low coverage shift to high coverage) resulted in significant greening of vegetation. However, at the periphery of most cities, vegetation degraded, and productivity declined, indicating that urbanization has led to the loss of vegetation [62]. In addition, studies have shown that the interaction of temperature, precipitation, and solar radiation has different effects on vegetation greening [63–65], and sustained warming and decreased precipitation are key factors in limiting vegetation growth [66]. In the past 20 years, the overall climate has been dry and warm, with a significant upward trend in temperature (0.42 ◦C/10 year) but no significant change in precipitation in southwest China [36,48]. The occurrence of extreme weather events and natural disasters hinders the growth of vegetation, resulting in a decline in regional vegetation coverage [13,67]. In recent decades, severe droughts have occurred frequently in southwest China [68–71], which have had a significant impact on grassland and cropland. The average NDVI in the vegetation growing season has shown a decreasing trend (Figure 3b). However, the area of forest in this study region was about 2.4 times that of cropland, and the increase in the greening effect on the whole region compensated for the decreasing trend of NDVI.

The productivity levels of different LULC types had different responses to influencing factors. For example, cropland and grassland were vulnerable to extreme climate disasters, but the positive effects of human activities on the effective management of cropland weakened the negative effects of drought. From 2000 to 2019, the area of cropland decreased, but the average GPP value in the growing season showed an increasing trend in southwest China (*p* < 0.01), which may be related to the improvement in the productivity of cropland by agricultural management measures in recent years [24]. The forest ecosystem was relatively stable, and human activities, such as deforestation and afforestation, had a greater impact on the productivity of the forest ecosystem than the impact of extreme climates [72]. The average GPP of forest in the growing season showed a significant increasing trend, which was related to the increase in forest area and vegetation restoration in this region [73]. Studies have shown that vegetation productivity exhibits different growth patterns at different stages of forest age [74,75]. Due to the conversion of grassland to other land types, the amount of grassland has been decreasing continuously for 20 years. However, with the influence of ecosystem protection policies in recent years, the GPP of grassland still showed a fluctuating growth trend (Figure 4a). These results indicate that the overall carbon sequestration capacity of the southwest region is gradually increasing, and the environmental quality of the ecosystem is gradually improving, which corresponds to the significant increase in the vegetation coverage in southwest China in recent years.

#### *4.2. Migration Changes in NDVI and GPP Centroid*

Our study indicated that the migration direction of the NDVI center of gravity of cropland, forest, and grassland was the same as that of GPP during 2000–2019 in southwest China (Figure 5), and the spatial distribution pattern of vegetation cover and productivity gravity center changed with a change in LULC distribution. In southwest China, the center of gravity of cropland moved westward, and there were inflection points of the center of gravity in 2001 and 2005. In 2001, the project of returning cropland to forest, ecological protection in the east, and the natural forest protection plan in the west were fully launched. Therefore, the area of cropland increased, and the area of forest decreased during 2000–2005. After 2005, ecological engineering began to achieve results, which showed that the cropland area decreased and the forest area increased, which was consistent with the results of previous studies [76]. These results indicated that the LULC change caused by the ecological restoration project was the direct cause of the change in the vegetation growth and productivity center of gravity in southwest China over the past 20 years. The inflection point times of the change in the center of gravity were consistent with the planning and implementation time of the restoration project. The grassland continued to migrate to the northwest, which may be related to the conversion of grassland to other land types in the southeast, while the northwest region had a higher altitude, less human activity, and fewer changes in land cover types [36]. The study also found that the change in GPP's gravity centers of different vegetation NDVI grades was highly similar to the migration direction of NDVI's gravity center changes of different grades (Figure A1). This indicated that the increase or decrease in vegetation productivity in southwest China was related to the restoration or degradation of vegetation; the productivity increased in the region of vegetation restoration, and the productivity decreased in the region of vegetation degradation. From 2000 to 2019, the change in the center of gravity of the lowest coverage was relatively stable, and the direction and distance of the center of gravity were not large. The center of gravity of low coverage and high coverage mainly shifted to the southeast, and the center of gravity of medium coverage and high coverage mainly shifted to the northwest. This was mainly because of the spatial distribution, composition, and climatic characteristics of LULC types of different vegetation grades at different stages, which affected the change in vegetation cover and the center of gravity of productivity. Therefore, reasonable planning of regional cover-type composition is of great significance to effectively improve regional vegetation NDVI and GPP.

The shift in vegetation types in the direction of the center of gravity indicates that the expansion of vegetation in the direction of migration or the degradation of vegetation in the opposite direction, and the long-term shift in the center of gravity in a single direction, lead to the imbalance of ecosystem structure and function. For example, the southern and southwestern regions of southwest China are the main regions of returning farmland to forest. Due to the favorable ecological conditions in this region, the vegetation coverage has shifted from medium coverage to high coverage. Recent studies have also confirmed that afforestation measures have achieved good results in improving vegetation cover and promoting carbon sequestration [58,77]. However, it was also found that the excessive growth of large regions of forest consumes surface water, resulting in a shortage of regional water resources [78,79]. In addition, local negative effects have also occurred due to the planting of unsuitable tree species [80]. To sum up, large-scale afforestation may rapidly improve the vegetation greening degree in the region in the short term, but it cannot guarantee the long-term stable development of the ecological environment. In addition, because of the heterogeneity of the growth of different vegetation types, it is understood that the composition structure of land cover types will affect the ecosystem balance and sustainable development in the southwest region of southwest China under different topographic and landform conditions.

#### *4.3. Implications for Future*

Our study found that the LULC change was beneficial to the improvement in regional greening and plant productivity in southwest China. Karst landforms are widely

distributed in the study area. Due to its unique binary structure, precipitation is rapidly lost, which leads to low utilization of precipitation by vegetation [81]. Moreover, the karst regions have thin soil layers and poor water storage capacities, and the climate warming trend may have an inhibitory effect on the growth of forests [14]. Studies have shown that ecological restoration and LULC pattern change not only ameliorate land degradation [40] but also affect local and regional climate [82]. For example, land surface temperature decreases significantly when forest is converted to cropland [83]. The global warming trend has a greater impact on ecosystems [84,85], which has led to dramatic changes in land cover types and plant biomass in the high northern latitudes and promoted the expansion of woody shrubs and forest areas [86]. Meanwhile, the rise in temperature has removed local environmental boundaries, allowing alpine plant species to move to higher altitudes [87]. These results indicate that in future implementations of ecological projects, it is necessary to consider the structural composition and environmental carrying capacity of LULC types under the background of climate warming so that they can more effectively serve the maintenance of local, sustainable ecosystem balance.
