Trends in vegetation productivity related to climate change in China’s Pearl River Delta

Trends in land cover specific vegetation productivity Fig 4 shows the trends of the phenology and productivity metrics, grouped by the individual LULC class, for those metrics showing significant trends over the study period (Table 2). These are the Mean Growing Season NDVI (MGS), the maximum NDVI in the season (Peak), and the minimum NDVI in the season (Trough). When measured over the whole 1982–2015 period, all have increased, and the End of Season (EOS) has become significantly delayed. Spa- tial distribution of trends in monthly NDVI and phenometrics, as well as graphical summaries of phenometrics, are given in S2–S4 Figs. For cropland, MGS is the most meaningful indicator of crop production and greenness, and this has increased with a significant upward trend of NDVI from 0.55 to 0.62, (r = 0.46) during 1982–2015 (Fig 4). Notably, Trough for cropland areas has increased very significantly from 0.38 to 0.48 (r = 0.78) (Fig 4), suggesting increasing opportunities for crops to be grown in some areas in the winter season. Both forest and cropland show significantly increasing NDVI for those phenology variables MGS, Peak and Trough, which measure vegetation greenness and/or productivity. Both also show significant lengthening of the growing season over the study period, with a delayed EOS by 15–20 days (all ecosystem types), with advanced SOS by 6–7 days (for forest) and advanced SOS by 10 days for cropland (Table 2). The effective EOS in the southern China region is Feb- ruary to March, the coldest month and dry season. As discussed above (Fig 2), winter and

spring temperatures have increased significantly thus these increased winter and spring tem- peratures are already significantly affecting vegetation productivity of the PRD region. The loess fitting of these four significant phenology metrics (Fig 4) shows an interesting reversal in the upward trend for the NDVI variables MGS and Peak, for all ecosystem types in the middle period from 1995 to 2004, followed by a continued rise up to 2015. It is also notable that the the Maximum NDVI in season Peak shows a much greater trend reversal than the Minimum NDVI in season Trough, suggesting that the reduced NDVI in this middle period 1995 to 2004, may be related to summer, rather than winter conditions. The time of Peak NDVI (POP) for all ecosystem types was also delayed by approximately 5 days in the middle period (Table 1, S5 Fig). Fig 5B shows the loess fitting of the average trend of sunshine hours in the PRD region, which also indicates a significant decline during the late 1990s and early 2000s. This was accompanied by increased summer rain (Fig 5A), and by a period of increased cloudiness over China [42] and southwest China [43]. A decrease in sunshine hours reported Fig 3. Urbanization and farmland reclamation. a) trend of urban area increase (1992–2012) in the PRD based on nightlight imagery b) annual trend of seasonally aggregated NDVI time series. Orange represents browning pixels while greening pixels are in green, and grey colour shows pixels with insignificant trends (p > 0.05). https://doi.org/10.1371/journal.pone.0245467.g003 PLOS ONE

Climate and vegetation trends in the Pearl River Delta PLOS ONE | https://doi.org/10.1371/journal.pone.0245467 February 24, 2021 8 / 20