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

between temperature and NDVI, which shrunk from 32% in 1982, and NDVI trends were sup- ported by tree ring evidence. Since nowhere in the PRD region is too cold for plant growth, ris- ing temperatures would not necessarily generate increased productivity, and the observed negative correlations may, as suggested by Corlett, (2011) [3] be a sign of an upper temperature threshold being reached. These may be reached earlier under conditions of moisture stress from increased evapotranspiration in a warming climate. The result is reduced carbon uptake with decreased productivity, rather than direct damage to leaf tissues. The correlation and lag time maps could provide inputs in developing specific models for vegetation response to climate predictors by considering temporal lag dynamics. For example, during the spring drought during 2009–10 in Yunnan, the southwestern province of China, it was observed that lag time response of vegetation to rainfall deficiency was 90 days for the evergreen forest while Cropland and Shrubland became stressed earlier (64 days). This indi- cates varying resistance capacities of different vegetation regimes in an ecosystem [60,61]. However, resistance to extreme climate events may also change due to species composition of plant communities within a particular land cover type. For example, Abbas et al., (2017) [12] observed that tropical species suffered more than sub-tropical species during the extreme cold event in January 2016 in the degraded tropical secondary forest of Hong Kong. Apart from

temporal lag, vegetation response to climate variability can also vary in an ecosystem. For example, Be´gue´ et al. (2011) [62] divided the study area into three eco-climatic zones based on the NDVI and rainfall relationship (positive, negative and neutral). Thus, the nature of vegeta- tion response to climate variability and inputs from lag time maps may enhance mapping and understanding of intricate responses of dynamic land cover regimes to climate indicators. The observed reversal in the upward trend in productivity during the decade 1995–2004 corresponds to a period of rapid urbanization and industrialization, deteriorating air quality and exceptionally warm years. Studies from other regions using GIMMS NDVI have also reported a reversal in greening trends around the same period, from the mid-1990s to mid- 2000s. For example, Liu et al (2005) [63] found trends of greening-browning-greening for the periods 1982–1994, 1995–2004, and 2005–2012 respectively over the globe, with browning in the middle period in many regions. While they do not provide any overall explanation, other researchers working in specific regions give climatic or other explanations. For example, Park and Sohn (2010) [64] observed a browning period from 1997 to 2006 in northern East Asia around 50˚N, explained as reduced rainfall combined with continued warming creating mois- ture stress. Piao et al. (2014) [65] also attributed a weakened relationship between temperature and productivity after 1997 in northern Eurasia, to increased drought under rising tempera-

tures. Weakening relationships between productivity and temperature over recent decades were also observed by de Jong et al. (2013) [46], attributed to large scale climatic catastrophes accompanying climate change. The complexity of the PRD region precludes a simple explana- tion for a decadal decline in productivity. However, ongoing massive urbanization and indus- trialization in the PRD during the study period suggests that our observations of declining sunshine hours in the late 1990s to early 2000s are at least partly due to dimming from atmo- spheric aerosols. Higher temperatures during this period may also be partly responsible for the productivity decline, as agricultural and forest plants meet upper temperature thresholds. This is supported by our correlation analysis (section 3.4), with negative correlations between tem- perature and MGS and Peak. In particular, the main staple crop rice may not support the higher summer temperatures observed over recent decades. Others have noted trends in solar radiation, rather than in temperature and rainfall, affect- ing plant productivity. Nemani et al., (2003) [66] observed an increase in Net Primary Produc- tivity (NPP) in tropical ecosystems between 1982 and 1999, which was primarily attributed to decreased cloudiness and increased solar radiation. Wei et al. (2018) [67] found that reduced PLOS ONE Climate and vegetation trends in the Pearl River Delta PLOS ONE | https://doi.org/10.1371/journal.pone.0245467 February 24, 2021 14 / 20