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

Lag time analysis The lag times of NDVI to precipitation and temperature were determined by performing cor- relation analysis. For this, the datasets were transformed to Vegetation Condition Index (VCI) [34], Precipitation Condition Index (PCI) [35], and Temperature Condition Index (TCI) (Eqs 1–3). These transformations are pixel-based normalization of the datasets to control local dif- ferences for an integrated raster-based analysis in which short signal of changes in the variables are filtered by separating them from the long-term ecological and climate signals. The cross- correlation coefficients of VCI with PCI and TCI were determined using Pearson’s cross-cor- relation function (Eq 4) at different lag-time scales (0, 15, 30, . . ., 150 days). The analysis was performed up to the 10th lag, in 15-day intervals up to 150 days [36]. For each lag, two rasters comprising correlation coefficients and significance values of the correlation were obtained. Each correlation raster layer was masked by pixels with significant correlation (p < 0.05), and insignificant pixels were removed for subsequent analysis. Overlay analysis of all the correlation raster layers was performed to find the highest correlation value of each pixel, then gave the lag time corresponding to the maximum correlation value for every pixel [37]. This resulted in two layers representing maximum correlation, as well as lag time for the maximum correlation. VCI ¼ ðNDVI NDVIminÞ ðNDVImax NDVIminÞ ð1Þ PCI ¼ ðRainfall RainfallminÞ ðRainfallmax RainfallminÞ ð2Þ TCI ¼ ðTemperature TemperatureminÞ

ðTemperaturemax TemperatureminÞ ð3Þ where max and min represent the maximum and minimum values of corresponding variables during the study period. PCC¼ X L 1 t¼0 TSa t TSa t � � TSb t lag TSb � � h i X L 1 t¼0 ðTSa t TSa t Þ � �2 !0:5 � X L 1 t¼0 ðTSb t lag TSbÞ h i2 !0:5 ð4Þ where TSa t and TSb t b corresponds to VCI and PCI (or TCI), respectively; L is the length of time series, lag is the length of temporal shift applied before the correlation, and t refers to the single time step of the time series. Results Spatial differences in trends in annual temperature and rainfall during 1982–2015 Temporal trend maps of temperature and rainfall indicate that over the study period, the whole PRD region has undergone a significant increase in temperature, with an increase in mean annual temperature of 0.05–0.06˚C per year, amounting to 1–1.2˚C over the study period (Fig 2). Spring temperatures show the greatest increase, of almost 2˚C. Winter and summer temperatures show an increase of approximately 0.7˚C over the study period. Rainfall, on the contrary, shows little change except in the northeast forested region of ZhaoQuing PLOS ONE Climate and vegetation trends in the Pearl River Delta PLOS ONE | https://doi.org/10.1371/journal.pone.0245467 February 24, 2021 6 / 20