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

from 1981 to 2015. Similarly, temperature data were derived from the gridded ERA5, the 5th generation Reanalysis data of the European Centre for Medium-Range Weather Forecasts (ECMWF), temperature product (mean_2m_air_temperature) provided by the Copernicus Climate Change Service (C3S) [26]. The daily average temperature datasets are available from 1979 to ‘within 5 days of real time’ at a spatial resolution of 0.25˚. To examine the functional relationship between the climate variables and the phenology metrics as well as to explain the year-to-year variation in vegetation dynamics, the climate variables (temperature and precipi- tation) were converted to seasonal and annual composites. Layers of annual and seasonal (four seasons) variables for precipitation (accumulative) and temperature (average) were created. The winter season comprises December, January, and February (DJF), spring comprises March, April, May (MAM); summer is June, July and August (JJA), and autumn includes Sep- tember, October and November (SON). Land use land cover data. The LULC map of the study area was prepared from a MODIS land cover type product (MCD12Q1, version 6 product) for 2001 and 2015, at 500 m spatial resolution. The data was retrieved from the NASA EOSDIS Land Processes Distributed Active Archive Center (LP DAAC, https://lpdaac.usgs.gov). Both LULC maps of 2001 and 2015 were overlaid to remove pixels showing change over the period, and the remaining persistent pixels were resampled to 8 km resolution using a majority filter to match the spatial resolution of the

GIMMS NDVI data. The analysis was performed on pure pixels i.e. those of the 8 km pixels having over 80% of internal sub-pixels belonging to the same class [31], out of the four general- ized vegetation classes in the study area, viz., Forest (F), Shrubland (SH), Cropland (CL), and mosaic of Farmland/Cropland/Forest (Mix). Temporal trend in phenology and climatology Temporal trends in phenology metrics over the 34 years 1982–2015, were determined on a pixel by pixel basis. The trends in NDVI series were determined using linear regression on sea- sonally adjusted time-series as well as on annually aggregated time series [32]. The values of slope (trend) were masked where the p-value was greater than 0.05 (95% confidence level) to obtain only significant NDVI time series trends. For a comprehensive overview of the LULC specific changes in the vegetation and climate variables, all variables were spatially averaged over the LULC classes and linearly regressed along time. In addition to LULC classes, the vari- ables were also averaged for pixels indicating greening and browning trends over time. From the fitted linear models, correlation coefficients, slope, the significance of slope and the total amount of change (multiplying slope of the fitted model by the length of the time series) were determined [6]. These trends were also fitted using a Local Polynomial Regression Fitting (loess) algorithm to show the overall changes over time. Spatial correlation of climate variables with seasonal phenology metrics

The functional relationships between NDVI-based phenology metrics and the ten environ- mental variables (seasonal and annual measures of precipitation and temperature) were exam- ined to explore inter-annual variation in phenology. Partial correlation analysis was used to explain the variation in each phenology metric against changes in temperature and precipita- tion variables. Thus the confounding effects of temperature were eliminated while computing the partial correlation with each of the precipitation variables, and vice versa for temperature [33]. Insignificant pixels (p > 0.05) were masked out. The results of partial correlations analy- sis were also spatially aggregated by homogenous patches of LULC classes to understand the vegetation responses across different LULC classes. PLOS ONE Climate and vegetation trends in the Pearl River Delta PLOS ONE | https://doi.org/10.1371/journal.pone.0245467 February 24, 2021 5 / 20