A Data-driven Upscale Product of Global Gross Primary Production, Net Ecosystem Exchange and Ecosystem Respiration

The product includes 10-day means of global gross primary production (GPP), net ecosystem exchange (NEE), and ecosystem respiration (RECO) of 1999 to 2019 in 0.1x0.1 degree spatial resolution. Random Forest (RF) was used to upscale observations of FLUXNET 2015 (https://fluxnet.fluxdata.org/data/fluxnet2015-dataset/fullset-data-product/) to the globe from 60°S to 80°N. The daily GPP was extracted from GPP_NT_VUT_REF, NEE from NEE_VUT_REF and RECO from RECO_NT_VUT_REF in the FULLSET of FLUXNET 2015. Predictor variables include leaf area index (LAI), fraction of absorbed photosynthetically active radiation (FAPAR), air temperature (T), relative humidity (RH), downward solar radiation (DSR) on the surface, the minimum and maximum of LAI, and the number of LAI larger than the means of the minimum and maximum. The later three were derived to indicate the plant functional type (PFT) in each year and grid. They were expected to represent the temporal and spatial variations of PFT better than such product as BIOME4 (https://pmip2.lsce.ipsl.fr/synth/biome4.shtml). So far as we were aware, all data-driven upscaling products used MODIS. We extracted LAI and FAPAR from the Copernicus Global Land Service (https://land.copernicus.eu/global/); and T, RH and DSR from ERA5 of the European Centre for Medium-Range Weather Forecasts. This product using a different approach is expected to provide new information for studying global GPP, NEE, and RECO.