Validating satellite based FluxSat v2.0 Gross Primary Production (GPP) trends with FluxNet 2015 eddy covariance observations

Gross Primary Production (GPP) is a measurement of the amount of organic carbon produced during photosynthesis. The ability to detect global and regional temporal variations in GPP, including trends, is critical to our understanding of the carbon cycle and the biosphere. FluxSat v2.0 (Joiner and Yoshida, 2020) is a globally gridded satellite based terrestrial GPP dataset that is trained using Machine Learning and MODIS reflectances calibrated to GPP derived from eddy covariance (EC) observations from a number of FluxNet 2015 and other EC sites. We validate the FluxSat v2.0 trends over the period 2001-2020 by conducting an analysis at collocated sites and show the trends are in good agreement (R > 0.70) for sites with overlapping time periods greater than 3 years with a sufficient number of quality daily data points. We then examine GPP trends regionally and globally. FluxSat v2.0 demonstrates a positive global trend in GPP, that is potentially underestimated when compared to FluxNet 2015 sites (2 upper bounds of slope < 1). Positive trends are found in Central North America, Europe, India, Eastern Asia, Southern Brazil, and the African Savanna, and negative trends are found in Eastern Brazil, Madagascar, and Northern Australia.