Abstract: The Qinghai-Tibet Plateau is a hotspot in the research of terrestrial carbon cycling. In the context of global climate change, gross primary production (GPP) plays a crucial role in regional carbon cycle. Model simulation combined with remote sensing data aids in understanding the changes of ecosystem productivity at the regional scale across the Qinghai-Tibet Plateau and predicts its future trends. With six common remote sensing GPP products (GLASS, MODIS MOD17A2, FLUXCOM, VODCA2, improved EC-LUE data, and VPM data) being validated by eddy covariance flux observation data from three sites (Haibei shrub, Haibei wetland, and Damxung), we used them to analyze the spatial distribution pattern and temporal change trend of GPP in the Qinghai-Tibet Plateau ecosystems from 2001 to 2015. We found significant differences in annual mean GPP, annual total GPP spatial distribution pattern, and the trends across different remote sensing GPP datasets. The trends calculated based on the six products are -0.77 g C·m^-2·a^-1 (GLASS), 3.63 g C·m^-2·a^-1 (MOD17A2), -1.21 g C·m^-2·a^-1 (FLUXCOM), 1.53 g C·m^-2·a^-1 (VODCA2), 4.73 g C·m^-2·a^-1 (VPM), and -0.81 g C·m^-2·a^-1 (improved EC-LUE). The overall spatial distribution of multi-year mean GPP generally displays a pattern of higher values in the southeast and lower values in the northwest, with significant regional variations. The GLASS product shows the highest multi-year average annual GPP (827.78 Tg C·a^-1), while the MOD17A2 product shows the lowest (484.04 Tg C·a^-1). The other products (VODCA2, improved EC-LUE, VPM, and FLUXCOM) report 827.20 Tg C·a^-1, 714.55 Tg C·a^-1, 634.00 Tg C·a^-1, and 587.86 Tg C·a^-1, respectively. Over the 15 years from 2001 to 2015, the regions with the most significant changes in GPP are consistent across the plateau, with the most intense changes in the southeast and the most stable in the west. Site data validation indicates that MOD17A2 performs relatively better on an 8-day scale, whereas the FLUXCOM dataset shows superior results on a monthly scale. Combined with its performance at the regional scale, the MOD17A2 dataset is more suitable for the Qinghai-Tibet Plateau region.

Key words: eddy covariance-light use efficiency (EC-LUE) model, VPM, VODCA2, FLUXCOM, MODIS