Improved global estimations of gross primary productivity of natural vegetation types by incorporating plant functional type

Abstract Satellite-based light use efficiency (LUE) models are important tools for estimating regional and global vegetation gross primary productivity (GPP). However, all LUE models assume a constant value of maximum LUE at canopy scale (LUEmaxcanopy) over a given vegetation type. This assumption is not supported by observed plant traits regulating LUEmaxcanopy, which varies greatly even within the same ecosystem type. In this study, we developed an improved satellite data driven GPP model by identifying the potential maximal GPP (GPPPOT) and their dominant climate control factor in various plant functional types (PFT), which takes into account both plant trait and climatic control inter-dependence. We selected 161 sites from the FLUXNET2015 dataset with eddy covariance CO2 flux data and continuous meteorology to derive GPPPOT and their dominant climate control factor of vegetation growth for 42 natural PFTs. Results showed that (1) under the same phenology and incident photosynthetic active radiation, the maximal variance of GPPPOT is found in different PFTs of forests (10.9 g C m−2 day−1) and in different climatic zones of grasslands (>10 g C m−2 day−1); (2) intra-annual change of GPP in tropical and arid climate zones is mostly driven by vapor pressure deficit (VPD) changes, while temperature is the dominant climate control factor in temperate, boreal and polar climate zones; even under the same climate condition, physiological stress in photosynthesis is different across PFTs; (3) the model that takes into account the plant trait difference across PFTs had a higher agreement with flux tower-based GPP data (GPPflux) than the GPP products that omit PFT differences. Such agreement was highest for natural vegetation cover sites (R2 = 0.77, RMSE = 1.79 g C m−2 day−1). These results suggest that global scale GPP models should incorporate both plant traits and their dominant climate control factor variance in various PFT to reduce the uncertainties in terrestrial carbon assessments.