Uncertainty in simulating regional gross primary productivity from satellite-based models over northern China grassland

Large-scale estimation of regional terrestrial gross primacy production (GPP) can improve our understanding of carbon cycle. However, model based estimates are subject to uncertainty. In this study, eight satellite-based models (i.e. VPM, TG, GR, VI, CFIX, ECLUE, VPRM and MODIS-GPP) were compared for GPP simulation in northern China grassland based on 17 site-year eddy covariance measurements, meteorological data and satellite data. Also, the regional spatial-temporal GPP patterns during 2001-2013 in northern China grassland were simulated and their uncertainties were quantified. The results showed that the model simulations exhibited significant correlations with observed GPP across these eight models and R2 or pseudo R2 ranged between 0.64 and 0.89 (p < .001), ECLUE model performed best. The annual grassland GPP had been growing in fluctuations from 2001 to 2013, with the averaged value of 241.8 g C m−2 a−1. Substantial spatial heterogeneity existed in grassland GPP, increasing from the west to the east. The disparities of satellite-based model structures resulted in the overall 49% relative uncertainty in regional simulation of GPP, which was high in area with arid dry climate. Our study highlighted the uncertainty traced back to model approaches under different environmental stresses (photosynthetically active radiation, soil water content and air temperature). For the accurate simulation of grassland GPP, uncertainty in alpine grassland and arid cold area on regional grassland GPP should be focused.