Sensitivity of Satellite-Derived Gross Primary Productivity to Environmental and Canopy Inputs in a Subtropical Mangrove Forest

Abstract

Accurate estimates of mangrove gross primary productivity (GPP) are needed to quantify coastal carbon dynamics, but the performance of satellite-based GPP models in mangrove forests remains uncertain. We compared eight models—four light-use-efficiency models (MOD17, EC-LUE, VPM, and MVPM) and four vegetation-index models (GR, VEI, AVM, and PCM)—with eddy-covariance observations from a subtropical mangrove forest in southern China. VPM showed the closest agreement with tower-based GPP in the calibration comparison (R2 = 0.58, RMSE = 1.38 g C m−2 d−1, and MAE = 0.72 g C m−2 d−1). We then applied VPM to nine cloud-minimized Sentinel-2 acquisitions from 2018–2020. The grouped means for these selected acquisitions were 7.47 g C m−2 d−1 in the post-monsoon period, 7.41 g C m−2 d−1 in summer, and 4.57 g C m−2 d−1 in winter. Sobol’ sensitivity analysis identified PAR, EVI, maximum light-use efficiency, LAI, and temperature as influential inputs over the prescribed ranges. These results provide a site-level comparison of satellite-based GPP models and identify priority inputs for improving future mangrove GPP assessments.