Evaluation and Fusion of Multi-Source Gross Primary Productivity Products over China under Widespread Greening

Abstract. Vegetation greening has been widely observed across China, yet whether this structural change has translated into stable terrestrial carbon uptake remains uncertain because gross primary productivity (GPP) estimates differ substantially among products. Here we systematically evaluated 12 widely used multi-source GPP products at a common 0.05° monthly resolution over 2001–2020, using paired observations from 33 ChinaFLUX eddy-covariance towers spanning the major vegetation types in China. We also characterized the greening signal with six vegetation indicators derived from satellite leaf area index (LAI), dynamic global vegetation model (DGVM) LAI, and near-infrared reflectance of vegetation (NIRv). Inter-product discrepancies were largest in high-productivity regions and wetlands, where about half of the products systematically underestimated monthly GPP exceeding 300 g C m⁻². Differences in greenness inputs and model structures further amplified the divergence among product categories. Among individual products, VPM, PML-V2, and GOSIF showed the strongest overall agreement with flux-tower observations. Using inter-product consistency as a selection constraint, we tested three fusion approaches on multiple subsets of source products. The arithmetic mean of five products (GLASS, VPM, TL-LUE, PML-V2, and GOSIF) achieved the best overall agreement with tower observations (r = 0.87; RMSE = 53.9 g C m⁻² month⁻¹) and produced a national multi-year mean GPP of 7.72 ± 2.03 Pg C yr⁻¹. Joint analysis of the GPP and LAI ensembles indicates that approximately 27.21% of greening areas showed an effective carbon-uptake response across China, but the relationship is spatially heterogeneous. These findings demonstrate that China’s greening is functionally meaningful at the national scale, but its carbon-sink effectiveness depends strongly on regional environmental constraints and structure-function coupling.