Simulation of carbon fluxes from Tibetan Plateau grasslands by integrating data assimilation and parameter inversion within the Biome-BGC model

Abstract

The Qinghai‒Tibet Plateau (QTP) grassland ecosystem is highly sensitive to climate change, but significant uncertainties caused by model parameters and state variables remain in its simulated carbon fluxes. This study integrates data assimilation and parameter inversion within the Biome-BGC model to improve simulation accuracy. By simultaneously optimizing both model parameters across multiple sites and multiple objectives—including gross primary production (GPP), ecosystem respiration (RECO), soil carbon, and aboveground biomass—as well as key state variables, such as the leaf area index and soil moisture, this approach addresses limitations in the radiation and soil moisture modules of the Biome-BGC for the QTP. At the site scale, the optimized model improved the GPP simulation accuracy, with an average increase of 0.42 in R ² from 0.41 to 0.83 compared with the original model, whereas the RECO simulation accuracy improved, with an average R ² increase of 0.32 from 0.42 to 0.75. The mean carbon sink of Tibetan Plateau grasslands was 41.47 Tg C yr⁻¹ during 2000–2022, with the eastern region acting as a strong carbon sink, whereas the western region presented weaker carbon sinks or carbon sources. Over these 23 years, the QTP has shown a continuous greening trend, partly because the increase in RECO was smaller than that in GPP. This study presents a new model optimization framework for research on carbon neutrality on the QTP.