An improved estimate of soil carbon pool and carbon fluxes in the Qinghai-Tibetan grasslands using data assimilation with an ecosystem biogeochemical model

Abstract

The accurate estimation of soil carbon (C) pool and fluxes is a prerequisite to better understand the terrestrial C feedback to climate change. However, recent studies showed considerable uncertainties in soil C estimates. To provide a reliable C estimate in the grasslands of the Qinghai-Tibet Plateau (QTP), we calibrated key parameters in a process-based ecosystem model (the CENTURY model) through data assimilation based on 570 soil samples and 21 sites of eddy covariance measurements. Two assimilating strategies (Opt1 – assimilating C pool observations; Opt2 –assimilating both C pool and C flux) were examined. Compared to default parameterization, our results showed both Opt1 and Opt2 improved the soil organic carbon density (SOCD) estimation, with R2 increasing from 0.59 to 0.75 and 0.73, respectively. Opt2 was superior to Opt1 in constraint of parameters dominating aboveground processes and yield a better estimation of net ecosystem production (NEP). Based on different parameterization, the spatial variability of SOCD and NEP across the QTP grassland were generated. Both Opt1 and Opt2 ameliorated the overestimation of SOCD by the default model, estimating a total soil C of 6.63 Pg and 6.48 Pg C for the topsoil (0–30 cm) of the QTP grasslands, respectively. Opt2 showed lower uncertainties in the NEP estimation and predicted a net sink of 14.33 Tg C annually. Compared with existing datasets, our study provided a more reliable estimation of carbon storage and fluxes in the QTP grassland with the calibrated ecosystem model. The results highlight that data assimilation with multiple observational data sets is promising to constrain process-based ecosystem models and increase the robustness of model predictions for terrestrial C cycle feedback to future climate change.