Stepwise calibration of age-dependent biomass in the Integrated Biosphere Simulator (IBIS v2.6) model

The carbon sequestration capacity of forest ecosystems is closely related to their stand age. However, land surface models (LSMs) usually omit the impact of stand age and calibrate the carbon fluxes from equilibrated states after the model spin-up followed by a perturbation of biomass only resulting from environmental factors such as CO₂ and climate. The mismatch between modelled and real forest stand ages will bring large uncertainties in the simulation of carbon stocks and the projection of future carbon sequestration potential. In this study, we implemented and calibrated age-dependent biomass in the Integrated Biosphere Simulator (IBIS) model using forest age and biomass information at 13 representative forests across the world. Specially, to avoid error compensation in model processes, we developed a stepwise optimization framework that integrates remotely sensed gross primary productivity (GPP), leaf area index (LAI), and age-dependent biomass curves into the IBIS model in three calibration steps. The modified adaptive surrogate modelling optimization (MASM) algorithm was applied in our framework to accelerate the parameterization based on surrogate modelling. Compared with the original model, our improved model leads an average error reduction (AER) in GPP, LAI and biomass by 23.7%, 28.6% and 65.7%, respectively, after each calibration step. The new parameters decreased the mean annual net biomass productivity (NBP) during 2000-2020 by 23.1 and 35.7 gC/m²/year in the mixed forests and deciduous broad-leaved forests, respectively, and increased NBP by 36.1-68.7 gC/m²/year in coniferous forests and evergreen broad-leaved forests. Our work highlights the necessity of considering forest age in LSMs, and provides a new framework for better calibrating LSMs under the constraints of multiple satellite products.