Impacts of LULCC on Land Carbon Cycle in CAS-ESM2.0: Historical Benchmarking, Uncertainty, and CMIP6/7 Comparisons

This study develops and evaluates an enhanced Land Use and Land Cover Change (LULCC) scheme integrated within the Common Land Model (CoLM) of CAS-ESM2.0, which is coupled with the dynamic vegetation model IAP-DGVM. The updated model captures carbon fluxes and storage changes from key LULCC activities (e.g., deforestation, afforestation, wood harvest) and simulates the dynamic responses of both natural and human-managed vegetation to climate. Historical simulations conducted for CMIP6 LS3MIP/LUMIP demonstrate that the updated model reasonably reproduces the evolution of land-use emissions and trends in ecosystem carbon storage, showing significant improvements over the standard CoLM in simulating the surface climate and terrestrial carbon cycle.

We further quantify multi-source uncertainties in simulated historical LULCC carbon emissions (1850–2014) through ensemble experiments. Results indicate that model parameterization is the dominant source of uncertainty (contributing >50% and exceeding 80% in some comparisons), followed by climate forcing. Initial state uncertainty is significant only in the first few decades, while differences among CMIP6 land-use pathways contribute least to the total uncertainty, highlighting the priority of refining model parameterizations.

Finally, a comparison of simulations forced by CMIP6 (LUH2) and CMIP7 (LUH3) land-use datasets reveals that while global totals of carbon sink and LULCC emissions are similar, notable spatial discrepancies exist in key regions. Emissions in the LUH3-driven simulation show a marked increase (~40%) over the last three decades, attributable to accelerated forest-to-cropland conversion and intensified management in the updated dataset. This work underscores the critical roles of model process representation and input data in assessing the carbon impacts of land-use change.