Evaluating Modeled Terrestrial Carbon Cycles in Earth System Models and Offline Models Using Multiple Observation-based Estimates

Improvement of terrestrial models in earth system models (ESMs) is important to reduce uncertainties in future projections of global carbon cycle and climate. To test performances of terrestrial models in ESM, many studies from ESM communities evaluated terrestrial model outputs from ESMs. On the other hand, communities of offline terrestrial models, which run with observed climate data, also have evaluated using observation data. Therefore, evaluation of terrestrial models have been mostly separately conducted by ESM community and offline model community. In this study, we evaluated the terrestrial outputs from ESMs and offline models with the multiple remote sensing based datasets. We used CMIP-5 outputs as ESM outputs, and TRENDY outputs as offline model outputs. As observation datasets, we used various satellite-based datasets, in particular, data-driven estimation of terrestrial CO₂ and H₂O fluxes (e.g. FLUXCOM products). As a result, for example, regarding the gross primary productivity (GPP), model-by-model differences in offline models were smaller than those in ESMs. On the other hand, for net biome productivity (NBP), ESM outputs has less uncertainty than offline models. It might be because ESMs focus on the phenomenon on the global scale and is modeled so as to better match the change in CO₂ concentration at global scale. On the other hand, with regard to the offline model, emphasis is placed on the reproducibility of individual processes, and it is considered that there are many variations among models as to global totals. From these results, we can expect model improvements by bringing advantages in both ESM and offline model.