Traceability Analysis of Ensemble Modeling Results with A Unified Diagnostic System to Improve Land Carbon Cycle Predictions

Wednesday, December 12, 2018 - 13:40
Add to Calendar

Ensembles of land models become a major means in many model inter-comparison projects (MIPs) to assess model performance skills. This ensemble modeling approach is to learn about model performance mainly through comparison of model outputs. The latter has ubiquitously revealed great uncertainty that individual models perform so differently among and within MIPs. Scientists have made great effort in an attempt to identify causes of model uncertainty from MIPs. However, the output-based ensemble modeling is not very effective to identify the causes of uncertainty as targets of model improvement. We have recently developed a unified diagnostic system for uncertainty quantification of land carbon storage dynamics. The diagnostic system includes three elements: one dynamical equation to unify all land carbon cycle models, one three-dimensional space to evaluate all model outputs, and five traceable components to pinpoint uncertainty sources. The dynamical equation is expressed in a matrix form or called a matrix approach. We have converted a couple dozen of existing models, such as ORCHIDEE, CLM 3.5, 4.0, 4.5, and 5.0, to the unified matrix form. Meanwhile, we have developed a matrix-based land model uncertainty analysis platform based on the super-computing cluster in Wuxi, China and a new community-based ESM, Community Integrated Earth System Model (CIESM). For the first step, we focus on the soil C, the largest land C pool but is poorly predicted by current ESMs. Specifically, multiple soil C decomposition schemes from different land models (CABLE, CLM-CENTURY, CLM-BGC, ORCHIDEE and JULES) have been incorporated into the unified matrix form. Thereafter, we analyzed the uncertainty sources (carbon input, residence time and storage potential) within the traceability framework and quantitatively attributed them into more specific ecological components. Those identified major uncertain components were further constrained by a matrix-based data assimilation system. The platform will thus provide a unique unified uncertainty assessment tool for both data-model and inter-model comparisons.

Link for More Information: