GC105-01
Machine Learning Based Process-oriented Earth System Model Evaluation

Tuesday, 15 December 2020: 05:30
Virtual
Veronika Eyring, German Aerospace Center (DLR), Oberpfaffenhofen, Germany, Gustau Camps-Valls, Image Processing Laboratory, Universitat de València, Paterna, Spain, Pierre Gentine, Columbia University, Earth and Environmental Engineering, New York, NY, United States, Markus Reichstein, Max Planck Institute for Biogeochemistry, Department of Biogeochemical Integration, Jena, Germany, Jakob Runge, German Aerospace Center, Jena, Germany and Manuel Schlund, German Aerospace Center DLR Oberpfaffenhofen, Earth System Model Evaluation and Analysis, Oberpfaffenhofen, Germany
Recorded Presentation
Abstract:
Earth system models (ESMs) are complex and represent a large number of processes, resulting in a persistent spread across climate projections for a given future scenario. The Coupled Model Intercomparison Project Phase 6 (CMIP6) is confronted with a number of new challenges. Compared to the former phase (CMIP5), an increased number of institutions participate in CMIP6, many with multiple model versions. The latest generation of climate models feature increases in spatial resolution, improvements in physical parameterizations (e.g., in the representation of clouds) and inclusion of additional Earth system processes (e.g., nutrient-limitations on the terrestrial carbon cycle). These additional processes are needed to represent key feedbacks that affect climate change, but also increase the spread of climate projections across the multi-model ensemble. This increases the need for innovative and comprehensive model evaluation approaches. An exciting opportunity is provided by the Earth System Model Evaluation Tool (ESMValTool) that addresses big data challenges and for the first time in CMIP allows rapid, quantitative comparisons of model results to a wide range of observations as soon as the model output is submitted to the CMIP archive. Such rapid and comprehensive feedback on model performance will help address the causes of long-standing systematic errors and facilitate a shift towards more process-oriented diagnostics, while ensuring continuity with more ‘traditional’ diagnostics applied in previous CMIP phases. Other promising diagnostic developments on the horizon include the application of innovative data science methods in Earth system science, in particular from deep learning and causal discovery. Results from causal model evaluation, emergent constraints and weighting of climate model projections from the European Research Council (ERC) Synergy Grant “Understanding and Modelling the Earth System with Machine Learning (USMILE)” will be presented.
Previous Abstract | Next Abstract >>