Building an Australian community framework for Evaluation of Earth System Models
Romain Beucher1, Michael Tetley1, Rhaegar Zeng1, Dougal Squire1, Sean Bryan1, Kelsey Druken1 1ACCESS-NRI, Canberra, ACT, Australia
Abstract
Climate models represent an integral part of international efforts to assess the likelihood of climate change scenarios and are fundamental to developing policies around climate actions and investments. Simulations rely on accurate modelling of climate physics and realistic forcing conditions (e.g., CO2 emissions). However, model development is a lengthy iterative process that requires constant evaluation of both model performance and accuracy in matching observations. Additionally, users typically want to compare model simulations produced from various configurations and evaluate how closely they match observations. These analyses are achieved by running “hindcasts” of variables and comparing them with historical records, requiring dedicated tools, workflows, and access to diverse data sources working seamlessly in high-performance computing and data environments.
The Australian Earth System Simulator (ACCESS-NRI) is a national research infrastructure created to support the development and research of the Australian Community Climate and Earth System Simulator (ACCESS) suite of models. One of ACCESS-NRI’s strategic goals is to improve quality and performance when running the ACCESS suite of model configurations, including the open development and release of a Model Evaluation and Diagnostics (MED) framework for the Australian Earth system modelling community. Here we present some community-driven strategies to streamline the evaluation of the ACCESS suite of models. We discuss community needs and constraints with a particular focus on the World Climate Research Programme’s Coupled Model Intercomparison Project phase 6 (CMIP6) and its future iterations. We also address the need to incorporate model evaluation in code development cycles.
Biography
Romain is the Leader of the ACCESS-NRI Model Evaluation and Diagnostics. He is highly experienced in data exploration and analysis where he provides technical support, develops open-source software solutions and provides training to facilitate effective research and development of workflows.
Romain has 12 years of research experience working on world-leading computational geoscience research groups in France, Norway and Australia.