Title: “Leveraging Subgrid-Scale Spatial Organization and Variability for Improved Cloud Fraction Parametrization Using PINACLES Simulations”

Format: Hybrid

Virtual: Zoom link provided upon registration

Abstract: Given the coarse resolution of Earth System Models (on the order of 100 km), cloud processes must be parametrized. As a consequence, cloud processes are a major source of uncertainty in climate projections, with particular intermodel disagreement in the representation of low marine boundary layer clouds. Cloud fraction parametrizations have long relied upon subgrid-scale information of the total water distribution using various probability density functional forms. However, the importance of subgrid-scale spatial organization, including coherent updraft and downdraft, for cloud fraction has not been fully examined. In this study, we use machine learning to implicitly learn the subgrid-scale organization of shallow clouds and assess the information gained from including these subgrid features in a neural network-based parametrization. For this work, we use data from the Predicting INteractions of Aerosol and Clouds in Large Eddy Simulations (PINACLES) model for domains in the Eastern North Atlantic, Northeastern Pacific, Southern Great Plains, and Southern Ocean. Particular focus is given to the varying importance of subgrid-scale variability across shallow cloud regimes.

Bio: Margaret (Maggie) is interested in improving the parameterization of marine boundary layer clouds using machine learning and high-fidelity numerical simulations. As a DOE Computational Science Graduate Fellow, she is excited to apply high-performance computing methods within atmospheric science. Maggie has previously worked as a data scientist at a climate-tech startup and as a researcher at an environmental consulting firm. She received her A.B. in Earth & Planetary Sciences from Harvard University, where she researched Arctic methane emissions.

Time: 12:00-1:30 pm EST

Free!

Registration