Upcoming Event:

Understanding Geoscience Processes using Nonlinear Data Assimilation and Information Flows

Peter Jan van Leeuwen, Professor, Colorado State University

Abstract

Many systems in the geosciences are high-dimensional and highly nonlinear. Examples are Hurricane rapid intensification, atmospheric cloud processes, ocean sub-mesoscale circulations and their interaction with the atmosphere, sea-ice dynamics, and subsurface flows. In this talk, I will demonstrate our latest efforts to combine observations and models using nonlinear data assimilation, extract nonlinear causal information from observations of complex geophysical systems, and infer information flows in complex numerical models. And, yes, all high-dimensional. I will explain nonlinear data assimilation using Particle Flow Filters and their application to extreme weather events, including how machine learning is incorporated. Then, we dive into nonlinear causal discovery of cloud processes and their radiative effect on Earth's climate and show how much richer nonlinear causal discovery is from its linear counterpart. This is followed by how we can infer how (Shannon) information flows in complex systems with an application to the atmosphere. I will briefly touch upon the influence of rain layers on the ocean heat budget and new insights into cloud aggregation. Finally, I elaborate on future plans.

Biography

Peter Jan van Leeuwen is a professor in the Department of Atmospheric Science at Colorado State University. His research interests focus on the development and use of data assimilation (including machine learning), causality (cause and effect relations), and information theory for better understanding geophysical fluids, with emphasis on the atmosphere, the ocean, and the cryosphere. Application areas include ocean dynamics, air-sea interaction (rain layers, momentum transfer), sea-ice dyamics and prediction, cloud microphysics, aerosol-cloud-precipitation interactions, cloud aggregation, shallow-cumulus initiation, and hurricane dynamics and prediction.