Title: Exploring physical and machine learning approaches for stochastic modeling and ensemble prediction of weather and climate
Speaker: Aneesh Subramanian, Asst. Prof., Atmospheric and Oceanic Sciences, CU Boulder

Abstract:
Convection and cloud processes play a key role in the dynamics of the atmosphere just as mesoscale turbulence and deep convection do in ocean dynamics and ice-shelf processes do in ice sheet dynamics. Yet, even today our shortcomings in parameterizing these subgrid scale processes in global climate models (GCMs) are limiting our ability to simulate and understand the climate and weather of the planet. Recent innovative ideas on convection parameterization such as super-parameterization (embedding cloud-resolving models within the GCM grid), stochastic-parameterization or machine learning emulation in weather and climate models have helped improve its representation of the climate and weather systems. These approaches in parameterization have emerged as new paths forward and complement the conventional approaches rather than replace them. We study the impact of these approaches on forecasts from weather to climate timescales. Results from studies using stochastic parameterization in ensemble forecasting systems as well as machine learning approaches for causal discovery, feature detection and ensemble prediction post-processing will be presented. In addition, results from using machine learning approaches for stochastic parameterization of subgrid-scale variability in an idealized system will be presented to motivate future studies in this direction with the weather and climate forecasting systems. This has implications on improving conventional parameterization using hybrid approaches as we await the exascale computing systems of the future to resolve key processes in climate models.

Speaker Bio:
Dr. Aneesh Subramanian’s research interests are mainly in climate processes, global and regional weather and climate prediction and coupled ocean-atmosphere data assimilation.  He is a climate scientist who studies the predictability of global weather systems such as atmospheric rivers on medium-range to sub-seasonal timescales and oceanic influences on prediction in these timescales using both computational models and observational analysis. He is also focused on improved understanding of the physical processes and their representation in computer models for helping improve the predictions. Sub-seasonal to seasonal prediction is a key area of research highlighted in the “Weather Research and Forecasting Innovation Act of 2016”.  

Dr. Subramanian received his Ph. D. (2012) from the Scripps Institution of Oceanography and completed a post-doctoral appointment in SIO from 2012-2014. He then went on to become a Post-Doctoral Research Scientist and Lecturer in the Physics Department at the University of Oxford from 2014-2017. During his term as a Postdoctoral Research Scientist in the University of Oxford, Dr. Subramanian has studied advanced weather modeling techniques and their impact on predictability of the El Niño Southern Oscillation, the Madden-Julian Oscillation, tropical cyclones, and atmospheric rivers (ARs). Importantly, his skills in weather prediction and expertise in sub-seasonal to seasonal prediction is key to pursuing and achieving the current challenges in extending weather forecasting.  He then joined the Center for Western Weather and Water Extremes as a Project Scientist in 2017 to lead efforts on subseasonal prediction and data assimilation to help improve Western US extreme weather predictions. Dr. Subramanian then joined the Atmospheric and Oceanic Sciences Department at the University of Colorado Boulder in 2019 as an Assistant Professor and continues his research on prediction of our earth system from weather to climate timescales as well as improving our understanding of climate processes. Dr. Subramanian is the author or co-author of more than 80 refereed publications.