Fall Meeting 2022

December 12 - 16, 2022

Coarse-resolution outcomes of climate models often need to be downscaled to finer spatial or temporal scales to be suitable for regional climate change impact assessment. The downscaling of climate variables is central to numerous applications: hydrological modeling, weather forecasting, energy optimization, natural hazard risk assessment, and mitigation, to name a few. Whereas numerous physical (using regional numerical models) and statistical downscaling methods exist, the fundamental understanding of effective scaling laws and their efficacious computation remains elusive. One community response has been the rapidly emerging interest in coupling learning machines and systems with numerical "physics-based" models to do better than either source alone. A second direction is to discover effective equations of the scaling dynamics, from data and a few detailed simulations, for repeated rapid subsequent use. This session aims to share progress and recent developments in foundational methodology and computational advances that couple learning and physics for tractable and interpretable downscaling with application to natural hazards and climate.