Can Machine Learning Diagnose Extremes

Abstract

Project AbstractApproved for Public ReleaseDeep-Learning-Weather-Prediction (DLWP) has shown great promise as a method for improvi,ng subseasonal-to-seasonal (S2S) forecasts, as well as weather forecasts at shorter lead times. Two key aspects of DLWP forecast mod,els are their extreme computational efficiency and their abil- ity to learn physical parameterizations in a holistic sense as part o,f the model training. DLWP?s computational efficiency allows the generation of ensemble forecasts of unprecedented size, with thousa,nds of members. Such large ensembles, if well-calibrated, could be very useful in identifying extreme events. This is a proposal for, a short project to fund an incoming graduate student and the PI to assess a crucial property of such DLWP ensembles: their ability,to accurately capture extremes. In particular, we will test the ability of ML to diagnose extreme surface temperatures when trained,only on less extreme events.

Document Details

Document Type
DoD Grant Award
Publication Date
Oct 06, 2022
Source ID
N000142212807

Entities

People

  • Dale R. Durran

Organizations

  • Office of Naval Research
  • United States Navy
  • University of Washington

Tags

Fields of Study

  • Environmental science

Readers

  • Neural Network Machine Learning.
  • Ocean-Atmosphere Mesoscale Modeling, Data Assimilation, and Flux Boundary Layers
  • Systems Analysis and Design

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference
  • AI & ML - Neural Networks