Physics Constrained Stochastic-Statistical Models for Extended Range Environmental Prediction
Abstract
Using a combination of theoretical, observational, and high-order model resources, we aim to develop and improve the modelling and forecasting capabilities for crucial problems such as long range weather forecasting of planetary scale convection patterns in the tropics and short term climate change in the polar regions such as sea ice reemergence. To develop simplified physics constrained stochastic statistical models and techniques for long range environmental forecasting by blending novel ideas from mathematics, statistics and physics and validating the skill of these new models on a suite of tests ranging from observational data to data output from high order models like GCM s to synthetic data from instructive toy models. These objectives include 1) the systematic development of low order (few dimensional) stochastic statistical models, 2) intermediate models with hundreds of variables, as well as 3) novel strategies for improvement of higher order models like GCM s. These objectives include the development and application of new techniques for 1) finding and assessing the intrinsic prediction skill in crucial variables associated with massive data outputs from observation or high order models, 2) the development of multi-scale data assimilation and parameter estimation algorithms, and 3) the crucial understanding of the role of model errors in data assimilation and prediction both to reveal intrinsic information barriers in model classes and to develop strategies to mitigate such errors.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 30, 2014
- Accession Number
- ADA615968
Entities
People
- Andrew J. Majda
- Dimitrios Giannakis
- Duane E. Waliser
- John Harlim
- Samuel N. Stechmann
Organizations
- New York University