Seminar - Incorporating model discrepancy into a data assimilation framework

12/03/2013 - 12:00pm
Mesa Laboratory - Chapman Room

Incorporating model discrepancy into a data assimilation framework

Cari Kaufman and Linda Tran, University of California Berkeley

Data assimilation allows us to incorporate observations into a physical model as the observations come online. Applications include weather forecasting and GPS systems; methods include extended Kalman filters, particle filters, variational assimilation, and many others. One key assumption of the data assimilation framework is that the physical model is the correct representation of reality. However, this is often not true -- the physical model is an approximation and can therefore be discrepant with reality. We examine inter-battery factor analysis, a Bayesian generative model for canonical correlation analysis (CCA), as a method to predict model discrepancy and show some preliminary results using Lorenz 2005 as a toy example.

Tuesday, December 3, 2013

Mesa Lab Chapman Room

Time: 12:00 PM