Event Detail

Event Type: 
Thursday, June 3, 2021 - 10:00 to 12:00
Zoom - If you are interested in attending this presentation, please send an email to Nikki Sullivan - nikki.sullivan@oregonstate.edu - to request Zoom log in details.

Many geophysical phenomena exhibit complicated dynamics that, due to a variety of factors, diverge quickly from physical models. The arrival of new observations allows researchers to combine the model estimate with measurements in a statistical process called data assimilation to produce a revised estimate of the phenomenon. This assimilation of observations with geophysical models, particularly hyperbolic wave-like processes, can be frustrated by the combination of sparse, infrequent observations and model error misspecification that leads to unphysical artifacts in the forecast. Furthermore, sparse but highly accurate measurements render the task of calibrating the uncertainties of the model forecast to reality difficult. In this presentation, I detail the capabilities of the dynamic likelihood approach to data assimilation (DLF), a Bayesian data assimilation scheme that leverages the dynamic characteristics of geophysical wave-like phenomenon to propagate observation information forward in time to attenuate problems that appear in traditional data assimilation strategies. This methodology is particularly effective when observations have small inherent measurement errors and are sparse in space and time -as is often the case in geophysical wave problems. I will demonstrate the capability and flexibility of the DLF against traditional data assimilation approaches on test problems in one and two dimensions. The DLF is not only applied to stationary observation networks that are common for atmospheric processes but also moving observation systems commonly used for ocean processes. For these experiments the DLF not only improves mean estimate errors but also dramatically improves the quantification of the forecast uncertainty.