Dynamics of Data Assimilation

I attended the colloquim titled, Dynamics of Data Assimilation, by Liz Bradley in the computer science department. The talk was actually part of the applied math department. Dr. Bradley talked about numerical solvers and the difficulties posed by unmodelled effects like noise. She focused primarly on complex nonlinear systems like fluid flows and how to address problems in these systems. She proposed a method of using observations of the system being modelled to periodically correct the simulation. However, changing the state variables of a running solver can generate additional difficulties, which were also discussed in the talk. Issues were discussed in the context of a planar air jet.

I enjoyed this talk. The math was a little difficult to follow at some points, but the overall concept was interesting. Discussions of non-linear dynamics make it easy to see why it's so difficult to predict the path of a hurricane, or any system subject to unknown perterbations. From an AI perspective, understanding non-linear dynamics can help explain why AI algorithms have not able to model every problem.

Last modified 10 December 2007 at 2:54 pm by RhondaHoenigman