Recent Results from Application of the Implicit Particle Filter to High-dimensional Problems

We present our most recent results on the application of the implicit particle filter to a stochastic shallow water model of nearshore circulation. This highly nonlinear model has approximately 30,000 state variables, and, in our twin experiments, we assimilate 32 observed quantities. Application of most particle methods to problems of this size are subject to sample impoverishment. In our implementation of the implicit particle filter, we have found that manageable size ensembles can still retain a sufficient number of independent particles for reasonable accuracy.