A space-time Kalman filter for combining satellite radiance data with a sediment transport model

Jonathan Stroud
University of Pennsylvania

Satellite data provide a rich source of information about the earth and its atmosphere due to their high spatial resolution and global coverage. An important use of satellite data is calibration of physical models. The Kalman filter (KF) provides a natural framework for combining these two sources of information - satellite images and physical models. However, because of the high-dimensional state and observation vectors, the exact KF recursions cannot be implemented in practice.

In this talk, we propose a dimension-reduced KF algorithm for satellite data assimilation based on the idea of covariance tapering. We apply the algorithm to combine daily images of visible reflectance from the SeaWiFS sensor with a two-dimensional sediment transport model in Lake Michigan. Sensitivity of the approximation to the choice of the tapering radius is assessed by comparing the results to an ensemble Kalman filter (EnKF). We find that a fairly small tapering radius provides an excellent approximation to the EnKF solution at a substantial reduction in computational cost.

Joint work with Michael Stein, Barry Lesht, Dmitry Beletsky and Dave Schwab.

Back to agenda