Speaker: Brian Hunt, Mathematics, University of Maryland, College Park

Title: Determining the initial conditions for a weather forecast

Abstract: A key factor limiting the accuracy of weather forecasts is the accuracy of the initial conditions used. Global weather forecast models require as input meteorological variables (temperature, wind speed, etc.) at a regular grid of points throughout the atmosphere. Accurate measurements of these variables are not available at nor even near many of the grid points. In practice, the initial conditions are formed by a statistical interpolation between the available measurements and a prior forecast. This interpolation, involving millions of variables, must be done in a computationally efficient manner. I will describe an approach to this problem that we have developed at the University of Maryland, which we call a Local Ensemble Transform Kalman Filter. As with other Ensemble Kalman Filters, we track an ensemble of solutions of the forecast model in order to assess the likely states of the atmosphere at a given time, and periodically adjust the ensemble to select the states that are most likely in light of newly collected measurements. Unlike other methods, we do the adjustment in a manner that is explicitly geographically local, which both improves the accuracy of the results and allows for a massively parallel implementation. Our method can be used more generally to estimate the state of a spatiotemporally chaotic system for which a forecast model is known but only limited measurements are available.

Time: Friday, Feb. 20, 2009, 1:30-2:30 p.m.

Place: Science and Tech I, Room 242

Department of Mathematical Sciences
George Mason University
4400 University Drive, MS 3F2
Fairfax, VA 22030-4444
http://math.gmu.edu/
Tel. 703-993-1460, Fax. 703-993-1491