Plasma Physics Seminar ( Phys 769)

Dr. Brian Hunt, Department of Matematics, University of Maryland

Determining the Initial Conditions for a Forecast

While it is generally agreed that the earth's atmosphere is chaotic, applications of the theory of chaotic dynamical systems are most successful for low-dimensional systems -- those that can be modeled with only a few state variables. By contrast, current numerical weather models keep track of over 1,000,000 variables. Nonetheless, our group at the University of Maryland has found that in a local region (say the Eastern United States) over a time span of a few days, the model behaves much like a low-dimensional system. I will describe how we are using this perspective to attack a fundamental problem in weather forecasting -- how to set the initial conditions for the forecast model. Our methodology can be used more generally to forecast spatiotemporally chaotic systems for which a model is available but initial conditions cannot be measured accurately.