HR: 0830h
AN: NG51A-0827    [PDF]
TI: Space-time multifractal models of rain, predictability and nowcasting
AU: * Lovejoy, S
EM: shaun@physics.mcgill.ca
AF: Dpt of Physics, McGill University, Montreal, Qc H3A2T8 Canada
AU: Schertzer, D
EM: Daniel.Schertzer@cereve.enpc.fr
AF: CEREV, Ecole Nationale des Ponts et Chaus‚es, Marne-la vallee, 77455 France
AB: Rainfall nowcasting is currently performed with the help of sequences of radar or satellite derived rainfall surrogate fields, the usual assumption being that the field is "frozen" and is simply advected i.e. without any development. In this talk, we argue that space-time multifractal processes provide appropriate frameworks for modeling rain and that this can be used to forecast both the advection and the development. Necessary ingredients include the correct degree of space-time stratification, vertical-horizontal stratification as well as estimates of the basic (universal mulitfractal) parameters. In addition, special care must be given to numerical stability, and to structures larger than the model domain, and to small subpixel scale structures. We show theoretically and numerically that states with a shared past diverge algebraically (rather than exponentially) in time, and study the dependence on spatial scale. We also show how forecasts can be made using conditional expectations and quantify the error growth.
DE: 1854 Precipitation (3354)
DE: 3220 Nonlinear dynamics
DE: 3250 Fractals and multifractals
DE: 3354 Precipitation (1854)
SC: Nonlinear Geophysics [NG]
MN: 2003 Fall Meeting