There is little doubt that hybrid mesoscale model and image extrapolation s
ystems will eventually become widespread for the nowcasting of precipitatio
n for hydrology and aviation safety. Arguably the final forecast output tha
t is required for these applications needs to be probabilistic in nature, t
herefore appropriate forecast schemes should statistically account for both
the chaotic and intermittent nature of the intrinsic fields. In principle,
the detailed knowledge that we have about the high-order statistics of rai
n patterns should also allow improved extrapolation schemes for nowcasting.
A well-known technique that preserves space-time statistics is presented a
s an example of this. It is argued that the relative scoring of forecast sy
stems is highly space-time resolution dependent, and that even when the sam
e meteorological event is being considered this resolution dependency must
be taken into account. (C) 2000 Elsevier Science B.V. All rights reserved.