Hybrid short-range weather forecast systems

It has been clear for some time that on large space and/or time scales numerical weather prediction models show the greatest forecast skill. Conversely, on small spatial and temporal scales image extrapolation techniques tend to preserve the observed small scale characteristics, many of which cannot be represented reliably in current mesoscale models. Such 'nowcasting' techniques can maintain practical levels of forecast skill out to tens of minutes. The challenge presented by this state of affairs is deciding how we should go about performing real-time weather prediction on the space and time scales where both systems exhibit comparable skill. The first step in developing a hybrid NWP-extrapolation weather forecast system of this type is to compare the forecasts made by the individual techniques, assuming that they have access to the same meteorological data. We will present the results of such an analysis, using forecasts of precipitation, and discuss the results in the context of recent work performed by other groups in this field.