Combining dynamical and statistical ensembles

Abstract

Aprediction accompanied by quantitative estimates of the likely forecast accuracy is inherently superiorto a single “best guess” forecast. Such estimates can be obtained by “dressing” a single forecast usinghistorical error statistics. Dressing ensemble forecasts is more complicated, as one wishes to avoiddouble counting forecast errors due, for example, to uncertainty in the initial condition when thatuncertainty is explicitly accounted for by the ensemble (which has been generated with multiple initialconditions). The economic value of dressed forecasts has been demonstrated by previous studies. Thispaper presents a method for dressing ensembles of any size, thus enabling valid comparisons to bemade between them. The method involves identifying the “best member” of an ensemble in a multidimensionalforecast space. The statistics of the errors of these best members are used to dress individualforecasts in an ensemble. The method is demonstrated using ECMWF ensemble forecasts, which arecompared with the ECMWF high-resolution best guess forecasts. It is shown that the dressed ECMWFensembles have skill relative to the dressed ECMWF best guess, even at the maximum lead time ofthe ECMWF forecasts (10 days). The approach should be applicable to general ensemble forecasts(initial condition, multi-model, stochastic model etc.), allowing better informed decisions on forecastaquisition and forecast system development.