ENSEMBLE FORECASTING AT NMC - THE GENERATION OF PERTURBATIONS

On 7 December 1992, The National Meteorological Center (NMC) started o perational ensemble forecasting. The ensemble forecast configuration i mplemented provides 14 independent forecasts every day verifying on da ys 1-10. In this paper we briefly review existing methods for creating perturbations for ensemble forecasting. We point out that a regular a nalysis cycle is a ''breeding ground'' tor fast-growing modes. Based o n this observation, we devise a simple and inexpensive method to gener ate growing modes of the atmosphere. The new method, ''breeding of gro wing modes,'' or BGM, consists of one additional, perturbed short-rang e forecast, introduced on top of the regular analysis in an analysis c ycle. The difference between the control and perturbed six-hour (first guess) forecast is scaled back to the size of the initial perturbatio n and then reintroduced onto the new atmospheric analysis. Thus, the p erturbation evolves along with the time-dependent analysis fields, ens uring that after a few days of cycling the perturbation field consists of a superposition of fast-growing modes corresponding to the contemp oraneous atmosphere, akin to local Lyapunov vectors. The breeding cycl e has been designed to model how the growing errors are ''bred'' and m aintained in a conventional analysis cycle through the successive use of short-range forecasts. The bred modes should thus offer a good esti mate of possible growing error fields in the analysis. Results from ex tensive experiments indicate that ensembles of just two BGM forecasts achieve better results than much larger random Monte Carlo or lagged a verage forecast (LAF) ensembles. Therefore, the operational ensemble c onfiguration at NMC is based on the BGM method to generate efficient i nitial perturbations. The only two methods explicitly designed to gene rate perturbations that contain fast-growing modes corresponding to th e evolving atmosphere are the BGM and the method of Lorenz, which is b ased on the singular modes of the linear tangent model. This method ha s been adopted operationally at The European Centre for Medium-Range F orecasts (ECMWF) for ensemble forecasting. Both the BGM and the ECMWF methods seem promising, but since it has not yet been possible to comp are in detail their operational performance we limit ourselves to poin ting out some of their similarities and differences.