Ensemble Forecasting at NCEP and the Breeding Method

作者: Zoltan Toth , Eugenia Kalnay

DOI: 10.1175/1520-0493(1997)125<3297:EFANAT>2.0.CO;2

关键词:

摘要: The breeding method has been used to generate perturbations for ensemble forecasting at the National Centers Environmental Prediction (formerly known as Meteorological Center) since December 1992. At that time a single cycle with pair of bred forecasts was implemented. In March 1994, expanded seven independent cycles on Cray C90 supercomputer, and were extended 16 days. This provides 17 global valid two weeks every day. For efficient forecasting, initial control analysis should adequately sample space possible errors. It is shown like cycle: it acts nonlinear perturbation model upon evolution real atmosphere. (i.e., error), carried forward in first-guess forecasts, ‘‘scaled down’’ regular intervals by use observations. Because this, growing errors associated evolving state atmosphere develop within dominate subsequent forecast error growth. simulates development cycle. A difference field between (and scaled down intervals) atmospheric fields. By construction, vectors are superpositions leading local (timedependent) Lyapunov (LLVs) An important property all random assume structure LLVs after transient period, which large-scale processes about 3 When several performed, phases amplitudes individual regional) random, ensures quasi-orthogonality among from cycles. Experimental runs 10-member (five cycles) show mean superior an optimally smoothed randomly generated compares favorably medium-range double horizontal resolution control. Moreover, potentially useful relationship spread also found both spatial domain. improvement skill 0.04‐0.11 pattern anomaly correlation beyond 7 days, together potential estimation skill, indicate this system operational tool. methods so far produce forecasts—that is, adjoint (or ‘‘optimal perturbations’’) technique applied European Centre Medium-Range Weather Forecasts—have significant differences, but they attempt estimate subspace fast perturbations. provide estimates fastest sustainable growth thus represent probable optimal perturbations, other hand, future. practical simpler less expensive than technique.

参考文章(40)
Lev Semenovich Gandin, Objective Analysis of Meteorological Fields ,(1963)
Grant Branstator, The Variability in Skill of 72-hour Global-Scale NMC Forecasts Monthly Weather Review. ,vol. 114, pp. 2628- 2639 ,(1986) , 10.1175/1520-0493(1986)114<2628:TVISOH>2.0.CO;2
Carolyn A. Reynolds, Peter J. Webster, Eugenia Kalnay, Random Error Growth in NMC's Global Forecasts Monthly Weather Review. ,vol. 122, pp. 1281- 1305 ,(1994) , 10.1175/1520-0493(1994)122<1281:REGING>2.0.CO;2
P. L. Houtekamer, Jacques Derome, Prediction Experiments with Two-Member Ensembles Monthly Weather Review. ,vol. 122, pp. 2179- 2191 ,(1994) , 10.1175/1520-0493(1994)122<2179:PEWTME>2.0.CO;2
Giancarlo Benettin, Luigi Galgani, Jean-Marie Strelcyn, Kolmogorov entropy and numerical experiments Physical Review A. ,vol. 14, pp. 2338- 2345 ,(1976) , 10.1103/PHYSREVA.14.2338
P. L. Houtekamer, Louis Lefaivre, Jacques Derome, Harold Ritchie, Herschel L. Mitchell, A System Simulation Approach to Ensemble Prediction Monthly Weather Review. ,vol. 124, pp. 1225- 1242 ,(1996) , 10.1175/1520-0493(1996)124<1225:ASSATE>2.0.CO;2
Anna Trevisan, Roberto Legnani, Transient error growth and local predictability: a study in the Lorenz system Tellus A. ,vol. 47, pp. 103- 117 ,(1995) , 10.3402/TELLUSA.V47I1.11496
P. L. Houtekamer, Jacques Derome, Methods for Ensemble Prediction Monthly Weather Review. ,vol. 123, pp. 2181- 2196 ,(1995) , 10.1175/1520-0493(1995)123<2181:MFEP>2.0.CO;2