4.3
NCEP Global Ensemble Forecast System in Past 20 Years

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Wednesday, 7 January 2015: 4:30 PM
229A (Phoenix Convention Center - West and North Buildings)
Yuejian Zhu, NOAA/NWS/NCEP/EMC, College Park, MD

NCEP Global Ensemble Forecast System in Past 20 Years

Yuejian Zhu

Environmental Modeling Center, NCEP, NWS/NOAA

5830 University Research Ct. College Park, MD 20740 Tel: 301-683-3709; Fax: 301-683-3703; E-mail: Yuejian.Zhu@noaa.gov

NCEP Global Ensemble Forecast System (GEFS) was introduced in daily operation since 1992 by Kalnay and Toth (Toth and Kalnay, 1993; 1997). It was started once per day for T62L28 (about 200km) resolution, initiated from breeding vector (BV) method with 24 hour cycling, integrated to 12 days for two members only. Through the development of 20 more years, NCEP GEFS is an major system in daily operation, which is running 4 times (totally 84 members) per day, at T254L42 (about 50km) resolution, initiated through BV method and ensemble transform with rescaling (ETR) of 6-hour cycling, integrated to 16 days. Later of 2015, it will be upgraded to T574L64 (33km) resolution, initiated from EnKF data analysis directly.

Currently, NCEP GEFS is fully assimilating initial uncertainties through BV-ETR (will be replaced by EnKF later 2015), and model uncertainties from Stochastic Total Tendency Perturbation (STTP). In addition, tropical storm (TS) relocation has been introduced in 2005 to reduce unrealistic initial uncertainty. With the improvement of global data assimilation, global forecast system and ensemble techniques, NCEP GEFS improved forecast skill (60% of yearly average Northern Hemisphere 500hPa geopotential height anomaly correlation) from about 6 days to 9.5 days. In recent years, NCEP GEFS hurricane track forecast is one of the major guidance for NHC forecasters.

In order to improve NCEP GEFS forecast uncertainty and skill of surface elements, many stochastic physics schemes have been tested for future implementation. Meanwhile, stochastic parameterization for deep convection, cloud physics, radiations, boundary layer, land surface, SST and others are in plan to test for further improvement of precipitation, temperature and tropical storm forecasts.