22nd Conference on Weather Analysis and Forecasting/18th Conference on Numerical Weather Prediction

J4.2

12 Years of NCEP Short-Range Ensemble Forecasting (SREF) System from 1995 to 2006

Jun Du, NOAA/NWS/NCEP, Camp Springs, MD; and G. DiMego, J. McQueen, Z. Toth, S. Tracton, B. Zhou, D. Jovic, G. Manikin, B. Ferrier, E. Rogers, H. Juang, H. Y. Chuang, D. Stensrud, S. J. Weiss, R. H. Grumm, P. Manousos, and S. Silberberg

Encouraged by the success of global medium-range ensemble practice at NCEP and ECMWF, a prototype Short-Range Ensemble Forecasting (SREF) system with regional models was setup and run at NCEP during late 1995 to early 1997 roughly once or twice per week for targeted weather events. A new version of it with "regional breeding" was setup in 1997 and run once per day in real-time mode during the SAMEX'98 experiment. Based on the promising results, this new version SREF started to run twice per day (but in a delayed mode) from April 2000 for case studies and systematic evaluation and became part of the NCEP's operational suite in May 2001 with 10 members from two regional models over the North America domain. This was the first real-time, operational regional ensemble prediction system in the world at the time.

From the very beginning of its development stage, the NCEP SREF system has strategically emphasized both initial condition (IC) and model/physics uncertainties by using "multi-analysis, multi-LBCs, perturbed ICs and multi-model (different dynamics and physics)" this mixed approach. The SREF was pioneering in the multi-model ensembling approach at the time. After the NCEP SREF system became real-time in production in May 2001, three major upgrades were made to further address model/physics uncertainties. In September 2003, another set of five Eta members but with a different version of convective scheme was added for a total of 15 members. The date of August 17, 2004 marks a new level of addressing physics diversity by implementing a total of six various convective schemes to the system besides perturbing ICs in 12 out of its 15 members with breeding method. As WRF modeling framework made available to NWP community, six WRF-based members were added to the NCEP SREF system on December 6, 2005. Within, three members use the NCEP NMM core, while other three members use the NCAR ARW core to add diversity. This implementation marks the beginning of transition of NCEP SREF into WRF era and also the first use of WRF infrastructure in real-time NWP operation over the entire US. By then, there are a total of 21 members in the NCEP SREF system with four different regional models. It runs four cycles (03, 09, 15, 21z) per day to 87 hours. It has various horizontal resolutions among models from 32 to 45 km and covers almost the entire North America including Alaska and Hawaii. Numerous real-time ensemble products are available at http://wwwt.emc.ncep.noaa.gov/mmb/SREF/SREF.html.

The NCEP SREF now becomes an integrated and crucial part of daily weather forecasting process in the US National Weather Service. How it was utilized by NCEP service centers and field forecasters will be illustrated besides basic score statistics showing how the system performs. Besides NWP, other possible applications of the SREF such as aviation, energy community, air quality, dispersion, hydrology as well as major societal events (e.g. Olympic Games) are also briefly mentioned. To be more informative to users, the data source, format and products from the NCEP SREF are listed. At last, some outstanding issues related to the NCEP SREF are discussed. Current efforts (2007) and short-term plan are outlined for the interest of NWP community.

Joint Session 4, Modeling Systems II
Monday, 25 June 2007, 3:30 PM-4:30 PM, Summit AB

Previous paper  Next paper

Browse or search entire meeting

AMS Home Page