VAREPS (VAriable Resolution EPS)

Between 2002 and 2005 , three ensemble configurations have been tested to improve severe weather warning and weather-risk management: a plain higher-resolution ensemble (HEPS), an incremental ensemble system (INCEPS) and a VAriable Resolution EPS (VAREPS). Following extensive testing, in 2005 it was decided to upgrade the EPS with VAREPS, an approach that will provide ensemble users with more accurate predictions of small-scale features in the medium-range and will extend the EPS forecast range from 10 to 15 days.

VAREPS was implemented in the ECMWF operational suite in September 2006, with a TL399L62 resolution up to forecast day 10, and a TL255L62 resolution between forecast day 10 and 15 (VAREPS-TD10). VAREPS was merged with the monthly ensemble system in March 2008. Since then it has been running:

• at 00 UTC with persisted sea surface temperature anomalies up to day 10, and with a coupled ocean model from day 10 to day 15 everyday but thursdays, when it is extended to day 32
• at 12 UTC with persisted sea surface temperature anomalies for the whole 15 days

On the 26th of January 2010, the resolution of VAREPS was increased to TL639L62 between day 0 and 10, and to TL319L62 afterwards.

Hereafter, a pre-print of a recent manuscript (Buizza 2010, MWR, in press) that discusses the value of using a variable resolution approach to numerical weather prediction, and an internal report that investiagets the impact of truncation on variable resolution forecasts (Buizza 2010, ECMWF RD Tech. Memo. 614) can be accessed:

• The Value of a Variable Resolution Approach to Numerical Weather Prediction (MWR)
• Impact of truncation on variable resolution forecasts (ECMWF RD TM 614)

Furthermore, two manuscripts that describe the VAREPS system as it was implemented in 2006 and is currently run with a 10-day truncation (VAREPS-TD10) can be accessed using the following links:

• An earlier version of an article published in the ECMWF Newsletter that describes t he ECMWF VAriable Resolution Ensemble Prediction System (VAREPS)
• A document that describes some technical charactersitics of VAREPS-TD10, the system planned to implemented in the ECMWF operational suite in the 2nd half of 2006

VAREPS-TD10 has been run with model cycle 30r2 for 38 cases, from 5 February 2006 to 10 August, every 5 days. Hereafter, some plots with VAREPS-TD10 38-case average performance over the Norther Hemisphere can be accessed:

• Error of EPS-control, perturbed members and ensemble mean, and ensemble spread, measured in terms of 500 hPa geopotential height
• Error of EPS-control, perturbed members and ensemble mean, and ensemble spread, measured in terms of 850 hPa temperature
• Error of EPS-control, perturbed members and ensemble mean, and ensemble spread, measured in terms of total precipitation
• Ranked probability score, area under the relative operating characteristic curve and Brier score for the probabilistic prediction of 500 hPa geopotential height
• Ranked probability score, area under the relative operating characteristic curve and Brier score for the probabilistic prediction of 850 hPa temperature
• Ranked probability score, area under the relative operating characteristic curve and Brier score for the probabilistic prediction of total precipitation

Hereafter, copies of few old presentations, one discussing the planned 32-day VAREPS/monthly extension and few summarizing results obtained with a VAREPS system with day-7 truncation (VAREPS-TD7), can be accessed:

• VAREPS/Monthly extension (Expert Meeting 8/9 May 2007)
• VAREPS-TD7: February 2006 update
• VAREPS-TD7: September 2005 update
• VAREPS-TD7: May 2005 update
• VAREPS-TD7: January 2005 update
• VAREPS-TD7: February 2004 update

(site map)