The GFS 4D hybrid EnVar data assimilation has been approved for implementation in May 2016. Upgrades to the operational GFS include:


4D hybrid ensemble-variational data assimilation: The ensemble provides an updated estimate of situation dependent background error every hour as it evolves through the assimilation window. This flow dependent statistical estimate is combined with a fixed estimate.

Improved use of satellite radiances

Improved use of satellite winds and aircraft observations

Corrections to land surface to reduce summertime warm, dry bias over Great Plains

Two of the retrospectives were run by NCO. Retrospectives are now finished. There was an EMC CCB on March 8 and an NCEP directors’ briefing on March 17. A 30 day parallel for IT checkout purposes only is scheduled to begin in April. The improved system is scheduled to be implemented May 17.

A consistent parallel feed of gridded data is being planned for paraNOMADS. 

Files from the real time parallel and 1 degree files from the retrospectives are on WCOSS.

Verification of the real time parallel can be found at:

Official Evaluation

Presentation to NCEP Director March 17

WPC presentation to NCEP Director

SPC presentation to NCXEP director

Presentation to EMC CCB

Official evaluation forms were completed by:

NCEP Centers

Aviation Weather Center

Climate Prediction Center

Additional power points

National Hurricane Center

Additional power point

Ocean Prediction Center

Storm Prediction Center

Additional power point

Space Weather Prediction Center

Weather Prediction Center

NWS Regional Headquarters







Other NWS

Meteorological Development Laboratory

National Water Center

Dowstream testing progress report

Synoptic evaluation

Near the top on the left is a link labelled Precip QPF. This will take you a page where you can find verification plots over CONUS for precipitation forecasts by the operational and experimental (GFSX ) GFS for a wide range of dates. If you go to the bottom on the left, there are 2D maps showing the current forecasts from the operational and parallel GFS. By 00z there is a link to arch. This will take you to an archive over the past few years for synoptic maps for different regions for the operational and parallel GFS. Note that the parallel GFS is whatever was running that day. Only since Nov. 1 2015 have we been running the current GFSX with the land surface correction.

Another site for precipitation verification is

These sites are subject to the availability of the computers to EMC and may not be up to date every day.

GEMPAK plots are being produced from the real time parallel and can be seen at:

and can be compared to the operational gfs at:

Western Region has a side by side display of the operational and parallel GFS for North America and the North Pacific:

Ocean Prediction Center displays both the GFSX and operational GFS 9950 sigma winds and has difference fields for wind and EMSL. Click on GFS Evaluation.

A real time comparison of soundings from the GFS and GFSX can be found at:

A plume diagram comparing the GFS and GFSX forecasts of near-surface fields for selected US stations


EMC Objective Verification

An extended version of the real time parallel can be found at:

Please note that an additional analysis change from the operational not in the current parallel was in pr4devb until the forecast initialized Oct. 31 18 GMT. It was suspected that this change may have hurt the forecasts and it was removed and is not in any of the retrospectives below.

Verification covering all retrospectives from 2013 to 2016 can be found at (The page is a work in progress):

Verification for each of the seven retrospectives covering Sandy, 3 summers and 2 winters can be found at: covering Sandy.

(This covers from the 13 km implementation to the end of the winter.)

(This covers from Nov. 2014 up to the 13 km implementation in January 2015.)

A comparison of hurricane track and intensity skill

Case studies

The global branch has offered to plot maps for specific cases from the retrospectives and has solicited specific cases and specific fields to look at. Western, Central, Southern and Alaska Regions have sent in cases to look at. NHC and WPC are working with the global branch to enable their evaluation of the new GFS.

MEG review of case studies proposed by WPC, Western Region and Central Region

MEG review of additional case studies

Presentation to WPC on case studies

Global Branch Meeting Summary of Western Region case studies and one Central region case study

Global Branch Meeting Summary of another Central Region case study, Alaska case study and Southern Region case study

Four case studies requested by Western region are in:

A case study requested by Central Region for Jan. 29 2015 is in:

A case study for Central Region for March 23, 2015

A case study for Central Region for April 2, 2015

A case study for Central Region for June 4-5, 2015

A case study for Central Region for July 6, 2015

A case study of the Nov. 16-17, 2015 tornado outbreak in Texas and Oklahoma

Hurricane Joaquin and South Carolina flooding

Blizzard of January 22-23, 2016

A case study for Dec. 5-6, 2013 requested by Southern Region

Precipitation cases for WPC

A comparison of operational and experimental GFS forecasts for Atsani is shown in

Forecast tracks for Sandy plotted by initial time can be found at

WPC documented a dry bias over the southeast US in the GFS and GFSX

Case study of GFS and GFSX cold bias over snowpack

Documents and evaluations

The official letter asking for evaluation of the GFSX

Data assimilation changes in the new GFS

updated version of data assimilation changes

Hybrid 4D EnVar

Preliminary evaluation presentation to EMC

Warm dry bias over Great Plains in summer (case study)

MEG presentations reviewing the new GFS

MODE evaluations of new GFS

Extratropical storm tracks


Comparison of systematic errors in the GFS and GFSX

This is a new implementation procedure and very much work in progress. The global branch wants a thorough evaluation of the GFSX and values your insight. Please be patient and persistent with us.

Please contact or if you have any comments, concerns or questions.