SREF-32 km Operational Change Log



Jif #

12Z June 28, 2006
Implementation of SREF 4x/day runs
In addition to the current 09Z and 21Z runs, the script is being modified to add off-time cycles 03Z and 15Z for SREF 21 member 87 hour forecasts.
12Z June 20, 2006
Addition of GEMPAK NCO Graphical products
SREF_GEMPAK_GIF.  (Sager, NP12).  New script creates five new sets of Short Range Ensemble Forecast (SREF) model GEMPAK graphics for the NCEP Model Analysis and Forecast website.  The new graphics being generated are Predominate Precipitation Type, Probability of CAPE greater than 500, Probability of 850mb Temperatures less than 0 Celsius, Probability of 2-meter winds greater than 25 Knots, and Probability of 6 hour precipitation greater than .25 inches.    Script are also being changed to create graphics for every three rather than six forecast hours and to extend the graphics out to 87 hours.
exsref_post.  (Du, NP2).  This program produces ensemble products (mean, spread and prob) for the SREF.  The cloud ceiling is being added as required by AWIPS.


12Z June 13, 2006
Eta and WRF member initialization from new NAM-WRF GSI analysis
SREF_PREP_GET_NDAS.SH - IBM Job  (Du and Chuang, NP2).  This program reads in high-resolution regional analysis to covert it into lower-resolution for the SREF Eta members.  The program is being modified to interperate the WRF restart file instead of the Eta restart file
 EXSREF_WRFPERTURB.SH.SMS - IBM Job  (Du and Jovic, NP2).  This program extracts land surface states from ndas to replace GFS's ones.  The program is being modified to match the regional analysis (ndas) change (from Eta-based to WRF-nmm based).    
12Z June 6, 2006
Generation of GRIB2 ensemble products
SCRIPT -  (Vuong, NP11).  Script produces the (mean, spread and probability) of selected variables and writes them out in GRIB format.  The script is being modified to produce grids 212, 216 and 243 in GRIB2 format for AWIPS from the SREF model.
12Z May 30, 2006
Use of  GEFS forecasts for init and Lateral boundary conditions for all members

Jobs changed to point to Global Ensembl Forecast System (GEFS) directory and file name structure and for WRF:  (Du, NP2).  This program runs WRF standard initialization (SI) for the WRF-NMM and WRF-EM model in the SREF.  The following modifications are being implemented: (1) Global ensemble's directory name and file names are being modified to match the new glb ensemble output format.
(2) a new pair of glb ensemble members is being assigned as LBCs to the 3 WRF ARW members for more diversity.
12Z April 25, 2006
PARM FILE - IBM Job  (Du, NP2).  Parm file defines variables and product type that are calculated in the SREF ensemble product generator.  Helicity, CIN and vorticity fields have been added back after they were mistakenly removed in the previous code upgrade for AWIPS.  
Crisis 356
12Z April 18, 2006
Inclusion of additional ensemble product fields for AWIPS
EXSREF_POST.  (Du, NP2).  This program calculates ensemble products i.e. mean, spread and probability output GRIB files.   The program is being modified to (1) add all missing variables required by AWIPS (requested by NCO); (2) change units from K to C for lifted index to avoid all zero probability values; and  (3) correct an error to output the  probability of snow only once, not twice.
335, 336
12 Z March 21, 2006
WRF and RSM members precip fix
SREF_WRFPOST - (Du, NP2).  The program calculates ptype for RSM and WRF members in the SREF.  Two errors are being corrected: (1) use instanteous precip rate now instead of 3hr-accumulated precip previously and (2) remove tiny precip rate values in the calculation to avoid too wide spread ptype areas from RSM members (a characteristic of spectral model) observed by SPC's forecasters.
12 Z March 14, 2006
WRF & RSM member Precip type fix
SREF_WRFPOST_SOURCE -  (Du, NP2).  This program determines the precip type category such as rain, snow, sleet or freezing rain.  The program is modifying an error that caused incorrect dimensions in a few of the arrays, this lead to an increased bias for freezing rain category.
12 Z February 14, 2006

WRF ARW PBL fix, Precip type algorithm improvement
The YSU PBL used in the WRF-ARW members was corrected. This correction now allows ARW members to maintain numerical stability with longer time steps (currently, DT=108s, DX=45 km). Songyou Hong, KMA, provided the correction to limit variables such as uflxpbl, vflxpbl to realistic values. High fluxes can sometimes occur in response rare point instability caused by strong vertical shear in the YSU PBL..

An improved dominant precipitation type algorithm (developed by Geoff Manikin, EMC)  was implemented in the ensemble product generator.  The dominant type is derived by first computing precip types from various algorithms (eg: Baldwin, Rainer,...) and then determining the dominant precip type from each member and then for the whole ensemble.

12 Z February 7, 2006
WRF member optimization, ARW LSM fix,  NMM non-hydro option
NAMELIST - IBM Job exsref_wrffcst.  (Du, NP2).  This program defines how the WRF_arw and nmm members are executed in SREF.  The program was modified to  change the debug level from 5000 to 0 to speedup the model run.   Each WRF  NMM and ARW member now run in under 30 and 40 minutes respectively.

ARW_SOURCE - IBM Job exsref_wrffcst.  (Du, NP2).  This is the NOAH land surface model used by ARW (EM) members in SREF.  The program is being modified to; (1) to correctly initialize downward longwave radiation; and (2) to avoid dividing by zero in light snow condition.   This change was recommended by NCAR/MMM to help control instabilities noted in NMM in the NE portion of the ARW domain near Greenland.

SOLVE_NMM - IBM Job exsref_wrffcst.  (Du, NP2).  This program is the dynamical solver of WRF_nmm model used in SREF.  The program is being modified to turn on the Non-hydrostatic option for the vertical velocity field.


12 Z January 10, 2006
WRF-NMM nenbers V2.0 GFDL radiation and BMJ convective physics fixes
RADIATION SOURCE - IBM Job sref_wrffcst.  (Du, NP2).  This program is the radiation calculation code for the WRF_nmm model in SREF.  The program has been modified to add logic to prevent downward surface solar insolation and shortwave potential temperature tendencies from being overwritten when running the GFDL shortwave radiation.   

CUMULUS SOURCE - IBM Job sref_wrffcst.  (Du, NP2).  This is BMJ cumulus paramerization code for the WRF_nmm model in SREF.  The program has been modified to comment out 3 lines of code that inadvertently resulted in bypassing all of the shallow convection code. 

Vertical Motion fields added to SREF member GRIB files in addition to Omega
CRISIS 29, 30
12 Z December 28, 2005
WRF members script
 Script executes the WRF SI job for SREF.  The script has been modified to not use date information in the identification to real_input files (in order to avoid time confusion during the transition from one year to another). 

2005 JIFS

December 13, 2005
Eta Members CAPE postprocessing
SREFETA_ETAPOST - IBM Job sref_eta_post.    This program calculates the Convective Available Potential Energy (CAPE).  The surface-based CAPE calculation is being revised based on SPC's suggestion (now searches a depth based on surface pressure rather than a fixed lowest 70mb previously).    
  December 6, 2005
  • 6 WRF members added to SREF:
    • 3 WRF-NMM (1 ctl, 1 pert. pair) at 40 km w/ NCEP physics suites
    • 3 WRF-ARW  (1 ctl, 1 pert. pair) at 45 km with NCAR physics suites
  • Additional output to SREF individual member and products files for AWIPS
August 30, 2005
  • RSM: Updated and optimized version used.  Domain now covers all of Alaska and Hawaii
  • RSM resolution decreased from 40 to 45 km
  • RSM Bufr sounding files now output for all  RSM members
  • Implement WRF-POST: all derived fields will be based on model state variables in a common WRF model output format.
  • SREF Forecast Length to be extended to 87 hours.
  • LBCs: 3 hour old GFS ensemble forecasts now used for SREF member LBCS (instead of 9 hr old).
  • Extra Grib Grids to be outputted besides 212(CONUS-40): 216 (Alaska-45 ), 243 (Hawaii-45)

 August 16, 2005
ETA members SST initial condition change
The SREF program to read in all Eta related global and regional files including SST, snow, ndas and gfs data has been modified to use the Real-time Global SST analysis data instead of the 50km SST analysis, recently changed by NESDIS.  
June 28, 2005
Eta members BUFR sound postprocessing
The bufr sounding station list was modified to match the new operational RUC and NAM list.
May 31,  2005
All SREF members: Upgraded GFS ICs/LBCs
SREF system modified to initialize IC and  LBCs with upgraded GFS T382 system

May 3, 2005
 Eta members LSM physics upgrade
New Land surface initialization codes added to SREF system to initialize SREF members with  upgraded NAM-12 LSM using 1 km land and soil type fields.

March 14, 2005
 Ensemble mean FVS (Forecast Verif. System) Files added to NCO processing
March 8, 2005
The specific humidity(q) has been modified to reset to a small value, when q = 0 (absolute dry in atmosphere) to avoid a mathematical problem (divided by zero) in the initial condition breeding or forecast process.
February 22, 2005
NDAS scripts modified to save the 06z and 18z cycles of the NDAS initial conditions to meet the proposed running SREF 4 cycles per day.  
January 3, 2005
Eta members freezing level postprocessing: Fixed problems with freezing level computations.
2 Crisis

2004 Changes

  RSM-Post jobs changed to run post simultaneously with model and therefore have outputs  available earlier


SREF 32  improved physics diversity and scaled breeding system implemented:
Change  Notice
  • 5 Eta-BMJ & 5 Eta-KF member resolutions increased to 32km L60.  2 BMJ members modified to have moister profiles.  2 KF members modified to enhance cloud detrainment.
    All Eta members upgraded to July 2003 Eta model version.
  • 5 RSM member resolutions increased to 40 km L28.  2 RSM members modified to use RAS cloud physics.  All RSM member physics upgraded to current GFS physics.
  • Scaled breeding system for all members

SREF-48 km Operational Change Log: 2001-2003



Jif #

TheShort Range Ensemble Forecasting (SREF) system consisting of ten members (five Eta and five Regional Spectral Model (RSM) members) at a 48 km resolution covering the Eta model
domain with regional breeding  for generating initial conditionperturbations was implemented.  To enable product availability at approximately the same time as the operational Eta from 00Z and 12Z, the SREF runs will start three hours earlier (at 21Z and 9Z respectively) and run through 63 forecast hours with output available in Three hour intervals.  Lateral boundary conditions are provided by the respective nine hour old global ensemble runs.


Eta Members: Implemented new Ferrier Cloud physics



Eta Members:  Updated land-sfc model heat flux routines to prevent negative fluxes & precip



Eta members: Corrected truncation program that outputs Eta-12 at 48 km resolution to read from new format Eta restart files



SREF Summer-03 Upgrade (5 Eta-KF members added, additional fields, Eta sounding files
Change Notice

RSM Members: Modified RSM-Post to produce extra output levels (40) and storm relative humidity, storm grid motion,  2m dewpoint temps, and best cape fields in member grib files



Eta Members: Modified  Grib post to produce extra cloud and convective output fields in member grib files

Officially Add 5 Eta-KF members to operational SREF system (now 15 members)



Eta Members:  Create individual Eta Bufr member files



Common Products: Use of NCEP standard grib extension definitions for all SREF probabilistic products


Corrected RSM member winds to be relative to true north.

Corrected RSM Helicity sign error



SREF-32 Physics Diversity/scaled breeding Parallel System Change Log 

Summer 04 Implementation






Began Testing on  Enhanced Physics Diversity System

1 Control run:  Eta-Ferrier Shallow Convective param (FSC)

7 perturbed pairs(p,n) run with operational SREF unscaled breeding:

32 km: (PAR-04A)

2 Eta-BMJ, 2 Eta-KF,  2 RSM-SAS

2 Eta-SAT, 2 Eta-DET, 2 RSM-RAS, 2 Eta-RAS

Too much spread, esp 1/2/04 run


Reduced RSM member resolutions to 40 km to fit in operational window


1/22/04 09Z

Turned breeding perturbation off for all 10 Eta perturbed members.  LBC perturbations kept on.

RSM members, breeding turned off, but different  Ics used from GFS ensemble


Spread too small

1/30/04 09 Z

Turned breeding perturbation on using GFS & WRF scaled perturbation system.  Used average 850 mb T standard deviation (0.5 C)  to scale IC perturbations. 

IC perturbation scale = 0.5/ D

Where D=Fneg-Fpos  of the 12 hour domain avg 850 mb T forecast.

Reasonable spread (48h forecast) :

1/30/04 BMJ-SAT, RSM-RAS too wet

1/31/04 RSM-RAS too wet, BMJ-SAT


2/3/04 21 Z

Reduced membership diversity: (Par-O4B): 4 perturbed pairs

3 Eta-BMJ, 3 Eta-KF, 3 RSM-SAS, 3 RSM-RAS

1 Eta-Sat, 1 Eta-DET, 1 Eta-RAS

2/4/04 21z 2/5/21z : did not complete.

2/13-2/16/04 Did not complete

Not enough spread..esp on low precip amts

3/3/04 21Z

PAR-04C less control runs: 6 perturbed pairs

3 Eta-BMJ, 3 Eta-KF, 3 RSM-SAS

2 Eta-SAT, 2 Eta-DET, 2 RSM-RAS

Improved forecast variance with small effects on skill

3/12/04 21z

Increased Eta nphysics frequency call to every other time step (180 sec) Except for KF members where physics are called every time-step

 Reduced Eta RH bias at upper levels

4/5/04 21z

Fixed RSM wind output to be true rather than grid relative.

Corrected helicity sign error.


4/12/04 21z

Changed BMJ-SAT saturated moisture profile to a dryer moist profile


Done to account for very high BMJ-SAT precip biases and low KF-Det precip scores

6/10/04 21z

Corrected error in BMJ-SAT run where convective param was turned off

Changed KF-DET to allow for instead of full convective cloud detrainment


Improved BMJ-SAT over-prediction biases.

Click here for more info.














2 (p,n)

3 (c,p,n)



























1 (c)







Eta-BMJ type memb 5

4 (2 sat)


5 (2 sat)

Eta-KF type memb    5

4 (2 det)


5 (2 det)

RSM-SAS                 5








# perturbed members 12