Recent Changes to AMPS Configuration

Recent Changes to AMPS Configuration

2016-05-25/12 Switched over to hybrid DA.
2016-05-11/12 Not an AMPS change per se, but NCEP implemented an upgrade to the GFS with this forecast.
2016-05-05/12 Switched to background error statistics computed from Apr 2016 forecasts
2016-04-04/12 Switched to background error statistics computed from Mar 2016 forecasts
2016-03-21/12 Remove James Ross Island grid from NZ/Palmer run.
2016-03-02/12 Put in a re-interpolation step which reduces the vorticity couplet that Metgrid tends to give us over the pole.
2016-03-02/12 Switched to background error statistics computed from Feb 2016 forecasts
2016-02-08/12 Added James Ross Island grid (1.8km) to the NZ/Palmer run.
2016-02-04/12 Switched to background error statistics computed from Jan 2016 forecasts.
2016-01-19/12 Upgrade from WRF version 3.3.1 to WRF version 3.7.1
2016-01-04/12 Switched to background error statistics computed from Dec 2015 forecasts.
2015-12-07/12 NZP run upgraded to 3.7.1 (with extra output for ORCAS).
2015-12-04/00 Switched to background error statistics computed from Nov 2015 forecasts. (NZP stats updated on 03/12.)
2015-12-02/12 Now using 0.5-degree GEFS fields for our ensemble initialization.
2015-11-04/12 Switched to background error statistics computed from Oct 2015 forecasts.
2015-10-02/00 Switched to background error statistics computed from Sep 2015 forecasts.
2015-09-02/12 Switched to background error statistics computed from Aug 2015 forecasts.
2015-08-20/12 Incorporated ensemble products into AMPS web page.
2015-08-20/12 Use an alternate blending, to prevent negative humidities (and very cold temperatures) over sea ice.
2015-08-03/12 Switched to background error statistics computed from Jul 2015 forecasts. (switch for NZP grids on 04/00).
2015-07-07/12 Switched the NZP runs to using the larger, ORCAS-inspired 27-km grid for D1.
2015-07-02/12 Switched to background error statistics computed from Jun 2015 forecasts.
2015-06-04/12 Switched to background error statistics computed from May 2015 forecasts.
2015-05-03/12 Switched to background error statistics computed from Apr 2015 forecasts.
2015-04-04/00 Switched to background error statistics computed from Mar 2015 forecasts.
2015-03-03/00 Switched to background error statistics computed from Feb 2015 forecasts.
2015-02-19/12 Turned on assimilation of AIRS retrievals in the RT forecasts.
2015-02-03/00 Switched to background error statistics computed from Jan 2015 forecasts.
2015-01-20/12 Additional output for ORCAS test period created for the NZP runs; Palmer grid (Drake Passage) files archived hourly, hours 00 through 36.
2015-01-14/12 Not an AMPS change per se, but NCEP implemented a major upgrade to the GFS with this forecast, so there may be some effects on AMPS. AMPS handled the transition pretty smoothly.
2015-01-06/12 Switched to background error statistics computed from Dec 2014 forecasts.
2014-12-03/12 Switched to background error statistics computed from Nov 2014 forecasts.
2014-11-14/00 Corrected an adjustment to the snow-cover fields that put a blanket of snow over non-Antarctic land in our New Zealand and Palmer grids.
2014-11-04/12 Switched to background error statistics computed from Oct 2014 forecasts.
2014-10-06/12 Updated coastal features – landmask should match RAMP2 fields now. Updated the Ross Ice Shelf edge (from recent MODIS satellite images). Updated map backgrounds for Ross Ice Shelf.
2014-10-02/12 Switched to background error statistics computed from Sep 2014 forecasts. Keep Aug BE stats for our NZP run, because of several missing forecasts in Sep.
2014-09-05/12 Switched to background error statistics computed from Aug 2014 forecasts.
2014-08-02/00 Switched to background error statistics computed from Jul 2014 forecasts. Problem with d1, and wrfvar did not run for d1 at this time. Fixed for 02/12 forecast.
2014-07-18/12 Switched to background error statistics computed from Jun 2014 forecasts.
2014-06-17/12 Shut off PIG nest
2014-06-07/12 Switched to background error statistics computed from May 2014 forecasts.
2014-05-08/12 Got wrfvar gen_be working again, so I've got new background error statistics computed from Apr 2014 forecasts.
2014-03-03/12 Switched to background error statistics computed from Feb 2014 forecasts. BE Stats not working. Reverted to Jan 2014 be stats.
2014-02-12 Update to rip4 version of 2011-04-05 (with AMPS modifications).... This has a corrected vertical interpolation of geopotential height, so we get some slightly different results.
2014-02-06/12 Update to rip4 version of 2009-04-10 (with AMPS modifications)... Should have identical results (images, GRIB, and time series) to previous version.
2014-02-05/12 Update to rip4 version of 2008-04-04 (with AMPS modifications)... some tiny differences in results from the previous version.
2014-02-03/12 Switched to background error statistics computed from Jan 2014 forecasts.
2014-02-03/12 Update to rip4 version of 2006-12-21 (with AMPS modifications)... should have identical results to previous version.
2014-02-01/00 Update to rip4 version of 2005-08-11 (with AMPS modifications)... should have identical results to previous version, but adds a 10m wind speed option we may want to take advantage of in the future.
2014-01-31/12 Update to rip4 version of 2005-05-20 (with AMPS modifications).... beginning an effort to bring the rip program used in AMPS up to date. tsfc fields have been fixed in this version to use 2m temperature.
2014-01-03/12 Switched to background error statistics computed from Dec 2013 RT forecasts.
2013-12-04/12 Switched to background error statistics computed from Nov 2013 RT forecasts.
2013-11-04/12 Switched to background error statistics computed from Oct 2013 RT forecasts.
2013-10-02/12 Switched to background error statistics computed from Sep 2013 RT forecasts. (Got the D2 file wrong, so WRFVAR didn't go for D2 today.)
2013-09-30/12 Revert to using fine_input_stream = 0 for domain 2 in wrf namelist.
2013-09-30/12 Set use_crtm_kmatrix to .TRUE. in the wrfvar namelists, on Tom's recommendation.
2013-09-04/12 Switched to background error statistics computed from Aug 2013 RT forecasts.
2013-08-15/00 Recompiled WRFVAR (3.4.1) using updated bufrlib so I can assimilate AMSU data again
2013-08-02/12 Switched to background error statistics computed from July 2013 RT forecasts.
2013-07-08/12 For NZP jobs, switched to Domain 1 background error statistics computed from Jun 2013 RT forecasts.
2013-07-03/12 Switched background error statistics computed from Jun 2013 RT forecasts.
2013-06-04/12 Was able to generate Domain 2 background error statistics (Thanks, Gael Descombes!) so I've turned on var for domain 2. We only got satellite data for this forecast (LDM server being upgraded), but it seems to have worked.
2013-06-03/12 Switched to Domain 1 background error statistics computed from May 2013 RT forecasts. Still no var on domain 2.
2013-05-30/12 Expanded Domain 5, now running at 613 x 718 points. See updated grids. Our grids are now:
  • 30-km Antarctica and Southern Ocean ( 330 x 436 grid points )
  • 10-km Antarctic continent ( 667 x 628 grid points )
  • 3-km Ross Sea ( 538 x 826 grid points )
  • 1-km Ross Island ( 613 x 718 grid points )
  • 3-km Antarctic Peninsula ( 520 x 454 grid points )
2013-04-02/00 Switched to Domain 1 background error statistics computed from April 2013 RT forecasts. Still no var on domain 2.
2013-04-15/12 Recommence assimilation of AMSU-A radiances. Still no var on domain 2.
2013-04-09/12 Now running the Palmer grid, and a new New Zealand grid, in a separate, independent 2-way nested run.
2013-04-08/12 Removed the South Pole nest (domain 4), and expanded the area of the Ross Sea nest (domain 3) to encompass the region of the late domain 4. Domain 3 now running with 538 x 826 grid points. See updated grids. Also, extended our domains 3, 5, and 6 out to 39 hours.

Our grids are now:
  • 30-km Antarctica and Southern Ocean ( 330 x 436 grid points )
  • 10-km Antarctic continent ( 667 x 628 grid points )
  • 3-km Ross Sea ( 538 x 826 grid points )
  • 1-km Ross Island ( 325 x 346 grid points )
  • 3-km Antarctic Peninsula ( 520 x 454 grid points )
2013-04-05/00 Switched to Domain 1 background error statistics computed from March 2013 RT forecasts. Still no var on domain 2.
2013-03-01/12 Switched to Domain 1 background error statistics computed from February 2013 RT forecasts. Still no var on domain 2.
2013-02-05/12 Switched to Domain 1 background error statistics computed from January 2013 RT forecasts. Still no var on domain 2.
2013-01-14/12 Transition from bluefire to erebus: New computer, new model version, new configuration:

Update model version from 3.2.1 to 3.3.1 (with polar modifications).
Higher vertical resolution: switch from 43 layers to 60 layers

Higher horizontal resolution:
  • 30-km Antarctica and Southern Ocean ( 330 x 436 grid points )
  • 10-km Antarctic continent ( 667 x 628 grid points )
  • 3-km Ross Sea ( 538 x 526 grid points — domain 3 much expanded )
  • 3-km South Pole ( 256 x 256 grid points )
  • 1-km Ross Island ( 325 x 346 grid points )
  • 3-km Antarctic Peninsula ( 520 x 454 grid points )

Also,
  • WRFVAR running only on domain 1 for now.
  • Assimilation of AMSU radiances not activated for now.
  • BE Statistics for VAR computed from 2012 Oct experimental forecasts.
2012-12-07/00 Switched to BE statistics computed from Nov 2012 RT forecasts. Also, restarted the VARBC cycling for 00Z initializations after several WRFVAR failures.
2012-11-11/00 After a few WRFVAR failures for d2, I restarted the VARBC cycling for 00Z initializations.
2012-11-02/12 Switched to BE statistics computed from Oct 2012 RT forecasts.
2012-11-01/00 After a couple of 00Z wrfvar failures, restarting the VARBC cycling for 00Z initializations
2012-10-17/00 Restarted the VARBC cycling again for 00Z initializations.
2012-10-08/12 After several WRFVAR jobs failed, restarting the VARBC cycling from cold start.
2012-10-04/00 Switched to BE statistics computed from Sep 2012 RT forecasts.
2012-09-10/12 Put in place an improved sea-ice interpolation program, which does a better job of filling in missing data around the coastlines. This should prevent the spurious open-water regions that show up around the coast.
2012-09-07/00 Switched to BE statistics computed from Aug 2012 RT forecasts.
2012-08-20/00 Reduced the maximum timestep, and increased the epssm damping namelist values, to keep the numerics stable in a strong wind event over the peninsula.
2012-08-03/00 Switched to BE statistics computed from Jul 2012 RT forecasts.
2012-07-31/12 Fixed the adaptive timestep logic, so that radiation is called at the proper intervals. Before this change (and probably since we've been running adaptive timesteps), the code was skipping some of our radiation calls.

Also, made a namelist change to set CUDT to 0, so that the cumulus parameterization schemes are called for every time step. This is so that the logic for the adaptive time step does not bite us in the cumulus.
2012-07-07/00 Switched to BE statistics computed from Jun 2012 RT forecasts
2012-05-15/00 Switched to BE statistics computed from the Apr 2012 RT forecasts. -- Change reverted starting with the 15/12Z forecast. WRFVAR for domain 2 started indicating NaNs in the printout, and did not create a usable analysis for domain 2.
2012-04-06/12 Switched to BE statistics computed from the Mar 2012 RT forecasts.
2012-03-03/00 Switched to BE statistics computed from the Feb 2012 RT forecasts.
2012-02-03/00 Switched to BE statistics computed from the Jan 2012 RT forecasts.
2012-01-07/00 Switched to BE statistics computed from the Dec 2011 RT forecasts.
2011-12-09/00 Start the VARBC cycling from coldstart again, this time turning off both PW and Tsk predictors.
2011-12-03/00 Switched to BE statistics computed from the Nov 2011 RT forecasts.
2011-11-29/12 Turned on one-way nest for PIG.
2011-11-16/00 Start the VARBC cycling form coldstart again, this time turning off PW predictor.
2011-11-11/00 Switched to BE statistics computed from the Oct 2011 RT forecasts.
2011-10-06/12: Switched to BE Statistics computed from the Sep 2011 RT forecasts.
2011-09-09/00: Turn off use of METOP-2 AMSUA data to see if our WRFDA analyses stay sane. Moved to a 24-hour (rather than a 12-hour) offset time when searching for VARBC files from previous analyses.
2011-09-07/12: Switched to BE statistics computed from the Aug 2011 RT forecasts. (Skipped the Jun 2011 and Jul 2011 updates, when the genbe program decided not to work for some reason. Mysteriously, the program worked again for Aug 2011.)
2011-06-10/12: Switched to BE statistics computed from the May 2011 RT forecasts. (Skipped the Apr 2011 updates when I had problems with the new version of the genbe program. Worked through the problems for the May 2011 updates).
2011-04-27/12: Upgraded to WRFv3.2.1. Source code changes, lots of script changes. Other significant changes:
  • -- Changed to RRTMG, longwave scheme.
  • -- Moved from 10-minute radiation timestep to 15-minute radiation timestep (test for efficiency).
  • -- For grids 4-6, switched to fine_input_stream=2, in order to maintain more consistent initial conditions between nest levels.
2011-04-05/00: Switched to BE statistics computed from Mar 2011 RT forecasts
2011-03-18/12: Shut off the LARISSA one-way nest.
2011-03-03/12: Shut off CTAM one-way nest.
2011-03-02/12: Switched to BE statistics computed from Feb 2011 RT forecasts
2011-02-04/00: Switched to BE statistics computed from Jan 2011 RT forecasts. (Skipped the Dec 2010 updates when the genbe program failed for that month's forecasts.)
2010-12-03/00: Switched to BE statistics computed from Nov 2010 RT forecasts
2010-11-11/00: Switched to BE statistics computed from Oct 2010 RT forecasts
2010-10-08/12: Switched to BE statistics computed from Sep 2010 RT forecasts
2010-10-01: Added CTAM one-way nest. Turned LARISSA one-way nest back on.
2010-09-02/12: Switched to BE statistics computed from Aug 2010 RT forecasts. We missed the July statistics because of incomplete data (several missing forecasts for the month). Stopped using the adaptable time step, in favor of constant time steps again. (for how long? when did we go back to adaptable steps?)
2010-07-28/12: Not an AMPS change in itself, but it should be noted that NCEP has implemented an upgrade to the GFS.
2010-07-07/00: Switched to BE statistics computed from Jun 2010 RT forecasts
2010-06-04/12: Switched to BE statistics computed from May 2010 RT forecasts
2010-05-05/00: Switched to BE statistics computed from Apr 2010 RT forecasts
2010-04-26/00: Implemented Steven Cavallo's modification to RRTM, which addresses a cold bias at the model top.
2010-04-06/00: Switched to BE statistics computed from Mar 2010 RT forecasts
2010-04-01/00: Began assimilating AMSU-A radiances (with variational bias correction, VARBC) in the WRF Data Assimilation step.
2010-03-08/12: Turned off the Larissa one-way nest
2010-03-04/00: Switched to BE statistics computed from Feb 2010 RT forecasts
2010-02-06/00: Switched to BE statistics computed from Jan 2010 RT forecasts.
2010-01-08/00: Switched to BE statistics computed from Dec 2009 RT forecasts.
2010-01-04/12: Turned off the Marie Byrd Land one-way nest (mbbig) and started the Larissa one-way nest. Will fine-tune Larissa jobs over the next few days.
2009-12-19/00: Revert to the Goddard SW, RRTM LW, at least through the holidays, and perhaps until we can fix the random crashes that the CAM radiation seems to trigger.
2009-12-04/00: Switched to BE statistics computed from Nov 2009 RT forecasts
2009-11-06/12: Switched to BE statistics computed from Oct 2009 RT forecasts
2009-10-17/00: Switched from Goddard SW, RRTM LW, to CAM SW CAM LW, with snow albedo varying from 75% at elevations of 250m and below, to 80%, at elevations of 2500m and above.
2009-10-05/12: Switched to BE statistics computed from Sep 2009 RT forecasts
2009-09-13/00: WRF has been crashing (strong winds over steep terrain) so I've increased damping slightly, increased the depth over which damping is applied, and reduced the maximum CFL allowed with the adjustable time step, and reduced the maximum timestep allowed on domain2. These didn't seem to help these crash cases (20090911/00Z, 11/12Z, 12/00Z, 12/12Z), but may keep things a little more stable under extreme conditions.
2009-09-03/00 Switched to BE statistics computed from Aug 2009 RT forecasts
2009-09-01/00 Sea-ice product SSMIF13 from NSIDC has been discontinued. We have switched to using product SSMISF17 from NSIDC
2009-08-24/12 Began archiving WRF input files in the misc.<YYYYMMDDHH>.tar.bz2 tar files.
2009-08-13/12: Switched to BE statistics computed from Jul 2009 RT forecasts
2009-07-06/12: Switched to BE statistics computed from Jun 2009 RT forecasts
2009-06-05/00: Switched to BE statistics computed from May 2009 RT forecasts
2009-05-20/00: Corrected a couple of fields in the geogrid files, that had incorrect values at the pole. This is a slight modification to geo_em.d01.nc, geo_em.d02.nc, and geo_em.d04.nc
2009-05-07/00: Switched to BE Statistics computed from Apr 2009 RT forecasts. Also switched to computing these statistics with CV Options 5
2009-04-16/00: Turn QSCAT assimilation on
2009-04-08/12: Switched to BE Statistics computed from the Mar 2009 RT forecasts
2009-03-25/12: Turned on positive-definite advection for moisture
2009-03-03/00: Switched to the newly-created Feb 2009 BE statistics for WRFVAR
2009-01-22/00: Boosted the albedo over the glacial ice from 0.7 to 0.8 -- try to mitigate the summertime warm bias.
2009-01-15/00: Modified the model initialization procedures to adjust the initialization of the deeper nests above the surface. This is a hack to the share/mediation_integrate.F source code file, blending the fields generated by independently preprocessed nests with fields interpolated from the parent domain (which has had WRFVAR applied). It removes spurious 300mb vorticity signatures which outlined the deeper nests in the domain 2 initialization.
2008-12-18/12: Cap the timestep on domain 5 to ( 6 * deltaX ), i.e., 10 seconds. This adds about 20 minutes to the wallclock time, bringing us to a run time of about 4:50
2008-12-18/00: Turned the WRF 3d-var analysis back on, using Oct 2008 BE statistics
2008-11-14: Switched to Goddard SW Radiation (which we meant to be using all along) from the Dudhia short wave. This adds about 30 minutes to the wallclock time, bringing us to a run time of 4:30
2008-11-13: Applied the official WRF bug fix to the RRTM radiation scheme (starting with the 00Z forecast)
2008-11-07: Correct the wind rotation in the time series output
2008-11-03: Running exclusively WRFV3.0.1.1 in the new higher-resolution setup (mss:/AMPSRT/WRF_45)
  • Grid 1 :: 45 km :: 220 x 290 points :: 0-120 hours
  • Grid 2 :: 15 km :: 442 x 418 points :: 0-120 hours
  • Grid 3 :: 5 km :: 157 x 190 points :: 0-36 hours
  • Grid 4 :: 5 km :: 169 x 169 points :: 0-36 hours
  • Grid 5 :: 1.67 km :: 214 x 229 points :: 0-36 hours
  • Grid 6 :: 5 km :: 346 x 301 points :: 0-36 hours
2008-11-02: Last forecasts of WRFV2.2 in the 60/20-km configuration.
2008-10-01: Begin experimenting with WRFV3.0.1.1 with BPRC Polar mods.
Horizontal resolution of 45/15/5/1.7 km
44 Vertical levels

Major changes over the longer term (reconstructed from the AMPS archives)

2008-06-30/00: Last MM5 forecast for AMPS. Using WRF exclusively after this date.
2006-03-06/00: Begin WRF archives, with the 60/20/6.6/2.2-km configuration (mss:/AMPSRT/WRF_60)
2005-12-19/12: Last of the 90/30/10/3.3-km jobs.
2005-09-21/00: Begin splitting Domain 1 and Domain 2 MMOUT files in the MM5_60 archive, into 00-24, 27-48, and 51-72 hour chunks
2005-09-17/00: Begin archiving the enhanced resolution jobs, with 60/20/6.6/2.2-km configuration (mss:/AMPSRT/MM5_60)
  • Grid 1 :: 60 km :: 165 x 217 points :: 0-120 hours (archive only out to 72 hours)
  • Grid 2 :: 20 km :: 331 x 313 points :: 0-120 hours (archive only out to 72 hours)
  • Grid 3 :: 6.6 km :: 118 x 142 points :: 6-36 hours
  • Grid 4 :: 6.6 km :: 127 x 127 points :: 6-36 hours
  • Grid 5 :: 2.2 km :: 160 x 172 points :: 6-36 hours
  • Grid 6 :: 6.6 km :: 259 x 226 points :: 6-36 hours
2004-08-05/12: Extend forecast time for domains 1 and 2 out to 120 hours.
2003-09-18/12: Grid 6 added, 10-km grid over Antarctic Peninsula (Archives for grid 6 begin 2003-09-19/12Z)
  • Grid 6 :: 10 km :: 172 x 151 points :: 6-36 hours
2003-05-20/12: Raise model top from 100 hPa to 50 hPa. Implement top boundary nudging toward GFS forecast fields
2002-11-04/12: Grid 2 expanded to get the nest borders a little farther away from the continent.
  • Grid 2 :: 30 km :: 220 x 208 points :: 0-72 hours
2002-07-02/12: Grid 1 expanded to reach S. Africa.
  • Grid 1 :: 90 km :: 110 x 145 points :: 0-72 hours
2001-12-13/12: Grid 5 added, 3.33-km grid over Ross Island and vicinity
  • Grid 5 :: 3.3 km :: 85 x 115 points :: 6-36 hours
2001-11-28/12: Grid 4 added,10-km grid over Pole (archives for this grid start on 2001-11-29 / 00 Z)
  • Grid 4 :: 10 km :: 85 x 85 points :: 6-36 hours
2001-10-31/12: Forecast time for grids 1 and 2 extended to 72 hours; Forecast time for grid 3 extended to 36 hours
2001-01-05/00: AMPS archives begin, 3 domains only, 90/30/10-km grids (mss:/MM5RT/ARCHIVE)
  • Grid 1 :: 90 km :: 110 x 115 points :: 0-48 hours
  • Grid 2 :: 30 km :: 178 x 187 points :: 0-48-hours
  • Grid 3 :: 10 km :: 79 x 94 points :: 6-30 hours
e-mail: Kevin Manning
Last modified: Wednesday 25 Feb 2015