Mesoscale Parallel Experiment Change Log

Experiment Name:  North American NEMS/NMMB test

When Who What
2011/03/01/00 Rogers
  • Turned off digital filter initialization
  • Restarted with cycled land-states from NDASX, which was restarted from operational NDAS land states at 18z 2/23/2011
  • 2011/02/15/00 Gayno, J. Wang
  • Changed the NEMS code to read the base or snowfree albedo from the NEMSIO input file created by NPS at the start of each NDAS run. The code was prevously only reading the dynamic albedo, using it as the base albedo. This led to erroneously high albedo over shallow/patchy snow cover.
  • 2011/02/05/18 Ferrier
  • A slight change is made to the new cloud fraction calculation introduced into the NAMX in November 2010. Cloud fractions are set to 0 for QCLD (cloud water and total ice mixing ratios) <1.e-4 g/kg. A minimum cloud fraction of 1% is assumed for QCLD=1.e-4 g/kg. Cloud fractions increase by the square root of QCLD until a maximum cloud fraction of 100% is reached when QCLD>=0.01 g/kg. In other words, the QCLD thresholds are reduced by an order of magnitude compared to the November 2010 code.
  • 2011/02/01/00 S. Liu, Wu
  • Changes to the GSI analysis:
    1. Assign a sub-type to level 2, level 2.5, and level3 radar data
    2. Limit use level 2.5 data in CONUS domain
    3. Increase observation error for level 3 data over CONUS domain.
  • 2011/01/19/18 Janjic, Pyle
  • To remove computational noise in the 1.333/1.5 km Fire Weather nest, the NMMB model (in ./atmos/nmm/module_DYNAMICS_ROUTINES.F90) was changed to gradually reduce dw/dt near the top of the model, by assuming attenuation of dw/dt of the form cos**2 in the top 15 mb of the atmosphere, i.e. in the top 1.5% of the mass of the atmosphere. The problem occurred in the fire weather run since it has high horizontal resolution, combined with the high model top (2 mb), so the horizontal resolution near the top becomes higher than the vertical resolution, which is opposite to the situation lower in the atmosphere.
  • 2011/01/06/00 Pyle
  • NDAS forecast now running with a digital filter with a +/- 40 minute window.
  • 2010/12/09/12 Ferrier, H.-M. Lin
  • Forecast cloud fractions associated with high, thin, cold cirrus clouds were too high, and so a change was made in the grid-scale cloud fractions (Fc) calculated from the GFDL radiation. Values of Fc were estimated from the total relative humidity (RHtot) with respect to water (>0C) or ice (<0C), and this was replaced by estimating Fc as a simple function of the cloud water and ice mixing ratios. Cloud amounts were set to 1.0 in the internal GFDL calculations wherever Fc>0 so as to preserve the long- and short-wave radiative fluxes and temperature tendencies. This 9-slide PDF file shows an example of the old and new Fc fields,
  • 2010/11/27/00 Rogers
  • When the MODIS-IGBP land-use definitions were implemented in the NAMB control run on 11/4, the AVHRR weekly greenness fraction was inadvertently put into all NAM parallels at the same time. The weekly greenness fraction data set was not intended to be implemented in the NAM-NEMS-NMMB change package, so both NAMB, NAMX, and all nests will revert to using the operational monthly greenness fraction as of this cycle.
  • 2010/11/22/18 Ferrier, H-M. Lin
    1. Enhance diffusion for specific humidity and cloud water by 4x (equivalent to setting smag2=0.8)
    2. Radiation code changed to calculate the mean value for the cosine of the zenith angle every time step instead of every physics timestep
    2010/11/10/18 Wong
  • These IGBP roughness length values were changed:
    1. Mixed forests : changed from 0.563 to 0.8
    2. Grasslands : changed from 0.035 to 0.08
    3. Croplands : changed from 0.035 to 0.07
    4. Cropland/Natural Vegetation Mosaics : changed from 0.035 to 0.07
  • 2010/11/09/18 Pyle, Janjic, Rogers
  • This change put into NAMB on 4/29/10:

    "Slight off-centering in time in the Crank-Nicholson scheme (parameter w1 in /atmos/dyn/nmm/module_DYNAMICS_ROUTINES.F90 changed from 1.0 to 0.9). This change was made to eliminate a failure in a test of a 4 km nested CONUS run, which was caused by large temperature oscillations in the vertical over the Sierra Nevada mountains as a strong jet streak was impinging on the mountains."

    was inadvertently removed from the code (i.e., parameter w1 in subroutine ADV1 in ./atmos/dyn/nmm/module_DYNAMICS_ROUTINES.F90 reverted back to 1.0), probably when the new code structure was put in the trunk in June 2010. w1 has been changed back to 0.9 as intended.

  • 2010/11/04/12 Wong, Ek, Ferrier
  • Change from USGS to MODIS_IGBP land-use definitions.
  • 2010/11/03/18 Ferrier
  • Bug fix in module_BL_MYJPBL: convert QSFC array to specific humidity as intended over land when CHKLOWQ=0 (95% RH).
  • 2010/10/20/18 Wu
  • Changes to the GSI analysis:
    1. Update to new version (r9527), reduces CPU consumption and now uses 2 parallel threads (~25 second speedup)
    2. New observation errors and retuned background errors
    3. Use of new data types : RASS virtual temperature, MAP winds, ASCAT winds
  • 2010/10/09/00 Rogers, DiMego, Gayno
  • Replace the 1/2 deg RTG_SST analysis with the high-resolution (1/12th degree) RTG_SST analysis.
  • 2010/09/21/18 Janjic
  • Changed the routine that computes the pressure gradient force (pgforce): zeroed contributions to the "divergence correction" term along lateral boundaries.
  • 2010/09/16/18 Rogers
  • Both NDASB and NDASX restarted from satellite bias corrections from 12z 9/14 NDASX run to ensure cleaner comparison of the MODIS_IGBP land use (NAMX) vs USGS land use (in NAMB)
  • 2010/09/14/12 Ferrier
  • The following microphysics changes were put into NAMB:
    1. Uses expanded rain look up tables, increasing the maximum size of raindrops from 0.45 to 1.0 mm.
    2. Parameterizes sedimentation of 50-micron ice crystals.
    3. New cloud water to rain autoconversion (PRAUT) scheme following work of Liu & Daum et al. This new scheme no longer requires an autoconversion threshold, and the functional dependence of cloud water mixing ratio is quite different from the original Kessler-based formulation.
    4. NCW increased from 100 to 200 cm^-3.
    5. T_ice_init=0 results in faster initiation of ice. In 1D column runs it produced higher maximum surface rainfall rates.
    6. FLARGE2=0.07 based on tests by JMA visiting scientist Nakagawa in his global runs. The parameter determines the number fraction of large precipitation ice particles.
    7. VEL_INC~VRIMEF**2 – even faster rimed ice fall speeds than in the HAIL microphysics.
    8. These last two changes make the code faster and/or easier to read.
      1. Uses SQRT intrinsic function more, slight code optimization.
      2. New GET_INDEXR function to get an estimate for the mean drop size in microns (INDEXR).
  • 2010/08/16/12 Pyle, Rogers
  • Begin direct use of GDAS spectral coefficients in NPS (rather than a GDAS 1 deg lat/lon grid) to create the first guess at the start of the NDAS.
  • 2010/07/29/12 Pyle, Rogers
  • Begin use of modified vertical level distribution with more layers in the stratosphere
  • 2010/07/14/00 Wu, Pondeca
  • Begin use of dynamic reject list (from RTMA) for mesonet data in the GSI analysis.
  • 2010/05/17/06 Pyle
  • Boundary condition treatment changed to specify 5 boundary rows (LNSH, LNSV=5) instead of one row in the control run.
  • 2010/04/29/12 Pyle
  • Slight off-centering in time in the Crank-Nicholson scheme (parameter w1 in /atmos/dyn/nmm/module_DYNAMICS_ROUTINES.F90 changed from 1.0 to 0.9). This change was made to eliminate a failure in a test of a 4 km nested CONUS run, which was caused by large temperature oscillations in the vertical over the Sierra Nevada mountains as a strong jet streak was impinging on the mountains.
  • 2010/04/05/18 Wu, Parrish
  • Changes to the GSI analysis
    1. Update to new version (r7244)
    2. Turn on assimilation of GPS radio occultation (COSMIC) data
    3. GSI analysis will now update the first guess 2-m temperature/specific humidity and 10-m wind
  • 2010/03/30/12 Ferrier
  • In the radiation parameterization, the absorbtion coefficients for water and ice have been halved to 800 and 500, respectively.
  • 2010/03/18/18 Ferrier
  • Bug fix in the radiation code to remove spikes in the vertical profile of shortwave heating
  • 2010/03/09/00 Ferrier
  • Bug fixes to the radiation code:
    1. Fixed error with EQUIVALENCE statements for SAVEd arrays which were zeroing out lookup tables
    2. Removed two modifications to longwave radiation made for the June 2006 WRF-in-NAM implementation, that were removed from the ops NAM in March 2008:
      1. Remove averaging of longwave temperature tendencies from the lowest two model layer
      2. Compute the upward LW at the surface based on the ground (skin) temperature instead of the average of the skin and lowest model layer temperatures.
  • 2010/02/17/12 Rogers
  • Due to an oversight, the long-term precipitation budget from the operational NDAS has been used in every parallel NDAS run since 12z 2/2/10, instead of having it cycle independently. This has been fixed as of this cycle.
  • 2010/02/02/12 Rogers, Lin, Gayno
  • The b-grid copygb utility is used to 1) interpolate the daily CPC precip analysis to the b-grid (for the long-term precip budget used to bias-correct the precip analysis assimilated in the NDAS) and 2) interpolate the merged Stage II/IV precipitation analysis used in the NDAS forecast as forcing to the land-surface model. On 25 January, the b-grid copygb binary was changed on cirrus only, which corrupted these interpolations to the b-grid. This lasted until 27 January, when the parallel runs switched to stratus, where the copygb binary was unchanged. Because of this finding, it was decided to restart the NDASB land-states and long-term precipitation budget from the operational NDAS.
  • 2010/01/27/12 Janjic
  • Turned on Eulerian passive tracer advection of scalars
  • 2010/01/09/00 Treadon
  • Turned off assimilation of NOAA-19 AMSUA channel 8 radiances due to excessive noise.
  • 2009/12/18/06 Janjic
  • Removed the factor sqrt(2) from the denominator in all the definitions of hdac* in ./src/nmm_libutil/module_CONTROL.F90. This will lead to increased horizontal diffusion.
  • 2009/12/08/00 Wu, Keyser
  • "WDSATR" : superob'ed Scatterometer winds over oceans, were not used in the assimilation until this cycle because the prepbufr parm card did not set the observation error for this ob type, so they were given a quality mark=9, and therefore not used in the GSI analysis.
  • 2009/11/25/18 Wu
  • Additional observation types turned on in GSI analysis:
    1. ACARS humidity
    2. "WDSATR" : superob'ed Scatterometer winds over oceans
    3. HIRS and AMSU-A radiances from NOAA-19
    4. IASI radiances from METOP
    5. AMSU-A radiances from AQUA
  • 2009/11/10/18 Ferrier
  • Bug fix to NEMS configure file: changed the logical switch with turns on gravity wave drag/mountain blocking (GWDFLG) from ".true." to "true"; the code did not accept .true. as a valid setting and therefore reverted to the default setting which was "false". Thus, gravity wave drag/mountain blocking was not turned on in the NAMB parallel until this cycle.
  • 2009/10/07/00 Rogers
  • Because hourly post-processing was turned off to reduce disk space usage and the number of jobs on the system, the 00z NAMB 12-36 h precip for LSPA outside of the CONUS (OCONUS) was only processed at 3 hourly intervals. To fix this, hourly post-processing will be turned back on for the 00z NAMB cycle only as of this cycle.
  • Bug fix in NAM PRDGEN driver script; only run the job processing the 00z NAMB 12-36 h precip for OCONUS LSPA if forecast hour = 84. The script was being run for all forecast hours; this did not corrupt the OCONUS LSPA fields but it is a waste of resources to run it for all forecast hours.
  • Restarted cycled land states from the ops NDAS.
  • 2009/10/02/18 Parrish
  • Fixed a bug in the subroutine that reads the Level 2 radar data, it was using the incorrect analysis valid date when the input first guess was a NEMSIO restart file. This bug resulted in no level 2 radar data being used by the GSI analysis except in the first (tm12) NDAS analysis of each cycle.
  • 2009/09/29/18 Rogers, Treadon, Wu
  • Revert back to using operational NAM CRTM binary coefficients in the GSI analysis
  • Reset satellite bias corrections using files from the WRF-NMM (NDASX) parallel
  • 2009/09/28/12 Janjic, Pyle, Jovic, Rogers
  • New b-grid domain :
    • IM = 954
    • JM = 835
    • DX = .126
    • DY = .108
    • WBD = -60.039
    • SBD = -45.036
    • PHI0 = 54 (North latitude)
    • LAM0 = -106 (West longitude)
    • Southwest corner point = -7.491214752N 144.1336823W
    • Northeast corner point = 44.53958130N 14.80149937E
  • NDAS forecast component now runs with 5x divergence damping and a 20 second timestep, time step for free forecast still 26 2/3 seconds.
  • Since new b-grid domain is larger than the ops NAM e-grid domain, switch to using 00z NAMB 12-36 h precip for LSPA outside of the CONUS (OCONUS)
  • 2009/09/19/00 Vasic
  • Fix in NEMS/NMMB gravity wave drag module; since the gravity wave drag-specific fields are read into the NEMS model every time it runs, apply the radians to degrees conversion for the angle of the mountain range w/r/t east for all forecast runs, whether the logical switch RESTART is TRUE or FALSE. The code was only applying the conversion if RESTART=FALSE, since in the WRF-NMM the GWD fields were passed through via the cycled restart file, and only needed to be converted if RESTART=FALSE.
  • 2009/09/16/06 Gayno, Rogers
  • Sea ice update in NDASB sfcupdate script restored.
  • Fix in sfcupdate code: update the "THS" and "TSEA" arrays in the NEMSIO input file with the latest sea-surface temperature analysis. This change ensures that sea ice points that switch to open water are assigned a valid SST that is recognized as such in the forecast model. Failure to do so was causing the model to fail in the first (tm12) forecast of each 06z NDAS cycle since the 9/8/2009 change was made.
  • Restarted cycled land states from the ops NDAS.
  • 2009/09/15/06 Rogers
  • Temporarily disabled the sea ice update in the NDASB sfcupdate script since it is causing failures in the first (tm12) forecast of each 06z NDAS cycle.
  • 2009/09/08/06 Gayno, Rogers
  • Fixed a namelist error in the NDASB sfcupdate script; the output_file_type was not defined, so the code failed silently and did not insert into the NEMS input file the updated snow, sea-ice, and SST analyses.
  • Restarted cycled land states from the ops NDAS.
  • 2009/09/04/18 Ferrier, W. Wang, Parrish, Pyle, Rogers
  • Restored the first microphysics change made on 8/5: enforce a minimum number concentration for large ice particles (1 per liter) at all temperatures, rather than not enforcing it at >0C in the operational code where a fixed mean diameter is assumed.
  • Fixed an error in the GSI analysis : a bug in the interpolation of b-grid wind points had an error which caused the analysis to set the latitude of the North Pole to 39N.
  • 2009/08/31/12 NEMS/NMMB Team
  • Now running the NEMS code in the repository trunk, only significant change from the early August version running previously is that the accumulator variables were changed from scalars to 2-d arrays (ESMF "feature") and code fixes were installed so that time-averaged fields were correctly computed for forecast hours in between times when the accumulatiors are reset to zero (every 6-h in 00z/12z NAM, every 3-h in 06z/18z NAM and in NDAS)
  • 2009/08/29/00 Pyle
  • Changes to the lateral boumdary generation code (MKBND): refines a change to the boundary flux correction in SIG2HYB.f. Uses the GFS interface pressures surrounding the input wind point to get the true mass of the layer. Previously had been approximating the mass from differences in mid-layer pressure. Forecast impact minor relative to last change, but "more correct".
  • 2009/08/28/00 Pyle
  • Two changes to the lateral boumdary generation code (MKBND):
    1. Changed an aspect of the boundary wind flux correction to match what is done in the NPS. Only the winds that can impact the mass flux in/out of the domain are matched to the source data. On the W/E boundaries, only the U component is touched; on the N/S boundaries, only the V component is touched
    2. Changed the limits of inclusion on the source wind data. All input data that falls over PT to PSFC is considered. Previously only data falling between the mid-layer pressures at the top and bottom would be used.
  • 2009/08/18/18 Janjic, Rogers
  • Changed the sigma-pressure vertical coordinate interface pressure parameter (ptsgm) from 420 mb to 300 mb, the recommended value for the ivcoord=1 setting ("vc"). If ptsgm is > 300 mb with the vc option and if the surface pressure reaches 500 mb, the coordinate degenerates and layer thicknesses become negative.
  • 2009/08/17/12 Rogers
  • Due to degraded upper tropospheric height/temperature skill scores in the current NAMX (WRF-NMM) parallel, the first and third microphysics bug fixes (those involving large ice particles) added on 12z 8/5 were temporarily pulled.
  • 2009/08/11/12 Rogers
  • Begin making 00z and 12z 84-h forecasts every day, begin generation of verification files, designated "NAMB" in FVS database.
  • 2009/08/07/18 Lin
  • Fixed an error in ./src/nmm_phys/module_PRECIP_ADJUST.F90; the I,J dimensions in read statement for the precipitation analysis were each +1 more than they should have been; land-states restarted from ops NDAS.
  • NDAS Forecast code is now outputting the land-sfc precipitation accumulation (LSPA) and the NDAS long-term precipitation budget is now updated using the NDASB output.
  • 2009/08/05/12 Ferrier
  • Fix for the value for DTPH in subroutine Ferrier_init in module_MICROPHYSICS, which was set to a factor of 60 too large. The variable DTPH is locally derived from the input physics time step (GSMDT), which was in units of minutes for the Eta and WRF but in seconds in the NMMB. This had the effect of speeding up by roughly the same factor (60) those microphysical processes that convert cloud water to rain and to ice.
  • Addition of subroutine UPDATE_WATER near the start of PHY_RUN, used to update the WATER array from the CWM, F_rain, F_ice arrays after dynamics when the Ferrier microphysics is run.
  • Begin running with these three microphysics bug fixes from the NAM 2009 "bug-fix" parallel:
    1. Enforce a minimum number concentration for large ice particles (1 per liter) at all temperatures, rather than not enforcing it at >0C in the operational code where a fixed mean diameter is assumed.
    2. When initiating ice in water saturated conditions, growth rates are calculated by combining the number concentrations of ice initiation from eq. (2.4) of Meyers et al. (JAM, 1992) with estimated mass growth rates of ice crystals using the calculations of Miller and Young (JAS, 1979). The Miller-Young calculations were in cgs units (g per unit time), so the values in the model need to be multiplied by 1e.-3 to convert to mks units (kg per unit time) in subroutine MY_GROWTH_RATES.
    3. A dummy variable representing the mean mass of large ice particles (snow) was instead the first-guess mean diameter of the ice particles. The bug occurs only when the the first-guess number concentrations of large ice particles are outside of a predefined range between NLImin an
  • Restarted cycled land states from the ops NDAS on cirrus.
  • 2009/08/05/00 Parrish, Wu, Rogers
  • Begin use of GSI code and binary coefficients from GSI subversion trunk, no appreciable differences from previous version.
  • 2009/08/01/00 Wang, Vasic
  • Fixed an error near line 2130 of ./src/nmm_phys/module_PHYSICS_INIT_READ.F90 which caused the model top pressure (PT) to be set to zero instead of the number defined in the input NPS namelist.
  • 2009/07/31/18 Rogers, Lin
  • Glitch in script that created NDAS precipitation analysis files fixed; land-states restarted from ops NDAS.
  • 2009/07/30/12 V. Wong
  • Under stable conditions, modify roughness length for heat so that it is only a function of surface-layer bulk Richardson number, removing the dependence on surface elevation. (19 June 2007 NAM change)
  • 2009/07/29/18 NEMS/NMMB Team
  • Start of NDAS/NAM NEMS/NMMB parallel test: run 4 NDAS cycles per day, run one free forecast every 30-h
  • The following components/settings in the operational NDAS/NAM are not yet working/running in this parallel:
    1. Land-sfc precipitation accumulation (LSPA) is not yet output from the NDAS forecast, so the operational NDAS long-term precipitation budget is used to adjust the precipitation analysis. LSPA added to output and began recalculaution of long-term precipitation budget at 18z 8/17/2009/
    2. The operational NAM 00z 12-36 h precip is used for LSPA outside of the CONUS (OCONUS). Switched to 00z NAMB 12-36 h precip at 12z 9/28/2009.
    3. 19 June 2007 NAM Change: Under stable conditions, modify roughness length for heat so that it is only a function of surface-layer bulk Richardson number, removing the dependence on surface elevation (Added this change at 12z 7/30/2009)
    4. The NDAS forecast runs with the same amount of divergence damping as the NAM forecast; in operations, WRF-NMM NDAS forecast has 5x higher divergence damping than NAM forecast. (NOTE : NDASB forecast runs w/5x damoing as of 12z 9/28/2009)
    5. Two model changes in the 16 December 2008 NAM change package:
      1. The PBL/turbulance schemes were modified to mix each hydrometeor species in the vertical.
      2. To apply vertical diffusion for separate water species, the model was changed so that (a) it can apply vertical diffusion to an arbitrary number of species, (b) the counter gradient option can be applied to some or all of the species if desired, and (c) option to set to zero some or all of the surface fluxes is also made available.
  • The assumed fraction of large, precipitating ice particles to the total number concentration of ice particles (FLARGE2) is reduced from 0.2 to 0.03, which acts to increase the number of small, nonprecipitating ice particles.