LOG OF OPERATIONAL ETA MODEL CHANGES

COMING SOON...

MARCH 04

• 16/12Z: The following changes were made to the operational EDAS/Eta:
  • Use of daily gauge data for precipitation assimilation bias adjustment.
  • Assimilation of GOES cloud top radiances.
  • Land-sfc model precip type now based on model microphysics.
  • Other post-processing changes.
  Further information can be found here.

JULY 03

• 08/12Z: The following changes were made to the operational EDAS/Eta:
  • Extension of the "off-time" (06/18Z) runs to 84 hours.
  • Addition of hourly ouput on selected grids to 36 hours.
  • Modifications to the cloud physics and radiation.
  • Assimilation of GOES cloud top pressure, Stage IV precipitation data, and super-observations of NEXRAD radial wind data.
  Further information can be found here .

JUNE 02

• 19/12Z: The following changes were made to the operational EDAS/Eta:
  • The physics time step was reduced from 500 seconds to 300 to mitigate a pathological warm 2-meter and skin temperature bias.
  • The radiation driver was modified to eliminate very small values of cloud water and very thin clouds to prevent model crashes caused by the computation of shortwave radiative fluxes.
  • The land-surface physics were changed such that under very dry soil conditions, the soil moisture availability can not become negative.

May 02

• 21/12Z: The following changes were made to the operational EDAS/Eta, in order to
  • Correctly pass the adjusted convective precipitation rates in the precipitation assimilation to radiation
  • Properly write convective cloud properties used in radiation calculations to the model restart files
  • Modify a number of routines to improve model efficiency
  • Lower the limits of turbulent kinetic energy (Q2 array) in the advection routines to 0.2

FEBRUARY 02

• 26/12Z: The following change was made to the operational EDAS/Eta:
  • Increased the treatment of thermal conductivity for both patchy snow and full snow cover in response to a cold bias in areas with snow cover. (NOAH LSM version 2.3.1)
  Further information can be found here .

NOVEMBER 01

• 27/12Z: The following changes were made to the operational EDAS/Eta:
  • Increased horizontal and vertical resolution from 22 km/50 levels to 12 km/60 levels
  • New cloud microphysics scheme
  • Improvements to the 3DVAR initialization
  Further information can be found here .

JULY 01

• 24/12Z: The following changes were made to the operational EDAS/Eta:
  • 3DVAR analysis modified for improved mass-wind balance constraint
  • Assimilation of the 4-km NCEP National Precipitation Analysis (stage II) in the EDAS
  • Extensive modifications to the Eta model land-surface physics (NOAH LSM version 2.3)
  Further information can be found here.

APRIL 01

• 17/12Z: The 00Z and 12Z cycles have been extended to 84 hours, and a correction to the computation of 10-meter winds has been made. Further information can be found here.

JANUARY 01

• 30/12Z: The sea-surface temperature data used in the model has changed. The Real Time Global SST RTG_SST) analysis is now being used; it has a superior analysis of the Gulf Stream and is able to resolve the colder shelf water as well. Further details on this change can be found here.

SEPTEMBER 00

• 26/12Z: The model is now run at a horizontal resolution of 22 km with 50 levels in the vertical. The domain has been expanded slightly, and the 3DVAR initialization has been improved. Vertical advection of cloud ice/water has been added, and slight modifications to the convective scheme have been made. Further details on this change can be found in the TPB.

MARCH 00

• 29/12Z: The 00Z and 12Z runs have been extended to 60 hours. The convective scheme has been modified to reduce the dry bias in the western U.S. and the moist bias along the Gulf and southeast U.S. coast. VAD winds are now being used again with a new quality control code, and balloon drift is now accounted for in the processing of radiosonde data. Further details on this change can be found in the TPB.

JUNE 99

• 22/12Z: The BUFR output was modified to increase the number of variables created as output and to increase the number of stations. More info is available here.

MAY 99

• 13/12Z: Changes have been made to the 3DVAR to improve the mass/wind balance (at slight expense to the fit to radiosonde data). Tests of the new 3DVAR analysis produced much higher forecast accuracy. Further details on this change can be found in the TPB.

FEBRUARY 99

• 16/12Z: Puerto Rico grid #237 has been added to the Eta output. (filename on cray is eta.TXXZ.AWP237HH.tm00)

DECEMBER 98

• 23/12Z: The 03 Nov 98 change to the Eta post processor involving computing RH has been removed. RH will now always be computed with respect to water saturation. Expect to see, as before 3 Nov, the same lower RH values at upper levels or wherever cloud ice is found. This change has not been implemented for all RH computations (Eta boundary layers, NGM and LFM layer calculations that get into FOUS and AFOS) those will be fixed as soon as possible.

NOVEMBER 98

• 09/12Z: The control cards for the product generatory were modified to remove all smoothing for fields with discontinuities or sharp gradients. (such as topography, cloud base/top height, max wind level, etc.) This is done to prevent unphysical values such as negative cloud base heights, for example.
• 03/12Z: The Eta package has been upgraded to use fortran 90 and to be Y2K compliant. Changes have also been made to several codes to use the same 2-D grid indexing that the model has been using for some time. This also means that 'native' Eta grid output has changed, the new grid number is 192 and the GDS has also changed to include more information (type 203). Modifications have been made to the 3DVAR to fix an error that was resulting in the exclusing of almost all surface data, and to improve the analysis fit to data.

  From this set of corrections to the 3DVAR, forecasters should expect to see improved Eta surface and low-level analyses which better reflect the observations. In terms of impact on forecasts, these will be generally minor. What we saw from tests were less in magnitude than the impact we saw from the AVN going from T170 back to T126.

SEPTEMBER 98

• 30/12Z: The control cards for the product generator have been modified to include the Shuell sea level pressure reduction in the AWIPS grid 211 file.

JUNE 98

• 03/18Z: The Early Eta is running 4 times a day, at 00, 03, 12, and 18Z. The 03 and 15Z Meso Eta runs have been replaced by the 03 and 18Z 32km Eta runs. Output from the "off-time" runs keeps the Meso Eta naming convention and model id in the GRIB PDS until further notice.

APRIL 98

• 27/12Z: The GOES-PW input to the Early Eta 3DVAR EDAS was temporarily turned off.

FEBRUARY 98

• 9/12Z: The Early Eta was upgraded to 32km/45lyrs, and the EDAS now uses 3DVAR instead of OI. The domain is slightly smaller, but still covers Alaska and Hawaii. For more details, check out the 32km TPB.

JANUARY 98

• 12/12Z: BUFR tables for the hourly output from the MESO are being modified accommodate the assignment of numbers and mnemonics for parameters in the NCEP BUFR tables. This change will make the MESO Eta parameter numbers consistent with the new Eta 32 and the other NCEP models.

SEPTEMBER 97

• 9/12Z: The station list for the hourly BUFR sounding forecasts was changed. Several new stations were added, mainly in the west, and a few stations near water were moved to the closest water point. A detailed list of changes is available here .

AUGUST 97

• 19/12Z: Model code was changed to remove an error in the computation of the depth of the planetary boundary layer. This error reduced vertical moisture transport out of the lowest model layers and led to a reduction in the prediction of heavy precipitation events.

JULY 97

• 29/12Z: Several new output fields were made available in the post processor, cloud water was fixed in the Early Eta, and several changes were made to the station list for hourly BUFR output. Details can be found in the FAQ .

JUNE 97

• 27/03Z: The horizontal domain for the Eta-10 was reduced. This was necessary in order to continue running a 10km Eta while testing the new 32km Early Meso.
• 10/12Z: Changes were made to the PDS for boundary layer fields and precipitable water to follow the GRIB definitions more exactly. Look here for a more detailed list of these changes.

MARCH 97

• 18/12Z: Two masking arrays (one that masks underground points and another that masks the outer two rows of the domain) were using an improper index in the part of the radiation code that handles cloud/radiation interaction. The result was relatively small errors occurring on even rows just to the east of the step topography and just to the east of the western boundary. The proper index has been inserted.

FEBRUARY 97

• 18/12Z: A "bundle" of changes were implemented in both the Early and Meso (15Z). Changes to the model physics include addition of a form-drag scheme, changes to the radiation scheme, changes to the snow melt physics and bare soil evaporation. Several post-processed fields have been modified, mainly to fix errors with header information in GRIB. Some files are now output with a 3h frequency from the Early Eta. Look here for a more detailed list of changes to codes and output files.
• 11/12Z: The Meso and Early Eta networks will now use observational data which is decoded on the UNIX workstations, ingested into the Cray BUFR /dcom database, and time-windowed and duplicate-checked in the "dump script" utility programs. This change disconnects the Meso and Eta networks from the HDS decoders and Office Note 29 processing. However, the HDS Office Note 29 data sets will continue (for the short term) to provide a backup to the Cray BUFR data dumps in the event that there is a failure at some point in the new UNIX data processing stream.
• 06/12Z: Array lengths in the OI quality control code have been enlarged to accommodate more observations.

JANUARY 97

• 14/12Z: The station list for the output sounding data has been modified to add Houlton, ME as station number 00150.
• 02/12Z: An array dimension in an input parameter file for the Eta/Meso OI analysis code was increased to allow more observations to be processed by the analysis without failure.

DECEMBER 96

• 27/12Z: The Meso Eta forecast model was modified to correct an error occurring in the code where the boundary conditions are created for the eastern 10km Eta nest. Since that nest is not being used, the creation of those boundaries has been commented out.

JULY 96

• 24/12Z: The green vegetation fraction in the Early Eta was modified, it was found to be too low.

JUNE 96

• 28/21Z: The faster Meso post-processor caused problems with AWIPS processing and file transfer to OSO, since the output files are now produced in a span of about 15 min rather than spaced out in time. AWIPS processing is now started after the 33h forecast is complete.
• 25/12Z: The Early Eta will output a new file (AWPICF, AWIPS grid 212), which is input to the Interactive Computer Worded Forecasts (ICWF) code. After the ICWF files are created, an alert will be sent to the Olympic weather forecasters in Atlanta.
• 18/12Z: The source code for the Meso-Eta was modified to eliminate known out-of-bounds references that can be dangerous with multitasking, eliminate a huge scratch COMMON that contributed to ambiguity in the code, and extract conditions for the nested 10km Eta forecast over the eastern U.S. (the Olympic domain). Also, the faster post-processor was also implemented. Run time has decreased substantially. The script which processes observational data into quality controlled "prepbufr" format for subsequent assimilation into the Eta/EDAS, Eta/MESO (and RGL/RDAS) analyses has been modified to point to the "best available guess" at off-time cycles 3, 9, 15 and 21Z. Prior to this, the 9-hour forecast from the FNL/GDAS, initiating 9-hours prior to cycle time, was the only guess used. This caused a failure in the 15Z Meso Eta on 12 June 1996 because the 06Z GDAS did not run on that day.

MAY 96

• 07/12Z: A faster version of the Eta Model and post is being implemented into the Meso. This code produces identical results as the current version of the code but utilizes the multiple CPUs much more effectively and runs much faster.
• 02/12Z: The previous change to the analysis increment code had a problem in implementation which was subsequently fixed as of this cycle.

APRIL 96

• 30/12Z: The code that computes analysis increments was changed to remove inconsistencies near higher terrain. This fixes a problem with the initial fields noted by many forecasters where the data and analyzed quantites were quite different.
• 02/12Z: The surface topography file in the early Eta was modified to correct some problems. The station lists for the BUFR output for the Meso and early Eta were brought into agreement and the lat/lon of Buffalo, NY was changed. In the Meso, the analysis increment code was modified to allow for TOVS soundings, and problems found in the code that interpolates the GDAS guess surface fields were fixed.

MARCH 96

• 27/12Z: The data preprocessing jobs were modified to allow in more satellite wind and marine data. The code that computes increments between the model guess and data was also corrected to remove inconsistencies found in the data format. These changes went into the early Eta on 00Z 3/28.
• 05/12Z: The meso directory structure on the nic server will be changed from meso.00z and meso.12z to meso.03z and meso.15z. This changes is being made in order for the directory structure to match the actual forecast time of the model runs. There will be no change in the model output.

FEBRUARY 96

• 09/09Z: First successful run of the 09Z Meso. Regular 09Z data dump has begun, the 09Z Meso Eta run will in parallel testing mode. This is to support the 1996 Olympics forecasting effort.
• 08/12Z: The 40km AWIPS files on the OSO server should now have the precip, SLP, CAPE, and CINH.
• 06/12Z: Files that contain the correct tables for the hourly BUFR output were put into the Early Eta as of this cycle.

JANUARY 96

• 31/12Z: The Early Eta has been upgraded to match the code currently running in the Meso Eta. This includes a new 2-layer soil model, new turbulence, parallelized radiation code, BUFR hourly sounding output performed as an outside post-processor, and 2-D (I,J indexing) code. Look here for the NTIM on this change.
• 22/18Z: The time window for choosing upper air data for the Meso Eta had been specified incorrectly, allowing for day-old data to exist in the data dumps. This was fixed as of this cycle.
• 03/12Z: BUFR prepda files are now used, which allows for new observations such as the SSM/I precipitable water.

DECEMBER 95

• 21/12Z: AWIPS processing is now performed as soon as the files are available rather than waiting until the entire forecast is finished. This should make files available quicker on the OSO server.

• 19/12Z: The Meso post change made on 12/12 did not update the control file, resulting in 8-bit packing for each GRIB variable. This was fixed as of this cycle. The solar geometry constants were updated to include 1996. This was done in the Early and the Meso.

• 12/12Z: The post was modified to have greater control on the precision of the output fields in GRIB. The operational early and meso Eta were changed as of this cycle. Routines affected are AVBLFLDS, GRIBIT, GTBITS, OPTIONS, READCNTRL2, RQSTFLD, and all control files. SCAL was added to each variable in the control files.

NOVEMBER 95

• 09/00Z: The error fixed in the meso on 6/15/95 (cloud water reset to zero at the beginning of each eta model forecast w/RESTRT=TRUE) was not fixed in the /dm/wd22tb/standard/fcst version (1-d) which was implemented in the 48 km EDAS/Early Eta. The JIF submitted to fix this problem was implemented on this date.

OCTOBER 95

• 25/12Z: Changes JIF'd into the MESO run:
  • (1) The sounding post processing was put into separate jobs external to both the forecast and the regular post. BFRIZE,BFRHDR,ANEMLV6,SHLTRX,CALHEL,CALWXT and WETBULB were removed from the model as a result.
  • (2) The turbulence now uses Paulson functions.
  • (3) The 2m 10m variables are diagnosed inside of SFCDIF.
  • (4) An error was removed in MIXLEN where AUBR should have been multipled by CUBR.
  • (5) Common block SRFDSP was INCLUDE'd in INIT and in CRST12 as it should have been.
  • (6) SFCDIF was modified to incorporate the Zilitinkevitch fix for Z0T using C=0.10.
  • (7) ALL smoothing in the post-processor was taken out.
  • (8) Accumulated precip type was removed from the post and several GRIB parameter numbers were updated to reflect changes being made to the documentation.
  • (9) The station list for the hourly output was modified to add GCIP stations and match the one used in the early Eta.
• 12/12Z: The Early Eta was upgraded : (a) 48km resolution; (b) explicit cloud water added; (c) 12-hr EDAS for initialization. Also, SSM/I data over water now being used.

AUGUST 95

• 21/12Z: A small error was removed in MIXLEN. In the computation of QOL2ST, AUBR needs to be multiplied by CUBR.
• 14/12Z: In BOCOH.f, the common block PVRBL2 was replaced with CLDWTR and the array W was replaced with CWM. This has no effect on the integration but will should reduce the memory size by eliminating the 3-D array W.

JULY 95

• 31/12Z: A bug in rotating the winds for Lambert conformal grids was fixed in the post. E2OUT and EUVGUV were changed to include the constant of the cone in computing the rotation angle. /dm/~/standard/post, /dm/~/MESO/post, and /c90disk1/~/MESO/post were changed and the 40km post was recompiled. Also, makefiles were changed in ~/fcst directories to point to the /ntprod/util/bufr/obj.bufr/bufr.o library. The previous library (wd20jw) had been removed.
• 13/12Z: The 2-D version was again put into the operational Meso run but with NCPUS reduced to 8 from 12 in order to reduce the size of the executable.
• 12/00Z: The 2-D version was removed from the operational Meso run because of swapping/thrashing problems that are probably directly related to the size of the executable (made very large by the explicit parallelization of the radiation). The standard 1-D code was put back in place.
• 10/12Z: The new 2-D version of the Eta code with the improved soil and turbulence packages was put into the operational MESO run.

JUNE 95

• 15/12Z: A major fix was put into the operational MESO. The cloud water had been reset to zero at the beginning of the free forecast and Zhao's critical humidity interpolating factors were also reset. Both of these errors were corrected.