[meg.infolist.announce] Status update on GFSv15 (FV3-based GFS)

Sat Mar 30 19:20:38 UTC 2019

Dear FV3 Community,

On May 17, 2018, EMC announced
<https://docs.google.com/document/d/1eTDPAPmpZQxNPIkX0v9x0vidAVu47bO-lhCEC2ljDI4/edit?usp=sharing>the
beginning of the evaluation period for version 15 of the Global Forecast
System (GFS), featuring the replacement of the spectral core with the FV3
core as well as the following additional model upgrades:

   -

   replacement of the Zhao-Carr microphysical scheme with the more
   sophisticated GFDL scheme.
   -

   an updated parameterization of ozone photochemistry with additional
   production and loss terms
   -

   a newly introduced parameterization of middle atmospheric water vapor
   photochemistry
   -

   a revised bare soil evaporation scheme

Three full years of fully-cycled data assimilation retrospective model
simulations (May 2015 through May 2018) were completed, and a real-time
parallel simulation with fully-cycled data assimilation was launched on
April 1, 2018. The validation and verification of the results are
summarized in the EMC  FV3 Evaluation Page
<https://www.emc.ncep.noaa.gov/users/Alicia.Bentley/fv3gfs/>. In October
2018, the NCEP Office of the Director approved of proceeding with plans to
implement FV3-based GFS in the Winter of 2019 (the briefing material can be
found here
<http://www.emc.ncep.noaa.gov/users/Alicia.Bentley/fv3gfs/updates/OD_Brief_10-01-18_FV3GFS.pptx>
).

During the evaluation phase two bugs were identified and addressed in  the
summer of 2018:

   1.

   Snow was not adequately melting under warm conditions. A change to model
   parameters associated with communicating snow amounts to the land surface
   was introduced in July 2018 to address this.
   2.

   The calculation of the solar zenith angle was incorrect. A fix was
   introduced in September 2018.

During the fall of 2018,  EMC identified excessively cold temperature
increments in the data assimilation system in polar regions. This was the
result of an incorrect parameter associated with supersaturation over ice.

As the real-time parallel run and model evaluation by the community entered
the Winter of 2018, it became clear the change made to the snow accounting
(item 1 above) inadvertently caused excessive accumulating snow in
marginally cold environments in mid-latitude storms, and the fix to the
solar zenith angle (item 2 above) inadvertently exacerbated an existing
cold bias at lower levels in the atmosphere. The increased cold bias in the
lower atmosphere compounded the excessive accumulated snow issue noted
above in item 1.

Once these issues were identified, NWS paused
<https://www.weather.gov/media/notification/scn19-12gfsv15.pdf>the FV3
implementation plans on February 26, 2019, and EMC began exploring
corrective actions to alleviate the excessive snow and cold bias:

   1.

   We changed the way snow amounts were calculated and communicated to the
   land surface model, basing it on the fraction of frozen precipitation
   falling on the ground rather than basing it on the total precipitation in
   cold conditions.
   2.

   We refined the interaction of radiation with cloud particles, allowing
   for each type of hydrometeor (convective rain, stratiform rain, snow,
   graupel, and ice) to assume its own physical characteristics as calculated
   by the GFDL microphysics scheme (like particle radius) and interact
   accordingly with the radiation scheme.
   3.

   We updated the supersaturation parameter over ice in the data
   assimilation system.

All three of these modifications were introduced into a new model
configuration with fully cycled data assimilation. This simulates the full
prediction system starting from mid-December 2018; it is now running in
real time. In addition,  a hurricane season (from August 26, 2018, through
October 31, 2018) was also simulated in a fully-cycled DA configuration
with these three modifications.

The results are promising. The three new modifications are significantly
mitigating the cold bias and excessive snow in the FV3GFS, especially in
the shorter lead times (0-96 hrs), while retaining the benefits of the new
FV3 dynamical core documented in the brief to the NCEP Director in October
2018. These benefits include:

   -

   (Significantly)  Improved 500-hPa anomaly correlation (NH and SH)
   -

   Intense tropical cyclone deepening in GFS not observed in FV3-based GFS
   -

   FV3-based GFS tropical cyclone track forecasts improved (within 5 days)
   -

   Warm season diurnal cycle of precipitation improved
   -

   Multiple tropical cyclone centers generated by GFS not seen in FV3-based
   GFS forecasts or analyses
   -

   New simulated composite reflectivity output is a nice addition
   -

   Some indication that FV3-based GFS can generate modest surface cold
   pools from significant convection
   -

   Improved ozone and water vapor physics and products
   -

   Improved precipitation ETS score (hit/miss/false alarm)

Even so, case studies of the eastern February storm and January arctic air
outbreak illustrate that the new model configuration is still too cold, and
we still see excessive snow in the medium range. EMC will continue to
explore ways to decrease this cold bias.

Additional material relevant to the performance of the new model
configuration is available here
<https://docs.google.com/presentation/d/1EafmMwFpg7V8yZwqQ0WT7WIxzGRKSvBOMbLc7hAukck/edit?usp=sharing>
.

The experiment that includes the three modifications and other results from
our recent explorations will be made available to the community as soon as
possible and will be incorporated into the real-time parallel experiment.
The NWS is still paused in its implementation plans pending additional
testing and verification that the benefits of FV3-based GFS are sustained.
The investigation results will be used by NWS leadership to determine if
the implementation schedule will remain in a hold status or be allowed to
proceed. In the latter case, we will conduct a planned 30-day technical
test which is the last step in our implementation process. The final
decision to implement the upgrade will depend on successful completion of
this final technical test.

Cheers,

Brian D. Gross
Director, NOAA/NWS/NCEP/Environmental Modeling Center

5830 University Research Court, Room 2050
College Park, MD 20740

Email: Brian.Gross at noaa.gov
Phone: 301-683-3748
Cell:  609-439-9702
-- 
Sent from my mobile device
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://www.lstsrv.ncep.noaa.gov/pipermail/meg.infolist/attachments/20190330/b33584ec/attachment.html 