Decision Brief: Upgrade to HiResWindow

1
Decision Brief Upgrade to HiResWindow
Mesoscale Modeling Branch Geoff DiMego Matt Pyle
Eric Rogers 4 September 2007
where the nations climate and weather services
begin
2
Why Did SPC Start Looking at High Resolution WRF
Models?

• Severe weather types (tornadoes, hail, wind
  damage) can be closely related to convective mode
• Tornadoes (discrete supercells)
• Damaging wind (bow echoes and QLCSs)
• SPC working to increase lead time of watches and
  provide probabilistic information about tornado,
  hail, and wind threats in Day 1 Severe Weather
  Outlooks
• Accurate forecasts require knowledge about
  where and when storms will develop and how
  they will evolve
• There is a need to better predict convective mode
  and character of storms (stormscale details)
• Environmental clues (CAPE/shear, etc.) may not be
  sufficient
• Operational mesoscale models often lack smaller
  scale details

Courtesy of SPC Weiss et. al.
3
There were some striking successes from the very
beginning.

• Beginning in April 2004, Matt Pyle began running
  offline (on development/backup side of NCEPs
  computer) and providing to SPC
• 00z run each day to 36 hours
• High resolution WRF-NMM
• Initially 4.5 km, ultimately 4.0 km resolution
• Initially central CONUS, ultimately east-central
  CONUS
• SPC so pleased with the guidance, they would
  not let Matt turn the run off after initial
  Spring Program 2004 and he has been making the
  run every day since.

4
Example of Explicit 4.5 km WRF-NMM courtesy of
Jack Kain
WRF 24 hour 4.5 km forecast of 1 hour accumulated
precipitation valid at 00Z April 21, 2004 (better
than 12 hour forecasts by operational models)
4.5 km WRF-NMM
Verifying 2 km radar reflectivity
5
Over the Years, SPC Finds Value in HiRes Guidance

• Primarily in Convective Mode
• To lesser extent in location evolution
• Not surprising, then, they have requested
• Higher resolution
• Larger domain
• More frequent runs

6
HiResWindow Upgrade12/19/2006 Request from SPC

• Elevate Matt Pyles daily offline 00z run
• Twice per day runs at 00z 12z
• Expanded (East-Central) domain
• Higher (4 km) resolution WRF-NMM
• Will get 48 hr guidance instead of 36 hr
• Will get both WRF-NMM and WRF-ARW
• New requirements incorporated into plans for 2007
  HiResWindow upgrade

7
HiResWindow Upgrade 2007

• Model Upgrades
• Upgrade WRF-NMM
• From version 1.3 to 2.2
• IJK loop storage ordering (5-10 faster)
• Increase resolution from 5.2 km to 4.0 km
• Upgrade WRF-ARW
• From version 1.3 to 2.2
• Increase resolution from 5.8 km to 5.1 km
• Expand large domains
• Roughly double area covered
• 3 CONUS nests reduced to 2 overlapping
  (West-Central East Central)
• Allows schedule adjustment to run SPCs preferred
  domain (East-Central CONUS) at both 00z and 12z
• Post-Processing and New Products
• Add ability to output simulated reflectivity

8
HIRES Window 2007 UpgradeDomain Size Changes
New Large Domains 4.0 km for WRF-NMM 5.1 km for
WRF-ARW Small domain size is unchanged
Current Large Small Domains 5.2 km for
WRF-NMM 5.8 km for WRF-ARW
9
Old vs New Schedule
Retain FOUR routine runs made at the same time
every day. Both cores run for one large domain
and one small domain.

• Old Schedule
• 00Z Alaska Hawaii
• 06Z Western 1/3 CONUS Puerto Rico
• 12Z Central 1/3 CONUS Hawaii
• 18Z Eastern 1/3 CONUS Puerto Rico

• New Schedule
• 00Z Eastern 2/3 CONUS Hawaii
• 06Z Western 2/3 CONUS Puerto Rico
• 12Z Eastern 2/3 CONUS Hawaii
• 18Z Alaska Puerto Rico

Everyone gets daily high resolution runs if
only if hurricane runs are not needed.
10
HiResWindow WRF Configurations
(No Parameterized Convection)
WRF-NMM WRF-ARW
Horizontal Grid Spacing (km) 4.0 5.1
Vertical Levels 35 Sigma-Pressure 35 Sigma
PBL/Turbulence MYJ YSU
Microphysics Ferrier WSM3
Land-Surface NOAH NOAH
Radiation (SW/LW) GFDL/GFDL Dudhia/RRTM
11
Online Sites

• Matt Pyles online site http//www.emc.ncep.noaa.
  gov/mmb/mmbpll/cent4km/v2/
• Online Display at NCEPs Parallel Model Analysis
  and Forecast page established 2 July 2007
  http//www.nco.ncep.noaa.gov/pmb/nwpara/analysis/
• EMCs verification 7-day time-series. However,
  with the lack of runs due to tropical cyclone
  activity, these are very sparse
  http//www.emc.ncep.noaa.gov/mmb/mmbpll/mmbverif.h
  iresw_new2007/

12
Pyle Webpage Now Displaying Simulated
Reflectivityhttp//www.emc.ncep.noaa.gov/mmb/mmbp
ll/cent4km/v2/
13
Mesoscale Detail this 12 h forecast captures
the reflectivity minimum in eastern NY in the lee
of the Berkshires/Green Mountains, and some of
the enhanced reflectivites in northern CT and
extending into SE New York.
Albany, NY Radar Image
HRW Simulated Radar Reflectivity
14
Spring Program 2007
NMM4
ARW4
Circles denote locations of rotating updrafts
where updraft helicity is at least 50 m2s-2
Observed Composite Reflectivity Courtesy Kain,
Weiss Bright
15
Forecast Fit-to-Sfc Temperature Obs
Observed Operational Parallel
ARW for 23-30 August
NMM for 18-30 August
16
Forecast Fit-to-Sfc Wind Obs
Observed Operational Parallel
ARW for 23-30 August
NMM for 18-30 August
17
Eastern Western QPF scores
Western (WRF-NMM), 8/18 to 8/28
minimal
precip to verify during this period
Eastern (WRF-NMM), 8/18 to 8/28
18
Longer Duration QPF Verification12 Dec, 2006
29 Aug, 2007 (less Feb/Mar)

• SPC 4km run serving as proxy for Para HRW because
  Opnl Central CONUS run is a 12Z run while the SPC
  run is a 00Z run and they didnt cover exactly
  the same domain - so
• Each was independently verified against the NAM
  (these individual comparisons to the NAM are
  clean)
• The ratio of (ops HRW / NAM) is applied to get a
  relative score for the ops HRW against the NAM
  verifying the SPC run. Not perfect, but it
  levels the playing field.

19
QPF Verification Summary

• Parallel HiResW precip (WRF-NMM core) has
  comparable to slightly better ETS and a higher
  bias for heavier precipitation than the current
  production version.
• Improving the high warm season precip bias in the
  WRF-NMM will become an even more important
  priority between now and when there are 4 km
  Alaska CONUS nests run within the NAM in 2010.

20
14 August 2007 Example Case

• The HiresW runs (WRF-ARW and particularly the
  WRF-NMM) provide useful guidance of a convective
  line sagging south on the morning of 14 August
  2007.
• Both HiresW runs miss the timing by several hours
  (too slow), but highlight the possibility of
  heavy rain that is absent in the NAM forecast.
• Ops HiresW WRF-NMM run appears to capture event
  as well, but without instantaneous precipitation
  intensity information a direct comparison is
  impossible.

21
09 Z
22
10 Z
23
11 Z
24
12 Z
25
13 Z
26
14 Z
27
14 Z
11 Z
28
NAM forecast valid 12Z
NAM forecast valid 15Z
29
NMM para hiresw forecast valid 12Z
NMM para hiresw forecast valid 15Z
30
ARW para hiresw forecast valid 12Z
ARW para hiresw forecast valid 15Z
31
NMM ops hiresw forecast valid 12Z 3 h
precip
NMM ops hiresw forecast valid 15Z 3 h
precip
32
Eastern Region Interest

• 21 June, Jeff Tongue wrote Geoff - This is
  wonderful information! I'm still not clear on the
  timeline for the change to the 00 and 12 runs for
  the Eastern HiRes WIndow? We can wait for OB 8.3
  or what ever to get the data in AWIPS, but
  we're wondering when they'll be available on the
  web? (P.S. I'll subscribe to the ModelEval list
  server - thanks).
• EMC will attempt to add the hourly BUFR sounding
  data Jeff requested here in the near future when
  we add the hourly gridded output capability for
  SPC.

33
SPC Feedback

• List of NWS forecasters who participated in the
  2007 Spring Experiment for one week each
• Central Region - Matt Bunkers (SOO Rapid City SD)
  
• Western Region - Randy Graham (SOO Salt Lake City
  UT)
• Southern Region - Eric Platt (Forecaster Midland
  TX)
• Southern Region - Ken James (Forecaster Norman
  OK)
• Eastern Region - Steve Zubrick (SOO Sterling VA)
• Eastern Region - Paul Sisson (SOO Burlington VT)
• Southern Region - Bernard Meisner (SSD SRH Fort
  Worth TX)

34
SPC Evaluation Steve Weisspart 1

• SPC had only limited opportunities to examine the
  parallel runs of the HRW WRFs during the summer
  30 day evaluation period. 
• More importantly, however, we have been utilizing
  the once daily experimental WRF-NMM4 in both the
  annual Hazardous Weather Testbed Spring
  Experiment activities and daily operational
  severe weather forecasting at SPC since April
  2004. 
• This convection-allowing model was first tested
  during the 2004 Spring Experiment, which focused
  on exploring the utility of high resolution
  versions of the WRF-NMM, as well as WRF-ARW
  models from NCAR and OU-CAPS, to provide useful
  guidance for severe weather forecasters. 
• Spring Experiments examining WRF-NMM and WRF-ARW
  models were also conducted in 2005 and 2007,
  where intensive tests and evaluations were again
  conducted daily during severe weather episodes
  across the central and eastern U.S. 

35
SPC Evaluation Steve WeissPart 2

• Since 2004, the model has been periodically
  upgraded, including in early 2007 when NMM
  physics were standardized, grid length was
  decreased to 4 km, the domain was expanded, and
  model run time was optimized.  Most importantly,
  the early success of the WRF-NMM4 in spring 2004
  resulted in the model output being distributed to
  SPC operational forecasters in real-time, and
  over the last three years they have evaluated its
  performance and utility for a wide variety of
  geographic, seasonal, and synoptic regimes over
  the central and eastern states.  Most of our
  attention has focused on the ability of the
  WRF-NMM4 to resolve smaller scale features such
  as convective storms through the use of simulated
  reflectivity fields.  Compared to operational
  mesoscale models that incorporate convective
  parameterization, the WRF-NMM4 has been found to
  provide unique information about details of
  convective initiation, intensity, evolution, and
  mode (or morphology), which are critical aspects
  to severe weather forecasting.  The accurate
  prediction of convective mode (such as discrete
  cells, lines, or multicell complexes) is directly
  related to the occurrence of  tornadoes, large
  hail, and damaging winds, which tend to occur
  preferentially with specific convective modes. 
  In this area alone, SPC forecasters have found
  that the near-stormscale resolution of the
  WRF-NMM4 has substantially contributed to more
  confident and skillful forecasts of specific
  severe weather threats.  This has been aided by
  the extraction of gridded information concerning
  the dynamic structure of model-generated
  supercell thunderstorms, which are convective
  storms with persistent rotating updrafts.  The
  value of the WRF-NMM4 has been especially evident
  during strongly forced situations that are
  associated with severe weather outbreaks. (See
  attached powerpoint file entitled NWA 2006
  WRF-NMM4 Outbreaks that contains a presentation
  given at the NWA Meeting in fall 2006.) 

36
SPC Evaluation Steve WeissPart 3

• Since late 2006, SPC has had access to a
  real-time 4 km version of the ARW that is run
  once daily at NSSL.  SPC forecasters have been
  using both models in their severe forecasting
  mission, and have identified various strengths
  and weaknesses with both models but have learned
  how to use these performance characteristics to
  their advantage.  For example, the NMM appears to
  display a high convective storm/precipitation
  bias whereas the most recent ARW versions have a
  slight low bias.  Of course, the
  convection-allowing WRF models are relatively
  early in their development and considerable work
  remains, such as improving physics and
  development of cutting-edge data assimilation
  systems that are specifically designed for high
  resolution models with grid lengths lt 4 km.  But
  we have also seen considerable evidence during
  the last three years that operational severe
  weather forecasters at SPC have benefited from
  the daily availability of the WRF-NMM4 output,
  and more recently WRF-ARW4 output, and these
  benefits are reflected in improved SPC forecast
  products on important severe weather days.  See
  the attached powerpoint file entitled WRF-NMM
  and WRF-ARW Use at SPC, which was part of a
  presentation given at the 2007 AMS WAF/NWP
  Conference in June. Furthermore, the increasing
  number of WFOs running local experimental
  limited-area WRF models and viewing the WRF-NMM4
  online web page strongly suggests that local
  forecasters believe there are advantages to
  having routine access to operational high
  resolution model output. 

37
SPC Evaluation Steve WeissPart 4

• After using the WRF-NMM4 in operational severe
  weather forecasting for more than three years,
  the SPC strongly supports upgrading the High Res
  Window slot with the latest versions of the
  WRF-NMM4 and WRF-ARW5, and running them twice
  daily at 00z and 12z over an expanded
  central/east region, and once daily at 06z over
  an expanded west/central region.
•  
• Recommendation Implement as proposed XXX
•  
•  
• The need for hourly grids slipped through the
  cracks for the HRW upgrade, as we had assumed
  that the hourly output frequency of Matt Pyles
  experimental run would automatically transfer to
  the HRW.  We agree with the advice of NCO to
  implement the proposed version in September and
  then provide hourly grids early in FY08.  We are
  concerned, however, that with the approach of the
  secondary fall severe weather season over the
  southeast/Ohio Valley region from late October
  into December, there may be a need to continue
  Matts experimental run until the hourly grids
  can be implemented in the HRW.  This has been
  discussed in several email exchanges earlier
  today with Brent Gordon.
• EMC has agreed to continue running Matt Pyles
  experimental run with hourly output until this
  capability can be added to the Production run.

38
OSIP / SREC Progress at Silver Spring

• OSIP Project 07-012
• Cleared Gate 2
• Directed to SREC
• Worked with regions to pare SBN output by 23 May
  to just over 900mb per cycle
• Made list for AWIPS Build OB8.3, then dropped due
  to resource limitations
• Accepted as part of OB9