Northwest AIRQUEST 12406

1
Northwest AIRQUEST12/4/06

• Cliff Mass
• University of Washington

2
Topics

• PBL Parameterizations
• The Current and Future State of the System
• The Data Assimilation Revolution

3
Boundary Layer Parameterizations

• It is now recognized both locally and nationally
  that the boundary layer parameterizations in
  current mesoscale models (e.g., WRF, MM5) have
  substantial weaknesses when run at typical
  resolutions of 4-15 km horizontal grid spacing.
• These problems are most profound for stable
  boundary conditions, which are unfortunately
  important for the air quality community.
  Example, maintaining a shallow (few hundred meter
  thick cold layer near the surface).
• Another issue is what type of parameterization is
  appropriate for grid spacings below 1 or 2-km,
  where we start to explicitly model large BL
  circulations

4
Shallow FogNov 19, 2005

• Held in at low levels for days
• MM5 could produce and maintain the inversionbut
  generally without the shallow mixed layer of cold
  air a few hundred m deep
• MM5 could not maintain the moisture at low levels

5
(No Transcript)
6
(No Transcript)
7
(No Transcript)
8
(No Transcript)
9
(No Transcript)
10
(No Transcript)
11
Where are the problems?

• PBL parameterizations themselves?
• Land surface models or surface specifications?
• Model dynamical cores (too much diffusion, etc)?
• Other physics (such as radiation?)

12
Who is working on PBL parameterizations and
evaluating current ones?

• Generally a hollowing out of the U.S. efforts
  (e.g., UW Energy Transfer Group)
• For MM5/WRF mainly Korean effort (Yonsei
  University)
• Still activity in Europemainly Dutch?
• Eric Grimits activities and Chris Brethertons
  group
• No one at NCEP actively working on PBL
  parameterizations for mesoscale models.

13
Important Note

• At the WRF annual workshop at NCAR, PBL was
  recognized as number one physics problem.
• I chaired the Physics working group and several
  individuals wanted to organize a PBL effort.

14
Eric Grimit Research Before Leaving

• He generated two months of output
  (November-December 2005) using 3 model
  configurations in WRF
• Default WRF with YSU-PBL
• YSU-PBL with Garratt SL
• UW-PBL with Garratt SL
• Note YSU is Yonsei PBL Schemean improved
  version of the current MRF scheme
• Did initial comparison/verification using tower
  data at Hanford.
• Some encouraging results.

15

• the average theta difference from the surface
  value (2-m actually) at 12 Z for each
  configuration from the model column closest to
  Hanford, WA

16
Where do we go from here?

• Chris Bretherton is now supervising the further
  analysis by an undergraduate student (Alfredo
  Arroyo) of these comparison runs.
• Eric Grimit has left for 3-tier, but would like
  to continue working on the problem at some
  (small) level.
• Chris is willing to lead a renewed PBL effort if
  there was some hope of funding.
• There was a meeting last January at Hanford to
  talk about a regional boundary layer initiative.

17
The NW has many of the pieces

• Considerable intellectual resources in NW (UW,
  WSU)
• Real-time modeling system infrastructure
• Substantial observational assets (e.g. PNNL and
  wonderful mesoscale network)
• Engaged user community (e.g., NW modeling
  consortium)
• Active partners north of the border
• Developing regional data assimilation system that
  could be a major tool.

18
Computer forecast models were nearly perfect for
the onset and amount of snow over Puget Sound.
The Missing Element Leadership and around
Portland. and we know why.
19
Current State of the System
20
High-Resolution Forecasts

• MM5-GFS 36, 12 and 4-km runs have been highly
  stable without any significant down times.
• The MM5-NAM 36-12 km has been moved from the old
  Tahoma to Linux machines..earlier availability
• Did very well for both the snow storm and
  flooding events.
• Grid-based bias correction for temperature and
  dew point are now online.

21
The Switch to WRF

• Testing the current version of WRF and some
  components of the new system.
• MM5 is verifying better still.

22
WRF

• Waiting on new version of WRF with nudging and
  better radiation schemes.
• Will evaluate for an extended period. If equal
  or superior will switch..after the consortium
  provides approval. Probable switch..next summer.

23
Ensemble Probabilistic Prediction

• Our 36-12 km mesoscale ensemble system is highly
  stable.
• Running on new Linux processors and includes
  extended physics ensemble.
• Substantial effort is going into development of
  post processing (bias correction and Bayesian
  model averaging)--which greatly enhances the
  probabilistic skill.
• The ensembles and many products are online.

24
(No Transcript)
25
(No Transcript)
26
The New Revolution Mesoscale Analysis using
Ensemble Techniques

• New mesoscale data assimilation approaches based
  on ensembles should allow great improvement in
  using all observational assets to produce a
  highly realistic 3D description of the regional
  atmosphere.
• Will be a boon for weather and air quality
  prediction, both for verification and for
  initializing our models.
• Will allow us to get the most out of the rapidly
  increasing number of surface observations and
  aircraft reports.
• We are building the system right now at the UW to
  do it.

27
How it works

• You have a collection of ensemble members, each a
  little different.
• You can correlate an observed parameter with 3D
  atmospheric structures produced by the model.
• Example for the current synoptic/mesoscale
  configurationmore precip at Sea Tac would be
  associated with more convergence in wind field
• Using the model as a key part of the data
  assimilation cycle.

28
The Start

• We have purchased and installed a new clusterthe
  AOR clusterto be the computational resources (13
  nodes with 4 processors each)
• Have purchased a new RAID array to serve it.
• Have hired a very good post-doc..Brian Ancelto
  do the workand he has begun to build the system.
• This will be the highest resolution attempt to do
  this in the country.
• Based on our successful coarse (45 km) resolution
  system.
• Probably 60- 90 members.

29
The Impact

• If this works, we will have extraordinarily high
  quality mesoscale analyses suitable for air
  quality and NWS needs
• If will produce high quality forecasts and will
  greatly enhance the 0-12 hr forecasts of the
  real-time system. Maybe much longer in some
  cases.
• Can overcome some of the PBL problems by pushing
  the model hard with observational assets.

30
NewAOR Cluster