Consortium Meeting
June 14th 2012
Cliff Mass
University of Washington
Web Statistics
Hits
168… is WA School Information Processing Coop
Data (KB)
Finally…Canadians in Check
Rainshadow Verification
West minus East Precipitation
• East: Yakima,
Ellensburg and
Wenatchee
• West: Spada
Lake, Gold Bar
ES, South Fork
Tolt Reservoir
Observed versus WRF 1.3 km
Model Improvement
• The expanded 1.3 km domain is completely
reliable.
• But it has revealed/accentuated some issues
we knew where there and now should deal
with it.
• One example: how we handle the upper
portion of our domain.
Problems with our model top
• In the past we have had a relatively low top (100
hPA) and used the simple diffusive upper layer to
attenuation reflections off the model top.
• 100 hPA is too low for a number of reasons and
the diffusive upper layer was not effective
enough.
• We had done some experiments before that did
not suggest a huge impact on 12 and 4 km
domains..that is why we had left it alone.
4km
1.3 km
700 hPa Vertical Velocity
Solution
• Completed an extensive series of experiments,
working with Dale Durran.
• The Solution
– Move the model top to 50 hPa
– Switch to the WRF Rayleigh damping
parameterization
– Move enough levels to the upper troposphere to
insure 5-7 levels over a vertical wavelength of
upwardly propagating waves (roughly 6 km
wavelength)
Test Summary Web Page
Current
Proposed
Before
After
Dealing with the stable PBL
problem
• Would hyper-resolution help?
• Version 3.3 of WRF and later allows adding
more levels in PBL without it going unstable
• Inspired by overmixing last December and
January.
• Tried an extra 10 levels below 200 m and
1000m.
• Tried a few December 2011 dates and several
in January 2012.
Conclusion
• We can do this now without issues.
• Hyperresolution helps a small amount in
getting very shallow inversions.
• Keeps a bit more moisture at low
levels…better low level fog
• No big fix and is probably not worth slowing
down the system.
New Surface Drag Parameterization
• For over a year we have had in place a new
surface drag parameterization for the 36, 12
and 4 km domains, in which zo depends on
the variance of subgrid scale terrain.
• Substantial overall improvement in surface
winds and temperature.
• Even our past critics at the NWS were
satisfied!
But we were still not satisfied…
• Although we profoundly helped in stable
conditions, we hurt when wind speed were
strong (lots of mechanical mixing) or during
the day during summer (lots of convective
mixing).
• This made sense since local terrain or
roughness features are less important when
there is a large amount of mixing of higher
momentum down from aloft.
Solution (or at least a partial one)
• Fade out the sub-grid scale drag when the
vertical lapse rate is large or when vertical
wind shear is large. Or just one of them.
• We have tried all the combinations.
Bottom Line
• During the winter, all the approaches are
similar at low wind speeds, but current is
more harmful at high winds (> 20 knots)
• In summer during day, new approach is better
during the day at all wind speeds, but similar
at night.
An Issue to Work On: Shallow
Cumulus Convection
• We appears to underplay the development and
precipitation from shallow convection.
• Classic: destabilization over the Cascades, which
sometimes drifts over the lowlands.
• Problem is seen in all domains.
• Origin: lack of shallow convective
parameterization? PBL problems? Convective
parameterizations on outer domains suppressing
in 1.3 km? Try Grell or other convective
schemes?
Adding Requested Graphics
Trajectories: Close to Being
Ready…next week
Low Level Stability Plots
Something to Look at Over the
Summer
Other tasks for the next several
months
• Evaluate new release of WRF 3.4, which has a
number of potentially useful additions and
fixes (we are using 3.1.1-July 2009 version), .
Some examples:
– NOAH MP Land Surface Model: lots of
enhancements including multilevel snowpack
– A number of fixes and enhancements to PBL
schemes…need to test them again. Also new
schemes in 3.2 and 3.3, many we had tested
already.
EnKF
• Greg will talk about recent advances and UW
research.
• Dave Ovens has taken over the real-time EnKF
and stabilized it.
• Next steps:
– Improve web page (Mark Albright and Dave Ovens
working on this.
– move research improvements (e.g., pressure
assimilation) over to real time system
– Hourly data assimilation and adding more
observation types (e.g., radar)
Testing the analog/reforecast
approach
The end