Consortium102012final

advertisement
Consortium
October 4, 2012
Hits: Canadians Don’t Give Up
Who is scansafe.net?
What has not happened in a long
time?
Hardware Failures!
• Only a few disk failures…but those taken care
of by the RAID system.
• Enhanced back scripts as well.
• Other hardware issue: hoping to get a tentimes increase in bandwidth into dept server
by running optical all the way into computer
room (6-12 months out).
Ready for Some Major Changes
• Fixing the upper boundary of the model to
reduce reflections from the top, producing more
robust and less noisy forecasts—particularly in
the upper troposphere and lower stratosphere.
• Do this by:
– raising the model top from 100 hPa to 50 hPa.
– Adding Rayleigh damping to the upper seven layers.
– Much smoother transitions in model resolution.
http://www.atmos.washington.edu/~ovens/upp
erboundary/
Testing
• Have evaluated over extended period in
winter and summer: little impact in lower
troposphere on key fields.
• Does lessen excessive waves in upper
troposphere and even cleans up midtroposphere somewhat.
Old Vertical Levels without
damping
New
Old
New
Land Surface Model Testing
• We are NOT using the NOAH Land Surface
Model because it produced an unacceptable
cold bias during winter.
• A new version of the NOAH LSM has been
released: NOAH-MP (Multi-physics)
Noah-MP (Niu et al., 2011) (multi-physics)
• Uses multiple options for key land-atmosphere interaction processes.
• Noah-MP contains a separate vegetation canopy defined by a canopy top and
bottom with leaf physical and radiometric properties used in a two-stream
canopy radiation transfer scheme that includes shading effects.
• Noah-MP contains a multi-layer snow pack with liquid water storage and
melt/refreeze capability and a snow-interception model describing
loading/unloading, melt/refreeze, and sublimation of the canopy-intercepted
snow.
• Multiple options are available for surface water infiltration and runoff, and
groundwater transfer and storage including water table depth to an unconfined
aquifer.
• Horizontal and vertical vegetation density can be prescribed or predicted using
prognostic photosynthesis and dynamic vegetation models that allocate carbon
to vegetation (leaf, stem, wood and root) and soil carbon pools (fast and slow).
The Issue
• We found it doesn’t work in latest version of
WRF! (“Water budget problem in NOAHMP
LSM”)
• WRFhelp and others confirmed our findings.
• Hopefully, they will correct the issues this fall,
so we can test it and have the option of going
with it before winter snows.
• If this doesn’t work, we will experiment with
other LSMs (e.g., RUC LSM).
Transition to WRF 3.4.1
• We have been using a relatively old version of
WRF for a long time (WRF 3.1.1)—and wisely
so! (released July 2009).
• Our verification showed degradation of results
for 3.2, 3.3, and 3.4.
• Air Force Weather Agency Testing by the DTC
(Developmental Testbed Center) showed the
same thing.
• FINALLY, they found a key problem: an error in
the YSU PBL scheme that entered with 3.2
Early Summer Testing
• Before 3.4.1 was available, we tested a new
version of our drag parameterization (using
3.1.1) that pulled back the drag during well
mixed conditions. Looked good.
• But then 3.4.1 came out and we did
comprehensive tests on it.
51 different experiments:
http://www.atmos.washington.edu/~ovens/windbias/
Transition to 3.4.1
• We have completed extensive testing with
3.4.1 (which includes a number of bug fixes
and minor improvements as well).
• A change from 3.1.1 to 3.4.1 would produce a
minor, but perceptible improvement and we
recommend that the Consortium approve the
change.
• But there is an interesting wrinkle in all of
this…a surprise.
Comparisons of 3.1.1 with 3.4.1
WITHOUT our drag mods
10-m wsp: winter, 3.1.1
Wsp:winter, 3.4.1
2m temp, winter, 3.1.1
2-m temp winter 3.4.1
3.1.1
3.4.1
In Winter 3.4.1 is clear winner,
more mixed results in summer
2-m temp, 3.1.1 (12 z)
2-m temp, 3.41
2-m temp, 3.1.1, 00z
2m temp 3.4.1
Conclusions
• After looking at hundreds and hundreds of
comparisons, Dave and I believe that the
“vanilla, out of the box” 3.4.1 is superior to
3.1.1 and brings us absolutely up to date.
• But should we continue to use our current
drag parameteriztion?….there is a
complication here.
10 m wind– 3.1.1 (no drag)
3.1.1 with our drag mod
3.4.1 NO MODS
3.4.1 with drag mods
3.1.1 with drag mods
Big Question: Do we need to
recalibrate our drag mods? Might
we be slowing down the winds too
much?
Temperature a wash
Conclusions
• The drag generally helps…particularly for light
winds and terrain.
• Hurts for strong winds
• Probably should go with 3.4.1 with drag mods,
but continue experimentation of having it pull
back for well mixed situations.
December 2011
• Initial tests don’t reveal major impact of using
3.4.1 and fixed PBL scheme.
Download