the dtc winter forecast experiment (dwfe) dtc winter forecast experiment (dwfe): dave dempsey (san...

The DTCWinter Forecast Experiment (DWFE):

Dave Dempsey(San FranciscoState University;

DTC visitor)

William Skamarock(NCAR/MMM)

Analysis of Kinetic Energy Spectra

Acknowledgements:

• Ligia Bernardet (NOAA/FSL/DTC)

• Zavisa Janjic (NCEP)

• Bob Gall (Director, DTC/NCAR)

2006 DTC Reunion NCAR/DTC

• Kinetic energy spectra– background– backstory– accidental DWFE discoveries– ARW and NMM results– sensitivity study: NMM spectra and

model configurations

Outline:

• DTC Winter Field Experiment (DWFE)• Motivating observations

– mountain waves

• Conclusions and Further Work


• Two versions of WRF:– Two different dynamical cores

(NMM & ARW)– Two different physics packages

(NCEP & NCAR)

• 5 km resolution; 37 levels

• Initial and boundary conditions from Eta212 (40 km)

• 00Z initialization, daily 48-hour forecasts

• Explicit convection (no convective parameterization)

DTC Winter Forecast ExperimentJan 15 - March 31, 2005

ConUS Domain


My Context (a parable of howscience sometimes happens)

• Sabbatical leave (half pay), Jan-July ‘05• Invite self to visit DTC. Flexible agenda—will work for chance to be useful, learn something• Gall says “OK”, provides:

• office, computer, office supplies• no promises of anything else

• Office shared (1 day/wk) w/Steve Koch (NOAA)• Gall says, “Dempsey: DWFE forecasts.

See what you can see.”


Mountain WavesPotential Temperature Cross Sections

(12 hr forecasts valid 12Z Jan 13, 2005)ARW NMM


The Parable (cont’d)• I print copies of mountain wave plots.• NCAR is big place—lots of printers. Can’t find

the printer. Plots sit by printer 3 days.• It’s NCAR/MMM’s printer, near Bill Skamarock’s

office.• Skamarock notices plots. He sees possible

support for his (contentious) work on ARW andNMM KE spectra (work he’d given up).

• I finally find plots. Bill happens to be there. Bill revives interest in ARW/NMM spectra.

• He computes some.


• Kinetic energy spectra– background– backstory– accidental DWFE discoveries– ARW and NMM results– sensitivity study: NMM spectra and

model configurations

• DTC Winter Field Experiment (DWFE)

Coming Up Next:

• Motivating observations– mountain waves

• Conclusions and Further Work


• Nastrom and Gage (1985):Spectrum computed from GASP observations(commercial aircraft)

• Lindborg (1999):Functional fit to MOZAIC observations (aircraft)

Kinetic Energy Spectra

mesoscale


Backstory• Skamarock and Baldwin

(white paper, Nov. 2003)– NMM’s KE spectra show that NMM has too much

dissipation– ARW’s KE spectra are fine

• Janjic and Black (NCEP) respond:– Not true, on both counts– Don’t know what mesoscale model spectra should

look like, anyway

• Skamarock (MWR, 2004) refines white paper arguments; veils references to NMM.


The Parable (cont’d)

• Skamarock’s NMM and ARW KE spectra of DWFE forecasts (Jan. 2005) resemble previous spectra. Dispute over KE spectra threatens to revive.

• DTC idea: have Dempsey act as neutralinvestigator, recalculate spectra, try to get to bottom of Skamarock/Janjic differences.

• Hmmm ….


Questions• Do DWFE NMM & ARW model spectra resemble

the Nastrom & Gage (1985) observed spectrum?• Should they?• If they should, and if they do, do they for the right

(physical) reasons? And how can we tell?

• If not, why not?


Computing model KE spectra• Interpolate u, v to pressure levels• Detrend u, v along each 1-D transect• Compute discrete Fourier transform of kinetic

energy along each E/W (or N/S or diagonal) transect

• Average spectra over:– multiple transects, and– pressure layer (100 mb deep)


• NMM “E” grid: compute KE spectraon transects:– diagonally;– east/west; or– north/south

• ARW “C” grid: compute spectra:– east/west, or– north/south


Accidental DWFE Discoveries

1. Tested rewritten KE spectra code on NMM and ARW DWFE topography

ARW(DWFE)

NMM(DWFE)

Smallest-scaleterrainsmoothed

Noterrainsmoothing


Accidental DWFE Discoveries

2. Bug in WRF-NMM-SI initialization

• Spike in initial energyat L = 2Δx at all levels(diagonal spectra only)

• Gone by 3 hr forecast

Contour plot: Initial u-component(2Δx pattern due to array index error)


DWFE Case: 3/27/05500 mb Height & w

(24 hr fcst)MSL Pressure &

3-hr Precip (24 hr fcst)


Spin-up of Kinetic Energy Spectra

ARW NMM

• Spectra in 300-200 mb layer• 6 hour intervals, 00 to 48 hrs

mesoscale mesoscale

• Note: get mature spectrum by ~< 24 hours

initial initial


“Mature” Kinetic Energy Spectra• Average over 24-48 hours. Three tropospheric layers shown• Note:

– NMM: less mesoscale energy aloft– NMM: more energy on smallest scales, esp. at lower levels

NMMARW

Uppertroposphere

Uppertroposphere


Questions:1. Why did NMM spectrum have less energy at

mesoscales aloft than observations and ARW, hence lacking transition to k-5/3 slope?

2. Why did NMM spectrum have more energy than ARW at smallest scales at low altitudes?

Initial Testing Strategy:1. Rerun NMM with various bug fixes2. Rerun with smoothed topography3. Rerun without horizontal-divergence damping4. Rerun with more horizontal diffusion (2nd order, 4th order)

Hypotheses:• NMM has unsmoothed topography, very low horizontal

diffusion, and horizontal-divergence damping


Some results of NMM reruns:1. Fixing bugs: little effect on spectra

• Original DWFE run• 3 levels shown; 24-48 hr fcst avg

• Rerun #1: Bug fixes• 3 levels shown; 24-48 hr fcst avg


modest (x 0.5) decrease in energy at small scales

• Rerun #1: Unsmoothed terrain• 3 levels shown; 24-48 hr fcst avg

• Rerun #2: Smoothed terrain• 3 levels shown; 24-48 hr fcst avg

2. Smoothed terrain:

Some results of NMM reruns:

NMM,smoothed

terrainARW,

smoothedterrain

NMM, unsmoothed terrain


Some results of NMM reruns:3. No divergence damping:

• Rerun #3: No divergence damping• Forecasts: 00, 03, 06, 09, 12 hrs; • 300-200 mb layer

Oops!


External modes!Surface pressure change after 03 hrs(contour interval = 2 mb)

ConUSdomain


External modes:Vertical Velocity Cross Section

• 03 hr forecast


Revised Testing Strategy:

4. Rerun with no horizontal-divergence damping but add a vertically-integrated horizontal-divergence damper (to damp external modes)

5. Rerun with more horizontal diffusion (2nd order, then 4th order)


Results:

• 18 hr forecast• 3 tropospheric levels

• Rerun #3: No divergence damping

External mode removed

• Rerun #4: No divergence damping; external-mode damping added

• 18 hr forecast• 3 tropospheric levels


Conclusions• ARW in DWFE:

– Kinetic energy spectra similar to Nastrom & Gage (1985) observations at mid-mesoscales (~7Dx) and larger

– Implicit damping and weak explicit diffusion, plus smoothed topography, reduce small scale energy

– Arguably desirable (Skamarock and others); or maybe not (Janjic)• NMM in DWFE:

– NMM spectra showed less energy at mesoscales than Nastrom& Gage (1985) observations, especially aloft

• Can apparently change this by replacing horizontal-divergence damping with sufficient external-mode damping

– Spectra show more energy at smallest scales than ARW (but less than observed)

• smoothing topography reduces small-scale energy somewhat• can debate whether need more small-scale damping


Further Work (NMM)• Fine-tune external-mode damping

• Try 4th-order diffusion to reduce energy selectively on smallest scales– NCEP: Adapt 4th-order diffusion to parallel

computing (doesn’t work yet)

• Evaluate effect of these alternative filters on meteorological fields and traditional verification statistics– NOAA: Test against T-REX data (mountain

waves) and Rapid Refresh runs


24-Hour Precipitation BiasesDWFE: Jan 15-Mar 31 2005

WRF-NMM

WRF-ARW

Eta1.0

6.0


Grid-Point Storms? (NMM)

• 500 mbverticalvelocity

• 24 hr forecast

• bull’s eyepatterns


A Bull’s Eye Close-up:500 mb Vertical Velocity• Data on native “E”-grid

(no interpolation orsmoothing beforecontouring)

• Depth of verticalmotion: sfc to 300 mb(max near 300 mb)

Grid pts.


WRF-NMM 24h Forecast (for 00Z 2/15/05)


WRF-NMM, 13 Feb 2005• 00Z init, 24h forecast• vertical velocity (cint = 0.2 m/s)• model level 10 (~ 900 mb)

the dtc winter forecast experiment (dwfe) dtc winter forecast experiment (dwfe): dave dempsey (san...

Documents