An Examination Of Interesting Properties Regarding A Physics Ensemble 2012 WRF Users’ Workshop...
-
Upload
opal-booth -
Category
Documents
-
view
216 -
download
2
description
Transcript of An Examination Of Interesting Properties Regarding A Physics Ensemble 2012 WRF Users’ Workshop...
An Examination Of Interesting An Examination Of Interesting Properties Regarding A Physics Properties Regarding A Physics
EnsembleEnsemble2012 WRF Users’
Workshop
Nick P. BassillJune 28th, 2012
IntroductionIntroduction• During the 2009 North Atlantic During the 2009 North Atlantic
hurricane season, a real-time ensemble hurricane season, a real-time ensemble was created locally once per daywas created locally once per day
• Using simple linear regression Using simple linear regression techniques, I will present a few techniques, I will present a few (potentially) interesting results (potentially) interesting results comparing a low resolution WRF-ARW comparing a low resolution WRF-ARW physics ensemble with the operation physics ensemble with the operation GFS ensembleGFS ensemble
Data Generation OverviewData Generation Overview
Initialization Time
Two Days Between Each Initialization (From GFS 00Z Forecast)
Spread Of 120 Hour Forecasts
76 Cases From Early
June Through October
Dynamical Core is WRF-ARW 3.0
Outer Domain: 90 km Grid Spacing
Inner Domain: 30 km Grid Spacing
Physics Ensemble vs. GFS Physics Ensemble vs. GFS EnsembleEnsemble
• The results shown were calculated as follows:The results shown were calculated as follows:- Tune 120 hour forecasts to 0 hour GFS - Tune 120 hour forecasts to 0 hour GFS analyses for both the physics ensemble* and analyses for both the physics ensemble* and GFS ensemble comprised of an equal number GFS ensemble comprised of an equal number of membersof members
- After this is done, it’s easy to calculate the - After this is done, it’s easy to calculate the average error per case for both ensemblesaverage error per case for both ensembles- The average error will be shown, along with - The average error will be shown, along with the difference between the two normalized the difference between the two normalized to the standard deviation of the variable in to the standard deviation of the variable in questionquestion
* Using outer 90 km grid* Using outer 90 km grid
• (Left): Yellow and red (Left): Yellow and red colors mean the colors mean the parameterization parameterization ensemble ensemble outperformed the GFS outperformed the GFS ensemble, blue colors ensemble, blue colors mean the oppositemean the opposite
2 Meter Temperature (2 Meter Temperature (°°C)C)
500 hPa Geopotential Height 500 hPa Geopotential Height (m)(m)
• (Right): Yellow and (Right): Yellow and red colors mean the red colors mean the parameterization parameterization ensemble ensemble outperformed the outperformed the GFS ensemble, blue GFS ensemble, blue colors mean the colors mean the oppositeopposite
• (Left): Yellow and red (Left): Yellow and red colors mean the colors mean the parameterization parameterization ensemble ensemble outperformed the GFS outperformed the GFS ensemble, blue colors ensemble, blue colors mean the oppositemean the opposite
10 Meter Wind Speed (10 Meter Wind Speed (m/sm/s))
ObservationsObservations• On average, the GFS ensemble “wins” by On average, the GFS ensemble “wins” by
~0.01 standard deviations for any given ~0.01 standard deviations for any given variablevariable
• Generally speaking, the parameterization Generally speaking, the parameterization ensemble performs better in the tropics, ensemble performs better in the tropics, while the GFS ensemble performs better in while the GFS ensemble performs better in the sub/extratropicsthe sub/extratropics
Let’s examine the 2 m temperature more Let’s examine the 2 m temperature more closely …closely …
• The fill is the The fill is the mean 2 m mean 2 m temperature at temperature at hour 120 for hour 120 for the the parameterizatiparameterization ensemble on ensemble membersmembers
• The contour is The contour is the 95% the 95% significance significance thresholdthreshold
2 Meter 2 Meter Temperature Temperature CompositesComposites
• The fill is the The fill is the mean 2 m mean 2 m temperature temperature anomaly at hour anomaly at hour 120 for the 120 for the parameterization parameterization ensemble ensemble membersmembers
• The contour is The contour is the 95% the 95% significance significance threshold as seen threshold as seen previouslypreviously
2 Meter 2 Meter Temperature Temperature CompositesComposites
• The fill is the The fill is the mean 2 m mean 2 m temperature temperature anomaly at hour anomaly at hour 120 for the 120 for the parameterization parameterization ensemble ensemble membersmembers
• The contour is The contour is the 95% the 95% significance significance threshold as seen threshold as seen previouslypreviously
2 Meter 2 Meter Temperature Temperature CompositesComposites
• The fill is the The fill is the mean 2 m mean 2 m temperature at temperature at hour 120 for hour 120 for the GFS the GFS ensemble ensemble membersmembers
• Note – no Note – no significance significance contour is contour is shownshown
2 Meter 2 Meter Temperature Temperature CompositesComposites
• The fill is the The fill is the mean 2 m mean 2 m temperature temperature anomaly at hour anomaly at hour 120 for the GFS 120 for the GFS ensemble ensemble membersmembers
• Note how Note how indistinguishable indistinguishable one member is one member is from anotherfrom another
• Unlike the Unlike the physics physics ensemble, ensemble, member member differences aren’t differences aren’t correlated from correlated from run to runrun to run
2 Meter 2 Meter Temperature Temperature CompositesComposites
Parameterization Ensemble Member Parameterization Ensemble Member ErrorsErrors
Green = MYJ PBL Members
Black = YSU PBL Members
At ~44 N, -108 W
GFS Ensemble Member ErrorsGFS Ensemble Member Errors
ConclusionsConclusions• Different parameterization combinations do Different parameterization combinations do
have certain (statistically significant) biaseshave certain (statistically significant) biases• Pure parameterization ensembles theoretically Pure parameterization ensembles theoretically
are better than pure initial condition are better than pure initial condition ensembles, since member differences are ensembles, since member differences are correlated from run to runcorrelated from run to run
• Using this information, a simple (read: dumb) Using this information, a simple (read: dumb) parameterization ensemble can perform parameterization ensemble can perform equivalently to a “superior” ensembleequivalently to a “superior” ensemble
• This is done by viewing parameterization This is done by viewing parameterization biases as a biases as a benefitbenefit, not a problem, not a problem
Future WorkFuture Work• Currently, the parameterization ensemble Currently, the parameterization ensemble
vs. GFS ensemble results are being vs. GFS ensemble results are being redone with the use of Global WRFredone with the use of Global WRF
• More advanced statistical techniques More advanced statistical techniques could be used to improve on these resultscould be used to improve on these results- Unequal weighting- Unequal weighting- Using more predictors- Using more predictors- Identifying regimes- Identifying regimes
Kain-Fritsch Cumulus
Parameterization
Betts-Miller-Janjic Cumulus
Parameterization
Grell-3 Cumulus Parameterization
MYJ Boundary Layer
Parameterization
YSU Boundary Layer
Parameterization
WSM3 Microphysics
Parameterization
Ferrier Microphysics
Parameterization
• Generically speaking, all of these models are Generically speaking, all of these models are pretty similarpretty similar
• For each forecast day, I ran an additional For each forecast day, I ran an additional model with completely different model with completely different parameterizations, which were purposely parameterizations, which were purposely chosen to be “bad” (at least in combination)chosen to be “bad” (at least in combination)
• Afterward, I used the original ten models to Afterward, I used the original ten models to predict the new one (i.e. a prediction of a predict the new one (i.e. a prediction of a prediction) using multivariate linear regressionprediction) using multivariate linear regression
Idea: Predicting PredictionsIdea: Predicting Predictions
For Reference: New Composites ofFor Reference: New Composites of500 mb Height500 mb Height
Case 28: July 28th 2009
Case 38: August 17th 2009