Upload
chester-robertson
View
217
Download
0
Tags:
Embed Size (px)
Citation preview
INTELLIGENT USE OF THE AVN/MRF 2000
Wes Junker
HYDROMETEOROLOGICAL PREDICTION CENTER
CAMP SPRINGS, MD
E-MAIL ADDRESS: [email protected]
Presented COMAP Symposim 00-2Wednesday, 29 March 2000
Understanding the performance of an operational model is critical to being able to forecast the sensible weather
All models have strengths and weaknesses.All models have strengths and weaknesses. All have trouble handling smaller scale features. All have trouble handling smaller scale features. All have problems with convection.All have problems with convection. All do a decent job in handling the short range All do a decent job in handling the short range
(0-36 hr) forecast of synoptic scale features. (0-36 hr) forecast of synoptic scale features.
Beware of the initial analysis errors. Beware of the initial analysis errors.
MRF 1ST GUESS VERSUS OBSERVATIONS
MRF INITIAL ANALYSIS VERSUS MRF INITIAL ANALYSIS VERSUS OBSERVATIONSOBSERVATIONS
The T170 version of the Avn does a better job depicting in the West than the T126 but the model still misses much of the detail in the West. NOTE THAT NEITHER THE AVN OR ETA PREDICT A MAX IN THE SISKIYOU MOUNTAINS OF NORTHERN CA.
12-36 HR AVN QPF 12-36 HR AVN QPF V.T. 22 FEB 2000V.T. 22 FEB 2000
12-36 HR ETA QPF 12-36 HR ETA QPF V.T. 22 FEB 2000V.T. 22 FEB 2000
VERIFYING VERIFYING ANALYSIS OF QPF ANALYSIS OF QPF V.T. 22 FEB 2000V.T. 22 FEB 2000
NEED TO KNOW THE TERRAINNEED TO KNOW THE TERRAIN
AVN/MRF APPROXIMATED PHYSICS
THE AVN/MRF USE A MODIFIED GRELL SCHEME THE AVN/MRF USE A MODIFIED GRELL SCHEME THIS USES THE CHANGE IN STABILITY TO THIS USES THE CHANGE IN STABILITY TO
DETERMINTE WHEN TO RELEASE ENERGY AS DETERMINTE WHEN TO RELEASE ENERGY AS CONVECTION.CONVECTION.
NO DIRECT MIXING BETWEEN THE CLOUDY AIR NO DIRECT MIXING BETWEEN THE CLOUDY AIR AND ENVIRONMENTAL AIR. AND ENVIRONMENTAL AIR. (except at the cloud top and bottom) (except at the cloud top and bottom)
NO CLOUD WATER EXISTS, THEREFORE ALL NO CLOUD WATER EXISTS, THEREFORE ALL WATER IS CONVERTED TO RAIN. WATER IS CONVERTED TO RAIN.
A NUMBER OF AVN/MRF PERFORMANCE CHARACTERISTICS
HAVE CHANGED IN THE PAST YEAR.
THE AVN IS NOW AGAIN A T170 VERSION (SINCE THE AVN IS NOW AGAIN A T170 VERSION (SINCE JAN 28) AND SOMETIMES EXHIBITS ITS OLD JAN 28) AND SOMETIMES EXHIBITS ITS OLD CONVECTIVE FEEDBACK PROBLEM. ITS BIAS CONVECTIVE FEEDBACK PROBLEM. ITS BIAS HAS ALSO GONE UP. . HAS ALSO GONE UP. .
ISOLATED PRECIPITATION “BULLSEYES” HAVE ISOLATED PRECIPITATION “BULLSEYES” HAVE BEEN A PROBLEM DURING MARCH.BEEN A PROBLEM DURING MARCH. THIS IS MORE LIKELY WHEN A SYSTEM HAS PLENTY THIS IS MORE LIKELY WHEN A SYSTEM HAS PLENTY
OF MOISTURE WITH IT. OF MOISTURE WITH IT. HEAVY PRECIPITATION OFTEN DOES VERIFY BUT NOT HEAVY PRECIPITATION OFTEN DOES VERIFY BUT NOT
WHERE THE MODEL PREDICTS IT. WHERE THE MODEL PREDICTS IT.
AVN/MRF Still Has Problems Handling Upslope Events. It generally overforecasts
precipitationAround 75% of the precipitation predicted by the AVN during this event was grid scale, rather than convective, precipitation. In these cases, the model QPF is often too far to the northwest. The maximum rainfall falls farther to the south along the surface front.
12-36 hr AVN QPF V.T. 12Z 27 APR 98 VERIFYING 24H PRECIPITATION V.T. 12Z 27 APR 98
4”3”
4”3”
5”
About 75% of the AVN Rainfall Over the OK Panhandle Was Grid-scale Precipitation (Not Convection).
36-HR AVN/MRF V.T. 12Z 27 APRIL 98 VERIFYING AVN/MRF
V.T. 12Z 27 APRIL 98
The overprediction of grid-scale precipitation may result in latent heat being released at too low a level in the atmosphere. This tends to cause pressures to lower, often resulting in the lows wrapping up too far to the west or northwest.
In the past the MRF has spun-up precipitation bombs and tropical systems erroneously at all time ranges. Improvements have been made that make such marked error less likely. However, a few spin-ups will probably occur.
10”+ bullseye
36-h MRFX v.t. 00Z 28 May 199824-h MRFX v.t. 12Z 27 May 1998
24-36-h MRFX v.t. 00Z 28 May 1998
SFC ANALYSIS v.t. 00Z 28 May 1998
When you see a precipitation bulleye. How can you tell when latent heat feedback may be causing problems
Look at the convective and total Look at the convective and total precipitationprecipitation
If there is little or no If there is little or no convective precipitation, convective precipitation, beware especially if the mass beware especially if the mass fields start looking funnyfields start looking funny
More reasons to suspect a feedback
Round bulleye in vertical motion field and Round bulleye in vertical motion field and pimple that forms at 700 mb. Sometimes will pimple that forms at 700 mb. Sometimes will affect surface forecastsaffect surface forecasts
Produced kink in isobarsProduced kink in isobars
What happened? A 3 inch max near Shreveport, LA.
Verifying analysesVerifying analyses
Note that the kink in the isobars has disappeared. Such a feature can Note that the kink in the isobars has disappeared. Such a feature can be induced by latent heat. The induced trough may be subtle but can be induced by latent heat. The induced trough may be subtle but can enhance the low level convergence and help lead to a erroneous enhance the low level convergence and help lead to a erroneous forecast. forecast.
A test. From the following charts try to decide where
you would predict the precipitation max
36hr 700 height and vertical velocity 36hr 700 height and vertical velocity forecast v.t. 12Z 23 Mar 2000. forecast v.t. 12Z 23 Mar 2000.
36hr 500 height and vorticity forecast 36hr 500 height and vorticity forecast v.t. 12Z 23 Mar 2000. v.t. 12Z 23 Mar 2000.
Do you think, the 3 inch area is correct in NM,
Why or Why not? Would you modify the axis of
heavy rain in TX. IF so how?
36hr surface and thickness forecast v.t. 36hr surface and thickness forecast v.t. 12Z 23 Mar 2000. 12Z 23 Mar 2000.
12-36hr AVN/MRF QPF v.t. 12Z 23 12-36hr AVN/MRF QPF v.t. 12Z 23 Mar 2000. Mar 2000.
How did you do? The axis was east of the model’s forecast. Note that the surface low was weaker. Did the possible feedback cause the lower pressure? Doen’t his case look like the one in 1998?
Axis of heaviest rains outlined Axis of heaviest rains outlined in redin red
Verifying surface analysis v.t. Verifying surface analysis v.t. 12Z 13 March 200012Z 13 March 2000
Despite the low bias for heavy precipitation during most of the winter. During March
The AVN twice forecast 5 inch The AVN twice forecast 5 inch precipitation maxes that were not observed. precipitation maxes that were not observed.
During each case almost all the During each case almost all the precipitation was grid scale not convective.precipitation was grid scale not convective.
When this occurs other fields, especially the When this occurs other fields, especially the 700h height field will be affected. 700h height field will be affected.
AVN PRECIPITATION CHARACTERISTICS. AVN has a large bias for the lighter AVN has a large bias for the lighter
precipitation thresholds.precipitation thresholds. Its bias grows with time. Its bias grows with time.
During Feb. 2000, the AVN had a During Feb. 2000, the AVN had a low bias for heavy amounts. In low bias for heavy amounts. In March the model has had a high bias March the model has had a high bias for amounts up to 2.00”for amounts up to 2.00”
Avn threat score and bias during Feb. 2000
12-36 hr forecast12-36 hr forecast bias for .50”=1.28, for 1.00”=1.06, for 2.00”=.36bias for .50”=1.28, for 1.00”=1.06, for 2.00”=.36 T.S. for .50”=.348, for 1.00”=.233, for 2.00”=.052T.S. for .50”=.348, for 1.00”=.233, for 2.00”=.052
36-60 hr36-60 hr bias for .50”=1.5, for 1.00”=1.3, for 2.00”=.60bias for .50”=1.5, for 1.00”=1.3, for 2.00”=.60 T.S. for .50”=.28, for 1.00”=.176, for 2.00”=.04T.S. for .50”=.28, for 1.00”=.176, for 2.00”=.04
60-84 hr60-84 hr bias for .50=1.54, for 1.00”=1.3, for 2.00”=1.07bias for .50=1.54, for 1.00”=1.3, for 2.00”=1.07 T.S. for .50=.275, for 1.00”=.178, for 2.00=.054T.S. for .50=.275, for 1.00”=.178, for 2.00=.054
AVN Bias for various thresholds for the 30 days ending 25 Mar 2000
Note the AVN threat scores are better usually better than the eta during winter and worse in summer
The AVN/MRF had a warm bias at mid levels over the Rocky Mountains and Plains (2 or 3oC at 700 mb). If the eta had lower 1000-500 thickness over Plains in summer but had
similar 500h height, it was usually right.
36 hr surface and thickness v.t. 12Z 17 36 hr surface and thickness v.t. 12Z 17 July 1999, <576 (shaded) July 1999, <576 (shaded)
Verifying surface and thickness analysis Verifying surface and thickness analysis v.t. 12Z 17 July 1999, <576 (shaded) v.t. 12Z 17 July 1999, <576 (shaded)
Note how dry AVN/MRF was over the Plains. Also note that it predicted too much rain across the South. This was fairly typical of its summer forecasts.
24 hr precipitation analysis v.t. 12Z 24 hr precipitation analysis v.t. 12Z 17 July 1999 17 July 1999
12-36 hr AVN QPF v.t. 12Z 17 July 12-36 hr AVN QPF v.t. 12Z 17 July 1999 1999
For .50” the avn threat scores were lower than the eta during Jan and Feb. Possibly because the AVN overpredicted the threshold.
One of the typical errors of the AVN is to predict convective systems too far north
12-36 HR QPF V.T. 1200Z 29 JAN 1999
12-36 HR QPF V.T. 1200Z 30 JAN 1999
ANALYSIS V.T. 1200Z 29 JAN 1999
ANALYSIS V.T. 1200Z 30 JAN 1999
The AVN/MRF may be right on the synoptic scale
features but cannot handle outflow boundaries, etc. 36 HR QPF V.T. 1200Z 29 JAN 1999
36 HR QPF V.T. 1200Z 29 JAN 1999
ANALYSIS V.T. 1200Z 29 JAN 1999
A SLOW MOVING 500 CLOSE LOW AND QUASISTATIONARY FRONT WERE WELL FORECAST BY THE MODEL. HOWEVER, THE EFFECTIVE BOUNDARY REMAINED
SOUTH OF THE MODEL FORECAST
LOWS TO THE LEE OF THE ROCKIES
THE AVN AND NGM USUALLY PREDICT THE AVN AND NGM USUALLY PREDICT THEM TO FORM TOO FAR NORTH. USE THE THEM TO FORM TOO FAR NORTH. USE THE 300 MB UPPER LEVEL JET. THE SURFACE 300 MB UPPER LEVEL JET. THE SURFACE
LOW IS USUALLY FOUND BENEATH THE LOW IS USUALLY FOUND BENEATH THE LEFT EXIT REGION OF THE JETLEFT EXIT REGION OF THE JET
IN ZONAL FLOW
The AVN broke continuity on this forecast being much faster than previous runs or models from other centers
When systems are digging into the west with no kicker evident upstream, it is usually smart to follow the lead of the slowest model. In this case the AVN was much to fast with the upper low in the west.
ANOTHER COMMON ERRORANOTHER COMMON ERROR
If the 500 forecast is poor, the surface forecast will also be corrupted. Note that surface low in the
plains is too deep and far north on the forecast.
When the avn is in error to the lee of the mountains. This is the typical error.
HPC IS NOW TRYING VERIFY 500 HEIGHT ERRORS BASED ON REGIME, THIS IS A VERFICATION OF 72 HR FORECASTS
ANOTHER PATTERN. NOTE SIMILARITIES IN WHERE THE BIGGEST ERRORS OCCUR
THE NGM AND AVN/MRF HAVE SERIOUS PROBLEMS WITH ARCTIC AIRMASSES
36 HR NGM V.T. 00Z APR 09, 1995
36 HR AVN V.T. 00Z APR 09, 1995
AVN ANALYSIS V.T. 00Z APR 09, 1995
L
TEMPERATURES ACROSS KANSAS WERE IN THE LOW TO MID 50s WITH STRONG NORTH WINDS. SOUTH OF THE FRONT TEMPERATURES WERE IN THE UPPER 70s TO LOW 90s.
Why models have problems with arctic airmasses Terrain is averagedTerrain is averaged Initialization process sometimes robs Initialization process sometimes robs
shallow airmass of its coldnessshallow airmass of its coldness Models have problems handling the strength Models have problems handling the strength
of the inversionof the inversion The leading edge of the ETA LI gradient is The leading edge of the ETA LI gradient is
often the best indicator of the frontal positionoften the best indicator of the frontal position
MRF PERFORMANCE FOR 3-5 DAY FORECASTS SHALLOW COLD AIR IS NOT HANDLED WELL. SHALLOW COLD AIR IS NOT HANDLED WELL.
THE MODEL IS SLOW TO TRANSPORT SHALLOW THE MODEL IS SLOW TO TRANSPORT SHALLOW COLD AIRMASSES, ESPECIALLY ARCTIC COLD AIRMASSES, ESPECIALLY ARCTIC AIRMASSES JUST TO THE EAST OF THE ROCKY AIRMASSES JUST TO THE EAST OF THE ROCKY MOUNTAINS OR APPALACHIAN CHAIN. THIS IS MOUNTAINS OR APPALACHIAN CHAIN. THIS IS DUE TO MODEL TERRAIN ERRORS.DUE TO MODEL TERRAIN ERRORS.
EASTERLY BOUNDARY LAYER WINDS ARE EASTERLY BOUNDARY LAYER WINDS ARE OFTEN OVERPREDICTED ALONG THE FRONT OFTEN OVERPREDICTED ALONG THE FRONT RANGE OF THE ROCKY MOUNTAINS.RANGE OF THE ROCKY MOUNTAINS.
MRF PERFORMACE FOR 3-5 DAY FORECASTS (CONT) MODEL TENDS TO PHASE SEPARATE MODEL TENDS TO PHASE SEPARATE
STREAMS TOO MUCH. STREAMS TOO MUCH. AT HIGH LATITUTES (NORTH OF 50AT HIGH LATITUTES (NORTH OF 50OO), THE ), THE
MODEL PREDICTS TOO MUCH MODEL PREDICTS TOO MUCH RETROGRESSIONRETROGRESSION
TENDS TO WEAKEN THE REMAINS OF TENDS TO WEAKEN THE REMAINS OF UPPER LOWS TOO QUICKLY THAT ARE UPPER LOWS TOO QUICKLY THAT ARE COMING OUT OF THE SOUTHWESTCOMING OUT OF THE SOUTHWEST
A VERIFICATION OF THE 500 H FORECASTS FOR DAY 4
-45 -35 -25 -15 -05-40 -30 -20 -1030
80
90
40
50
60
70
DAY (VALID TIME RELATIVE TO TODAY)
AN
OM
AL
Y C
OR
RE
LA
TIO
N
MEAN ECMWF 70.4ECMWF
MRF
MEAN MRF 70.5
UKMET
MEAN UKMET 62.7
DEPICTS MEDIUM RANGE MODEL PERFORMANCE FOR THE PAST 50 DAYS WORTH OF RUNS. NOTE THAT HOW WELL A MODEL PERFORMS APPEARS TO BE REGIME DEPENDENT.
HIGH DAILY VARIABILITY IS WEATHER PATTERN DEPENDENT. THE
ECMWF IS BEST AT 500 MB BUT CAN HAVE SOME BAD MSLP FORECASTS
HPC/MRF/ECMWF MSLP ANOMALY CORRELATIONS HPC/MRF/ECMWF MSLP ANOMALY CORRELATIONS
MEAN SCORES
You need to know the characteristics of the MRF
MOS guidance. Stations included in MOS
How HPC breaks up regional verification of MOS
MOS verification for Northwest (left) and northern Plains (right)
MOS POPS ARE USUALLY TOO HIGH DURING WINTER
Verification of MOS POPS for Great Lakes Region (top) and Northeast (bottom)
Southwest (left), Southern Plains (right), Southeast (bottom left) and Mid Atantic/Oh Valley (lower right)
AVN performance characteristics changed when T170 was implemented
The model now has a high bias for most thresholds at least The model now has a high bias for most thresholds at least during Spring. during Spring.
The model suffers from latent heat feedbackThe model suffers from latent heat feedback During Summer 1999, the AVN had a warm bias at 700 mb During Summer 1999, the AVN had a warm bias at 700 mb
and a low precipitation bias for .50” greater amounts over and a low precipitation bias for .50” greater amounts over the Plains and Intermountain region. Suspect the problem is the Plains and Intermountain region. Suspect the problem is still there but won’t know for sure until Summer.still there but won’t know for sure until Summer.
Whenever a model is changed to correct one problem it Whenever a model is changed to correct one problem it often changes the model performance characteristics in often changes the model performance characteristics in other ways. other ways.