Kein Folientitel

1
NOAA12, 03 June 2005, 1533 UTC (University of
Bern)
Predicting severe weather by EPS tools - current
results
Thomas Schumann, Deutscher Wetterdienst, Zentrale
Vorhersage D-63067 Offenbach, Germany E-Mail
Thomas.Schumann_at_dwd.de
2
Outline 1. Introduction - current
situation 2. EPS products used for severe
weather prediction 3. Case studies 4.
Preliminary verification results 5. Conclusions
3
1. Introduction - current situation
Forecaster a great variety of products
ECMWF EPS and derived products www.ecmwf.int
4
PEPS Udos model market (DWD Intranet, in future
included into NinJo)
Decision, which EPS tool will be used, depends
from

• Purpose of my forecast (overview,

• detailled view, ... severe weather)
• lead time
• expected scale of the event

SRNWP PEPS
Plots for global models
Available parameters Z500 T850 MSLP

• ECMWF
• GME
• GFS
• UKMO
• LFPW
• CMC (still not used)

COSMO-LEPS
5
Problems (or advantages?)
Forecaster has to keep in mind

• Clustering always provides a compromise.
  Different clustering
• methods could lead to different results.

• The EPS mean / the best populated cluster / the
  majority of global
• models not always shows the szenario that
  finally will happen.

• No model is perfect, models more or less
  inconsistent (jumping,
• caused from changes in initial and boundary
  conditions).

• selection of available models allows to create
  a EPS as well
• from global models as from LAMs (different
  model physics,
• parametrisation scheme and resolution, ...)

---gt SRNWP EPS
6
2. EPS Products used for severe weather
prediction
7
3. Case studies
A) Late frost - 20/21/22 April 2005 B) Heavy
precipitation - 14 May 2005 C) Hot day .... 03
June 2005 D) Thunderstorm - Squall line 03/04
June 2005

• How did COSMO-LEPS and SRNWP-PEPS perform
• against observations ?
• against probabilities of the pure
  (uncalibrated) ECMWF EPS?

SRNWP-PEPS (PEPS) http//www.dwd.de/en/FundE/Proj
ekte/ PEPS/index.htm (forecasts
password-protected)
8
A) Late frost - 20/21/22 April 2005
Tmin Observations, 20 April, 06 UTC (NE-Part of
Germany)
MSLP analysis, 20 April, 06 UTC
9
COSMO-LEPS, 17 April 05, 12 42 ... 66 H
COSMO-LEPS, 18 April 05, 12 18 ... 42 H
PEPS forecast, 19 Apr 05, 12 18 ... 30
H (EPS-mean)
SRNWP-PEPS Temp below 3C in the N-part of
Germany likely, frost 2 m above sfc not !
COSMO-LEPS Frost in the NE-part of Germany
likely !
10
ECMWF EPS probabilities for Tmin gt 0 C 19 April,
00 24 H
Frost in the NE-part of Germany likely !
ECMWF EPS probabilities for Tmin gt 0 C 18 April,
12 36 H
11
Two days later... The climax of the cold outbreak
Tmin Observations, 22 April, 06 UTC
MSLP analysis, 22 April, 06 UTC
12
COSMO-LEPS, 17 April 05, 12 90 ... 114 H
ECMWF EPS, 17 April 05, 12 96 ... 120 H
COSMO-LEPS, 18 April 05, 12 66 ... 90 H
18 April 05, 12 72 ... 96 H
Probabilities Tmin lt 0 C 22 April, 06 UTC
Event well predicted by COSMO-LEPS as well as by
the ECMWF EPS even in the early medium-range !
13
B) Heavy precipitation - 14 May 2005
Precip (obs, 24 hr-accumulated), 15 May 2005, 06
UTC
Precip locally above 50 mm / 24 h in the W-part
above 30 mm
MSLP analysis, 14 May, 18 UTC
14
PEPS, 14 May, 00 06 ... 30 H 24 hr-acc
precip probab gt 50 mm
PEPS, 14 May, 00 06 ... 30 H 24 hr-acc
precip probab gt 20 mm
PEPS, 14 May, 00 06 ... 30 H EPS mean
15
COSMO-LEPS forecasts probabs gt 50 mm / 24 h
(top) and gt 20 mm / 24 h (bottom) 11 May 12 66
... 90 H 12 May 12 42 ... 66 H
13 May 12 18 ... 42 H
16
12 May 00 60 ... 84 H
11 May 12 72 ... 96 H
12 May 12 48 ... 72 H
13 May 00 36 ... 60 H
13 May 12 24 ... 48 H
14 May 00 12 ... 36 H
ECMWF-EPS forecasts 15 May 12 UTC (24 hr accum
precip probabs gt 20 mm / 24 h Signal became
weaker with decreasing lead time and approaching
of the event
17
C) Hot day .... 03 June 2005
Tmax (Obs), 03 June, 18 UTC Few stations in
SW-Germany gt 30C !
MSLP analysis, 03 June, 15 UTC
18
PEPSmean, 02 June, 12 06 ... 30 H 03
June, 00 06 ... 30 H Probabilities not
available !
PEPS forecasts based on 02 June, 12 UTC and 03
June, 00 UTC didnt show Tmax above 30 C over SW
Germany (EPSmean) !
19
COSMO-LEPS probabilities T gt 30 C 30 May 12 90
... 114 H 31 May 12 66 ... 90 H
01 June 12 42 ... 66 H
02 June 12 18 ... 42 H
Tmax above 30 C over SW-Germany very likely
! Persistent signal
EPS (ECMWF) without of any signal !
20
ECMWF EPS probabs Tx gt 30 C (03 June, 18 UTC)
ECMWF EPS probabs Tx gt 25 C (03 June, 18 UTC)
Tmax underestimated caused from the lower
resolution of the EPS
02 June, 00 36 ... 42 H
02 June, 12 24 ... 30 H
21
D) Thunderstorm - Squall line 03/04 June 2005
Observed gusts (m/s), 03 June, 18 UTC
04 June, 00 UTC
22
Hail in Northern Germany (picture by Matthias
Jaenicke)
Damages by the storm in the forests near
Offenbach (picture by Klaus Paetzold)
23
prob fx gt 20 m/s 03 June, 00 18 ... 30 H
PEPS, prob fx gt 20 m/s 03 June, 00 06 ... 18 H
PEPS, prob fx gt 20 m/s, 03 June, 12 06 ... 18 H
Most severe gusts over NW and N-Part of Germany ?
24
COSMO-LEPS fx gt 20 m/s 30 May 12 96 ... 120
H 31 May 12 72 ... 96 H
01 June 12 48 ... 72 H
02 June 12 24 ... 48 H
Weak signals for gusts over the W- and S-part
of Germany Signal over N-Germany for lead time gt
48 h only
25
Signal over N-part of Germany not consistent
01 June, 00 66 ... 72 H
01 June, 12 54 ... 60 H
Increased probs later... No indications
over the central part of Germany !
02 June, 00 42 ... 48 H
02 June, 12 30 ... 36 H
ECMWF EPS probabilities for gusts gt 20 m/s (03
June, 18 ... 24 UTC)
26
4. Preliminary verification results

• COSMO-LEPS
• still subjective verification, carried out by
  the medium-range
• shift meteorologist
• Tables Verification against observations

• Main results
• Tmin, Tmax useful, able to add value to
  forecasts (improved)
• Wind gusts good, orographic effects
  overestimated
• Conv wind gusts signals mostly too weak
  (Improvement ?)
• Large-scale precip good, orographic effects
  overestimated
• Convective precip not useful
• CAPE still under evaluation
• Snow good, orographic effects overestimated

27

• PEPS
• operational experimental suite since beginning
  of this year
• Forecaster collecting experience
• Set of forecasts (weather parameter, leading
  time) will be increased

Parameter 24h-accumutated precipitation (00
UTC 06...30h) Observations Max. 217 synoptical
stationens of the DWD (06...06 UTC) N
Sample size POD Probability of Detection (hit
rate). Perfect score 1 FAR False Alarm
Ratio. Perfect score 0 HSS Heidke Skill
Score. Perfect score 1 TS Threat
Score. Perfect score 1
HSS100(ad-R)/(abcd-R)
. Obs yes Obs no fc yes a b fx no
c d
R((ab)(ac)(cd)(bd))/(abcd)
TS100a/(abc)
28
(No Transcript)
29
(No Transcript)
30
5. Conclusions

• Forecaster hat to deal with and to check a
  great collection of
• products in a limited time frame before making
  a decision

• EPS-products more and more accepted by the
  forecaster
• - content has to be condensed and
  compressed

- products valid for different temporal and
spatial scales
- tailored products predicting severe weather
in the meso-scale -----gt COSMO - LEPS,
SRNWP PEPS

• preliminary verification results of COSMO -
  LEPS and SRNWP PEPS
• encouraging (daily use by the forecaster, case
  studies)

- COSMO - LEPS quasi-operational use,
improvements to be seen - SRNWP PEPS
experimental, first results promisingly
Problem presenting EPS forecast customer-friendly
not been solved
31
Thats it ! Thank you for your attention!
Headquarter of the DWD in Offenbach
Central Forecasting