Limited Area Ensembles using targeted singular vectors NORLAMEPS

1
Limited Area Ensembles using targeted singular
vectors - NORLAMEPS

• Inger-Lise Frogner, Marit H. Jensen, Hilde
  Haakenstad and Ole Vignes

2
NORLAMEPS

• An ensemble prediction system using IFS and
  HIRLAM
• Running daily since mid Feb. 2005 (quasi
  operational)
• Focus Precipitation (large precipitation events
  of special interest)
• NORLAMEPS is a supplement to EPS from ECMWF
• NORLAMEPS has better resolution
• NORLAMEPS is designed for our area of interest
• For day 1 3

3
NORLAMEPS set-up

• NORLAMEPS is a combination of
• TEPS (global model) (80 km/40 lev.) as EPS,
  except
• TEPS is targeted to our area of interest
• 20 1 members (EPS 50 1)
• LAMEPS (regional model, 20 km / 40 lev.)
• HIRLAM 20 1 members
• Initial perturbations and lateral boundary
  conditions from TEPS (6 hourly updates on
  boundaries)
• NORLAMEPS a combination of LAMEPS and TEPS. 41
  1 members

4
Verification area

• Target area SVs

Integration area HIRLAM
5
Running TEPS to 90h 20 1 members
Generation of perturbations
ECMWF
30 minutes
12 UTC
2 hours
TEPS is automatically transferred to met.no
TEPS is made ready to run in the Norwegian model
environment
19 UTC
TEPS is transferred to Trondheim
Generation of HIRLAM perturbations
00 UTC
Combination of TEPS and LAMEPS - NORLAMEPS
Generation of probability products for customers
LAMEPS running 18 UTC, to 60h
50 minutes
6
Test periodsOlder version of system

• Test cases are randomly chosen in all four
  seasons (but excluding completely dry periods)
• October 2002 14 days, 18. 31.
• January 2003 21 days, 10. 31. (-16.)
• May 2003 15 days, 4. 21. (-7., 9.,10,)
• August 2003 4 days, 11. 14.
• including very heavy rain
• A total of 54 days/cases

7
Spread - precipitation
8
Verification for a test period in 2002/2003 (54
cases)
__ EPS __ LAMEPS __ NORLAMEPS
For small precip. rates EPS is the best system
NORLAMEPS clearly better for larger precipitation
amounts
10mm/24h
25 mm/24h
9
Autumn (14 days)
__ EPS __ TEPS __ LAMEPS __ NORLAMEPS

• Including TEPS is in this case clearly
  beneficial.
• NORLAMEPS is clearly the better system of the
  four
• Notice that TEPS (20 members) verifies better
  than EPS (50 members)

25 mm/24 h
10
BSS/Area ROC (total test set)
11
BSS/Area ROC autumn
12
BSS/Area ROC spring
13
Case study 100 year precipitation in the
middle part of Norway in August 2003

• 14. 15. August 2003
• Atnadalen
• 116,5 mm/24h, 156,2 mm/48h
• Sunndalsøra
• 102,5 mm/24h, 171,9 mm/48h

14
14. Aug. 06UTC - 15. Aug. 06UTC
Observed
15
42h
Operational HIRLAM 20 km
16
ECMWF TEPS ensemble
42h
17
42h
LAMEPS ensemble
18
Use of NORLAMEPS

• Forecasting
• Eurorisk
• Storm surge LAMEPS

19
Future developments

• Include perturbing of the model physics
• Increase the time resolution of the boundary
  fields
• Expand system to more parameters temperature,
  wind, .
• Develop probability products
• Compute SVs within HIRLAM
• Move to higher resolution