Embed Size (px)
Citation preview
Norwegian Meteorological Institute met.no
LAMEPS – Limited area ensemble forecasting in Norway, using targeted EPS
Marit Helene Jensen, Inger-Lise Frogner, Trond Iversen and Ole Vignes
Norwegian Meteorological Institute met.no
Limited area ensemble forecasting in Norway - outline
• Ensembles using Norway’s operational version of the HIRLAM model – LAMEPS
• Perturbing initial state and lateral boundary conditions with a dedicated version of EPS from ECMWF - TEPS
• Combining these two systems gives NORLAMEPS
• Main focus is precipitation, especially extreme precipitation events
• Running storm surge from EPS,TEPS and LAMEPS
Norwegian Meteorological Institute met.no
TEPS – Targeted EPS
• A dedicated version of ECMWF EPS. – 20 + 1 ensemble members, as opposed to 50
+1 for EPS– Target area Northern Europe and adjacent sea
areas, as opposed to NH north of 30°N– Starts at 12 UTC every day– Runs to +72h, as opposed to +240h for EPS
• T255L40(~80 km), Since Feb 1st: T399L62(~55 km)
Norwegian Meteorological Institute met.no
LAMEPS
• HIRLAM in ensemble set-up • Quasi-operational at met.no since 14 February 2005• 20 members + Control,
– Control based on Norwegian HIRLAM analysis– 20 Initial and open boundary perturbations from
TEPS– Starts at 18UTC every day (fresh HIRLAM analysis),
i.e. a 6 hour delay compared to TEPS and EPS– Forecast Range: +60h
• Resolution: 0.2 deg (20km), 40 levels
Norwegian Meteorological Institute met.no
NORLAMEPS
• Combines simply IFS TEPS and HIRLAM LAMEPS
•A simple multi-model, multi-initial-state ensemble
•41 + 1 ensemble members
Norwegian Meteorological Institute met.no
Verification
area
Target area for TEPS SVs
Integration areaHIRLAM
Norwegian Meteorological Institute met.no
Generation of perturbations
Running TEPS to +72h
20 + 1 members12 UTCTEPS is automatically transferred to met.no
TEPS is made ready to run in the Norwegian model environment
TEPS is transferred to Trondheim
Generation of HIRLAM perturbations
LAMEPS running 18 UTC, to +60hResults from LAMEPStransferred back to met.no
Generation ofproducts forusers/customers
~02 UTC
~22 UTC
50 minutes
ECMWF
Norwegian Meteorological Institute met.no
LAMEPS
Precip, + 12 - +36
Reliability diagrams, percentiles - precip
70% 80%
95%90%
Norw. synop sitesSep05-April06
Norwegian Meteorological Institute met.no
Wind +24
LAMEPS
Reliability diagrams, percentiles -wind
70%
95%90%
80%
Norw. synop sitesSep05-April06
Norwegian Meteorological Institute met.no
Verification of Precip.
• Verify precipitation against “super-observations”. (Ghelli and Lalaurette): – All precipitation sites in Norway inside the verification
area are aggregated using the method of Kriging
• Distribution of precipitation in Norway is dominated by sharp gradients – We verify in sub-regions with grossly different
precipitation climatology separately.– Inside each sub-region precipitation climatology is
”fairly” uniform
Norwegian Meteorological Institute met.no
1West
2East
3North
Average annual precipitation amounts in Norway from 1961 – 90 (mm).
Norwegian Meteorological Institute met.no
Reliability diagrams, 20mm, 60h
NORLAMEPS
TEPS
Annual Feb05-Feb06
EPS
LAMEPS
Norwegian Meteorological Institute met.no
BRIER SKILL SCOREall three regions
Annual Feb05-Feb06
Norwegian Meteorological Institute met.no
ROC and Value(C/L), all three regions5 mm/24h
LAMEPSTEPSEPS
NORLAMEPS
Annual Feb05-Feb06
Norwegian Meteorological Institute met.no
LAMEPSTEPSEPS
NORLAMEPS
ROC and Value(C/L), all three regions20 mm/24h
Annual Feb05-Feb06
Norwegian Meteorological Institute met.no
Area under ROC-curveall three regionsAnnual Feb05-Feb06
Norwegian Meteorological Institute met.no
PRECIP.
Rms error and bias errorof control forecast Jan.-April 2006
-1,00
0,00
1,00
2,00
3,00
4,00
5,00
6,00
jan.06 feb.06 mar.06 apr.06
rmse teps (18h-42h)
bias teps (18h-42h)
rmse lameps (12h-36h)
bias lames (12h-36h)
-1,00
0,00
1,00
2,00
3,00
4,00
5,00
6,00
7,00
jan.06 feb.06 mar.06 apr.06
rmse teps (42h-66h)
bias teps (42h-66h)
rmse lameps (36h-60h)
bias lames (36h-60h)
Norwegian Meteorological Institute met.no
Spread around ensemble mean (Jan-Apr 2006)
0,00
0,50
1,00
1,50
2,00
2,50
EPS TEPS LAMEPS NORLAMEPS
18-42h
42-66h
Norwegian Meteorological Institute met.no
“100 year precipitation event” in the middle part of Norway
30-31. January 2006
>90mm/24h
130mm/24h
Norwegian Meteorological Institute met.no
LAMEPSP24 > 50mm
TEPSP24 > 50mm
+12h - +36h
+18h - +42h
Norwegian Meteorological Institute met.no
LAMEPSP24 > 60mm
TEPSP24 > 60mm
Norwegian Meteorological Institute met.no
Future developments
• Contribute to HIRLAM Ensemble Predictions (GLAMEPS)
• Include perturbations of model physics in LAMEPS• Increase the time resolution of the boundary fields
(now every 6 hour)• Develop a range of probability products• Compute meso-scale initial perturbations in
HIRLAM-domain• Increase resolution (0.1 degrees or finer)• Non-hydrostatic downscaling
Norwegian Meteorological Institute met.no
Test LAMEPS on a new configuration for TEPS
• A system that combines targeted SVs and hemispheric SVs (Martin Leutbecher, ECMWF)– 10 leading targeted singular vectors– 40 leading hemispheric singular vectors
computed in the subspace orthogonal to the targeted singular vectors
– Ensemble size 20 + 1– Initial perturbations constructed with
(revised) Gaussian sampling• Results in increased spread for TEPS after day
2, without increasing the error of the ensemble mean
Norwegian Meteorological Institute met.no
