Download - SLEPS First Results from SLEPS A. Walser, M. Arpagaus, C. Appenzeller, J. Quiby MeteoSwiss
SLEPS
First Results from SLEPS
A. Walser, M. Arpagaus, C. Appenzeller, J. Quiby
MeteoSwiss
Andre
.Wals
er@
Mete
oSw
iss.
chC
OSM
O G
enera
l M
eeti
ng,
20
03
, La
ngen
2
SLEPSSLEPS
SLEPS: Short-range limited-area ensemble prediction system.
Within project ‘Extreme Events’ of the Swiss national research project NCCR.
Strategy bases on an adapted COSMO-LEPS for the short-range.
Should be considered as a first step towards a short-range EPS.
The smaller scales resolved in SLEPS implies the consideration of alternative growth mechanisms beyond the baroclinic growth accounted for by the SV analysis.
Strategies for the definition of optimal small-scale perturbations have to be found.
Andre
.Wals
er@
Mete
oSw
iss.
chC
OSM
O G
enera
l M
eeti
ng,
20
03
, La
ngen
3
SLEPSSLEPS: Short-range adaptation of COSMO-LEPS
Start of COSMO-LEPS integrations
? members
COSMO-LEPS clustering times
short-range clustering times
number of RMs?
horizontal resolution?
day:
12
n-112
n+212
n+312
n+412
n+512
n+112
n0000 00 00 0000
moist? EPS
moist? EPS
moist? EPS
Andre
.Wals
er@
Mete
oSw
iss.
chC
OSM
O G
enera
l M
eeti
ng,
20
03
, La
ngen
4
SLEPSSLEPS: current setup
50+1 members
5 representative members (RMs)
5 Lokal Modell (limited-area) integrations nested into 5 RMsSLEPS: Short-range limited-area Ensemble Prediction System5 Lokal Modell (limited-area) integrations nested into 5 RMs
SLEPS: Short-range limited-area Ensemble Prediction System
5 clusters
Hierarchical Cluster Analysisarea: Europe
fields: 4 variables (U,V,Q,Z) at 3 levels (500, 700, 850) for 3 time steps (24h, 48h, 72 h),
number of clusters: fixed to 5
Hierarchical Cluster Analysisarea: Europe
fields: 4 variables (U,V,Q,Z) at 3 levels (500, 700, 850) for 3 time steps (24h, 48h, 72 h),
number of clusters: fixed to 5
Representative Member Selection one per cluster:
member nearest (3D) to the mean of its own cluster AND most distant to the other clusters’
means
Representative Member Selection one per cluster:
member nearest (3D) to the mean of its own cluster AND most distant to the other clusters’
means
Global ECMWF EPS ensembles with moist singular vectorsGlobal ECMWF EPS ensembles with moist singular vectors
Andre
.Wals
er@
Mete
oSw
iss.
chC
OSM
O G
enera
l M
eeti
ng,
20
03
, La
ngen
5
SLEPSSLEPS: model domain with topography
10 km mesh-size horizontally, 32 vertical levels.
Andre
.Wals
er@
Mete
oSw
iss.
chC
OSM
O G
enera
l M
eeti
ng,
20
03
, La
ngen
6
SLEPSMoist vs. operational singular vectorsCoutinho et al. (2003)
‚opr‘ SVs (T42L31, OTI 48 h): linearized physics package with
surface drag
simple vertical diffusion
‚moist‘ SVs (T63L31, OTI 24 h): linearized physics package includes additionally:
gravity wave drag
long-wave radiation
deep cumulus convection
large-scale condensation
moist SVs: use of moist processes during SV evolution, but same norm (‚total energy norm‘) no humidity perturbations.
Andre
.Wals
er@
Mete
oSw
iss.
chC
OSM
O G
enera
l M
eeti
ng,
20
03
, La
ngen
7
SLEPS2 case studies:
Storm Lothar (“Christmas Storm”): 26 December 1999
moist SV ECMWF EPS SLEPS 19991224 00 UTC, + 72 h
opr SV ECMWF EPS SLEPS 19991224 00 UTC, + 72 h
Storm Martin: 27/28 December 1999
moist SV EPS ECMWF SLEPS 19991226 00 UTC, + 72 h
opr SV EPS ECMWF SLEPS 19991226 00 UTC, + 72 h
Andre
.Wals
er@
Mete
oSw
iss.
chC
OSM
O G
enera
l M
eeti
ng,
20
03
, La
ngen
8
SLEPSMean RMS error Z@850 hPa of ECMWF EPS to
opr analysis in SLEPS domainR
MS
EN
S [m
]
Forecast time [h]
red: opr SVs EPS
blue: moist SVs EPS
solid: Storm Lothar
dotted: Storm Martin
N
n
I
iiiENS oy
INRMS
1 1
211 y: predicted value
o: values of analysis
N: number of members
I: number of grid points
Andre
.Wals
er@
Mete
oSw
iss.
chC
OSM
O G
enera
l M
eeti
ng,
20
03
, La
ngen
9
SLEPSStorm Lothar: probability forecast of SLEPS for 10 m wind gusts
moist SVs opr SVs
ECMWF opr analysis
Lothar
Andre
.Wals
er@
Mete
oSw
iss.
chC
OSM
O G
enera
l M
eeti
ng,
20
03
, La
ngen1
0
SLEPSLothar: 6 hours later…
moist SVs
opr SVs
opr SVs
Andre
.Wals
er@
Mete
oSw
iss.
chC
OSM
O G
enera
l M
eeti
ng,
20
03
, La
ngen1
1
SLEPSLothar: again 6 hours later…
moist SVs opr SVs
Andre
.Wals
er@
Mete
oSw
iss.
chC
OSM
O G
enera
l M
eeti
ng,
20
03
, La
ngen1
2
SLEPSLothar: again 6 hours later…
moist SVs opr SVs
Andre
.Wals
er@
Mete
oSw
iss.
chC
OSM
O G
enera
l M
eeti
ng,
20
03
, La
ngen1
3
SLEPSLothar: probability forecast for 10 m wind gusts over the 24 hours storm period
moist SVs opr SVs
Both ensemble quite similar. Moist SVs SLEPS predicts a higher risk for strong wind
gusts over northern France.
Andre
.Wals
er@
Mete
oSw
iss.
chC
OSM
O G
enera
l M
eeti
ng,
20
03
, La
ngen1
4
SLEPSLothar: ECMWF EPS using only LEPS representative members over the 24 hours storm period
moist SVs opr SVs
Rather reduced probabilities compared to SLEPS in particular over the Atlantic. SLEPS downscaling seems to be beneficial in this case.
Andre
.Wals
er@
Mete
oSw
iss.
chC
OSM
O G
enera
l M
eeti
ng,
20
03
, La
ngen1
5
SLEPSLothar: ECMWF EPS using all 51 members over the 24 hours storm period
moist SVs opr SVs
Very similar Effect of moist SVs not obvious.
Andre
.Wals
er@
Mete
oSw
iss.
chC
OSM
O G
enera
l M
eeti
ng,
20
03
, La
ngen1
6
SLEPSStorm Martin: probability forecast of SLEPS for 10 m wind gusts
opr SVsmoist SVs
Martin
Andre
.Wals
er@
Mete
oSw
iss.
chC
OSM
O G
enera
l M
eeti
ng,
20
03
, La
ngen1
7
SLEPSStorm Martin: 6 hours later…
moist SVs opr SVs
Andre
.Wals
er@
Mete
oSw
iss.
chC
OSM
O G
enera
l M
eeti
ng,
20
03
, La
ngen1
8
SLEPSStorm Martin: 6 hours later…
moist SVs opr SVs
Andre
.Wals
er@
Mete
oSw
iss.
chC
OSM
O G
enera
l M
eeti
ng,
20
03
, La
ngen1
9
SLEPSStorm Martin: 6 hours later…
moist SVs opr SVs
Andre
.Wals
er@
Mete
oSw
iss.
chC
OSM
O G
enera
l M
eeti
ng,
20
03
, La
ngen2
0
SLEPSStorm Martin: probability forecast for 10 m wind gusts over the 24 hours storm period
moist SVs opr SVs
moist SVs SLEPS predicts a risk in south-western France and northern Mediterranean Sea.
opr SVs SLEPS misses the storm.
Andre
.Wals
er@
Mete
oSw
iss.
chC
OSM
O G
enera
l M
eeti
ng,
20
03
, La
ngen2
1
SLEPSMartin: ECMWF EPS using only LEPS RMsover the 24 hours storm period
moist SVs opr SVs
Provided risk from moist SVs ECMWF EPS lower over southwestern France and northern Mediterranean sea.
Opr SVs ECMWF EPS also misses the storm.
Andre
.Wals
er@
Mete
oSw
iss.
chC
OSM
O G
enera
l M
eeti
ng,
20
03
, La
ngen2
2
SLEPSMartin: ECMWF EPS using all 51 members over the 24 hours storm period
moist SVs opr SVs
Using all EPS members, differences between moist and opr SVs EPS smaller moist SVs SLEPS additionally favoured by the clustering procedure.
Andre
.Wals
er@
Mete
oSw
iss.
chC
OSM
O G
enera
l M
eeti
ng,
20
03
, La
ngen2
3
SLEPSConclusions from first SLEPS simulations
Wind gust prediction for storm Lothar:
Effect of moist SVs in SLEPS and ECMWF EPS small.
Downscaling effect with SLEPS beneficial but not crucial.
Wind gust prediction for storm Martin:
Effect of moist SVs clearly positive both in SLEPS and ECMWF EPS.
opr SVs SLEPS does not provide a storm warning.
Downscaling effect with moist SVs SLEPS beneficial.