SREPS Priority Project: final report
description
Transcript of SREPS Priority Project: final report
SREPS Priority Project:final report
C. Marsigli, A. Montani, T. Paccagnella
ARPA-SIMC - HydroMeteorological Service of Emilia-Romagna, Bologna, Italy
F. Gofa, P. Louka
HNMS – Hellenic National Meteorological Service, Athens, Greece
Last FTEs of Chiara were used for this
TASK
Last FTEs of Chiara were used for this
TASK
ROMEO
Outline
COSMO-SREPS methodology
system set-up
analysis of the results on MAP D-PHASE DOP (JJA 2007+SON 2007)
role of different kind of perturbations
error vs spread
boundaries from mm vs different physics
ranking of different driving models and of different physics
conclusions
analysis of the results on MAP D-PHASE DOP (JJA 2007+SON 2007)
role of different kind of perturbations
error vs spread
boundaries from mm vs different physics
ranking of different driving models and of different physics
conclusions
System set-up
16 COSMO runs 10 km
hor. res.40 vertical
levels
COSMO at 25 km on IFS
IFS – ECMWF globalb
y I
NM
S
pain
COSMO at 25 km on GME
GME – DWD global
COSMO at 25 km on UM
UM – UKMO global
COSMO at 25 km on GFS
GFS – NCEP global
P1: control (ope)
P2: conv. scheme (KF)
P3: tur_len=1000
P4: pat_len=10000
00 UTCJJA=53 SON=54
DOP
SON=54 runsJJA=53 runs
COSMO observations
intra-group distance
Z500 COSMO analysis
JJA 2007 - 50 days
SON 2007 - 49 days
Same driving model
Different model
parameters
Same model parameters
Different
driving model
role of different kind of perturbations
intra-group distanceJJA 2007 - 50 days
t850
COSMO analysis
COSMO-SREPS
intra-group distanceSON 2007 - 49 days
t850
COSMO I7 analysis
COSMO-SREPS
intra-group distance
2mT Northern Italy
JJA 2007 - 50 days
SON 2007 - 49 days
SYNOP over D-PHASE area - Nearest grid point
JJA07 SON07
2m T - relationship between error and spread
COSMO I7 analysis
error vs spread Underdispersive
Synop stations on the Alpine area
218 stations
spread/skill relationship
2mT Alpine area (synop stations)
+12h +24h +36h
+48h +60h +72h
spread/skill relationship
2mT Alpine area (synop stations)
+12h +24h +36h
+48h +60h +72h
Underdispersive
t850 relationship between EM error and
spread
COSMO-I7 interpolated on SYNOP stations over the Alpine area
JJA07
t850 relationship between error and
spread
COSMO-I7 interpolated on SYNOP stations over the Alpine area
SON07
spread/skill relationship
t850 Alpine area (COSMO analyses)
+12h +24h +36h
+48h +60h +72h
spread/skill relationship
t850 Alpine area (COSMO analyses)
+12h +24h +36h
+48h +60h +72h
Z500 relationship between error and
spread
COSMO-I7 interpolated on Northern Italy stations and SYNOP stations over the Alpine area
JJA07
Z500 relationship between error and
spread
COSMO-I7 interpolated on SYNOP stations over the Alpine area
SON07
TP 24h – ave 0.5x0.5 JJA07
+30h
noss 1400 700 400 150 50
IT + CH
+54h
Same driving models
Same driving models
Same perturbation
Big impact of multi model BCs!!!!!!!
Looking at the right column it is evident that even with few members the skill does not decrease too much when the driving models are different.
Same perturbation
TP 24h – ave 0.5x0.5
noss 800 500 300 100 50
SON07
+30h
IT + CH
+54h
Same driving model
Same driving model
Same perturbation
Same perturbation
JJA07 +30h
IT
IT + CH
Same driving model
Same perturbation
-0.80
-0.40
0.00
0.40
0.80
1.20
1.60
2.00
2.40
0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72
forecast range (h)
bia
s (K
)
m1
m2
m3
m4
m5
m6
m7
m8
m9
m10
m11
m12
m13
m14
m15
m16
2m temperature - BIAS
Synop observations over Alpine area
Nearest grid point – lsm + altitude correction
JJA07
ecmwfgmencepukmo
p4
2m temperature - MAE
Synop observations over Alpine area
Nearest grid point – lsm + altitude correction
JJA07
2.60
3.00
3.40
3.80
4.20
0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72
forecast range (h)
mae
(K
)
m1
m2
m3
m4
m5
m6
m7
m8
m9
m10
m11
m12
m13
m14
m15
m16
ecmwfgmencepukmo
p4
2m temperature - BIAS
Synop observations over Alpine area
Nearest grid point – lsm + altitude correction
SON07
-3.20
-2.80
-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72
forecast range (h)
bia
s (K
)
m1
m2
m3
m4
m5
m6
m7
m8
m9
m10
m11
m12
m13
m14
m15
m16ecmwfgmencepukmo
p4
2m temperature - MAE
Synop observations over Alpine area
Nearest grid point – lsm + altitude correction
SON07
2.60
3.00
3.40
3.80
4.20
0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72
forecast range (h)
mae
(K
)
m1
m2
m3
m4
m5
m6
m7
m8
m9
m10
m11
m12
m13
m14
m15
m16
ecmwfgmencepukmo
p4
Deterministic scores – ave 0.5 x 0.5 IT
1mm/24h
5mm/24h
10mm/24h
father
Deterministic scores – ave 0.5 x 0.5 IT
1mm/24h
5mm/24h
10mm/24h
father
Deterministic scores – ave 0.5 x 0.5 IT
pert
1mm/24h
5mm/24h
10mm/24h
Deterministic scores – ave 0.5 x 0.5 IT
1mm/24h
5mm/24h
10mm/24h
pert
Test of more parameter perturbations (same father)
16 LM runs at 10 km
P1: control (ope)
P2: conv. scheme (KF)
P3: parameter 1
P4: parameter 2
P5: …
IFS – ECMWF global
SON 07
T2m deterministic scores – npo IT
-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0 3 6 9 12 15 18 21 24
forecast range (h)
bia
s (K
)
-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0 3 6 9 12 15 18 21 24
m1
m2
m3
m4
m5
m6
m7
m8
m9
m10
m11
m12
m13
m14
m15
m16
ctrl
KFtur_len=150tur_len=1000pat_len=10000rat_sea=1rat_sea=60qc0=0.001crsmin=50crsmin=200c_soil=0c_soil=2c_lnd=1c_lnd=10rlam_heat=0.1rlam_heat=10
2.00
2.10
2.20
2.30
2.40
2.50
2.60
2.70
2.80
0 3 6 9 12 15 18 21 24
forecast range (h)
mae
(K
)
T2m deterministic scores – npo IT
-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0 3 6 9 12 15 18 21 24
m1
m2
m3
m4
m5
m6
m7
m8
m9
m10
m11
m12
m13
m14
m15
m16
ctrl
KFtur_len=150tur_len=1000pat_len=10000rat_sea=1rat_sea=60qc0=0.001crsmin=50crsmin=200-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0 3 6 9 12 15 18 21 24
m1
m2
m3
m4
m5
m6
m7
m8
m9
m10
m11
m12
m13
m14
m15
m16
c_soil=0c_soil=2c_lnd=1c_lnd=10rlam_heat=0.1rlam_heat=10
Preliminary ConclusionsPerturbations -Multi Model ICs/BCs & Perturbations on
Ph. Params:
the use of different driving models seems to dominate with respect to physics parameter perturbations as regards the contribution to the spread; these contributions are different in the two seasons (2mT)
the selected parameters produce a detectable spread among members with the same father (driving model)
spread-skill relationship:
a correlation between error and spread exists, but the
system is under-dispersive -> a better representation of model error is needed
the different driving models contribute differently to the ensemble
skill, but there is a strong dependence on forecast range, season, verification area
the different perturbations can contribute differently to the ensemble skill as well
On-going activities and future plans
continue the analysis over the DOP MAP D-PHASE :
statistical analysis of the system
comparison with the other available mesoscale ensemble systems
verification carried out by HNMS
introduce the new parameter perturbations
analyse the impact of adding soil perturbations
COSMO-SREPS methodology
i.c. and b.c. perturbations -> INM multi-model multi-boundary ensemble (SREPS)
LAM perturbations -> physics parameter perturbations
LAM perturbations - smaller scale errors
driving model perturbations (ics and bcs) - larger scale errors
System set-up
16 COSMO runs 10 km
hor. res.40 vertical
levels
COSMO at 25 km on IFS
IFS – ECMWF globalb
y I
NM
S
pain
COSMO at 25 km on GME
GME – DWD global
COSMO at 25 km on UM
UM – UKMO global
COSMO at 25 km on GFS
GFS – NCEP global
P1: control (ope)
P2: conv. scheme (KF)
P3: tur_len=1000
P4: pat_len=10000
MAP D-PHASE DOP testing period
COSMO-SREPS was running during the DOP, at 00 UTC
107 runs out of 183 days, 53 in JJA and 54 in SON
ensemble verification
JJA07
intra-group distanceJJA 2007 - 50 days
2mT
Northern Italy
COSMO-SREPS
Same driving model
Different model parameters
Same model parameters
Different driving model
intra-group distanceJJA 2007 – 50 days
tp6h
Northern Italy
Same driving model
Different model parameters
Same model parameters
Different driving model
Mid Term comments
• Mid-upper troposphere: MULTI MODEL IC/BCs give the bigger contribution to the spread
• Surface/lower troposphere: model physics perturbations “gain ground”.
score evaluation
• +30h:
• ROC
• UKMO and GME the best, then ECMWF and NCEP
• P2 (KF) the best, then P4, P3, P1 (similar)
• BSS
• UKMO the best, ECMWF and GME the worst; NCEP improves with threshold
• P are similar, P2 (KF) slightly better
JJA07tp24I
T
score evaluation
• +54 :
• ROC
• similar for low thresholds, NCEP the best for high thresholds, then ECMWF
• P similar, P3 slightly better
• BSS
• ECMWF the best, GME the worst; NCEP improves with threshold
• P2 (KF) the worst, P3 the best but similar to P1 and P4
JJA07tp24I
T
1 5 10 20 30
EC +30 ROC
BSS
+54 ROC
BSS
MO +30 ROC
BSS
+54 ROC
BSS
GME +30 ROC
BSS
+54 ROC
BSS
AVN +30 ROC
BSS
+54 ROC
BSS
1 5 10 20 30
EC +30 ROC **** ** ** ** *
BSS
+54 ROC
BSS
MO +30 ROC ** **** **** **** ****
BSS
+54 ROC
BSS
GME +30 ROC * *** **** *** ****
BSS
+54 ROC
BSS
AVN +30 ROC ** * * * ***
BSS
+54 ROC
BSS
TP 24h – ave 0.5x0.5 JJA07
+30h
noss 700 350 200 60 20
IT
+54h
father
father
pert
pert
TP 24h – ave 0.5x0.5 JJA07
+30h
IT
+54h
father
father
pert
pert
noss 700 350 200 60 20
Grosso impatto del MULTI MODEL ai boundaries!!!!!!! Looking at the right column it is evident that even with few members the skill does not decrease too much when the driving models are different.
score evaluation
• +30h:
• ROC
• UKMO > GME > ECMWF > NCEP
• P2 (KF) the best, the others are similar
• BSS
• NCEP better for high threshold; UKMO > ECMWF > GME
• P2 (KF) the worst, the others are similar
JJA07tp24IT+CH
score evaluation
• +54 :
• ROC
• similar for low thresholds, NCEP and GME best for high thresholds
• P similar; P2 (KF) slightly better for high thresholds
• BSS
• ECMWF the best, NCEP the worst
• P similar; P2 (KF) slightly worse
JJA07tp24IT+CH
TP 24h – ave 0.5x0.5 JJA07
+30h
noss 1400 700 400 150 50
IT + CH
+54h
father
father
pert
pert
TP 24h – ave 0.5x0.5 JJA07
+30h
noss 1400 700 400 150 50
IT + CH
+54h
father
father
pert
pert
The performances reverse eith the leading time.
Bad skill adding switzerlad.
Deterministic scores – ave 0.5 x 0.5 IT+CH
1mm/24h
5mm/24h
10mm/24h
father
Deterministic scores – ave 0.5 x 0.5 IT+CH
1mm/24h
5mm/24h
10mm/24h
father
Deterministic scores – ave 0.5 x 0.5 IT+CH
1mm/24h
5mm/24h
10mm/24h
pert
Deterministic scores – ave 0.5 x 0.5 IT+CH
1mm/24h
5mm/24h
10mm/24h
pert
Deterministic scores – ave 0.5 x 0.5 IT
1mm/24h
5mm/24h
10mm/24h
fathernoss 250 100 50
Deterministic scores – ave 0.5 x 0.5 IT
1mm/24h
5mm/24h
10mm/24h
fathernoss 250 100 50
Deterministic scores – ave 0.5 x 0.5 IT
pert
1mm/24h
5mm/24h
10mm/24h
noss 250 100 50
Deterministic scores – ave 0.5 x 0.5 IT
1mm/24h
5mm/24h
10mm/24h
pertnoss 250 100 50
6h accumulated precipitationrelationship between error and
spread
obs Northern Italy
L1
Relationship between error and spread
Northern Italy observations
Nearest grid point
applicata correzione per la quota LAPSE=0.7
eliminati dati minori di –10 e maggiori di +42
t2m
Relationship between error and spread
SYNOP over MAP area
Nearest grid point
t2m
applicata correzione per la quota LAPSE=0.7
all (16 members)
Relationship between error and spread
SYNOP over MAP area
Nearest grid point
t2m
applicata correzione per la quota LAPSE=0.7
GME only (4 members)
Relationship between error and spread
SYNOP over MAP area
Nearest grid point
t2m
applicata correzione per la quota LAPSE=0.7
ECMWF only (4 members)
Relationship between error and spread
SYNOP over MAP area
Nearest grid point
t2m
applicata correzione per la quota LAPSE=0.7
NCEP only (4 members)
Relationship between error and spread
SYNOP over MAP area
Nearest grid point
t2m
applicata correzione per la quota LAPSE=0.7
UKMO only (4 members)
Relationship between error and spread
SYNOP over MAP area
Nearest grid point
t2m
applicata correzione per la quota LAPSE=0.7
2m temperature relationship between error and
spread
COSMO-I7 interpolated on Northern Italy stations and SYNOP stations over the Alpine area
SON07
6h precipitation - BSS
Northern Italy observations
Average over 0.5 x 0.5 deg boxes
6h precipitation – ROC area
Northern Italy observations
Average over 0.5 x 0.5 deg boxes
Daily precipitation - BSS
Northern Italy + Switzerland observations
Average over 0.5 x 0.5 deg boxes
Daily precipitationreliability and resolution
Northern Italy + Switzerland observations
Average over 0.5 x 0.5 deg boxes
Daily precipitation - ROC area
Northern Italy + Switzerland observations
Average over 0.5 x 0.5 deg boxes
Relationship between error and spread
Northern Italy observations
Nearest grid point eliminati dati minori di –10 e maggiori di +42
tp 6h
Relationship between error and spread
Northern Italy observations
Nearest grid point
t2m
applicata correzione per la quota LAPSE=0.7
eliminati dati minori di –10 e maggiori di +42
Relationship between error and spread
SYNOP over MAP area
Nearest grid point
t2m
applicata correzione per la quota LAPSE=0.7
2m temperature relationship between error and
spread
COSMO-I7 interpolated on SYNOP stations over the Alpine area
intra-group distanceSON 2007 - 49 days
2mT
Northern Italy
COSMO-SREPS
Same driving model
Different model parameters
Same model parameters
Different driving model
intra-group distanceSON 2007 - 49 days
tp6h
Northern Italy
COSMO-SREPS
score evaluation
• +30h:
• ROC
• crossing; UKMO and ECMWF slightly better, GME worse
• P2 (KF) the best, the others are similar
• BSS
• ECMWF the best, GME the worst
• P are similar, P2 (KF) slightly better
SON07
tp24I
T
score evaluation
• +54 :
• ROC
• NCEP the best, GME the worst
• P2 (KF) the best, the others are similar
• BSS
• similar
• similar
SON07
tp24I
T
TP 24h – ave 0.5x0.5 SON07
+30h
IT
+54h
father
father
pert
pert
noss 550 300 200 80 50
TP 24h – ave 0.5x0.5 SON07
+30h
IT
+54h
father
father
pert
pert
noss 550 300 200 80 50
score evaluation
• +30h:
• ROC
• similar; for high thresholds ECMWF is the best and GME the worst
• similar; P2 (KF) slightly better
• BSS
• ECMWF the best and GME the worst, , especially with increasing threshold
• P are similar, P4 (tur_len=1000) better for the last thresholds
SON07
tp24IT+CH
score evaluation
• +54 :
• ROC
• NCEP the best, GME the worst
• similar, P2 (KF) slightly better
• BSS
• ECMWF the best, UKMO the worst
• similar
SON07
tp24IT+CH
TP 24h – ave 0.5x0.5 SON07
+30h
IT + CH
+54h
father
father
pert
pert
noss 800 500 300 100 50
TP 24h – ave 0.5x0.5 SON07+30h
IT + CH
+54h
father
father
pert
pert
noss 800 500 300 100 50
Deterministic scores – ave 0.5 x 0.5 IT
1mm/24h
5mm/24h
10mm/24h
father
Deterministic scores – ave 0.5 x 0.5 IT
1mm/24h
5mm/24h
10mm/24h
father
Deterministic scores – ave 0.5 x 0.5 IT
pert
1mm/24h
5mm/24h
10mm/24h
Deterministic scores – ave 0.5 x 0.5 IT
1mm/24h
5mm/24h
10mm/24h
pert
Deterministic scores – ave 0.5 x 0.5 IT+CH
father
1mm/24h
5mm/24h
10mm/24h
Deterministic scores – ave 0.5 x 0.5 IT+CH
1mm/24h
5mm/24h
10mm/24h
father
Deterministic scores – ave 0.5 x 0.5 IT+CH
pert
1mm/24h
5mm/24h
10mm/24h
Deterministic scores – ave 0.5 x 0.5 IT+CH
1mm/24h
5mm/24h
10mm/24h
pert
Deterministic scores – ave 0.5 x 0.5 IT
1mm/24h
5mm/24h
10mm/24h
fathernoss 250 100 50
Deterministic scores – ave 0.5 x 0.5 IT
1mm/24h
5mm/24h
10mm/24h
fathernoss 250 100 50
Deterministic scores – ave 0.5 x 0.5 IT
pert
1mm/24h
5mm/24h
10mm/24h
noss 250 100 50
Deterministic scores – ave 0.5 x 0.5 IT
1mm/24h
5mm/24h
10mm/24h
pertnoss 250 100 50
Test of more parameter perturbations (same father)
16 LM runs at 10 km
P1: control (ope)
P2: conv. scheme (KF)
P3: parameter 1
P4: parameter 2
P5: …
IFS – ECMWF global
SON 07
run nr.parameter name parameter description range default used1 ctrl ope2 lconv convection scheme T or KF T KF3 tur_len maximal turbulent length scale [100,1000] m 500 1504 tur_len maximal turbulent length scale [100,1000] m 500 10005 pat_len length scale of thermal surface patterns [0,10000] m 500 100006 rat_sea ratio of laminar scaling factors for heat over sea [1,100] 20 17 rat_sea ratio of laminar scaling factors for heat over sea [1,100] 20 608 qc0 cloud water threshold for autoconversion [0.,0.001] 0 0.0019101112
13 c_lnd
surface area density of the roughness elements over land [1,10] 2 1
14 c_lnd
surface area density of the roughness elements over land [1,10] 2 10
15 rlam_heat scaling factor of the laminar layer depth [0.1,10] 1 0.116 rlam_heat scaling factor of the laminar layer depth [0.1,10] 1 10
crsmincrsminc_soilc_soil
Minimal stomata resistance 15015011 2
0
50200Minimal stomata resistance
Surface area index of the evaporating soilSurface area index of the evaporating soil
[50,200] s/m[50,200] s/m
]0,c_lnd[
]0,c_lnd[
T2m deterministic scores – npo IT
-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0 3 6 9 12 15 18 21 24
forecast range (h)
bia
s (K
)
-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0 3 6 9 12 15 18 21 24
m1
m2
m3
m4
m5
m6
m7
m8
m9
m10
m11
m12
m13
m14
m15
m16
ctrl
KFtur_len=150tur_len=1000pat_len=10000rat_sea=1rat_sea=60qc0=0.001crsmin=50crsmin=200c_soil=0c_soil=2c_lnd=1c_lnd=10rlam_heat=0.1rlam_heat=10
2.00
2.10
2.20
2.30
2.40
2.50
2.60
2.70
2.80
0 3 6 9 12 15 18 21 24
forecast range (h)
mae
(K
)
T2m deterministic scores – npo IT
-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0 3 6 9 12 15 18 21 24
m1
m2
m3
m4
m5
m6
m7
m8
m9
m10
m11
m12
m13
m14
m15
m16
ctrl
KFtur_len=150tur_len=1000pat_len=10000rat_sea=1rat_sea=60qc0=0.001crsmin=50crsmin=200-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0 3 6 9 12 15 18 21 24
m1
m2
m3
m4
m5
m6
m7
m8
m9
m10
m11
m12
m13
m14
m15
m16
c_soil=0c_soil=2c_lnd=1c_lnd=10rlam_heat=0.1rlam_heat=10
T2m deterministic scores – npo ITpia mon
1.40
1.80
2.20
2.60
3.00
3.40
3.80
0 3 6 9 12 15 18 21 24
forecast range (h)
mae
(K
)
1.40
1.80
2.20
2.60
3.00
3.40
3.80
0 3 6 9 12 15 18 21 24
forecast range (h)
mae
(K
)
-2.80
-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0.80
1.20
1.60
2.00
2.40
0 3 6 9 12 15 18 21 24
forecast range (h)
bia
s (
K)
-2.80
-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0.80
1.20
1.60
2.00
2.40
0 3 6 9 12 15 18 21 24
forecast range (h)
bia
s (
K)
Td2m deterministic scores – npo IT
-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0 3 6 9 12 15 18 21 24
m1
m2
m3
m4
m5
m6
m7
m8
m9
m10
m11
m12
m13
m14
m15
m16
ctrl
KFtur_len=150tur_len=1000pat_len=10000rat_sea=1rat_sea=60qc0=0.001crsmin=50crsmin=200c_soil=0c_soil=2c_lnd=1c_lnd=10rlam_heat=0.1rlam_heat=10
-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0.80
1.20
1.60
0 3 6 9 12 15 18 21 24
forecast range (h)
bia
s (K
)
-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0 3 6 9 12 15 18 21 24
m1
m2
m3
m4
m5
m6
m7
m8
m9
m10
m11
m12
m13
m14
m15
m16
ctrl
KFtur_len=150tur_len=1000pat_len=10000rat_sea=1rat_sea=60qc0=0.001crsmin=50crsmin=200
-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0 3 6 9 12 15 18 21 24
m1
m2
m3
m4
m5
m6
m7
m8
m9
m10
m11
m12
m13
m14
m15
m16
c_soil=0c_soil=2c_lnd=1c_lnd=10rlam_heat=0.1rlam_heat=10
2.20
2.40
2.60
2.80
3.00
3.20
3.40
3.60
3.80
0 3 6 9 12 15 18 21 24
forecast range (h)
mae
(K
)
Td2m deterministic scores – npo IT
-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0 3 6 9 12 15 18 21 24
m1
m2
m3
m4
m5
m6
m7
m8
m9
m10
m11
m12
m13
m14
m15
m16
ctrl
KFtur_len=150tur_len=1000pat_len=10000rat_sea=1rat_sea=60qc0=0.001crsmin=50crsmin=200
-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0 3 6 9 12 15 18 21 24
m1
m2
m3
m4
m5
m6
m7
m8
m9
m10
m11
m12
m13
m14
m15
m16
c_soil=0c_soil=2c_lnd=1c_lnd=10rlam_heat=0.1rlam_heat=10
Td2m deterministic scores – npo IT
-2.80
-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0.80
1.20
1.60
2.00
0 3 6 9 12 15 18 21 24
forecast range (h)
bia
s (K
)
-2.80
-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0.80
1.20
1.60
2.00
0 3 6 9 12 15 18 21 24
forecast range (h)
bia
s (K
)
pia mon
1.80
2.00
2.20
2.40
2.60
2.80
3.00
3.20
3.40
3.60
3.80
0 3 6 9 12 15 18 21 24
forecast range (h)
mae
(K
)
1.80
2.00
2.20
2.40
2.60
2.80
3.00
3.20
3.40
3.60
3.80
0 3 6 9 12 15 18 21 24
forecast range (h)
ma
e (
K)
0.9
1
1.1
1.2
1.3
1.4
6 12 18 24
forecast range (h)
bia
s sc
ore
TP deterministic scores – ave 0.5 x 0.5 IT
1mm/6h
-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0 3 6 9 12 15 18 21 24
m1
m2
m3
m4
m5
m6
m7
m8
m9
m10
m11
m12
m13
m14
m15
m16
ctrl
KFtur_len=150tur_len=1000pat_len=10000rat_sea=1rat_sea=60qc0=0.001crsmin=50crsmin=200c_soil=0c_soil=2c_lnd=1c_lnd=10rlam_heat=0.1rlam_heat=10
noss 452 479 470 447
0.44
0.46
0.48
0.5
0.52
0.54
0.56
0.58
6 12 18 24
forecast range (h)
thre
at s
core
TP deterministic scores – ave 0.5 x 0.5 IT
1mm/6h
-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0 3 6 9 12 15 18 21 24
m1
m2
m3
m4
m5
m6
m7
m8
m9
m10
m11
m12
m13
m14
m15
m16
ctrl
KFtur_len=150tur_len=1000pat_len=10000rat_sea=1rat_sea=60qc0=0.001crsmin=50crsmin=200c_soil=0c_soil=2c_lnd=1c_lnd=10rlam_heat=0.1rlam_heat=10 noss 452 479 470 447
0.28
0.3
0.32
0.34
0.36
0.38
0.4
0.42
6 12 18 24
forecast range (h)
fals
e al
arm
rat
eTP deterministic scores – ave 0.5 x 0.5
IT1mm/6h
-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0 3 6 9 12 15 18 21 24
m1
m2
m3
m4
m5
m6
m7
m8
m9
m10
m11
m12
m13
m14
m15
m16
ctrl
KFtur_len=150tur_len=1000pat_len=10000rat_sea=1rat_sea=60qc0=0.001crsmin=50crsmin=200-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0 3 6 9 12 15 18 21 24
m1
m2
m3
m4
m5
m6
m7
m8
m9
m10
m11
m12
m13
m14
m15
m16
c_soil=0c_soil=2c_lnd=1c_lnd=10rlam_heat=0.1rlam_heat=10
noss 452 479 470 447
0.9
1
1.1
1.2
1.3
1.4
6 12 18 24
forecast range (h)
bia
s sc
ore
TP deterministic scores – ave 0.5 x 0.5 IT
10mm/6h
-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0 3 6 9 12 15 18 21 24
m1
m2
m3
m4
m5
m6
m7
m8
m9
m10
m11
m12
m13
m14
m15
m16
ctrl
KFtur_len=150tur_len=1000pat_len=10000rat_sea=1rat_sea=60qc0=0.001crsmin=50crsmin=200-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0 3 6 9 12 15 18 21 24
m1
m2
m3
m4
m5
m6
m7
m8
m9
m10
m11
m12
m13
m14
m15
m16
c_soil=0c_soil=2c_lnd=1c_lnd=10rlam_heat=0.1rlam_heat=10
noss 73 81 69 70
0.2
0.24
0.28
0.32
0.36
0.4
6 12 18 24
forecast range (h)
thre
at s
core
TP deterministic scores – ave 0.5 x 0.5 IT
10mm/6h
-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0 3 6 9 12 15 18 21 24
m1
m2
m3
m4
m5
m6
m7
m8
m9
m10
m11
m12
m13
m14
m15
m16
ctrl
KFtur_len=150tur_len=1000pat_len=10000rat_sea=1rat_sea=60qc0=0.001crsmin=50crsmin=200c_soil=0c_soil=2c_lnd=1c_lnd=10rlam_heat=0.1rlam_heat=10
noss 73 81 69 70
0.44
0.48
0.52
0.56
0.6
0.64
0.68
0.72
6 12 18 24
forecast range (h)
fals
e al
arm
rat
eTP deterministic scores – ave 0.5 x 0.5
IT10mm/6h
-2.40
-2.00
-1.60
-1.20
-0.80
-0.40
0.00
0.40
0 3 6 9 12 15 18 21 24
m1
m2
m3
m4
m5
m6
m7
m8
m9
m10
m11
m12
m13
m14
m15
m16
ctrl
KFtur_len=150tur_len=1000pat_len=10000rat_sea=1rat_sea=60qc0=0.001crsmin=50crsmin=200c_soil=0c_soil=2c_lnd=1c_lnd=10rlam_heat=0.1rlam_heat=10
noss 73 81 69 70
t BIA t MAE td BIA td MAE tp1 BS
tp1 TS
tp1 FA tp10 BS
tp10 TS
tp10 FA
KF = = > = = < > <> > > < >
tur_len=150 = < = = > = = <> <> < = < =
tur_len=1000 = > = < = = = <> <> <> => <24
=
pat_len=10000
> <> = < = > = > > = < > > <24
< >24
rat_sea=1 = > = < > = > > <> <> > <> = >24
rat_sea=60 = < = > < = > < <> = < < = < = >
qc=0.001 = = = = = = < > = > <24
< >24
crsmin=50 = < = = > = < = = < = > = = =
crsmin=200 = = = = = = = < = = =
c_soil=0 > < < > < > < = < = <>
c_soil=2 < > > > > = < > = = <>
c_lnd=1 = < > > > = = = = = =
c_lnd=10 > < < > = < = = < = < = =
rlam_heat=0.1
<> = > > = > > <> > = > = =
rlam_heat=10
<> = < < <> < <> < = < = =
- -- +++
lconv=KF tur_len=150 tur_len=1000
pat_len=10000 rat_sea=1 rat_sea=60 qc0=0.001
crsmin=50 crsmin=200 c_soil=0 c_soil=2
c_lnd=1 c_lnd=10 rlam_heat=0.1 rlam_heat=10
ctrl
CSPERT fc-ctrl differences +12h04/05/07 06-12 UTC
mm/6h
CSPERT fc-obs differences +12hlconv=KF tur_len=150 tur_len=1000
pat_len=10000 rat_sea=1 rat_sea=60 qc0=0.001
crsmin=50 crsmin=200 c_soil=0 c_soil=2
c_lnd=1 c_lnd=10 rlam_heat=0.1 rlam_heat=10
ctrl
-5 55 20
20 50
-5 -20
-20 -50
mm/6h
CSPERT forecasts +12h
06-12 UTC 04/05/07
Comparisons
SYNOP over MAP area - Nearest grid point
JJA07 SON07
2m T - relationship between error and spread
COSMO-I7 analyses
intra-group distance
2mT Northern Italy
JJA 2007 - 50 days
SON 2007 - 49 days
intra-group distance
tp6h northern Italy
JJA 2007 - 50 days
SON 2007 - 49 days
intra-group distance
t850 COSMO analysis
JJA 2007 - 50 days
SON 2007 - 49 days
intra-group distance
Z500 COSMO analysis
JJA 2007 - 50 days
SON 2007 - 49 days
JJA07 SON07
pert
father
+30h
IT + CH
JJA07 SON07
pert
father
+30h
IT
JJA07
pert
father
+30h
IT
IT + CH
SON07
pert
father
+30h
IT
IT + CH
score evaluation
• 2m t:
• BIA > 0
GME the largest (+), then UKMO, ECMWF, NCEP mixed
P4 (tur_len=1000) the largest (+), for any father
• MAE
GME the smallest, then UKMO, ECMWF, NCEP mixed
P4 (tur_len=1000) the largest, especially for GME at night
JJA07
score evaluation
• 2m t:
• BIA < 0
ECMWF the largest (-), then GME, NCEP and UKMO
P4 (tur_len=1000) the more positive, for any father; P1, P2, P3 similar
• MAE
ECMWF the largest, then GME, NCEP and UKMO
P4 (tur_len=1000) the smallest; P1, P2, P3 similar
SON07