Slide 1ECMWF forecast User Meeting -- Reading, June 2006 Verification of weather parameters Anna Ghelli, ECMWF.

ECMWF forecast User Meeting -- Reading, June 2006 Slide 1

Verification of weather parameters

Anna Ghelli, ECMWF

ECMWF forecast User Meeting -- Reading, June 2006 Slide 2

Overview

Deterministic forecast performance for different weather parameters

Precipitation forecast: scores and their confidence

SYNOP on the GTSPrecipitation analysis

Ensemble prediction System: its performance relative to precipitation and 10m wind

ECMWF forecast User Meeting -- Reading, June 2006 Slide 3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

de

g C

-0.2

-0.1

0

0.1

0.2

0.3

0.4

J

1999A J O J

2000A J O J

2001A J O J

2002A J O J

2003A J O J

2004A J O J

2005A J O J

2006

fc error of indate findate wp step ntot BIAS STDV RMSE SKILL MAE SKILL CORR M-OB M-FC

skil 48h skil 54h skil 60h skil 66h

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

de

g C

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

J

1999A J O J

2000A J O J

2001A J O J

2002A J O J

2003A J O J

2004A J O J

2005A J O J

2006

fc error of indate findate wp step ntot BIAS STDV RMSE SKILL MAE SKILL CORR M-OB M-FC

skil 48h skil 54h skil 60h skil 66h

North America

Europe

2m TemperatureSkill (rmse) for different forecast ranges

Top panel: North America

Bottom panel: Europe

Strong seasonality in the skill with higher values during winters.

2mT skill have reached a plateau. The skills for different timesteps have levelled off. The two past winters were dominated by prolonged periods of cold weather.

ECMWF forecast User Meeting -- Reading, June 2006 Slide 4

1.8

1.9

2

2.1

2.2

2.3

2.4

2.5

2.6

2.7

2.8

m /

s

1.8

1.9

2

2.1

2.2

2.3

2.4

2.5

2.6

2.7

2.8

J

1999A J O J

2000A J O J

2001A J O J

2002A J O J

2003A J O J

2004A J O J

2005A J O J

2006

fc error of indate findate wp step ntot BIAS STDV RMSE SKILL MAE SKILL CORR M-OB M-FC

rmse 48h rmse 54h rmse 60h rmse 66h

-0.05

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

g /

kg

-0.05

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

J

1999A J O J

2000A J O J

2001A J O J

2002A J O J

2003A J O J

2004A J O J

2005A J O J

2006

fc error of indate findate wp step ntot BIAS STDV RMSE SKILL MAE SKILL CORR M-OB M-FC

skil 48h skil 54h skil 60h skil 66hPerformance for different forecast ranges Europe

Europe

Top panel: Specific humidity

Bottom panel: 10m wind speed

Higher skill (MAE) in winter. Last four winters have consistently kept higher level of

performance

RMSE: Changes in the forecasting model have improved the performance of the model in forecasting wind speed.

ECMWF forecast User Meeting -- Reading, June 2006 Slide 5

a b

c d

Observed yes Observed no

Forecast

yes

Forecast

no

1. FREQUENCY BIAS INDEX

caba

FBI

2. TRUE SKILL SCORE

3. HIT RATE

3. FALSE ALARM RATE

db

b

ca

aTSS

db

bF

ECMWF forecast User Meeting -- Reading, June 2006 Slide 6

0

0.2

0.4

0.6

TS

S

0

0.2

0.4

0.6

1992M

1993J S DM

1994J S DM

1995J S DM

1996J S DM

1997J S DM

1998J S DM

1999J S DM

2000J S DM

2001J S DM

2002J S DM

2003J S DM

2004J S DM

2005J S DM

2006

precipitation exceeding 15.0 mm/24h

t + 42 t + 90

15mm/24hEurope

24 hour accumulated precipitation verified against SYNOP on GTS

The forecast is reduced to a yes/no event by selecting thresholds. Confidence intervals have been plotted for each TSS value.

High thresholds have large confidence intervals, important to remember when assessing performance of the system

t+42

t+90

ECMWF forecast User Meeting -- Reading, June 2006 Slide 7

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

FB

I

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

S1993

DM1994

J S DM1995

J S DM1996

J S DM1997

J S DM1998

J S DM1999

J S DM2000

J S DM2001

J S DM2002

J S DM2003

J S DM2004

J S DM2005

J S DM2006

precipitation exceeding 1.0 mm/24h

t + 42 t + 90Europe24 hour accumulated precipitation verified against SYNOP on GTS

The forecasting system over-estimate the number of events for thresholds of 1 mm/24h. A decrease of FBI was observed when in the autumn 1999, when vertical resolution was increased and a new convection scheme was implemented. Further changes in the convection during 2003 have slightly increased FBI values

FBI measures the ratio between the frequency of the forecast events and the frequency of the observed events. FBI>1 over-estimateFBI<1 under-estimate

t+42: solid shadingt+90: dotted shading

ECMWF forecast User Meeting -- Reading, June 2006 Slide 8

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

FB

I

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

S1993

DM1994

J S DM1995

J S DM1996

J S DM1997

J S DM1998

J S DM1999

J S DM2000

J S DM2001

J S DM2002

J S DM2003

J S DM2004

J S DM2005

J S DM2006

precipitation exceeding 15.0 mm/24h

t + 42 t + 90

FBI decreases to values closer to 1 as we increase the threshold, but higher thresholds have larger confidence intervals!

24 hour accumulated precipitation verified against SYNOP on GTS

Europe 15mm/24h t+42: solid shadingt+90: dotted shading

ECMWF forecast User Meeting -- Reading, June 2006 Slide 9

0

0.2

0.4

0.6

TSS

0

0.2

0.4

0.6

1993M

1994M

1995M

1996M

1997M

1998M

1999M

2000M

2001M

2002M

2003M

2004M

2005 2006

precipitation exceeding 15.0 mm/24h

t + 42 t + 90

0.2

0.4

0.6

TSS

0.2

0.4

0.6

1993D

1994D

1995D

1996D

1997D

1998D

1999D

2000D

2001D

2002D

2003D

2004 2005

precipitation exceeding 15.0 mm/24h

t + 42 t + 90

15mm/24h

DJF

MAM

Green: t+42Light blue: t+90

15mm/24h

Stratifying the sample: selecting seasons as rule

ECMWF forecast User Meeting -- Reading, June 2006 Slide 10

40°N

50°N

60°N

70°N

40°W

40°W

20°W

20°W

0°

0°

20°E

20°E 40°E

40°E

1

2

5

10

15

20

25

40

40°N

50°N

60°N

70°N40°W

40°W

20°W

20°W

0°

0°

20°E

20°E 40°E

40°E

60°E

60°E

1

2

5

10

15

20

25

40

Precipitation analysis for Europe

•High density networks in Europe (Member and Co-operating states)•Upscaling (simple box averaging to obtain a areal precipitation value)

ECMWF forecast User Meeting -- Reading, June 2006 Slide 11

5

10

1515

20

50°N

60°N

0°

0°

1

2

5

10

15

20

25

40

Each grid box will contain a certain number of stations. The number of stations will not be constant every day.

The number of stations per grid box indicates how representative the analysis is for the specific grid point.

ECMWF forecast User Meeting -- Reading, June 2006 Slide 12

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

TS

S

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

J2002

S D M2003

J S D M2004

J S D M2005

J S

precipitation exceeding 1.0 mm/24h

t + 42 t + 42

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

TS

S

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

J2002

S D M2003

J S D M2004

J S D M2005

J S

precipitation exceeding 1.0 mm/24h

t + 90 t + 90

Green : proxy

t+42

t+90

1 mm/24h

Which reference system?

Use proxy: the shortest possible forecast verifying at 6UTC. We assume the proxy to be the best forecast. Its TSS values (verification against precipitation analysis) will give us an upper limit for the verification of other forecast ranges.

ECMWF forecast User Meeting -- Reading, June 2006 Slide 13

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

TS

S

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

J2002

S D M2003

J S D M2004

J S D M2005

J S

precipitation exceeding 15.0 mm/24h

t + 90 t + 90

15 mm/24h

t+90

Which reference system?

Use proxy: the shortest possible forecast verifying at 6UTC. We assume the proxy to be the best forecast. Its TSS values (verification against precipitation analysis) will give us an upper limit for the verification of other forecast ranges.

Green : proxy

ECMWF forecast User Meeting -- Reading, June 2006 Slide 14

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

2

FB

I

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

2

J2002

S D M2003

J S D M2004

J S D M2005

J S

precipitation exceeding 15.0 mm/24h

t + 66 t + 66 t + 66 t + 66

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

2

FB

I

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

2

J2002

S D M2003

J S D M2004

J S D M2005

J S

precipitation exceeding 1.0 mm/24h

t + 66 t + 66 t + 66 t + 66

Europe

FBI plotted for two thresholds (1mm/24h, and 15mm/24h)•Verification against SYNOP on the GTS (greenish shading), •Verification against precipitation analysis (blue dotted)

FBI values are higher (lower) in the verification against SYNOP on the GTS (analysis) for lower (higher) thresholds. Forecast range t+66

ECMWF forecast User Meeting -- Reading, June 2006 Slide 15

Europe

TSS (threshold 25mm/24h) plotted for t+66•Verification against SYNOP on the GTS (greenish shading), •Verification against precipitation analysis (blue dotted)

TSS values decrease as we increase forecast range.

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

TS

S

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

J2002

S D M2003

J S D M2004

J S D M2005

J S

precipitation exceeding 25.0 mm/24h

t + 66 t + 66 t + 66 t + 66

ECMWF forecast User Meeting -- Reading, June 2006 Slide 16

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1994A

1995A D A

1996A D A

1997A D A

1998A D A

1999A D A

2000A D A

2001A D A

2002A D A

2003A D A

2004A D A

2005A D

2006

Brier skill score (sample clim) fc step 96 24h-precipitation exceedingProbability forecastverification against obs ( 3-M. moving sample)

1 mm 5 mm 10 mm 20 mm

Timeseries of Brier Skill Score for Europe

The BSS is written as 1- BS/BSref Sample climate is the reference system

BS measures the mean squared difference between forecast and observation in probability space. Equivalent to MSE for deterministic forecast

Forecast vs. observations

Improvements back in Autumn 1999 – High thresholds performance down at the beginning of 2005 probably linked to drier conditions over Europe. Dec-Apr 2006 performance better than previous 3 years.

Increased resolutionT255 T399

ECMWF forecast User Meeting -- Reading, June 2006 Slide 17

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

OCT

2000FEB

2001JUN OCT FEB

2002JUN OCT FEB

2003JUN OCT FEB

2004JUN OCT FEB

2005JUN OCT

2006

Brier skill score (sample clim) fc step 96 24h-precipitation exceedingProbability forecastverification against obs ( 3-M. moving sample)

1 mm 5 mm 10 mm 20 mm

Forecast vs. observations

ECMWF forecast User Meeting -- Reading, June 2006 Slide 18

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

A

1996D A

1997A D A

1998A D A

1999A D A

2000A D A

2001A D A

2002A D A

2003A D A

2004A D A

2005A D

2006

Brier skill score (long term clim) fc step 96 10m wind speed exceedingProbability forecastverification against an ( 3-M. moving sample)

10 m/s 15 m/s

Timeseries of Brier Skill Score for Europe

The BSS is written as 1- BS/BSref Sample climate is the reference system

BS measures the mean squared difference between forecast and observation in probability space. Equivalent to MSE for deterministic forecast C

Forecast vs analysis

Increased resolutionT255 T399

10 m Wind speed

ECMWF forecast User Meeting -- Reading, June 2006 Slide 19

0.2

0.2

0.4

0.4

0.6

0.6

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

ob

s fr

equ

ency

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

forecast probability

67515918

3400

2853

2555

2301

2269

2322

2443

2777

5345

sample clim = 0.26 BS = 0.157 SSBS = 0.19 bias(ctr) = 0.2820041001-20050430 STEP 96 24-hour precipitation gt 1.0

0

0.1

0 0.2 0.4 0.6 0.8 1

rel FC distribution

sample clim

0.2

0.2

0.4

0.4

0.6

0.6

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

ob

s f

req

ue

nc

y

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

forecast probability

167569578

3778

2385

1665

1263

1047

841

643

519

459

sample clim = 0.11 BS = 0.079 SSBS = 0.21 bias(ctr) = 0.2820041001-20050430 STEP 96 24-hour precipitation gt 5.0

0

0.1

0.2

0.3

0.4

0.5

0 0.2 0.4 0.6 0.8 1

rel FC distribution

sample clim

EuropeRainy season (24h accum.): October to April Forecast range: t+96Verification against SYNOP on GTS

2005-20061mm/24h BS=0.079

2004-20051mm/24h BS=0.157

Consistent picture for the two seasons

0.2

0.2

0.4

0.4

0.6

0.6

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

ob

s fr

equ

ency

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

forecast probability

70746333

3457

2786

2431

2243

2100

2014

2163

2388

4970

sample clim = 0.13 BS = 0.071 SSBS = 0.39 bias(ctr) = 0.1020051001-20060501 STEP 96 24-hour precipitation gt 1.0

0

0.1

0 0.2 0.4 0.6 0.8 1

rel FC distribution

sample clim

0.2

0.2

0.4

0.4

0.6

0.6

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

ob

s f

req

ue

nc

y

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

forecast probability

172069045

3300

2092

1566

1206

982

784

627

541

610

sample clim = 0.06 BS = 0.036 SSBS = 0.33 bias(ctr) = 0.1020051001-20060501 STEP 96 24-hour precipitation gt 5.0

0

0.1

0.2

0 0.2 0.4 0.6 0.8 1

rel FC distribution

sample clim

ECMWF forecast User Meeting -- Reading, June 2006 Slide 20

0.2

0.2

0.4

0.4

0.6

0.6

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

ob

s fr

equ

ency

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

forecast probability

259208057

2096

1070

649410

268

183137

7965

sample clim = 0.05 BS = 0.040 SSBS = 0.16 bias(ctr) = 0.2820041001-20050430 STEP 96 24-hour precipitation gt 10.0

0

0.2

0.4

0.6

0.8

0 0.2 0.4 0.6 0.8 1

rel FC distribution

sample clim

0.2

0.2

0.4

0.4

0.6

0.6

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

ob

s f

req

ue

nc

y

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

forecast probability

346083414

477199

95

60

33

2220

4

2

sample clim = 0.01 BS = 0.012 SSBS = 0.12 bias(ctr) = 0.2820041001-20050430 STEP 96 24-hour precipitation gt 20.0

0

0.2

0.4

0.6

0.8

1

0 0.2 0.4 0.6 0.8 1

rel FC distribution

sample clim

EuropeRainy season (24h accum.): October to AprilForecast range: t+96Verification against SYNOP on GTS

2004-200510 mm/24h BS=0.04

2005-200610 mm/24h BS=0.019

Consistent picture for the two seasons. Higher thresholds better reliability

0.2

0.2

0.4

0.4

0.6

0.6

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

ob

s fr

equ

ency

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

forecast probability

255427522

1984

1007

612

423296

224

153

100

96

sample clim = 0.03 BS = 0.019 SSBS = 0.25 bias(ctr) = 0.1020051001-20060501 STEP 96 24-hour precipitation gt 10.0

0

0.1

0.2

0.3

0.4

0 0.2 0.4 0.6 0.8 1

rel FC distribution

sample clim

0.2

0.2

0.4

0.4

0.6

0.6

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

ob

s f

req

ue

nc

y

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

forecast probability

336453303

532

214

116

60

37

24

10

13

5

sample clim = 0.01 BS = 0.007 SSBS = 0.12 bias(ctr) = 0.1020051001-20060501 STEP 96 24-hour precipitation gt 20.0

0

0.1

0.2

0.3

0.4

0.5

0 0.2 0.4 0.6 0.8 1

rel FC distribution

sample clim

ECMWF forecast User Meeting -- Reading, June 2006 Slide 21

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

hit

rat

e

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

false alarm rate

20041001-20050430 24-hour precipitation gt 1.0

0

2000

4000

6000

8000

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

STEP 96

0

2000

4000

6000

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

STEP 120

t+ 96 A=0.864

t+120 A=0.830

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

hit

ra

te

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

false alarm rate

20041001-20050430 24-hour precipitation gt 5.0

0

5000

0.100 105

0.150 105

0.200 105

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

STEP 96

0

5000

0.100 105

0.150 105

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

STEP 120

t+ 96 A=0.864

t+120 A=0.832

EuropeRainy season (24h accum.): October to AprilForecast range: t+96Verification against SYNOP on GTS

2004-20055 mm/24h

2005-20065 mm/24h

Consistent picture for the two seasons.

Full symbol: T511/799Shape: T255/T399

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

hit

rat

e

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

false alarm rate

20051001-20060501 24-hour precipitation gt 1.0

0

2000

4000

6000

8000

0.100 105

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

STEP 72

0

2000

4000

6000

8000

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

STEP 96

0

2000

4000

6000

8000

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

STEP 120

t+ 72 A=0.899

t+ 96 A=0.877

t+120 A=0.849

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

hit

ra

te

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

false alarm rate

20051001-20060501 24-hour precipitation gt 5.0

0

5000

0.100 105

0.150 105

0.200 105

0.250 105

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

STEP 72

0

5000

0.100 105

0.150 105

0.200 105

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

STEP 96

0

5000

0.100 105

0.150 105

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

STEP 120

t+ 72 A=0.909

t+ 96 A=0.889

t+120 A=0.860

t+96 : A=0.864

t+120: A=0.832

t+96 : A=0.889

t+120: A=0.860

ECMWF forecast User Meeting -- Reading, June 2006 Slide 22

0.5

0.55

0.6

0.65

0.7

0.75

0.8

0.85

0.9

0.95

1

RO

CA

0.5

0.55

0.6

0.65

0.7

0.75

0.8

0.85

0.9

0.95

1

J1996

S D M1997

J S D M1998

J S D M1999

J S D M2000

J S D M2001

J S D M2002

J S D M2003

J S D M2004

J S D M2005

J S D M2006

24 hour total precipitation verified against observations t+ 96

threshold = 1 threshold = 5 threshold = 10 threshold = 20

Increased resolution

EuropeROC Area Verification against SYNOP on GTS for t+96

T255 T399

ECMWF forecast User Meeting -- Reading, June 2006 Slide 23

Conclusion 2m Temperature: the skill over

Europe has reached a plateau. Over

N.America winters are more skilful

than summers.

Specific humidity shows consistent

skills. Winters more skilful than

summers

Wind: Recent changes in the model

have resulted in a decrease of the

RMSE over Europe.

TCC: New cloud scheme was

introduced in April 2005. Forecasts

of certain cloud types have

improved.

Importance of confidence intervals

Precipitation forecast

improvements are slow, but evident.

FBI indicates over-estimation of

small threshold events

verification against precipitation

analysis shows a better picture.

Precipitation analysis can be used

for verification in a delayed mode. It

can be used to define the

performance of a reference system.

ECMWF forecast User Meeting -- Reading, June 2006 Slide 24

Brier skill score and ROC area: In

recent year the system has

maintained its good performance.

The increase in horizontal

resolution has had a small but

positive impact on both

precipitation and wind speed.

Reliability diagrams for the last two

rainy seasons show a skilful

system, with an evident decrease of

Brier Score for the 2005-2006

season.

Conclusion