Slide 1 Forecast Products User Meeting June 2006 Slide 1 ECMWF medium-range forecasts and products David Richardson [email protected] Met Ops

Forecast Products User Meeting June 2006

Slide 1

Slide 1

ECMWF medium-range forecasts and products

David Richardson

[email protected]

Met Ops Section, Jean Bidlot, Mark Rodwell, Roberto Buizza


Slide 2

Slide 2

Atmospheric data assimilation

4-dimensional variational data assimilation at 25/80 km resolution and 91 levels

Atmospheric global forecasts

Forecast to ten days from 00 and 12 UTC at 25 km resolution and 91 levels

50-member ensemble to ten days from 00 and 12 UTC at 50 km resolution, 62 levels

Ocean wave forecasts

Global forecast to ten days from 00 and 12 UTC at 50 km resolution

European waters forecast to five days from 00 and 12 UTC at 25 km resolution

Monthly forecasts: Atmosphere-ocean coupled model

Global ensemble to one month (weekly) 125 km, 62 level atmosphere; 1°, 29 level ocean

Seasonal forecasts: Atmosphere-ocean coupled model

Global ensemble to six months (monthly): 200 km, 40 level atmosphere; 1°, 29 level ocean

ECMWF Operational Forecast System


Slide 3

Slide 3

Changes to the forecasting system 2005 – 2006

5 Apr 2005 Cy29r1New moist boundary layer scheme, wavelet

Jb, adaptive bias correction

28 Jun 2005 Cy29r2Wide range of changes including rain-

affected SSM/I radiances

1 Feb 2006 Cy30r1High resolution forecasting system

Harmonisation of data availability times

14 Mar 2006 4d-Var for all BC suite analyses


Slide 4

Slide 4

Medium-range forecast performance


Slide 5

Slide 5

Time series Acc=0.6 N Hemisphere


Slide 6

Slide 6

Time series Acc=0.6 S Hemisphere


Slide 7

Slide 7

Time series Acc=0.6 Europe


Slide 8

Slide 8

Spring 2006 (MAM) over Europe


Slide 9

Slide 9

Comparison with other centres


Slide 10

Slide 10

Time series Z500 N Hemisphere


Slide 11

Slide 11

Time series Z500 N Hemisphere


Slide 12

Slide 12

Other centres Z500 N hem 2006 v 2005


Slide 13

Slide 13

Other centres Z500 Europe 2006 v 2005


Slide 14

Slide 14

Consistency of high resolution forecast


Slide 15

Slide 15

Forecast consistency


Slide 16

Slide 16



Slide 17

Slide 17



Slide 18

Slide 18

Weather parameters – this afternoon


Slide 19

Slide 19

Waves


Slide 20

Slide 20

Wave analysis – scatter index


Slide 21

Slide 21

Wave forecast – scatter index


Slide 22

Slide 22

Waves – other centres


Slide 23

Slide 23

Ensemble Prediction System- this afternoon (Roberto Buizza)


Slide 24

Slide 24

Brier Skill Score for Europe for 96h EPS forecast of precipitation exceeding various thresholds


Slide 25

Slide 25

12UTCN.hem Lat 20.0 to 90.0 Lon -180.0 to 180.0

500hPa GeopotentialTime series curves

MAR2005

APR MAY JUN JUL4

6

8

10

12

14

16

18

20

22

24

ENSTD T+24 Mean forecast

ENSTD T+48 Mean forecast

ENSMN T+24 rmse forecast

ENSMN T+48 rmse forecast

A diagnostics for the EPS spread (standard deviation of Gaussian) is now available on the internal web site. Problems similar to the one in May 2005 could be identified and corrected

EPS spread/skill


Slide 26

Slide 26

0

3

6

9

12

15

18

ST

D

0

3

6

9

12

15

18

OCT

2005

12 14 16 18 20 22 24 26 28 30 1NOV

3 5 7 9 11 13 15 17 19 21 23 25 27 29 1DEC

3 5 7 9

initialStandard deviation of Gaussian -- exp: 0001

N.Hemisphere S.Hemisphere

Diagnostic tool:

Standard deviation of gaussian

problematic forecast


Slide 27

Slide 27

0 1 2 3 4 5 6 7 8 9 10

Forecast Day

516

528

540

552

564

576dam

GEOPOTENTIAL 500hPa - Probability for 2.5 dam intervals Range: 72dam

Exp 0001

0 1 2 3 4 5 6 7 8 9 10

Forecast Day

0

2

4

6

8

10

12

14mm

Ensemble members of TOTAL PRECIPITATION - Accum. rate mm/12h

0 1 2 3 4 5 6 7 8 9 10

Forecast Day

-5

0

5

10deg

TEMPERATURE 850hPa - Probability for 1.0 deg intervals Range: 20deg

DATE: 20051125 READING LAT: 51.45 LONG: -0.95ECMWF ENSEMBLE FORECASTS FOR: ECMWF 0.5 - 10 % 10 - 30 % 30 - 50 % 50 - 100 %

Oper T255 Anal EMem

Reading plume for 20051125 00UTC run


Slide 28

Slide 28

Future developments


Slide 29

Slide 29

VAREPS (2006)

Plan to extend forecast range of EPS to day 15

Two additional fixed-resolution calibration forecasts will be run

Current products to day 10 will be unaffected

T799 deterministic will remain as now (days 1-10 only)

# fc day 0-10 day 10-15

VAR-CF 1 TL399 TL255

VAR-PF 50 TL399 TL255

CL-HRES 1 TL399

CL-LRES 1 TL255


Slide 30

Slide 30

VAREPS

Initial implementation No changes required to access products to day 10

Individual EPS members archived to day 15

Straightforward to access days 10-15 (note resolution change at day 10!)

No change required to MARS requests or EMOSLIB (current version: cy281)

First post-processed products Ensemble mean and spread (standard deviation) to day 15

• Calculate on model grid (was 1.5 degree lat-lon)

Probabilities extended to day 15 (same thresholds as now)

Longer term EPS products to day 15 will be developed

15-day VAREPS will be combined with the monthly system (coupling with ocean model)


Slide 31

Slide 31

Products


Slide 32

Slide 32

ECMWF Forecast Products

Global forecasts (deterministic)

Fields (Model variables, GRIB)

Time series (BUFR Meteograms)

Direct Model Output (DMO) only

EPS

Fields (GRIB) + Time Series (BUFR EPSGRAMs)

Post-processed products (Clusters, Tubes, EFI)

Monthly and seasonal forecasts: atmosphere-ocean coupled model

Anomalies (Field + Time series)


Slide 33

Slide 33

EPS forecasts: timeseries (EPSgram)

EPSgram for Reading

Base Tue 14/06/06 00UTC


Slide 34

Slide 34

EPS forecasts (post-processed products)

Extreme forecast index for 2m temperature

Base Sat 10/06/06 00UTC, Valid Wed 14/06/06 12UTC


Slide 35

Slide 35

Medium-range forecast products

Global Extreme Forecast Index

Indicates areas where today’s EPS has signal of extreme relative to model climate distribution

New EFI climatology introduced with high-resolution system

Verification against observations has started

More this afternoon

Tropical cyclone tracks

More realistic structure in higher resolution system

Indication of reduction in position and intensity errors

2006 Atlantic season started (Alberto) …

EPSgrams – combine deterministic and EPS forecasts


Slide 36

Slide 36

Tropical Storm Alberto


Slide 37

Slide 37

948

T511 L60 20050828 00UTC ECMWF FC t+36 VT: 20050829 12UTC Surf: MSL

942


963


944


Forecasts of Katrina for Monday 29 Aug. 12 UTCobserved pressure 922 hPa

T511 T799

t+36 t+36

t+72 t+72


Slide 38

Slide 38

0

200

400

600

800

km

0 24 48 72 96 120

forecast step

0

40

80

120

160

sa

mp

le s

ize

0 24 48 72 96 120

forecast step

0001 vs. 002817 Jul until 18 Aug

years: 2005 oper

Direct Position Error

281

Position error in high resolution T799 is smaller at all forecast steps compared to the T511.

T511L60

T799L91

Deterministic Scores – T511 vs T799


Slide 39

Slide 39

-20

0

20

40

hP

a

0 24 48 72 96 120

forecast step

0

40

80

120

160

sa

mp

le s

ize

0 24 48 72 96 120

forecast step

0001 vs. 002817 Jul until 18 Aug

years: 2005 oper

Core Pressure Error

281

The intensity of tropical cyclones is more realistic in the T799.

T511L60

T799L91

Deterministic Scores – T511 vs T799


Slide 40

Slide 40

Combined prediction system developmentMark Rodwell


Slide 41

Slide 41

Highly useful product but…• What should I believe?

• At D+2• At D+5

CPS - Motivation

Thu 1 Fri 2 Sat 3 Sun 4 Mon 5 Tue 6 Wed 7 Thu 8 Fri 9 Sat 10June 2006

22

24

26

28

30

32

max

min

75%median25% T799 OPS T399 CTRL

2m Temperature reduced to station height (°C) 28m (T799) 26m (T399)

Magics++ 1.1.0

0

2

4

6

8

1010m Wind Speed (m/s)

0

2

4

6

8

10

12

14

16Total Precipitation (mm/6h)

0

2

4

6

8

EPS MeteogramMumbai (26m) 19.1°N 72.9°EDeterministic Forecasts and EPS Distribution Thursday 1 June 2006 00 UTC

Total Cloud Cover (okta)

Thu 1 Fri 2 Sat 3 Sun 4 Mon 5 Tue 6 Wed 7 Thu 8 Fri 9 Sat 10June 2006

22

24

26

28

30

32

max

min

75%median25% T799 OPS T399 CTRL

2m Temperature reduced to station height (°C) 28m (T799) 26m (T399)

Magics++ 1.1.0

0

2

4

6

8

1010m Wind Speed (m/s)

0

2

4

6

8

10

12

14

16Total Precipitation (mm/6h)

0

2

4

6

8

EPS MeteogramMumbai (26m) 19.1°N 72.9°EDeterministic Forecasts and EPS Distribution Thursday 1 June 2006 00 UTC

Total Cloud Cover (okta)

Mumbai


Slide 42

Slide 42

CPS – Concept

Combining a 10-member ensemble of equally likely members (orange squares) with a single more accurate forecast (yellow rectangle)


Slide 43

Slide 43

1 2 3 4 5 6 7 8 9 10Lead-time (days)

0.0

0.1

0.2

0.3

0.4

Bri

er S

kill

Sco

reBrier Skill Score for the event that P

p > x mm day-1

x=1

CPS

EPS

x=5

CPS

EPS

x=10

CPS

EPS

CPS > EPS

5% Significance

CPS – Results

2001-2005


Slide 44

Slide 44

1 2 3 4 5 6 7 8 9 10Lead-time (days)

0

5

10

15

20

Wei

gh

t (n

um

ber

of

EP

S m

emb

ers)

Weight of Deterministic Forecast within Combined Prediction System

Pp > 1 mm day-1

Pp > 5 mm day-1

Pp > 10 mm day-1

CPS – Weights

2001-2005


Slide 45

Slide 45

Thu 1 Fri 2 Sat 3 Sun 4 Mon 5 Tue 6 Wed 7 Thu 8 Fri 9 Sat 100

20

40

60

Total Precipitation (mm/day) Combined Probability DistributionsOptimized for the critical event the precipitation exceeds 1 mmday-1

Prob > 1 mm/day

Brier Skill Score

100

33

100

34

100

31

100

26

100

19

100

12

100

7

94

2

98

0

94

0

CPS – Meteogram

Mumbai


Slide 46

Slide 46

Thu 1 Fri 2 Sat 3 Sun 4 Mon 5 Tue 6 Wed 7 Thu 8 Fri 9 Sat 100

20

40

60

Total Precipitation (mm/day) Combined Probability DistributionsOptimized for the critical event the precipitation exceeds 10 mmday-1

Prob > 10 mm/day

Brier Skill Score

100

23

98

19

91

14

70

9

90

5

81

1

45

0

27

0

27

0

43

0

CPS – Meteogram

Mumbai


Slide 47

Slide 47

ECMWF visits to Member and Co-operating StatesAutumn 2005 – spring 2006


Slide 48

Slide 48

Member State visits – comments (1)

Generally very happy with ECMWF products

High resolution system introduced halfway through this round of visits

How is consistency of T799 model?

• No significant change from previous model

ECMWF boundary conditions for limited area models

Some requests for hourly updates (currently 3-hourly)

• Discussed at TAC BC subgroup and TAC 2005; agreed not to change

High-impact weather early warnings (medium-range)

We are still interested in feedback and examples, especially of how ECMWF products may be useful


Slide 49

Slide 49

Member State visits – comments (2)

Big range of use for the EPS (internal; specific users; general public)

Clusters –

• some use ECWF clusters, some do own clustering

• ECMWF clustering needs updating

• Request for extra area

EPSgrams – noted requests for additional parameters (H/M/L cloud, wind)

How should the EPS be used?

How can EPS be used together with T799?

Increasing interest in monthly forecasts

Interest in potential of seasonal forecasts

Documents

Slide 1 Forecast Products User Meeting June 2006 Slide 1 ECMWF medium-range forecasts and products David Richardson [email protected] Met Ops