Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 1

Operational Seasonal Forecast Systems:

a view from ECMWF

Tim Stockdale

The team: Franco Molteni, Magdalena Balmaseda, Kristian Mogensen,Frederic Vitart, Laura Ferranti

European Centre for Medium-Range Weather Forecasts

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 2

Outline

• Operational seasonal systems at ECMWF System 3 - configuration System 3 – products System 3 – skill measures

• EUROSIP ECMWF, Met Office and Météo-France multi-model system

• Some (relevant) issues in seasonal prediction Estimating skill and model improvement Cost effective systems Multi-model systems, data sharing policy

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 3

Sources of seasonal predictability

KNOWN TO BE IMPORTANT:o El Nino variability - biggest single signalo Other tropical ocean SST - important, but multifariouso Climate change - especially important in mid-latitudeso Local land surface conditions - e.g. soil moisture in spring

OTHER FACTORS:o Volcanic eruptions - definitely important for large eventso Mid-latitude ocean temperatures - still somewhat controversialo Remote soil moisture/ snow cover - not well establishedo Sea ice anomalies - local effects, but remote?o Dynamic memory of atmosphere - most likely on 1-2 monthso Stratospheric influences - solar cycle, QBO, ozone, …

Unknown or Unexpected - ???

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 4

ECMWF operational seasonal forecasts

• Real time forecasts since 1997 “System 1” initially made public as “experimental” in Dec 1997 System 2 started running in August 2001, released in early 2002 System 3 started running in Sept 2006, operational in March 2007

• Burst mode ensemble forecast Initial conditions are valid for 0Z on the 1st of a month Forecast is created typically on the 11th/12th (SST data is delayed up to 11

days) Forecast and product release date is 12Z on the 15th.

• Range of operational products Moderately extensive set of graphical products on web Raw data in MARS Formal dissemination of real time forecast data

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 5

ECMWF System 3 – the model

• IFS (atmosphere) TL159L62 Cy31r1, 1.125 deg grid for physics (operational in Sep 2006) Full set of singular vectors from EPS system to perturb atmosphere initial

conditions (more sophisticated than needed …) Ocean currents coupled to atmosphere boundary layer calculations

• HOPE (ocean) Global ocean model, 1x1 mid-latitude resolution, 0.3 near equator A lot of work in developing the OI ocean analyses, including analysis of

salinity, multivariate bias corrections and use of altimetry.

• Coupling Fully coupled, no flux adjustments, except no physical model of sea-ice

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 6

System 3 configuration

• Real time forecasts: 41 member ensemble forecast to 7 months SST and atmos. perturbations added to each member

11 member ensemble forecast to 13 months Designed to give an ‘outlook’ for ENSO Only once per quarter (Feb, May, Aug and Nov starts) November starts are actually 14 months (to year end)

• Back integrations from 1981-2005 (25 years) 11 member ensemble every month 5 members to 13 months once per quarter

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 7

APR2010

MAY JUN JUL AUG SEP OCT NOV DEC JAN2011

FEB MAR APR MAY JUN

-2

-1

0

1A

nom

aly

(deg

C)

-2

-1

0

1

Monthly mean anomalies relative to NCEP adjusted OIv2 1971-2000 climatologyECMWF forecast from 1 Oct 2010

NINO3.4 SST anomaly plume

Forecast issue date: 15 Oct 2010

System 3

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 8

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 9

Other operational plots for DJF 2010/11

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 10

Tropical storm forecasts

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 11

Rms error of forecasts has been systematically reduced (solid lines) ….

Performance – SST and ENSO

0 1 2 3 4 5 6Forecast time (months)

0.4

0.5

0.6

0.7

0.8

0.9

1

An

om

aly

co

rre

latio

n

wrt NCEP adjusted OIv2 1971-2000 climatology

NINO3.4 SST anomaly correlation

0 1 2 3 4 5 6Forecast time (months)

0

0.2

0.4

0.6

0.8

1

Rm

s e

rro

r (d

eg

C)

Ensemble sizes are 5 (0001), 5 (0001) and 5 (0001)192 start dates from 19870101 to 20021201

NINO3.4 SST rms errors

Fcast S3 Fcast S2 Fcast S1 Persistence

MAGICS 6.11 cressida - net Tue Apr 17 16:45:18 2007

.. but ensemble spread (dashed lines) is still substantially less than actual forecast error.

0 1 2 3 4 5 6Forecast time (months)

0.4

0.5

0.6

0.7

0.8

0.9

1

An

om

aly

co

rre

latio

n

wrt NCEP adjusted OIv2 1971-2000 climatology

NINO3.4 SST anomaly correlation

0 1 2 3 4 5 6Forecast time (months)

0

0.2

0.4

0.6

0.8

1

Rm

s e

rro

r (d

eg

C)

Ensemble sizes are 5 (0001), 5 (0001) and 5 (0001)192 start dates from 19870101 to 20021201

NINO3.4 SST rms errors

Fcast S3 Fcast S2 Fcast S1 Persistence Ensemble sd

MAGICS 6.11 cressida - net Tue Apr 17 16:41:30 2007

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 12

More recent SST forecasts are better ....

0 1 2 3 4 5 6 7Forecast time (months)

0.4

0.5

0.6

0.7

0.8

0.9

1

An

om

aly

co

rre

latio

n

wrt NCEP adjusted OIv2 1971-2000 climatology

NINO3.4 SST anomaly correlation

0 1 2 3 4 5 6 7Forecast time (months)

0

0.2

0.4

0.6

0.8

1

Rm

s e

rro

r (d

eg

C)

Ensemble size is 11156 start dates from 19810101 to 19931201

NINO3.4 SST rms errors

Fcast S3 Persistence Ensemble sd

MAGICS 6.12n cressida - net Mon Mar 9 11:58:09 2009

0 1 2 3 4 5 6 7Forecast time (months)

0.4

0.5

0.6

0.7

0.8

0.9

1

An

om

aly

co

rre

latio

n

wrt NCEP adjusted OIv2 1971-2000 climatology

NINO3.4 SST anomaly correlation

0 1 2 3 4 5 6 7Forecast time (months)

0

0.2

0.4

0.6

0.8

1

Rm

s e

rro

r (d

eg

C)

Ensemble size is 11168 start dates from 19940101 to 20071201

NINO3.4 SST rms errors

Fcast S3 Persistence Ensemble sd

MAGICS 6.12n cressida - net Mon Mar 9 11:59:56 2009

1981-1993 1994-2007

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 13

At longer leads, model spread starts to catch up

0 1 2 3 4 5 6 7 8 9 10 11 12 13Forecast time (months)

0.4

0.5

0.6

0.7

0.8

0.9

1

Ano

mal

y co

rrel

atio

n

wrt NCEP adjusted OIv2 1971-2000 climatology

NINO3.4 SST anomaly correlation

0 1 2 3 4 5 6 7 8 9 10 11 12 13Forecast time (months)

0

0.2

0.4

0.6

0.8

1

Rm

s er

ror

(deg

C)

Ensemble size is 5: predictability limit for finite sample324 start dates from 19810101 to 20071201

NINO3.4 SST rms errors

Fc f3yi/m3 Persistence

MAGICS 6.12n cressida - net Thu Jul 16 17:10:57 2009

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 14

How good are the forecasts?

Hindcast period 1981-2003 with start in November and averaging period 2 to 4Near-surface temperatureAnomaly Correlation Coefficient for CodOecmfE0001S003M001 with 11 ensemble members

-1 -0.9 -0.8 -0.7 -0.6 -0.4 -0.2 0.2 0.4 0.6 0.7 0.8 0.9 1

Hindcast period 1981-2003 with start in November and averaging period 2 to 4Near-surface temperaturePerfect-model Anomaly Correlation Coefficient for CodOecmfE0001S003M001 with 11 ensemble members

-1 -0.9 -0.8 -0.7 -0.6 -0.4 -0.2 0.2 0.4 0.6 0.7 0.8 0.9 1

Temperature: actual forecasts Temperature: perfect model

Deterministic skill: DJF ACC

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 15

How good are the forecasts?

Precip: actual forecasts Precip: perfect model

Deterministic skill: DJF ACC

Hindcast period 1981-2003 with start in November and averaging period 2 to 4PrecipitationAnomaly Correlation Coefficient for CodOecmfE0001S003M001 with 11 ensemble members

-1 -0.9 -0.8 -0.7 -0.6 -0.4 -0.2 0.2 0.4 0.6 0.7 0.8 0.9 1

Hindcast period 1981-2003 with start in November and averaging period 2 to 4PrecipitationPerfect-model Anomaly Correlation Coefficient for CodOecmfE0001S003M001 with 11 ensemble members

-1 -0.9 -0.8 -0.7 -0.6 -0.4 -0.2 0.2 0.4 0.6 0.7 0.8 0.9 1

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 16

How good are the forecasts?

Tropical precip < lower tercile, JJA NH extratrop temp > upper tercile, DJF

Probabilistic skill: Reliability diagrams

Threshold estimated with a kernel method for the PDFHindcast period 1981-2003 with start in May and averaging period 2 to 4Precipitation anomalies below the lower tercile over tropical band (land and sea points)Reliability diagram for CodOecmfE0001S003M001 with 11 ensemble members

0.0

0.2

0.4

0.6

0.8

1.0

Obs

erve

d fr

eque

ncy

0.0 0.2 0.4 0.6 0.8 1.0Forecast probability

1.0000

0.1000

0.0100

0.0010

0.0001

Sha

rp/R

el/R

es

0.2 0.4 0.6 0.8 1.0Forecast probability

-0.025

-0.020

-0.015

-0.010

-0.005

0.000

0.005

0.010

0.015

0.020

0.025

Brie

r sc

ore

Sharp Rel Res BS

( 0.316, 0.296)ROC skill score: 0.306 ( 0.257, 0.351)Sharpness: 0.104 ( 0.098, 0.110)Resolution skill score: 0.073 ( 0.051, 0.097)Reliability skill score: 0.916 ( 0.894, 0.933)Brier skill score: -0.012 (-0.054, 0.027)Skill scores and 95% conf. intervals ( 1000 samples)

Threshold estimated with a kernel method for the PDFHindcast period 1981-2003 with start in November and averaging period 2 to 4Near-surface temperature anomalies above the upper tercile over Northern extratropics (land and sea points)Reliability diagram for CodOecmfE0001S003M001 with 11 ensemble members

0.0

0.2

0.4

0.6

0.8

1.0

Obs

erve

d fr

eque

ncy

0.0 0.2 0.4 0.6 0.8 1.0Forecast probability

1.0000

0.1000

0.0100

0.0010

0.0001

Sha

rp/R

el/R

es

0.2 0.4 0.6 0.8 1.0Forecast probability

-0.025

-0.020

-0.015

-0.010

-0.005

0.000

0.005

0.010

0.015

0.020

0.025

Brie

r sc

ore

Sharp Rel Res BS

( 0.283, 0.258)ROC skill score: 0.271 ( 0.198, 0.336)Sharpness: 0.077 ( 0.073, 0.081)Resolution skill score: 0.061 ( 0.034, 0.091)Reliability skill score: 0.964 ( 0.936, 0.980)Brier skill score: 0.024 (-0.028, 0.072)Skill scores and 95% conf. intervals ( 1000 samples)

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 17

How good are the forecasts?

Europe: Temp > upper tercile, DJF

Probabilistic skill: Reliability diagrams

Threshold estimated with a kernel method for the PDFHindcast period 1981-2003 with start in November and averaging period 2 to 4Near-surface temperature anomalies above the upper tercile over Europe (land and sea points)Reliability diagram for CodOecmfE0001S003M001 with 11 ensemble members

0.0

0.2

0.4

0.6

0.8

1.0

Obs

erve

d fr

eque

ncy

0.0 0.2 0.4 0.6 0.8 1.0Forecast probability

1.0000

0.1000

0.0100

0.0010

0.0001

Sha

rp/R

el/R

es

0.2 0.4 0.6 0.8 1.0Forecast probability

-0.025

-0.020

-0.015

-0.010

-0.005

0.000

0.005

0.010

0.015

0.020

0.025

Brie

r sc

ore

Sharp Rel Res BS

( 0.113, 0.041)ROC skill score: 0.077 (-0.076, 0.232)Sharpness: 0.073 ( 0.065, 0.080)Resolution skill score: 0.013 ( 0.003, 0.058)Reliability skill score: 0.895 ( 0.771, 0.953)Brier skill score: -0.092 (-0.217, 0.005)Skill scores and 95% conf. intervals ( 1000 samples)

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 19

EUROSIP

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 20

Single model Multi- model

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 21

Reliability diagrams (T2m > 0)1-month lead, start date May, 1980 - 2001

DEMETER: multi-model vs single-model

0.0390.8990.141

0.0390.8990.140

0.0950.9260.169

-0.001 0.877 0.123

0.0650.9180.147

-0.064 0.838 0.099

0.0470.8930.153

0.2040.9900.213

multi-model

Hagedorn et al. (2005)

BSSRel-ScRes-Sc

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 22

Some (relevant) issues in seasonal prediction

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 23

60°S60°S

30°S 30°S

0°0°

30°N 30°N

60°N60°N

135°W

135°W 90°W

90°W 45°W

45°W 0°

0° 45°E

45°E 90°E

90°E 135°E

135°E

Global rms acc: 0.612 NH:0.331 TR:0.783 SH:0.389Z500 Anom. correlation S3(11)-ERA Int 1989-2008 DJF

ACC

-0.9

-0.8

-0.6

-0.4

-0.2

0.2

0.4

0.6

0.8

0.9

60°S60°S

30°S 30°S

0°0°

30°N 30°N

60°N60°N

135°W

135°W 90°W

90°W 45°W

45°W 0°

0° 45°E

45°E 90°E

90°E 135°E

135°E

Global rms acc: 0.613 NH:0.294 TR:0.793 SH:0.387Z500 Anom. correlation ffcf(11)-ERA Int 1989-2008 DJF

ACC

-0.9

-0.8

-0.6

-0.4

-0.2

0.2

0.4

0.6

0.8

0.9

60°S60°S

30°S 30°S

0°0°

30°N 30°N

60°N60°N

135°W

135°W 90°W

90°W 45°W

45°W 0°

0° 45°E

45°E 90°E

90°E 135°E

135°E

sigma: 0.343 mean: -0.0154Fisher z transform diff S3(11)-ffcf(11) 1989-2008 DJF

z

-1

-0.8

-0.6

-0.4

-0.2

0.2

0.4

0.6

0.8

1

Tentative results from ECMWF S4

System 3

Cy36r4 - T159L62

(11 members, 20 years)

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 24

60°S60°S

30°S 30°S

0°0°

30°N 30°N

60°N60°N

135°W

135°W 90°W

90°W 45°W

45°W 0°

0° 45°E

45°E 90°E

90°E 135°E

135°E

Global rms acc: 0.624 NH:0.346 TR:0.804 SH:0.371Z500 Anom. correlation ffky(6)-ERA Int 1989-2008 DJF

ACC

-0.9

-0.8

-0.6

-0.4

0.2

0.4

0.6

0.8

0.9

60°S60°S

30°S 30°S

0°0°

30°N 30°N

60°N60°N

135°W

135°W 90°W

90°W 45°W

45°W 0°

0° 45°E

45°E 90°E

90°E 135°E

135°E

Global rms acc: 0.605 NH:0.288 TR:0.795 SH:0.332Z500 Anom. correlation ffn5(6)-ERA Int 1989-2008 DJF

ACC

-0.9

-0.8

-0.6

-0.4

-0.2

0.2

0.4

0.6

0.8

0.9

60°S60°S

30°S 30°S

0°0°

30°N 30°N

60°N60°N

135°W

135°W 90°W

90°W 45°W

45°W 0°

0° 45°E

45°E 90°E

90°E 135°E

135°E

sigma: 0.343 mean: 0.0512Fisher z transform diff ffky(6)-ffn5(6) 1989-2008 DJF

z

-1

-0.8

-0.6

-0.4

-0.2

0.2

0.4

0.6

0.8

1

Alternate stochastic physics

0.346 vs 0.294

A real improvement, now

scoring better than S3

60°S60°S

30°S 30°S

0°0°

30°N 30°N

60°N60°N

135°W

135°W 90°W

90°W 45°W

45°W 0°

0° 45°E

45°E 90°E

90°E 135°E

135°E

Global rms acc: 0.629 NH:0.342 TR:0.819 SH:0.338Z500 Anom. correlation fg4m(5)-ERA Int 1989-2008 DJF

ACC

-0.9

-0.8

-0.6

-0.4

-0.2

0.2

0.4

0.6

0.8

0.9

60°S60°S

30°S 30°S

0°0°

30°N 30°N

60°N60°N

135°W

135°W 90°W

90°W 45°W

45°W 0°

0° 45°E

45°E 90°E

90°E 135°E

135°E

Global rms acc: 0.614 NH:0.313 TR:0.812 SH:0.287Z500 Anom. correlation fg2d(5)-ERA Int 1989-2008 DJF

ACC

-0.9

-0.8

-0.6

-0.4

-0.2

0.2

0.4

0.6

0.8

0.9

60°S60°S

30°S 30°S

0°0°

30°N 30°N

60°N60°N

135°W

135°W 90°W

90°W 45°W

45°W 0°

0° 45°E

45°E 90°E

90°E 135°E

135°E

sigma: 0.343 mean: 0.0506Fisher z transform diff fg4m(5)-fg2d(5) 1989-2008 DJF

z

-1

-0.8

-0.6

-0.4

-0.2

0.2

0.4

0.6

0.8

1

T159L91, plus revised stratospheric physics

Only 5 members, but score of 0.342 is much better than L62

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 25

60°S60°S

30°S 30°S

0°0°

30°N 30°N

60°N60°N

135°W

135°W 90°W

90°W 45°W

45°W 0°

0° 45°E

45°E 90°E

90°E 135°E

135°E

Global rms acc: 0.627 NH:0.273 TR:0.819 SH:0.381Z500 Anom. correlation fgcn(11)-ERA Int 1989-2008 DJF

ACC

-0.9

-0.8

-0.6

-0.4

-0.2

0.2

0.4

0.6

0.8

0.9

60°S60°S

30°S 30°S

0°0°

30°N 30°N

60°N60°N

135°W

135°W 90°W

90°W 45°W

45°W 0°

0° 45°E

45°E 90°E

90°E 135°E

135°E

Global rms acc: 0.646 NH:0.39 TR:0.819 SH:0.409Z500 Anom. correlation fg79(11)-ERA Int 1989-2008 DJF

ACC

-0.9

-0.8

-0.6

-0.4

-0.2

0.2

0.4

0.6

0.8

0.9

60°S60°S

30°S 30°S

0°0°

30°N 30°N

60°N60°N

135°W

135°W 90°W

90°W 45°W

45°W 0°

0° 45°E

45°E 90°E

90°E 135°E

135°E

sigma: 0.343 mean: -0.0457Fisher z transform diff fgcn(11)-fg79(11) 1989-2008 DJF

z

-1

-0.8

-0.6

-0.4

-0.2

0.2

0.4

0.6

0.8

1

T255L91

Score is now 0.390, cf 0.294 for T159L62

60°S60°S

30°S 30°S

0°0°

30°N 30°N

60°N60°N

135°W

135°W 90°W

90°W 45°W

45°W 0°

0° 45°E

45°E 90°E

90°E 135°E

135°E

Global rms acc: 0.627 NH:0.273 TR:0.819 SH:0.381Z500 Anom. correlation fgcn(11)-ERA Int 1989-2008 DJF

ACC

-0.9

-0.8

-0.6

-0.4

-0.2

0.2

0.4

0.6

0.8

0.9

60°S60°S

30°S 30°S

0°0°

30°N 30°N

60°N60°N

135°W

135°W 90°W

90°W 45°W

45°W 0°

0° 45°E

45°E 90°E

90°E 135°E

135°E

Global rms acc: 0.646 NH:0.39 TR:0.819 SH:0.409Z500 Anom. correlation fg79(11)-ERA Int 1989-2008 DJF

ACC

-0.9

-0.8

-0.6

-0.4

-0.2

0.2

0.4

0.6

0.8

0.9

60°S60°S

30°S 30°S

0°0°

30°N 30°N

60°N60°N

135°W

135°W 90°W

90°W 45°W

45°W 0°

0° 45°E

45°E 90°E

90°E 135°E

135°E

sigma: 0.343 mean: -0.0457Fisher z transform diff fgcn(11)-fg79(11) 1989-2008 DJF

z

-1

-0.8

-0.6

-0.4

-0.2

0.2

0.4

0.6

0.8

1

T255L91, with alternate stochastic physics

Score is 0.273From the best to the worst!

(Also other fields)

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 26

Possible interpretations

Statistical testing suggests differences are real, for this 20 year period Different model configurations give different model “signals” in NH winter

Hope was that hemispheric averaging would increase degrees of freedom enough to make scores meaningful

Hypothesis 1: this is not true - a given set of signals gets a given score for the 20 year period, but this is of no relevance to expected model skill in the future, and cannot be used for model selection.

Hypothesis 2: Some model configurations really do better capture the “balance” of processes affecting NH winter circulation, even if it is via compensation of errors. Better to choose the model with the better score.

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 27

Choosing a model configuration

• Encouraging that some configurations give good results

• Higher horizontal and vertical resolution are consistently positive

• Model climate is much improved, again resolution clearly helps

• Forecast skill??

• How should we weight seasonal forecast skill?

• What other tests should we use for a model?

Links to extended/monthly forecast range??

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 28

Cost effective systems

• Back integrations dominate total cost of system System 3: 3300 back integrations (must be in first year) 492 real-time integrations (per year)

• Back integrations define model climate Need both climate mean and the pdf, latter needs large sample May prefer to use a “recent” period (30 years? Or less??) System 2 had a 75 member “climate”, System 3 has 275. Sampling is basically OK

• Back integrations provide information on skill A forecast cannot be used unless we know (or assume) its level of skill Observations have only 1 member, so large ensembles are much less

helpful than large numbers of cases. Care needed eg to estimate skill of 41 member ensemble based on past

performance of 11 member ensemble For regions of high signal/noise, System 3 gives adequate skill estimates For regions of low signal/noise (eg <= 0.5), need hundreds of years

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 29

Data policy and exchange issues

• Present data policy In Europe, constrains the free distribution/exchange of seasonal forecast

data Policy is not fixed in stone, and may evolve over time

• Science Want to make sure that scientific studies are hindered as little as possible CHFP is main research project on seasonal prediction; data policy has

been OK, resources for data exchange were long a sticking point. High level support for new projects may be helpful

• Real-time forecasts Some data can be used by /supplied to WMO Need to ensure that it is enough Need to ensure that important “public good” applications are supported

Workshop on Sub-seasonal to Seasonal Prediction, Exeter, 1-3 December 2010 30

Conclusions

• Seasonal prediction still exciting and challenging

• Mid-latitude skill and reliability still need much work

• Higher resolution seems helpful

• Testing/assessing/selecting models needs to cut across timescales

• Coordinated experimentation has potential to be valuable, beyond CHFP

• Careful design will make it easier for operational centre’s to participate