22
OceanObs 09, Venice 21-25 September 2009 1 THE ECMWF Seasonal Forecasting system

OceanObs 09, Venice 21-25 September 2009 1 THE ECMWF Seasonal Forecasting system

Embed Size (px)

Citation preview

Page 1: OceanObs 09, Venice 21-25 September 2009 1 THE ECMWF Seasonal Forecasting system

OceanObs 09, Venice 21-25 September 2009 1

THE ECMWF Seasonal Forecasting system

Page 2: OceanObs 09, Venice 21-25 September 2009 1 THE ECMWF Seasonal Forecasting system

OceanObs 09, Venice 21-25 September 2009 2

Overview

• OCEANOBS 09 & EUROBRISA Applications most welcome for the concept of End To End

Seasonal Forecasting Systems

• The ECMWF S4 Better skill in the Equatorial and South Atlantic Mixed results everywhere else (large biases)

• EUROBRISA PROJECT Subseasonal time scales (WCRP workshop in Exeter) Decadal time scales (Paco’s talk) NCEP part of EUROSIP Corean Centre?

Page 3: OceanObs 09, Venice 21-25 September 2009 1 THE ECMWF Seasonal Forecasting system

OceanObs 09, Venice 21-25 September 2009 3

End-To-End Seasonal forecasting System

EN

SEM

BL

E G

EN

ER

AT

ION

COUPLED MODEL Forecast PRODUCTS

Initialization Forward Integration Forecast Calibration

OCEAN

PR

OB

AB

ILIS

TIC

CA

LIB

RA

TE

D F

OR

EC

AS

T

JUL2006

AUG SEP OCT NOV DEC JAN2007

FEB MAR APR MAY JUN JUL AUG SEP

-1

0

1

2

Ano

mal

y (d

eg C

)

-1

0

1

2

Monthly mean anomalies relative to NCEP adjusted OIv2 1971-2000 climatologyECMWF forecast from 1 Jan 2007

NINO3.4 SST anomaly plume

Produced from real-time forecast data

System 380°S80°S

70°S 70°S

60°S60°S

50°S 50°S

40°S40°S

30°S 30°S

20°S20°S

10°S 10°S

0°0°

10°N 10°N

20°N20°N

30°N 30°N

40°N40°N

50°N 50°N

60°N60°N

70°N 70°N

80°N80°N

20°E

20°E 40°E

40°E 60°E

60°E 80°E

80°E 100°E

100°E 120°E

120°E 140°E

140°E 160°E

160°E 180°

180° 160°W

160°W 140°W

140°W 120°W

120°W 100°W

100°W 80°W

80°W 60°W

60°W 40°W

40°W 20°W

20°W

14.3 10.39.8 13.325.1 26.22 2.9

No Significance 90% Significance 95% Significance 99% Significance

Ensemble size = 40,climate size = 70Forecast start reference is 01/06/2005Tropical Storm FrequencyECMWF Seasonal Forecast

Significance level is 90%JASON

FORECAST CLIMATE

Page 4: OceanObs 09, Venice 21-25 September 2009 1 THE ECMWF Seasonal Forecasting system

OceanObs 09, Venice 21-25 September 2009 4

A decade of progress on ENSO prediction

•Steady progress: ~1 month/decade skill gain

•How much is due to the initialization, how much to model development?

S1 S2 S3

Relative Reduction in SST Forecast ErrorECMWF Seasonal Forecasting Systems

TOTAL GAIN

OC INI

MODEL

0

5

10

15

20

25

30

35

40

1

%TOTAL GAIN OC INI MODEL

Half of the gain on forecast skill is due to improved ocean initialization

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

0.4

0.5

0.6

0.7

0.8

0.9

1

Anom

aly

corr

ela

tion

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 err

or

(deg C

)

Ensemble sizes are 5 (0001), 5 (0001) and 5 (0001) 64 start dates from 19870401 to 20021201

NINO3.4 SST rms errors

Fcast S3 Fcast S2 Fcast S1 Persistence

MAGICS 6.11 windmill - neh Mon Sep 14 11:48:01 2009

OceanObs09 plenary paper

Page 5: OceanObs 09, Venice 21-25 September 2009 1 THE ECMWF Seasonal Forecasting system

OceanObs 09, Venice 21-25 September 2009 6

What is the value of a long historical record? Example from the Medium Range Weather Forecasts (TIGGI)

Impact of Increased ensemble size versus longer calibration period

(Continuous Rank Probability Skill Score, T-2m Europe)

A longer calibration period has larger impact than increasing the ensemble size. From Hagerdorn 2008

Page 6: OceanObs 09, Venice 21-25 September 2009 1 THE ECMWF Seasonal Forecasting system

OceanObs 09, Venice 21-25 September 2009 8

5-month lead fcst Obs Corr. skill

Prediction of Dengue Risk transmission: 5 month lead time

From EUROBRISAhttp://eurobrisa.cptec.inpe.br/

Numerical Model+ Calibration + Dengue model

Forecast issued in Nov 1997, valid for Apr 1998

Page 7: OceanObs 09, Venice 21-25 September 2009 1 THE ECMWF Seasonal Forecasting system

OceanObs 09, Venice 21-25 September 2009 9

ECMWF S4

•NEMO (ORCA1)+CY36R4

•Increased atmos resolution (to T255 + 91 levels) [S3 was T159+62 levels)

•Initial conditions with NEMOVAR, ERA-INTERIM, and…

Page 8: OceanObs 09, Venice 21-25 September 2009 1 THE ECMWF Seasonal Forecasting system

OceanObs 09, Venice 21-25 September 2009 10

ECWMF: COMBINE Ocean Re-AnalysisUsed to initialized EC-EARTH decadal forecasts

• It uses NEMO/NEMOVAR, ORCA1 configuration, 42 levels (ORCA1_Z42_v2)

• NEMO V3.0 + Local Modifications.• Forced by ERA40 (until 1989) + ERA Interim (after 1989)

• Assimilates Temperature/Salinity from EN3 (corrected XBT’s).

• Strong relaxation to SST (OI_v2)

• Offline+Online model bias correction scheme (T/S and pressure gradient): Offline bias term estimated from Argo Period Latitudinal dependence of the P/T/S bias: P strong at the Eq, weak at mid latitudes. Viceversa with

T/S

• 5 ensemble members (perturbations to wind, initial deep ocean, observation coverage)

Page 9: OceanObs 09, Venice 21-25 September 2009 1 THE ECMWF Seasonal Forecasting system

OceanObs 09, Venice 21-25 September 2009 11

Assessment of the COMBINE re-analysis

•Compared with the CONTROL (e.i., no data assim)

•Better fit to T/S profiles

•No degraded Equatorial Currents

•Spread in the deep ocean

•Improvement in ENSO forecasts

•Correlation with altimeter data as a measure of interannual variability: Improvements in the tropics, slight degradation at mid latitudes (especially North East Atlantic)

•Atlantic MOC?

Further developments for the next operational system (due end of this year):

Altimeter, revised assimilation parameters, partition of bias,SST,…

Page 10: OceanObs 09, Venice 21-25 September 2009 1 THE ECMWF Seasonal Forecasting system

OceanObs 09, Venice 21-25 September 2009 13

rms EQ2 Potential Temperature

0.0 0.5 1.0 1.5 2.0 2.5rms EQ2 Potential Temperature

-500

-400

-300

-200

-100

0

Depth

(m

)

rms EQIND Potential Temperature

0.0 0.5 1.0 1.5 2.0rms EQIND Potential Temperature

-500

-400

-300

-200

-100

0

Depth

(m

)

rms TRPAC Potential Temperature

0.4 0.6 0.8 1.0 1.2 1.4rms TRPAC Potential Temperature

-500

-400

-300

-200

-100

0

Depth

(m

)

rms GLOBAL Potential Temperature

0.6 0.8 1.0 1.2 1.4 1.6rms GLOBAL Potential Temperature

-500

-400

-300

-200

-100

0

Depth

(m

)

CONTROL ASSIM: T+S ASSIM: T+S+Alti

EQ Central Pacific EQ Indian Ocean

TROPICAL Pacific GLOBAL

Altimeter Improves the fit to InSitu Temperature Data

RMSE of 10 days forecast

Page 11: OceanObs 09, Venice 21-25 September 2009 1 THE ECMWF Seasonal Forecasting system

OceanObs 09, Venice 21-25 September 2009 14

Correlation with Altimeter

correl (1): fe5x sossheig ( 1993-2008 ) correl (1): fe5x sossheig ( 1993-2008 )

(ndim): Min= -0.37, Max= 1.00, Int= 0.02

100E 160W 60WLongitude

60S

40S

20S

0

20N

40N

60N

Latitu

de

0.40 0.44 0.48 0.52 0.56 0.60 0.64 0.68 0.72 0.76 0.80 0.84 0.88 0.92 0.96 1.00

RMSE (1): fe5x sossheig (1993-2008) RMSE (1): fe5x sossheig (1993-2008)

(m): Min= 0.01, Max= 0.15, Int= 0.01

100E 160W 60WLongitude

60S

40S

20S

0

20N

40N

60N

Latitu

de

0.02 0.04 0.06 0.08 0.10

rms/signal(1): fe5x sossheig (1993-2008) rms/signal(1): fe5x sossheig (1993-2008)

(N/A): Min= 0.10, Max= 3.18, Int= 0.20

100E 160W 60WLongitude

60S

40S

20S

0

20N

40N

60N

Latitu

de

0.00 0.40 0.80 1.20 1.60 2.00

sdv_fe5x/sdv_aviso (1): sossheig (1993-2008) sdv_fe5x/sdv_aviso (1): sossheig (1993-2008)

(N/A): Min= 0.22, Max= 2.74, Int= 0.20

100E 160W 60WLongitude

60S

40S

20S

0

20N

40N

60N

Latitu

de

0.00 0.40 0.80 1.20 1.60 2.00

fa9p_oref_19932008_1_sossheig_stats.ps) Aug 4 2010

correl (1): fa9p correl ( 19932008 ) correl (1): fa9p correl ( 19932008 )

(ndim): Min= -0.40, Max= 0.99, Int= 0.02

100E 160W 60WLongitude

60S

40S

20S

0

20N

40N

60N

Latitu

de

0.40 0.44 0.48 0.52 0.56 0.60 0.64 0.68 0.72 0.76 0.80 0.84 0.88 0.92 0.96 1.00

RMSE (1): fa9p sossheig (19932008) RMSE (1): fa9p sossheig (19932008)

(m): Min= 0.01, Max= 0.16, Int= 0.01

100E 160W 60WLongitude

60S

40S

20S

0

20N

40N

60N

Latitu

de

0.02 0.04 0.06 0.08 0.10

rms/signal(1): fa9p sossheig (19932008) rms/signal(1): fa9p sossheig (19932008)

(N/A): Min= 0.18, Max= 3.46, Int= 0.20

100E 160W 60WLongitude

60S

40S

20S

0

20N

40N

60N

Latitu

de

0.00 0.40 0.80 1.20 1.60 2.00

sdv_fa9p/sdv_aviso (1): sossheig (19932008) sdv_fa9p/sdv_aviso (1): sossheig (19932008)

(N/A): Min= 0.22, Max= 3.25, Int= 0.20

100E 160W 60WLongitude

60S

40S

20S

0

20N

40N

60N

Latitu

de

0.00 0.40 0.80 1.20 1.60 2.00

fa9p_oref_19932008_1_sossheig_stats.ps) Feb 19 2010

COMBINE ASSIM T+S+Alti

Page 12: OceanObs 09, Venice 21-25 September 2009 1 THE ECMWF Seasonal Forecasting system

OceanObs 09, Venice 21-25 September 2009 15

Impact of Ocean Assim in SST forecasts Prototype of S4: latest NEMOVAR+36r4

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

0.4

0.5

0.6

0.7

0.8

0.9

1

Anom

aly

corre

latio

n

wrt NCEP adjusted OIv2 1971-2000 climatology

EQ2 034a anomaly correlation

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

0

0.2

0.4

0.6

0.8

1

Rm

s er

ror

Ensemble sizes are 5 (feik) and 5 (ff5e) 64 start dates from 19930201 to 20081101

EQ2 034a rms errors

Fcast feik Fcast ff5e Persistence Ensemble sd

MAGICS 6.12 windmill - neh Fri Sep 3 10:59:05 2010

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

0.4

0.5

0.6

0.7

0.8

0.9

1

Anom

aly

corre

latio

n

wrt NCEP adjusted OIv2 1971-2000 climatology

ATL3 034a anomaly correlation

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

0

0.2

0.4

0.6

0.8

1

Rm

s er

ror

Ensemble sizes are 5 (feik) and 5 (ff5e) 64 start dates from 19930201 to 20081101

ATL3 034a rms errors

Fcast feik Fcast ff5e Persistence Ensemble sd

MAGICS 6.12 windmill - neh Fri Sep 3 10:59:05 2010

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

0.4

0.5

0.6

0.7

0.8

0.9

1

Anom

aly

corre

latio

n

wrt NCEP adjusted OIv2 1971-2000 climatology

EQIND 034a anomaly correlation

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

0

0.2

0.4

0.6

0.8

1

Rm

s er

ror

Ensemble sizes are 5 (feik) and 5 (ff5e) 64 start dates from 19930201 to 20081101

EQIND 034a rms errors

Fcast feik Fcast ff5e Persistence Ensemble sd

MAGICS 6.12 windmill - neh Fri Sep 3 10:59:05 2010

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

0.4

0.5

0.6

0.7

0.8

0.9

1

Anom

aly

corre

latio

nwrt NCEP adjusted OIv2 1971-2000 climatology

NSTRATL 034a anomaly correlation

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

0

0.2

0.4

0.6

0.8

1

Rm

s er

ror

Ensemble sizes are 5 (feik) and 5 (ff5e) 64 start dates from 19930201 to 20081101

NSTRATL 034a rms errors

Fcast feik Fcast ff5e Persistence Ensemble sd

MAGICS 6.12 windmill - neh Fri Sep 3 10:59:05 2010

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

0.4

0.5

0.6

0.7

0.8

0.9

1

Anom

aly

corre

latio

n

wrt NCEP adjusted OIv2 1971-2000 climatology

NSTRPAC 034a anomaly correlation

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

0

0.2

0.4

0.6

0.8

1

Rm

s er

ror

Ensemble sizes are 5 (feik) and 5 (ff5e) 64 start dates from 19930201 to 20081101

NSTRPAC 034a rms errors

Fcast feik Fcast ff5e Persistence Ensemble sd

MAGICS 6.12 windmill - neh Fri Sep 3 10:59:05 2010

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

0.4

0.5

0.6

0.7

0.8

0.9

1

Anom

aly

corre

latio

n

wrt NCEP adjusted OIv2 1971-2000 climatology

SSTRATL 034a anomaly correlation

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

0

0.2

0.4

0.6

0.8

1

Rm

s er

ror

Ensemble sizes are 5 (feik) and 5 (ff5e) 64 start dates from 19930201 to 20081101

SSTRATL 034a rms errors

Fcast feik Fcast ff5e Persistence Ensemble sd

MAGICS 6.12 windmill - neh Fri Sep 3 10:59:05 2010

ASSIM CONTROL

NEMOVAR consistent improves the forecast skill of SST at different lead times and different regions, at SEASONAL TIME SCALES. See Later for Decadal

Page 13: OceanObs 09, Venice 21-25 September 2009 1 THE ECMWF Seasonal Forecasting system

OceanObs 09, Venice 21-25 September 2009 16

Combine project –Strategies for dealing with systematic errors in a coupled

ocean-atmosphere forecasting systemProject concept

Nature climateNature climate

Model climateModel climate

Flux correction

Anomaly initialisationNormal initialisation

Linus Magnusson et al.

Page 14: OceanObs 09, Venice 21-25 September 2009 1 THE ECMWF Seasonal Forecasting system

OceanObs 09, Venice 21-25 September 2009 18

Momentum flux correction - rationale

Systematic wind error (example October)

Page 15: OceanObs 09, Venice 21-25 September 2009 1 THE ECMWF Seasonal Forecasting system

OceanObs 09, Venice 21-25 September 2009 19

Experiments

• Seasonal (14-month forecasts), 1989-1999, Start dates November and

May

• Decadal (10-year), 1960-2005, Start dates November every 5th year

• Control forecast

• Anomaly initialisation

• Momentum flux correction

• Heat and momentum flux correction

• Model cycle 36r1, Nemo version 3, sampled sea-ice

• 3 ensemble members

Page 16: OceanObs 09, Venice 21-25 September 2009 1 THE ECMWF Seasonal Forecasting system

OceanObs 09, Venice 21-25 September 2009 20

SST bias in decadal integrations (fc year 2-10)

Control (“or” Anomaly initialisation)

U-flux correction U- and H-flux correction

Page 17: OceanObs 09, Venice 21-25 September 2009 1 THE ECMWF Seasonal Forecasting system

OceanObs 09, Venice 21-25 September 2009 21

T bias cross section, equatorial Pacific (fc year 2-10)Control (“or” Anomaly initialisation)

U-flux correction U- and H-flux correction

Page 18: OceanObs 09, Venice 21-25 September 2009 1 THE ECMWF Seasonal Forecasting system

OceanObs 09, Venice 21-25 September 2009 22

Nino3.4 SST forecasts November 1995 – November 1998Control Anomaly Initialisation

U-flux correction U- and H-flux correction

96 97 98 99 96 97 98 99

Page 19: OceanObs 09, Venice 21-25 September 2009 1 THE ECMWF Seasonal Forecasting system

OceanObs 09, Venice 21-25 September 2009 23

Model drift during the first year (10 start dates, 3 members)

Page 20: OceanObs 09, Venice 21-25 September 2009 1 THE ECMWF Seasonal Forecasting system

OceanObs 09, Venice 21-25 September 2009 24

ENSO statistics – seasonal cycle (year 2-10)

• Re-analysis• Control• Anomaly Init.• U- flux corr.• U- and H-flux corr.

Nino 3.4 SST mean Nino 3.4 SST st. dev.

Page 21: OceanObs 09, Venice 21-25 September 2009 1 THE ECMWF Seasonal Forecasting system

OceanObs 09, Venice 21-25 September 2009 27

Regression of rainfall anomalies on NINO3.4

AnoIni

U+H flux corr

Page 22: OceanObs 09, Venice 21-25 September 2009 1 THE ECMWF Seasonal Forecasting system

OceanObs 09, Venice 21-25 September 2009 28

Opinions

• At WCRP there is “thirst” for examples of applications: EUROBRISA is very well placed! Should continue

• The FORECAST ASSIMILATION project is very powerful THERE is a lot of science to do.