THOR Annual Meeting - Bergen 9-11 November 2011 1/25

On the impact of initial conditions relative

to external forcing on the skill of decadal

predictions: preliminary results with the

ECMWF coupled system

Susanna Corti European Centre for Medium-Range Weather Forecasts

With contributions from Magdalena Balmaseda, Per Kållberg, Linus Magnusson and Tim Palmer

THOR Annual Meeting - Bergen 9-11 November 2011 2/25

Table 1 - Summary of experiments. Model Ensemble

Memebers Atmospheric Analysis

Ocean Analysis

Initialisation Forcing fields Flux Correction

Volcanic eruptions

Sea Ice Model

IFS36R1/NEMO3 T159L91

/1x1

7 ERA NEMOVAR/COMBI

NE

Full Initialisation

Varying GHG &

anthropogenic aerosols

NO No Sampled obse rved Sea Ice

IFS36R1/NEMO3 T159L91

/1x1

7 ERA NEMOVAR/COMBI

NE

Full Initialisation

Varying GHG &

anthropogenic aerosols

Heat &

Momentum

No Sampled obse rved Sea Ice

IFS36R1/NEMO3 T159L91

/1x1

7 ERA NEMOVAR/COMBI

NE

Anomaly Initialisation

Varying GHG &

anthropogenic aerosols

NO No Sampled obse rved Sea Ice

IFS36R1/NEMO3 T159L91

/1x1

3 ERA NEMOVAR/COMBI

NE

Full Initialisation

Varying GHG &

anthropogenic aerosols

NO YES Sampled obse rved Sea Ice

IFS36R1/NEMO3 T159L91

/1x1

3 ERA NEMOVAR/COMBI

NE

Full Initialisation

Varying GHG &

anthropogenic aerosols

Heat &

Momentum

YES Sampled obse rved Sea Ice

IFS36R4/NEMO3 T159L91

/1x1

5 ERA NEMOVAR/ORAS4

Full Initialisation

CMIP5 NO YES LIM2

2

1

THOR Annual Meeting - Bergen 9-11 November 2011 3/25

(EXP1) Surface Temperature Anomaly Correlation Coefficients YR 2-5 12-month average

THOR Annual Meeting - Bergen 9-11 November 2011 4/25

(EXP1) Near Surface Temperature Anomaly Correlation Coefficients YR 2-5 12-month average- Linear Climate trend out

THOR Annual Meeting - Bergen 9-11 November 2011 5/25

(EXP1) Near Surface Temperature Anomaly Correlation Coefficients

YR 2-5 12-month average

Climate trend in

Climate trend out

Anom. Init.

Flux Cor.

Full Init.

Flux Cor.Volc. in

Full Init.Volc. in

Full Init.

Anom. Init.

Flux Cor.

Full Init.Volc. in

Flux Cor.Volc. in

THOR Annual Meeting - Bergen 9-11 November 2011 6/25

(EXP1) Near Surface Temperature Anomaly Correlation Coefficients YR 6-9 12-month average

Climate trend in

Climate trend out

Anom. Init.

Flux Cor.

Full Init.

Flux Cor.Volc. in

Full Init.Volc. in

Full Init.

Anom. Init.

Flux Cor.

Full Init.Volc. in

Flux Cor.Volc. in

THOR Annual Meeting - Bergen 9-11 November 2011 7/25

(EXP2) Near Surface Temperature Anomaly Correlation Coefficient[Climate trend out]

YR 2-5 12-month average YR 6-9 12-month average

Grand Ensemble of all (EXP1) experiments - 27 Ensemble membersNear Surface Temperature Anomaly Correlation Coefficients

[Climate trend out]YR 2-5 12-month average YR 6-9 12-month average

THOR Annual Meeting - Bergen 9-11 November 2011 8/25

Grand Ensemble of all (EXP1) experiments - 27 Ensemble members

YR 2-5 12-month averageCor=0.95, SPREAD/RMSE=0.63

YR 2-5 12-month averageCor=-0.01, Climate forcing out

AMO index YR 2-5 12-month averageCor=0.78, SPREAD/RMSE=0.82

YR 2-5 12-month averageCor=0.9, Climate forcing out

T2m

THOR Annual Meeting - Bergen 9-11 November 2011 9/25

10-year integrations with 5-member ensembles from:

A 1965 initial conditions, observed forcing (GHG & aerosols) from 1965 (control1)

B 1995 initial conditions, observed forcing (GHG & aerosols) from 1995 (control2)

C 1965 initial conditions, observed forcing from 1995

D 1995 initial conditions, observed forcing from 1965

By comparing A with D, and B with C, we have two estimates of decadal predictability (arising from having different initial conditions and the same GHG forcing).

By comparing A with C, and B with D, we have two estimates of the impact of GHG forcings (since initial conditions are the same).

1965-1995 Experiment

THOR Annual Meeting - Bergen 9-11 November 2011 10/25

1yr Running mean - Global SSTs

FOR FOR

INI INI

1965-1975 1995-2005

THOR Annual Meeting - Bergen 9-11 November 2011 11/25

1yr Running mean - Indian Ocean SSTs

INI INI

FOR FOR

1965-1975 1995-2005

THOR Annual Meeting - Bergen 9-11 November 2011 12/25

1yr Running mean - North Atlantic SSTs

INI INI

FOR FOR

1965-1975 1995-2005

THOR Annual Meeting - Bergen 9-11 November 2011 13/25

Reanalysis Control

SSTs: [1995-2005] - [1965-1975]

Ctl95-FOR65

FOR95-Ctl65

Ctl95-INI65

INI95-Ctl65

THOR Annual Meeting - Bergen 9-11 November 2011 14/25

Reanalysis Control

SSTs: [1995-1996] - [1965-1966]

Ctl95-FOR65

FOR95-Ctl65

Ctl95-INI65

INI95-Ctl65

THOR Annual Meeting - Bergen 9-11 November 2011 15/25

Reanalysis Control

SSTs: [1995-1997] - [1965-1967]

Ctl95-FOR65

FOR95-Ctl65

Ctl95-INI65

INI95-Ctl65

THOR Annual Meeting - Bergen 9-11 November 2011 16/25

Reanalysis Control

SSTs: [1995-1998] - [1965-1968]

Ctl95-FOR65

FOR95-Ctl65

Ctl95-INI65

INI95-Ctl65

THOR Annual Meeting - Bergen 9-11 November 2011 17/25

Reanalysis Control

SSTs: [1995-1999] - [1965-1969]

Ctl95-FOR65

FOR95-Ctl65

Ctl95-INI65

INI95-Ctl65

THOR Annual Meeting - Bergen 9-11 November 2011 18/25

Rianalysis Control

SSTs: [1995-2000] - [1965-1970]

Ctl95-FOR65

FOR95-Ctl65

Ctl95-INI65

INI95-Ctl65

THOR Annual Meeting - Bergen 9-11 November 2011 19/25

Reanalysis Control

SSTs: [1995-2005] - [1965-1975]

Ctl95-FOR65

FOR95-Ctl65

Ctl95-INI65

INI95-Ctl65

THOR Annual Meeting - Bergen 9-11 November 2011 20/25

Reanalysis Control

SSTs: [1995-1996] - [1965-1966]

Ctl95-FOR65

FOR95-Ctl65

Ctl95-INI65

INI95-Ctl65

THOR Annual Meeting - Bergen 9-11 November 2011 21/25

Reanalysis Control

SSTs: [1995-1997] - [1965-1967]

Ctl95-FOR65

FOR95-Ctl65

Ctl95-INI65

INI95-Ctl65

THOR Annual Meeting - Bergen 9-11 November 2011 22/25

Reanalysis Control

SSTs: [1995-1998] - [1965-1968]

Ctl95-FOR65

FOR95-Ctl65

Ctl95-INI65

INI95-Ctl65

THOR Annual Meeting - Bergen 9-11 November 2011 23/25

Reanalysis Control

SSTs: [1995-1999] - [1965-1969]

Ctl95-FOR65

FOR95-Ctl65

Ctl95-INI65

INI95-Ctl65

THOR Annual Meeting - Bergen 9-11 November 2011 24/25

Reanalysis Control

SSTs: [1995-2000] - [1965-1970]

Ctl95-FOR65

FOR95-Ctl65

Ctl95-INI65

INI95-Ctl65

THOR Annual Meeting - Bergen 9-11 November 2011 25/25

Summary

The decadal prediction experiments show a positive forecast quality that can be statistically significant over several areas.

When the linear climate trend is subtracted some regions (common to all the experiments carried out) of more pronounced predictability have been identified.

Preliminary results of the 1965-1995 experiment indicate that over time scales longer than 5 years predictability arises mainly from the forcing.

The correct initialisation has a strong impact up to about 1 year on a global domain, but it seems to affect the predictability over the North Atlantic up to 5 years.