AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 1/25

Decadal climate predictions with

the ECMWF coupled system

Susanna Corti European Centre for Medium-Range Weather Forecasts

With contributions from Magdalena Balmaseda, Linus Magnusson, Franco Molteni, Frederic Vitart, Antje Weisheimer and Tim Palmer

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 2/25

Outline

•  Decadal (10-year), 1960-2005 (varying GHG, no volcanoes)

•  IFS cycle 36r1 (T159 91 vertical levels), Nemo 3 (42 vertical levels 1° average resolution), sampled sea-ice. Initialisation: ECMWF reanalysis ERA and NemoVar

•  Starting dates November every 5th year

•  Full initialisation

•  Anomaly initialisation

•  Heat and momentum flux correction

•  7 ensemble members (+3 with volcanoes for control and flux correction)

•  Overall Grand Ensemble of 27 members

Experiments:

Diagnostics: •  Heat content and heat transport (very preliminary diagnostics)

•  Nino3.4 plumes for 2 major El Nino events

•  Teleconnection patterns (i.e. Regressions between SSTs and Atmospheric fields)

•  Deterministic & Probabilistic scores

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 3/25

Momentum flux correction - rationale

� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �

Cross-section ocean bias U-current

Temperature

(Walker circulation feedback)

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 4/25

North Atlantic 300 m Heat Content

Black: Analysis Red: Full Ini. Purple: Anom. Ini Blue: Flux Correction

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 5/25

� � � T� � � � � � � � � � � � � � � T� s cc � � � � �� � � � � � � � � � � � � � � � � � � � � �   � � � � � � � � � � � � oTr c‐�

Control

U- and H-flux correction

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 6/25

MOC � � � � � � � � �

� � � S� � � � � � � � � � � � � � � � � � � � � � � �   � � � � � � � � r cTos ‐�

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 7/25

ATL50N 170 l$v$l= 4901.46

1960 1970 1980 1990 2000 2010 2020Tim$

0.55

0.60

0.65

0.70

0.75

0.80

0.85

North Atlantic Heat Transport –50N

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 8/25

Diagnostics – Regression SST in Nino3.4 vs. SST

� � � � � � � � � S� � � � � � � � � � � � � � � �

� � � � � � � � � � � �

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 9/25

Diagnostics – Regression SST in Nino3.4 vs. Precipitation vs. Precipitation vs. Precipitation � � � � � � � � � � � �

� � � � � � � � � S� � � � � � � � � � � � � � � �

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 10/25

Tropical storms - Flux corrected exp.

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 11/25

Diagnostics – NHE Winter -500hPa regional EOFs – Full Ini.

EOF1 DJF American-Pacific

EOF1 DJF Euro-Atlantic

EOF1 DJF Asian-West Pacific

Covariance with 2m temperature

Reanalysis Model

Covariance with 2m temperature

EOF1 DJF American-Pacific

EOF1 DJF Euro-Atlantic

EOF1 DJF Asian-West Pacific

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 12/25

Diagnostics – NHE Winter - MSLP 1st EOF – Full Ini.

EOF1 DJF NH

Cov 2Tm (NH,DJF)

EOF1 JJA SH

Reanalysis Model

EOF1 JJA NH

Cov 2Tm (NH,JJA)

EOF1 DJF SH

EOF1 DJF NH

Cov 2Tm (NH,DJF)

EOF1 JJA SH

EOF1 JJA NH

Cov 2Tm (NH,JJA)

EOF1 DJF SH

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 13/25

ACC – YR 2-6 - YRs 6-9 - 12-month average – T2m

Near surface Air temperature ACC yr 2-5 - full ini.

Near surface Air temperature ACC yr 6-9 - full ini.

Near surface Air temperature ACC yr 2-5 – flux cor.

Near surface Air temperature ACC yr 6-9 – flux cor.

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 14/25

ACC – YR 2-6 - YRs 6-9 - 12-month average – T2m

Near surface Air temperature ACC yr 6-9 - full ini.

Near surface Air temperature ACC yr 6-9 – Grand Ens.

Near surface Air temperature ACC yr 2-5 - full ini.

Near surface Air temperature ACC yr 2-5 – Grand Ens.

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 15/25

Time series - AMO index

2-5 yr - Cor=0.92 Spread/RMSE=0.67 RPSS=0.39

6-9 yr - Cor=0.45 Spread/RMSE=0.45 RPSS=-0.15

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 16/25

Time series - North Atlantic SSTs

2-5 yr - Cor=0.85 Spread/RMSE=0.58 RPSS=0.13

6-9 yr - Cor=0.93 Spread/RMSE=0.67 RPSS=-0.60

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 17/25

Time series - Global average 2-metre temperature

2-5 yr - Cor=0.95 Spread/RMSE=0.63 RPSS=0.52

6-9 yr - Cor=0.96 Spread/RMSE=0.73 RPSS=-0.67

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 18/25

Probabilistic scores - Mean Square Skill Score MSSS

Near surface Air temperature MSSS yr 2-5 – Grand Ens.

Near surface Air temperature MSSS yr 1 – Grand Ens.

Near surface Air temperature MSSS yr 6-9 – Grand Ens.

Near surface Air temperature MSSS yr 2 – Grand Ens.

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 19/25

Reliability diagram

Poor resolution Good resolution

Reliability score (the smaller, the better) Resolution score (the bigger, the better)

c c

Size of red bullets represents number of forecasts in probability category (sharpness)

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 20/25

Brier Skill Score & Reliability Diagram

UNCRELRES

UNCUNCRESREL −

=+−

−=1

• How to construct the area of positive skill?

perfect reliability

Obse

rved

Fre

quen

cy

Forecast Probability

line of no skill

area of skill (RES > REL)

climatological frequency (line of no resolution)

€

BSS =1− BSBSc

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 21/25

Reliability diagrams - T2m - 2-5 yr

T2m anomaly below the lower tercile - Global average

T2m anomaly above the upper tercile - Global average

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 22/25

Reliability diagrams - T2m - 6-9 yr

T2m anomaly above the upper tercile - Global average

T2m anomaly below the lower tercile - Global average

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 23/25

Reliability diagrams - T2m - 2-5 yr

T2m anomaly above the median over North America

T2m anomaly above the median over Europe

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 24/25

Summary

 The IFS/NEMO coupled model is affected by an overall cold SST bias, with a too strong Pacific equatorial cold tongue (especially in summer).

 Flux correction improves the mean state and ENSO variability.

 In spite of model drift and the fact that several climate processes, such as those related to sea-ice formation, export and melting, are not represented in the model, the decadal prediction experiments show a positive forecast quality that can be statistically significant over several areas.

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 25/25

Yr 6-10 Full Initialisation (bias correction)

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 26/25

Yr 6-10 Full Initialisation

AGCI Workshop on Decadal Predictions - Aspen 26 June -1 July 2011 27/25

Yr 6-10 Full Initialisation