decadal predictability of winter sst in the nordic seas and barents sea in three cmip5 models naclim...

Post on 04-Jan-2016

220 Views

Category:

Documents

1 Downloads

Preview:

Click to see full reader

TRANSCRIPT

Decadal predictability of winter SST in the Nordic Seas and Barents Sea

in three CMIP5 models

NACLIM ANNUAL MEETING, OCTOBER 14, 2014 BERLIN

H. R. Langehaug, K. Lohmann, T. Eldevik, D. Matei, Y. Gao

The research leading to these results has received funding from:

RCN-EPOCASAEnhancing seasonal-to-decadal Prediction Of Climate for the North Atlantic Sector and Arctic

EU-NACLIMNorth Atlantic climate

Main goal Assessment of decadal predictability wrt SST (winter) What do we expect from the models in this particular region?

3

Nordic Seas

Barents Sea

Region of interest

Map based on NASA Worldwind-globe

The region of interest has not been investigated in detail before wrt hindcast prediction experiments in CMIP5

Atla

ntic

dom

ain

Outline

› Introduction of the three models› How is the model initialized?

› What is the surface condition compared to HadISST?

› How well do the hindcasts reproduce the observed SST changes in the Atlantic domain?

› Hindcasts are compared with non-initialized runs and the persistence forecast

› Are there specific regions with high predictive skill compared to the rest of the region?

4

Nordic Seas

Barents Sea

Map based on NASA Worldwind-globe

Atla

ntic

dom

ain

Variance in winter SST in the period 1961-2010

HadISST MPI-ESM-LR

mean SST

MPI-ESM-LR

• Anomaly initialization• 3D assimilation of T & S from ocean reanalysis

initialized every fifth year

Variance in winter SST in the period 1961-2010

HadISST CNRM-CM5

mean SST

CNRM-CM5

• Full-field initialization Correct for the drift• 3D assimilation of T & S from ocean reanalysis

Variance in winter SST in the period 1961-2010

HadISST IPSL-CM5

IPSL-CM5

• Anomaly initialization• SST assimilation, where sea ice concentration is less than 50%

mean SST

8

Assessment of decadal predictability wrt winter SST

Time (years) after initialization

Lead time = 1-3 yrsPersistence: HadISST is correlated with itself with a time lag of 1-3 yrs

Hindcast CMIP5 model

Non-initialized CMIP5 model hindcast

non-initialized

9

MPI-ESM-LR

CNRM-CM5

IPSL-CM5

hindcast

non-initializedHow well do the models reproduce the observed SST changes in the Atlantic domain?

. Significant correlation

MPI-ESM-LR

. Significant correlation

. Significant correlation

. Significant correlation

. Significant correlation

. Significant correlation

CNRM-CM5

. Significant correlation

. Significant correlation

. Significant correlation

. Significant correlation

. Significant correlation

IPSL-CM5

. Significant correlation

. Significant correlation

. Significant correlation

. Significant correlation

Summary

› Introduction of the three models: similarities/differences?

› How well do the hindcasts reproduce the observed SST changes?

› Local and remote influence on the prediction skill?

25

Barents Sea

MPI-ESM-LR and CNRM-CM5 most similar to observations in terms of SST variance

MPI-ESM-LR

top related