Atlantic Multidecadal Variability and Its Climate Impacts in CMIP3 Models and Observations

Atlantic Multidecadal Variability and Its Climate Impacts in CMIP3

Models and Observations

Mingfang TingWith

Yochanan Kushnir, Richard Seager, Cuihua Li

June 8, 20102010 U.S. AMOC Annual Meeting

Miami, FL

Observed AMO (AMV) Indices

North Atlantic SST Index (7.5W-75W, 0-60N, ocean only) for Models and Observations

Linear de-trending

Global mean SST as forced signal

Signal to noise EOF PC1 as forced signal (Ting et al, 2009)

Forced Atlantic SST in models

Observed AMV

Climate Impacts of Forced versus Internal NASST Variability (Regression based on Annual

Mean) InternalForced

T sPr

ecip

.

(Stippled regions are for values significant at or above 95% confidence level)

What are the spatial and temporal characteristics and climate impacts of AMV in CMIP3 models for both the 20th and 21st Centuries?

Hoe do they compare to 20th Century observations?

Is the signal to noise maximizing EOF method successfully separating AMV from the externally forced component?

Questions to be Addressed:

In this study… Apply signal-to-Noise Maximizing EOF Analysis to

IPCC multiple model, multi-ensemble members of the 20th and 21st simulations

First apply EOF analysis to deviations from multi-model average to determine the spatial structure of the internal modes of variability (noise)

Apply a spatial pre-whitening transformation based on the internal EOFs to remove the spatial correlations in the internal atmospheric variability (i.e., “climate noise”) contained in the multi-model average

Apply EOF analysis to the signal to get forced component

S/N Maximizing EOF 1 for 19 IPCC Model Simulations

20th Century 21st Century

Obs.

Comparison of 20th and 21st Century AMV in CMIP3 Models

20th Century 21st Century

(Stippled regions indicate at least 15 out of 19 models have the same sign regression)

Forced Internal

Regression of Ts and Precip on Forced versus Internal NASST Variability (Annual Mean 20th Century)

Ts

Precip Precip

Ts

(Stippled regions indicate at least 15 out of 19 models have the same sign regression)

Forced Internal

Regression of Ts and Precip on Forced versus Internal NASST Variability (Annual Mean 21st Century)

Ts

Precip Precip

Ts

Ts Regression (Annual)Forced Internal (AMV)

Obs.

20th

21st

Obs.

20th

21st

Precip Regression (Annual)Forced Internal (AMV)

Obs.

20th

21st

Obs.

20th

21st

AMV in Pre-Industrial CMIP3 Runs Can we reproduce AMV spatial patterns

and climate impacts in a model without radiatively forced changes?

If so, what are the typical time scales and amplitude of the AMV?

What are the circulation features associated with the AMV?

AMV AmplitudePre-Industrial

20th Century

21st Century

AMV Surface Temperature Regression(Annual Mean)

Stippling indicates sign agreement for 15 out of the 20 models or 5% significance (obs.)

Pre-Industrial 21st Century

20th Century Observations

AMV Precipitation Regression(Annual Mean)

Stippling indicates sign agreement for 15 out of the 20 models or 5% significance (obs.)

Pre-Industrial 21st Century

20th Century Observations

AMV and AMOC

Atlantic Meridional Overturning Streamfunction

Summary A robust spatial pattern of AMV is identified for

observations, CMIP3 models’ pre-industrial, 20th and 21st Century simulations, despite the differing temporal scales between models and observations.

The AMV spatial pattern is characterized by a comma-shaped SST pattern over the North Atlantic with the largest amplitude in the sub-polar region which extends to the tropical Atlantic along the east side of the basin.

The precipitation patterns associated with AMV are remarkably similar for the pre-industrial and 20th Century model simulations, and to some extent for the 21st Century simulations and observations. The positive phase of AMV (warm North Atlantic) is associated with northward shifted Atlantic ITCZ, increased rainfall over Sahel and eastern tropical Pacific, and dry condition over tropical south Atlantic, North America and Australia.