1. Lamont-Doherty Earth Observatory
Detection of Forced and Natural Atlantic Multidecadal Variability in Coupled Models and Observations
Mingfang Ting
Yochanan Kushnir
Cuihua Li
Lamont-Doherty Earth Observatory
Columbia University

2. AMO Index (7.5W-75W, 0-60N, ocean only) for Models and Observations

3. Detrended AMO Index

4. Questions:
How much of the multi-decadal Atlantic SST variability is caused by internal dynamics and how much is externally forced?
Can model ensemble average be taken as forced signals?
How do one separate the natural and forced components in observations?

5. Focusing on IPCC models with multiple realizations…
NCAR CCSM – 8 members
GFDL CM2.1 – 5 members
GISS_EH – 6 members
GISS_ER – 9 members
MRI – 5 members
NCAR PCM – 4 members

6. AMO index for NCAR model (dotted line – ensemble average)

7. AMO for GFDL model (dotted line – ensemble average)

8. How much of the ensemble average is forced signal?
If infinite number of realizations are available, then ensemble average represents true forced signal
If only a small number of realizations are available, ensemble average contains internal variability as well as forced signal

9. EOF analysis of NCAR model’s ensemble average 20th century simulations
Correlation between PC and surface temperature
85% 5%
1.5%

10. GFDL model with 5 ensemble members
Correlation between PC and surface temperature
73%
14% 3%

11. PC1 of the ensemble average 20th Century simulations may be taken as forced signal

12. Natural and Forced Variability of AMO in NCAR Model
Run 7
Run 1
Run 4
Run 2
Run 5
Run 8
Run 3
Run 6
Forced

13. Natural and Forced Variability of AMO in GFDL Model
Run 1
Run 3
Run 5
Run 2
Run 4
Forced

14. What about observations…
Only one realization, no ensemble average is available
But, we can identify regions where most of the variability is forced

15. Ratio of variance
Total variance of all member ensembles lumped together, sT2
Variance of the ensemble average, sa2
Variance of internal variability,
sI2= sT2- sa2
Ratio of forced and total variance

16. Ratio of Variance for 20th Century IPCC Coupled Models
NCAR (8)
GFDL (5)
MRI (5)
GISS_ER (9)
GISS_EH (5)
PCM (4)
Ratio of Variance for 20th Century IPCC Coupled Models

17. For Indian Ocean, almost all models show that about 85% of the total variance is forced

18. Indian Ocean SST Index (50E-100E, 20S-10N)
CCSM
GFDL
GISS_ER
MRI
GISS_EH
PCM

19. Indian Ocean SST index (ensemble mean + observations)

20. Correlation between IO index and global TS in observations

21. Forced versus Natural AMO in Observations
Forced Signal
AMO Forced Signal
MDR Forced Signal
Natural Signal
AMO Natural Signal
MDR Natural Signal

22. Summary
Ensemble averages of a limited number of realizations do not necessarily represent forced signal.
EOF analysis of the ensemble average can be a useful way to separate the forced and natural components.
Indian Ocean responds primarily to radiative forcing and a large percentage of the total variance is forced. Thus it can be used as the footprint of the forced signal in observations.
Using Indian Ocean SST index to separate the forced and natural AMO and MDR, we found that the forced and natural components are of comparable magnitude in both cases.

24. Regression of global SST on PC1
NCAR CCSM
GFDL CM2.1
GISS EH
MRI
GISS ER
PCM

25. Spatial structure of natural AMO