Noise Forcing and Coupled Feedbacks in Low Frequency North Atlantic SST Variability

1
Noise Forcing and Coupled Feedbacks in Low
Frequency North Atlantic SST Variability

• Edwin K. Schneider
• George Mason University
• Climate Dynamics Program/Department
• COLA

2
Coauthors

• Zhaohua Wu
• COLA
• Meizhu Fan
• GMU
• Ben P. Kirtman
• GMU/COLA

3
Decadal Variability

• What is the potential predictability (perfect
  model and data) of the decadal variability?
• What mechanisms are responsible for observed
  decadal variability of SST?
• Filtering of weather noise (Hasselmann)
• Oscillations due to coupled atmosphere-ocean
  feedbacks (e.g. as in simple models of ENSO)
• External forcing (solar, volcanic, anthropogenic,
  cryosphere)

4
Modeling Decadal Variability

• In the pyramid of models, the coupled
  ocean-atmosphere general circulation model (which
  sits at the top) is
• the best tool for realistically simulating
  decadal variability
• the least useful for mechanistic diagnosis of
  observed decadal variability

5
Experimentation with a CGCM

• Change external forcing
• Change initial conditions
• No control over atmospheric noise (weather) due
  to chaotic nature of atmospheric dynamics
• No control over time evolution of SST or surface
  fluxes (part of solution)

6
Traditional Tools

• Forced response of models to observations of the
  evolution of the boundary conditions.
• Forms the basis of model verification,
  predictability, and dynamical understanding of
  the atmosphere and ocean (separately)
• AGCM
• Specify time evolution of SST from observations
• OGCM
• Specify time evolution of surface fluxes from
  observations
• Coupled model with evolving boundary conditions
• ICM
• Specify weather noise (because stable component
  models).

7
A New Class of Model

• A CGCM-class model has been designed which has a
  realistic representation of dynamics, physics,
  and coupled feedbacks, but which can be used to
  ask the same mechanistic questions as the ICM
• I-CGCM Intermediate CGCM
• FA(SST) N
• where A is an AGCM-class model without noise
  and noise N can be added externally

8
Our I-CGCM The Interactive Ensemble
9

• Each atmospheric model is forced by the same SST
  and produces surface fluxes
• FiA(SST)Wi(SST)
• Forced response A
• Weather noise Wi different for each model. Wi
  locally has properties like random noise Ni

10

• Ensemble mean flux F
• FA(SST)N
• As the number of atmospheres n becomes large, N?0
• If variance of the weather noise is ViV for each
  AGCM, then the variance of the ensemble mean
  noise V is
• V?V/n

11
A Question for the Interactive Ensemble

• Consider the observed North Atlantic decadal
  variability of SST 1950-present. What were the
  roles of atmospheric noise, coupled feedbacks,
  outside influences, in producing this
  variability?
• Was it entirely noise forced?
• Was it due to some unstable coupled air-sea mode?
  
• What was the role of the different coupled
  feedbacks?

12
Prior Application

• Diagnosis of the mechanism for low frequency
  North Atlantic SST variability in a free running
  CGCM simulation.
• I would have spoken about this at the 2004 CRCES
  Decadal Variability Workshop on the Big Island,
  but was unable to attend at the last minute.
• Wu, Z., E. K. Schneider, and B. P. Kirtman, 2004
  Causes of low frequency North Atlantic SST
  variability in a coupled GCM. Geophys. Res.
  Lett., 31, L09210, doi10.1029/2004GL019548.

13
Models

• COLA AGCM T42, 18 levels
• GFDL MOM3 OGCM
• Standard ARCs physics
• Medium resolution 1.5?, better near equator, 25
  levels
• World ocean (non-polar) 74?S - 65?N
• Climatological sea ice
• Anomaly coupled
• I-CGCM 6 copies of AGCM (initial conditions of
  each copy differ to produce uncorrelated weather
  noise)

14
Experiments

• Century long simulation with CGCM.
• 1000 year free simulation with the Interactive
  Ensemble.
• Low frequency SST variability due to coupled
  instabilities ENSO.
• Interactive Ensemble forced in the North Atlantic
  with a specific realization of the weather noise
• The evolution of weather noise of the real
  climate system as estimated from NCEP reanalysis
  1948-2001).
• This should have about 6 times the variance of
  the filtered internal noise in the I-CCGM.
• Results from I-CGCM can be compared with low
  frequency variability that actually occurred.

15
Quantitative Evaluation of Role of Noise Forcing

• Consider the ratio of SST variance
• RV(CGCM)/V(ICGCM)
• There are 6 members of the atmospheric model
  ensemble
• Therefore noise forcing of SST variability should
  be approximately 6x larger in CGCM than in ICGCM
• In regions where SST variability is force by
  atmospheric noise, R?6
• In regions where SST variability is due to
  coupled dynamics, or internal variability of the
  ocean, R?1.

16
Ratio of SST VarianceRV(CGCM)/V(ICGCM)
17
Noise-Forced Experiment

• Force Interactive Ensemble with observed
  atmospheric noise in the North Atlantic.
• That part of the SST variability forced by the
  noise will be reproduced in detail (except for
  errors in the analyzed fluxes and in the models).
• That part of the SST variability due to unstable
  coupled processes will not be reproduced.
• Maybe something else will happen. After all, this
  is a global model.

18
Outline of the History of the Theory of the Role
of Weather Noise in Forcing Low Frequency Climate
Variability

• Einsteins 1905 theory of Brownian motion.
• Hasselmann (1976) reinterpretation of (1) as the
  red noise response of a passive ocean forced by
  white noise.
• Barsugli and Battisti (1998) extension of (2) to
  include the coupled response of the atmosphere to
  ocean.

19
Determination of the Evolution of Atmospheric
Noise

• Subtract forced surface fluxes from NCEP
  reanalysis total surface fluxes.
• Residual is surface fluxes from weather noise
• Forced surface fluxes are from 10 member AGCM
  ensemble forced by observed SST evolution
  1950-2000.
• In the context of the simple model of Barsugli
  and Battisti (1998), this can be proved to be the
  correct procedure to evaluate the noise, given
  the total surface flux and a perfect coupled
  model.

20
WARNING !!!

• Preliminary Results Follow

21
Point Correlation Noise Forced Simulation with
Observed SST, JFM
22
NA SST Index JFM 30-40N, 70-50W
23
Correlations Noise Forced Simulation with its SST
Index
SLP
Ts
P
Z500
24

• There are 2 things going on.
• Local response to North Atlantic forcing
• ENSO
• There appears to be an association of North
  Atlantic pattern with ENSO-warm.
• In the North Atlantic, the SLP appears to be
  explicable by the thermal response to SSTA
  (warm/low, cold/high).

25
Correlation Reanalysis with Observed SST Index
SLP
Ts
P
Z500
26

• There are 2 things going on.
• North Atlantic SST variability
• ENSO
• There appears to be an association of North
  Atlantic pattern with ENSO-cold.
• In the North Atlantic, the SLP is not explicable
  by the thermal response. What is this SLP
  pattern?
• The noise that forced the SSTA?
• ENSO?

27
Correlation No Noise Simulation SST with Its
Index (100 years data)
SLP
Ts
P
Z500
28

• There is one thing going on (to the extent that
  we have filtered out noise).
• ENSO
• There appears to be an association of the North
  Atlantic pattern with ENSO-cold, as in
  reanalysis.
• In the North Atlantic, the SLP is not explicable
  by the thermal response. This SLP pattern is
  ENSO.

29
No Noise Simulation Correlations with NINO3.4
SSTA Pure ENSO (100 yrs. data)
SLP
Ts
P
Z500
30
A Story to Explain Results

• The North Atlantic SST variability is tied to
  (forced by) ENSO.
• By removing the forced component from the North
  Atlantic forcing, we have decoupled the North
  Atlantic from the remote SST forcing.
• If there we discover a serious error in our
  calculations and the results change, we will make
  up another story.