Transcript Interactive Ensemble

Disentangling the Link Between Weather
and Climate
Ben Kirtman
University of Miami-RSMAS
Noise and Climate Variability
• What Do We Mean By “Noise” and Why
Should We Care?
– Multi-Scale Issue
• How to Examine Noise within Context of
a Coupled GCM- Interactive Ensemble
– Typical Climate Resolution (T85, 1x1)
– Ex: Atmospheric Noise, Oceanic Noise,
ENSO Prediction, Climate Change
• Resolution Matters
– Noise Aliasing
• Quantifying Model Uncertainty (Noise)
Why is Noise an Interesting Question?
• Large Scale Climate Provides Environment
for Micro- and Macro-Scale Processes
– Local Weather and Climate: Impacts, Decision
Support
• Micro- and Macro-Scale Processes Impact
the Large-Scale Climate System
– Interactions Among Climate System Components
– Justification for High Resolution Climate Modeling
• But, this is NOT the Definition of Noise
– Noise Occurs on all Space and Time Scales
How Should Noise be Defined?
• Use ensemble realizations
– Ensemble mean defines “climate signal”
– Deviation about ensemble mean defines
Noise
– Climate signal and noise are not
Independent
– Examples:
• Atmospheric model simulations with prescribed SST
• Climate change simulations
SST Anomaly JFMA1998
SST Anomaly JFMA1989
Tropical Pacific
Rainfall (in box)
Different SST 
Different tropical atmospheric mean response
Different characteristics of atmos. noise
Modeling Weather & Climate
Interactions
• Previously, this required ad-hoc assumptions
about the weather noise and simplified
theoretically motivated models
• We adopt a coupled GCM approach
– Weather is internally generated
• Signal-noise dependence
– State-of-the-art physical and dynamical
processes
 Interactive Ensemble
Ensemble of N
AGCMs all
receive same
OGCM-output
SST each day
AGCM1
AGCM2
Sfc Fluxes1
Sfc Fluxes2
AGCMN
•••
Sfc FluxesN
average (1, …, N)
Average N
members’ surface
fluxes each day
Ensemble Mean Sfc Fluxes
OGCM receives ensemble average of
AGCM output fluxes each day
SST
OGCM
Interactive Ensemble Approach
Interactive Ensemble
• Ensemble realizations of
atmospheric component
to isolate “climate signal”
M=1
M=2
Ensemble mean = Signal + 
• Ensemble mean surface fluxes
coupled to ocean component
M=3
– Ensemble average only applied
at air-sea interface
– Ocean “feels” an atmospheric
state with reduced weather noise
M=4, 5, 6
M = number of atmospheric ensemble members
Control Simulation: CCSM3.0 (T85, 1x1)
300-year (Fixed 1990 Forcing)
Interactive Ensemble: CCSM3.0
(6,1,1,1)
Fixed
1990 GHG
COLA CCSM-IE run
Full CCSM
If all SST variability is
forced by weather
noise, the ratio of SST
variance (IE
CGCM)/(Standard
CGCM) is expected to
be 1/6 and the ratio of
standard deviations to
be 0.41.
Variability Driven
by Noise
Coupled Feedbacks?
Ocean Noise?
Ocean and Atmosphere Interactive
Ensemble
OGCMn Ensemble Member SST
AGCM1
●● ●
OGCM Ensemble Mean SST
Ensemble Mean
Fluxes
OGCM1
AGCMN
Ensemble Mean
SST
●● ●
OGCMM
AGCMn Ensemble Member Flux
AGCM Ensemble Mean Flux
Impact of Ocean Internal Dynamics with Coupled Feedbacks
SSTA Variability Due to Ocean Internal Dynamics
Reduced
Enhanced
Climate Change Problem
Interactive Ensemble
Control Ensemble
Interactive Ensemble
Climate of the 20th Century:
Interactive - Control Ensemble
Global Mean
Temperature
Regression
Control Ensemble
Interactive Ensemble
Local Air-Sea Feedbacks: Point Correlation SST and
Latent Heat Flux
“Best” Observational Estimate
Coupled Model Simulation
Why Does ENSO Extend Too Far To The West?
The Weather and Climate Link?
Conceptual Model
dTa
  (To  Ta )  F
dt
dTo
  (Ta  To )  To
dt
dTa
  (To  Ta )
dt
dTo
  (Ta  To )  To  F
dt
Atmos → Ocean
<HF,(dSST)/dt>
<HF,SST>
Ocean → Atmos
Conceptual Model
Atmos → Ocean
Atmosphere Forcing Ocean:
• <HF(t), SST(t) > < 0
• <HF(t), d(SST(t))/dt> <0
<HF,(dSST)/dt>
<HF,SST>
Ocean → Atmos
Ocean Forcing Atmospere:
• <HF(t), SST(t) > > 0
• <HF(t), d(SST(t))/dt> > 0
<HF,SST>
<HF,dSST>
GSSTF2 Observational Estimates
Area Averaged Fields Eastern
Equatorial Pacific from GCMs
Prescribed SST is Reasonable
In Eastern Equatorial Pacific
Conceptual Model: Ocean →Atmos
Conceptual Model: Atmos →Ocean
<HF,dSST>
<HF,SST>
GSSTF2 Observational Estimates
Area Averaged Fields Central/Western
Equatorial Pacific
CGCM Variability is too
Strongly SST Forced
Western Pacific Problem
• Hypothesis: Atmospheric Internal
Dynamics (Stochastic Forcing) is
Occurring on Space and Time Scales that
are Too Coherent
 Too Coherent Oceanic Response
 Excessive Ocean Forcing Atmosphere
 Test: Random Interactive Ensemble
Ensemble of N
AGCMs all
receive same
OGCM-output
SST each day
AGCM1
AGCM2
Sfc Fluxes1
Sfc Fluxes2
AGCMN
•••
Sfc FluxesN
average (1, …, N)
Average N
members’ surface
fluxes each day
Ensemble Mean Sfc Fluxes
OGCM receives ensemble average of
AGCM output fluxes each day
SST
OGCM
Interactive Ensemble Approach
Ensemble of N
AGCMs all
receive same
OGCM-output
SST each day
AGCM1
AGCM2
Sfc Fluxes1
Sfc Fluxes2
•••
AGCMN
Sfc FluxesN
rand (1, …, N)
Randomly select 1
member’s surface
fluxes each day
Selected Member’s Sfc Fluxes
OGCM receives output of single,
randomly-selected AGCM each day
SST
OGCM
Random Interactive Ensemble Approach
Nino3.4 Power Spectra
Moderate Stochastic
Atmospheric Forcing
Period (months)
Increased Stochastic
Atmospheric Forcing
Period (months)
Reduced Stochastic
Atmospheric Forcing
Period (months)
Increasing Stochastic Atmospheric
Forcing Increase the ENSO Period
Random IE
Control
4
4
3
3
2
2
1
1
0
0
-1
-1
-2
-2
-3
-3
-4
-4
Nino34 Regression on Equatorial Pacific SSTA
Random IE
Control
Nino34 Regression on Equatorial Pacific Heat Content
Contemporaneous Latent Heat Flux - SST Correlation
Observational
Estimates
Control
Coupled Model
Increased “Randomness”
Coupled Model
Random Interactive Ensemble:
Increased the Whiteness of the
Atmosphere forcing the Ocean
Noise and Climate Variability
• What Do We Mean By “Noise” and Why
Should We Care?
– Multi-Scale Issue
• How to Examine Noise within Context of
a Coupled GCM?
– Typical Climate Resolution (T85, 1x1)
– Atmospheric Noise, Oceanic Noise,
Climate Change Problem
• Resolution Matters
– Noise Aliasing
• Quantifying Model Uncertainty (Noise)
Equatorial SSTA Standard Deviation
Low Resolution:
IE
Control
Lower Resolution:
IE
Control
Understanding Loss of Forecast
Skill
• What is the Overall Limit of Predictability?
• What Limits Predictability?
– Uncertainty in Initial Conditions: Chaos within
Non-Linear Dynamics of the Coupled System
– Uncertainty as the System Evolves: External
Stochastic Effects
• Model Dependence?
– Model Error
CFSIE - Reduce Noise Version (interactive ensemble) of CFS
RMS(Obs)*1.4
CFSIE
RMSE
CFS
Spread
CFS
RMSE
CFSIE
Spread
CFSIE - Reduce Noise Version (interactive ensemble) of CFS
Predictability Estimates
Worst Case: Initial Condition
Error (A+O) + Model Error
Worst Case
Best Case
Best Case: Initial Condition
Error (A) + No Model Error
Better Case: Initial Condition
Error (A) + Model Error
Better Case
Best Case
Noise and Climate Variability
• What Do We Mean By “Noise” and Why
Should We Care?
– Multi-Scale Issue
• How to Examine Noise within Context of
a Coupled GCM?
– Typical Climate Resolution (T85, 1x1)
– Atmospheric Noise, Oceanic Noise,
Climate Change Problem
• Resolution Matters
– Noise Aliasing
• Quantifying Model Uncertainty
(Noise)
Multi-Model Approach to
Quantifying Uncertainty
• Multi-Model Methodologies Are a Practical
Approach to Quantifying Forecast Uncertainty
Due to Uncertainty in Model Formulation
• No Determination of Which Model is Better Depends on Metric
• Taking Advantage of Complementary or
Orthogonal “Skill”
• Taking Advantage of Orthogonal Systematic
Error
Time Mean Equatorial Pacific SST
COLA
COLA HF+CAM Winds
Obs
CAM
COLA Winds+CAM HF
ENSO Heat Content Anomalies
OBS
COLA
CAM
COLA HF + CAM Winds
COLA Winds + CAM HF
Noise and Climate Variability
• What Do We Mean By “Noise” and Why
Should We Care?
– Multi-Scale Issue
• How to Examine Noise within Context of
a Coupled GCM- Interactive Ensemble
– Typical Climate Resolution (T85, 1x1)
– Ex: Atmospheric Noise, Oceanic Noise,
ENSO Prediction, Climate Change
• Resolution Matters
– Noise Aliasing
• Quantifying Model Uncertainty (Noise)