Initialization Strategies for the CFS

1
Initialization Strategies for the CFS

• Ben Kirtman, Eric Altshuler, Kathy Pegion
• George Mason University and the Center for
  Ocean-Land-Atmosphere Studies

2
The Initial Condition Problem

• Best State Estimate
• Data Assimilation in the Separate Component
  Models
• Maybe Not Ready to Do the Coupled Assimilation
  Problem
• Coupled Model Climate ? Observed Climate
• Anomaly Initialization
• Coupled Modes of Coupled Model ? Observed
  Coupled Modes
• Initializing the Coupled Modes
• Identifying the Coupled Modes EOFs, SVDs,
• Context of the Forecast Environment

3
GODAS
CFS Simulation
Drift
4
GODAS
CFS Simulation
Drift
5
Forecast-GODAS
Forecast
GODAS
SST Climotologies
6
Forecast
Simulation
Simulation-GODAS
SST Climotologies
7
Forecast
GODAS
Forecast-GODAS
D20C Climotologies
8
Forecast
Simulation
Simulation-GODAS
D20C Climotologies
9
Coupled Ocean-Atmosphere Initialization
CFS Forecast
CFS Simulation
Observations
SSTA Standard Deviation as Function of Month
10
The Initial Condition Problem

• Best State Estimate
• Data Assimilation in the Separate Component
  Models
• Maybe Not Ready to Do the Coupled Assimilation
  Problem
• Coupled Model Climate ? Observed Climate
• Anomaly Initialization
• Coupled Modes of Coupled Model ? Observed
  Coupled Modes
• Initializing the Coupled Modes
• Identifying the Coupled Modes EOFs, SVDs,
• Context of the Forecast Environment

11
JAN 1997
JAN 1988
Anomaly Initialization
CFS Control
GODAS
12
Why Did the Forecasts Bust?

• Predictability Issue
• Not Enough Ensemble Members
• What Actually Happened was an Outlier
• Westerly Wind Burst
• Something Wrong with Initial Conditions
• Anomaly Initialization Did Not Help
• Coupled Modes of the Coupled Model

13
CFS Forecast
Ensemble Mean
GODAS
14
ENSO Precurser Three Month Lead
CFS Forecasts
GODAS
CFS Simulation
15
ENSO Precurser Six Month Lead
CFS Forecasts
GODAS
CFS Simulation
16
ENSO Precurser Nine Month Lead
CFS Forecasts
GODAS
CFS Simulation
17
The Initial Condition Problem

• Best State Estimate
• Data Assimilation in the Separate Component
  Models
• Maybe Not Ready to Do the Coupled Assimilation
  Problem
• Coupled Model Climate ? Observed Climate
• Anomaly Initialization
• Coupled Modes of Coupled Model ? Observed
  Coupled Modes
• Initializing the Coupled Modes
• Identifying the Coupled Modes EOFs, SVDs,
• Context of the Forecast Environment

18
Identifying the Coupled Modes

• Forecast Environment
• Analogue Approach
• Heat Content/D20C Analogues Dont Work
• Coupled Modes of Coupled Model are Not the Same
  as the Observations
• Use Long Simulation to Identify Best Fit SST
  Evolution (9 Months)
• Relate Analogue Initial State to Observed
  Initial State

19
Initial Condition D20C Analogues Do Not Work
SSTA GODAS
SSTA Forecast
SSTA Best D20C Analogue
20
Initial Condition D20C Analogues Do Not Work
D20C GODAS
D20C Forecast
D20C Best D20C Analogue
21
Forecast Evolution Analogues Do Work
GODAS SSTA
SSTA Best SST Evolution Analogue
22
Forecast Evolution Analogues Do Work
SSTA Best SSTA Evolution Analogue
GODAS SSTA
23
Forecast Evolution Analogues Do Work
GODAS D20C
D20C Best SSTA Evolution Analogue
24
Forecast Evolution Analogues Do Work
GODAS D20C
D20C Best SSTA Evolution Analogue
25
JAN1988
GODAS D20C
D20C Best SST Evolution Analogue
JAN1997
26
Initializing with the Coupled Modes

• Use Long Simulation to Identify Best Fit SST
  Evolution (9 Months)
• Relate Analogue Initial State to Observed
  Initial State
• Singular Value Decomposition
• Project, Reconstruct, Re-Forecast (Anomaly
  Initialization)

27
Equatorial Pacific SSTA Forecast
Initialized Coupled Modes
CFS Control
GODAS
28
Nino34 SSTA Evolution
CFS Control
Initialized Coupled Modes
GODAS
29
Equatorial Pacific D20C Forecast
Initialized Coupled Modes
CFS Control
GODAS
30
Equatorial Pacific SSTA Forecast
Initialized Coupled Modes
CFS Control
GODAS
31
Nino34 SSTA Evolution
Initialized Coupled Modes
CFS Control
GODAS
32
Equatorial Pacific D20C Forecast
Initialized Coupled Modes
CFS Control
GODAS
33
Conclusions

• Anomaly Initialization Not Much Help
• Coupled Modes in Observations are not the Same as
  the Coupled Modes in the Model
• ENSO Precurser, D20C Analogues Dont Work
• SSTA Evolution Analogues Show Promise
• Need More Longer Simulations (Easy)
• Need Many More Forecast Tests (Difficult)
• Predictability Issues
• WWB, Ensemble Size
• Expand the Approach to the Coupled System
• Blending or Nudging