Role of Atmosphere-Ocean Role of Atmosphere-Ocean InteractionInteraction
And Seasonal PredictabilityAnd Seasonal Predictability
International Workshop on Variability and Predictability of the Earth Climate System, 26-27 Jan 2005, Japan
In-Sik Kang and Kyung JinIn-Sik Kang and Kyung Jin
Climate Environment System Research CenterClimate Environment System Research CenterSeoul National UniversitySeoul National University
Contents Contents
I. Limitation of dynamic predictability in tier-two system
II. Local atmosphere-ocean interaction
Ⅲ. Local and remote influence in coupled system
Ⅳ. Examination of predictability in tier-one vs. tier-two
Prescribe SST as boundary condition
Atmosphere Atmosphere
OceanSST Prediction
Current activities of seasonal prediction: Tier-2 vs. Tier-1Current activities of seasonal prediction: Tier-2 vs. Tier-1
Tier-two system Tier-one system
SST is prescribed as boundary condition
Atmosphere-ocean interaction is embodied
SST predictionsystem
CGCMComponent
Feature
AGCM
Experimental Design and Participated Models Experimental Design and Participated Models
CLIVAR Asian-Australian Monsoon Atmospheric GCM Intercomparison Project CLIVAR Asian-Australian Monsoon Atmospheric GCM Intercomparison Project
Group Country Numerics Convection ParameterizationCOLA USA R40L18 Relaxed Arakawa-Schubert (RAS, Moorthi and Suarez, 92) DNM Russia 4o×5o, L21 Betts (86)
GEOS USA 2o×2.5o, L43 RAS (Moorthi and Suarez, 92) GFDL USA T42L18 RAS (Moorthi and Suarez, 92) IAP China R15L9 MCA (Manabe et al., 65)IITM India 2.5o×3.75o, L19 Mass flux penetrative convection scheme (Gregory and Rowntree, 90) MRI Japan 4o×5o, L15 Arakawa-Schubert, Tokioka et al. (88)
NCAR USA T42L18 Mass flux scheme (Zhang and McFarlane, 95)NCEP USA T42L28 RAS (Moorthi and Suarz, 92) SNU Korea T31L20 Simplified Arakawa-Schubert
SUNY USA 4ox 5o, L17 Modified Arakawa-Schubert
Institute Model Resolution Experiment Type Ensemble MemberJMA JMA T63L40 SMIP 10KMA GDAPS T106L21 SMIP 10NCEP NCEP T62L28 SMIP 10
NASA/NSIPP NSIPP 2ox2.5o L43 AMIP 9SNU GCPS T63L21 SMIP 10
APEC Climate Network (APCN) participantsAPEC Climate Network (APCN) participants
- 10 ensemble simulations from Nov1996 to Aug98
- 21 year simulation from 1979 to 1999
DJF97/98 Precipitation Anomaly for Each Model EnsembleDJF97/98 Precipitation Anomaly for Each Model Ensemble
CLIVAR Asian-Australian Monsoon Atmospheric GCM Intercomparison Project CLIVAR Asian-Australian Monsoon Atmospheric GCM Intercomparison Project
Climate signals caused by external forcing
Intrinsic transients due to natural variability
Forced Variance Free Variance Signal-to-noise
N
i
n
jiij XX
nN 1 1
2)()1(
1
N
ii XX
N 1
2)(1
1 Theoretical limit of
predictability
Analysis of Variance of 21-yr JJA Rainfall in Tier-Two systemsAnalysis of Variance of 21-yr JJA Rainfall in Tier-Two systems
Forced Variance Error Variance Forced/Error Variance
Error Variance of 21-yr JJA Rainfall in Tier-Two systemsError Variance of 21-yr JJA Rainfall in Tier-Two systems
AGCMs show systematic error over the western North Pacific during summer.
Area averaged correlation coefficientsArea averaged correlation coefficients
El Nino region (10oS-5oN, 80oW-180oW)
Western North Pacific (5-30oN, 110-150oE)
Predictability of JJA Precipitation in Tier-Two systemsPredictability of JJA Precipitation in Tier-Two systems
Correlation with JJA observed and simulated rainfall during 1979-99Correlation with JJA observed and simulated rainfall during 1979-99
(5 model composite)
Wrong model physics?Absence of air-sea
interaction?
Systematic error in tier-two system
Model Inability Modeling Strategy
Local Air-sea interactionLocal Air-sea interaction
Observed and simulated air-sea interaction
Local air-sea interaction processes
Climate Environment System Research CenterClimate Environment System Research CenterClimate Environment System Research CenterClimate Environment System Research Center
JJA SST-rainfall relationshipJJA SST-rainfall relationship
Correlation between JJA precipitation and SST during 1979-1999
(a) MME
(d) NCEP
(b) JMA
(e) NSIPP
(c) KMA
(f) SNU
Air-Sea InteractionAir-Sea Interaction
Lead-lag correlation between SST and rainfall pentad data during 1982-1999
Rainfall lead Rainfall lag> -20 -15 -10 -5 0 +5 +10 +15 +20 <
Only more than 95% significance level is shaded
Atmosphere forces the ocean where the correlation coefficients between rainfall and SST show negative.
JJA
-30 -20 -10 0 +10 +20 +30
days
Rainfall lead Rainfall lag
Western North Pacific (5-30N, 110-150E)
95% significance level
Seasonal March of Air-Sea Interaction and PredictabilitySeasonal March of Air-Sea Interaction and Predictability
(a) Observation
Correlation between observed and simulated rainfall
Month
La
titu
de
Time-latitude cross section averaged over 110-150oE during 1979-99
(b) SNU AGCM
Correlation between rainfall and SST
Contour denotes 95% significance level.
Experimental DesignExperimental Design
Atmosphere Atmosphere Atmosphere
Ocean(Full dynamics)
Perfect boundary Perfect boundary conditioncondition
Local air-sea Local air-sea interactioninteraction
Fully coupled Fully coupled systemsystem
SST
Slab ocean (No dynamics and
advection)
SSTObserved SST
heat flux, wind stress, fresh water flux
heat flux
AGCM(1950-1999, 4runs)
Mixed layer model+ AGCM(50 yrs, 4runs)
CGCM(75 yrs)
Experiment Integration Period
Runs
Resolution
Boundary Conditions Properties
AGCM 1950~1999(50 years) 4 T31L21
GISST and OISST and Sea
ice
Prefect boundary condition with observed SST
Mixed-layer Model
50 years 4 T31L21Climatological cycle OISST and Sea ice
Local air-sea interaction With slab ocean mixed-layer model
(Waliser et al. 1999)
CGCM 75 years 1 T42L21 NoFully coupled system
T42 SNU AGCM v2 (Kim, 1999)+MOM2.2 (Pacanowski et al., 1993)
Model Resolution Note
SNU AGCM T42L21 (2.8125oX2.8125o) No flux correction
MOM2.2 OGCM 1/3o lat. x 1o lon. over tropics(10S-10N), Vertical 32 levels
Ocean mixed layer model (Noh and Kim, 1999)
CGCM
Mixed-layer AGCM
Model Note
SNU AGCM T31L21 (3.75oX3.75o)
Slab ocean mixed-layer model
• Fixed depth slab ocean mixed-layer model without ocean dynamics and advection • Anomaly coupling per each time step (Waliser et al. 1999)
Model DescriptionModel Description SNU AGCM
Model Dynamics Physics
SNUAGCM
Spectral model using semi-implicit
method
• 2-stream k-distribution radiation scheme (Nakajima and Tanaka 1986)• Simplified Arakawa-Schubert cumulus convection scheme based on RAS scheme (Moorthi and Suarez 1992)• Orographic gravity-wave drag (McFarlane 1987)• Bonan’s land surface model (Bonan 1996)• Mon-local PBL/vertical diffusion (Holtslag and Boville 1993)• Diffusion-type shallow convection
THC
F
dt
Td
p
H : mixed layer depth = 50 m : density of sea water = 1022 kg/m3
Cp : heat capacity of sea water = 4000 J/kg·k : damping factor = (150day)-1
Model SST equation
Observation
Mixed Layer Model
Correlation between SST and PrecipitationCorrelation between SST and Precipitation
AGCM
JJA Atmosphere-Ocean InteractionJJA Atmosphere-Ocean Interaction
CGCM
Perfect boundary condition
Local air-sea
interaction without
dynamics
Air-sea interaction and Ocean dynamics
StrategyStrategy
Local air-sea interaction
• Thermodynamic processes • Except tropical eastern Pacific mixed-layer ocean model
Local air-sea interaction Remote forcing+
• Thermodynamic processes • Except tropical eastern Pacific mixed-layer ocean model
• Ocean dynamic processes • Tropical eastern Pacific Observed SST
Part Ⅰ
Part Ⅱ
What regulate the direction of air-sea interaction?
Part Ⅲ Fully coupled system Tier-two systemvs.
Influence on the extratropical circulation variability
Examination of real predictability
Consideration of radiative fluxesConsideration of radiative fluxes
COA anomalies by rainfall in mixed-layer model during 50 yearsCOA anomalies by rainfall in mixed-layer model during 50 years
(a) Surface short-wave fluxJJA DJF
(b) Surface long-wave flux
(c) (a) minus (b)
(d) Surface short-wave flux
(f) Surface long-wave flux
(g) (d) minus (f)
Positive for downward flux
COA = CORRELATION[A,B]*σB (Kang et al. 2001 JMSJ)
To measure an actual magnitude of quantity of B related to the reference data A
Consideration of radiative forcing Consideration of radiative forcing
JJA climatological cloud cover and ratiJJA climatological cloud cover and ratio of radiative fluxes o of radiative fluxes
Climatological total cloud cover
Rat
io o
f su
rfac
e lo
ng
-wav
e /
sh
ort
-wav
e fl
ux
Western North Pacific (5-30N, 110-170E) Eastern Pacific (15S-15N, 180E-80W) North Pacific (30-70N, 140E-120W)
Over the cloud heavy region having small climatological cloud cover such as western North Pacific, the ratio of surface long-wave flux by short-wave flux related with rainfall has smaller value than cloud free region. Rainfall cools the ocean surface well due to strong radiative cooling over those regions.
Y axis is ratio of radiative fluxes(COA of long-wave/short-wave flux)
Consideration of net surface fluxes Consideration of net surface fluxes
COA anomalies by rainfall in mixed-layer model during 50 yearsCOA anomalies by rainfall in mixed-layer model during 50 years
JJA DJF(a) Surface radiative flux (d) Surface radiative flux
(b) Surface latent heat flux (e) Surface latent heat flux
(c) (a) minus (b) (f) (d) minus (e)
Latent heat flux prevail Radiative flux prevail
Rainfall SSTSST
Summer hemisphere
Radiative flux > Latent heat flux
radiative cooling
Winter Hemisphere(DJF 10-30oN North Pacific, JJA Southern Indian Ocean)Radiative flux < Latent heat flux
Winter Hemisphere(DJF 30-50oN North Pacific)Radiative flux < Latent heat flux evaporative cooling
• Contour denotes net surface flux anomalies• Positive for downward flux
Opposite sign
Same sign
Shortwave flux has an important role to decrease the SST anomalies associated with increasing rainfall in summer hemisphere.
Except the region where ocean dynamics is important such as central and eastern Pacific, thermodynamic processes may work
AGCM cannot simulate the interaction atmosphere forces the ocean
Thermodynamic Processes of Local Air-Sea InteractionThermodynamic Processes of Local Air-Sea Interaction
Local air-sea interaction
• Thermodynamic processes • Except tropical eastern Pacific mixed-layer ocean model
Part Ⅰ
What regulate the direction of air-sea interaction?
Local and Remote Response in Local and Remote Response in Coupled SystemCoupled System
Characteristics of extratropical North Pacific variability as the air-sea coupled mode
Influence on the extratropical predictability
Climate Environment System Research CenterClimate Environment System Research CenterClimate Environment System Research CenterClimate Environment System Research Center
Consideration of predictability Consideration of predictability using coupled systemusing coupled system
Low potential predictability due to internal dynamics different from tropics Strong modal characteristics of SST anomalies North Pacific Ocean has a rich spectrum of interannual to interdecadal climate variability (Wallace et al. 1993; Trenberth and Hurrel 1994; Latif and Barnett 1996; Jin 1997).
Local coupling can influence on the atmospheric variability?
Tropical SST Anomaly
North Pacific SST Anomaly
Extratropicalcirculation
over North PacificDownstreamLocal air-sea
interaction
Remote influence
Influence from tropics and extratr
opics
Local air-sea interaction Remote forcing+
• Thermodynamic processes • Except tropical eastern Pacific mixed-layer ocean model
• Ocean dynamic processes • Tropical eastern Pacific Observed SST
Part Ⅱ
For the focus on the summertime extatropical North Pacific
Observed North Pacific modeObserved North Pacific mode
(a) 1st mode of EOF (b) PC time series (North Pacific Index)
(c) Lag Cor [NPI(JJA), NINO3.4(JJA- )]
ENSO Impact It has identical interannual variability different from NINO3.4 index, even though it has negative lag relation with previous spring NINO3.4 index.
Origin of North Pacific SST variability Air-sea coupled feedback (Frankignoul 1985; Norris et al. 1998; Lau et al. 2003) Tropical remote forcing (Pan and Oort 1990; Lau and Nath 2001) Stochastic atmospheric forcing (Blade, 1997; Barsugli and Battisti 1998) Delayed feedback provided by slow ocean dynamics (Latif and Barnet, 1996; Pierce et al. 1999)
Influence on the adjacent climate Summertime teleconnection patterns linking the rainfall anomalies over the North American to those of the East Asian monsoon and North Pacific SST are suggested by many authors (Nitta 1987; Huang and Sun 1992; Latif and Barnet 1996; Livezey and Smith, 1999; Lau and Weng 2000)
Experimental DesignExperimental Design
Observed SST
Interactive Ocean
AMIPAMIP(GOGA, Global (GOGA, Global Ocean Global Ocean Global Atmosphere)Atmosphere)
TOGA-MLTOGA-ML(Tropical Ocean (Tropical Ocean
Global Global Atmosphere-Mixed Atmosphere-Mixed
Layer)Layer)
MLML(Mixed Layer)(Mixed Layer)
AGCM(1950-1999, 4runs)
Mixed layer model (50 yrs, 4runs)
Extratropics
Tropics
Observed SST(Perfect
boundarycondition)
Mixed layer model+ Tropical SST
(1950-1999, 4runs)
Interactive Ocean(Local air-sea
interaction withimperfect SST)
+
Experiment Integration Period
Runs
Resolution
Boundary Conditions Properties
AMIP 1950~1999(50 years) 4 T31L21 GISST and OISST
and Sea icePrefect boundary condition
with observed SST
TOGA-ML 1950~1999(50 years) 4 T31L21 GISST and OISST
and Sea ice
Local air-sea interaction over extratropics + perfect tropics
With slab ocean mixed-layer model(Waliser et al. 1999)
ML 50 years 4 T42L21Climatological cycle of OISST and Sea
ice
Local air-sea interaction over whole globe
With slab ocean mixed-layer model(Waliser et al. 1999)
Observed and Simulated North Pacific modeObserved and Simulated North Pacific mode
TOGA-ML
Observation
NPI (North Pacific Index) is defined as the PC time series of 1st EOF mode of the 9-yr high filtered SST anomalies over the North Pacific
Simulated North Pacific local mode in TOGA-ML run shows similar relationship with ENSO, even though the interannual variability of NPI is different from observed with 0.3 correlation coefficients.
ML
Lag Cor [NPI(JJA), NINO3.4(JJA- )]
ObservationTOGA-MLML
NINO lead NINO lag
Most realistic reproducibility of North Pacific mode is simulated in TOGA-ML case with tropical forcing and local air-sea interaction.
7cases positive minus negative composite differences
Observed and Simulated North Pacific modeObserved and Simulated North Pacific mode
Observation AMIP ML
500 hPa geopotential height anomalies
Only local air-sea interaction
Perfect boundary condition
Local couplingTropical influence
TOGA-ML
7cases El Nino minus La Nina composite differences7cases El Nino minus La Nina composite differences
Observed and Simulated ENSO modeObserved and Simulated ENSO mode
Observation TOGA-ML
TOGA-ML minus AMIP
500 hPa geopotential height anomalies
AMIP
Most of AGCMs underestimate the intensity of PNA (Kang et al. 2003). Difference charts primarily portray the amplification of the signals: Affirmative characteristics of coupled system.
• Air-sea coupling effectively reduces the thermal damping of the atmosphere, thus amplifying the variability and enhancing the temporal persistence of extratropical atmospheric signals (Blade 1997; Barsugli and Battisti 1999; Lau and Nath 2001).
Observed and Simulated ENSO modeObserved and Simulated ENSO mode
Local coupling improves the amplitude and pattern of circulation over the North Pacific and the downstream, even though extratropical SST is imperfect.
Pattern Correlation with observed composite differencesPattern Correlation with observed composite differences
Extratropical Northern Hemisphere (0-360oE, 30-80oN)
Change of partial influence in coupled systemChange of partial influence in coupled system
Interactive ocean over extratropics enhances the local negative relationship.
Coupled system alleviates the overestimated remote influence in AMIP.
(a) Local SST
(b) NINO 3.4
(c) Local SST
(d) NINO 3.4
(e) Local SST
(f) NINO 3.4
Observation TOGA-ML
Partial Correlation between JJA SST and PrecipitationPartial Correlation between JJA SST and PrecipitationAMIP
Partial Correlation (Edward, 1979)
Calculate the partial effect of local SST and NINO 3.4 SST on the precipitation anomalies by removing relationship between local and NINO3.4 SST
223
213
2313123,12
11 RR
RRRR
Increased Potential Predictability: Perfect Model Correlation
North Pacific (120-280oE, 30-80oN)
North America (240-300oE, 30-60oN, land)
Perfect Model Correlation- Considering one member of the ensemble as an observation and making spatial correlation between the model observation and the ensemble mean of the other members. - Theoretical predictability limit using a hypothetical perfect model with no systematic error.
Perfect model pattern correlation of composite differencesPerfect model pattern correlation of composite differences
Local coupling increase the upper limit of theoretical potential predictability of atmospheric variability during ENSO years.
Local air-sea interaction Remote forcing+
• Thermodynamic processes • Except tropical eastern Pacific mixed-layer ocean model
• Ocean dynamic processes • Tropical eastern Pacific Observed SST
Part Ⅱ
Influence on the extratropical circulation variability
Influence of Air-sea Interaction Influence of Air-sea Interaction on the Real Predictabilityon the Real Predictability
The North Pacific SST variability has coupled feedback mechanism required both air-sea interaction and tropics-extratropics interaction. Accordingly, both local coupling and remote forcing is needed to simulation of circulation variability associated with this mode.
During ENSO years when strong remote influence and local air-sea interaction works together, the intensity and predictability of PNA is increased by local coupling.
In additions, PNA is potentially more predictable by increase of forced variance in coupled system. during ENSO years.
Without coupled process, the exact reproduction of extratropical atmospheric circulation such as PNA is impossible.
Examination of PredictabilityExamination of Predictabilityin Tier-One vs. Tier-Twoin Tier-One vs. Tier-Two
SNU SMIP/HFP (tier-two) vs. DEMETER (tier-one)
Climate Environment System Research CenterClimate Environment System Research CenterClimate Environment System Research CenterClimate Environment System Research Center
Model Experiments: Tier-one vs. Tier-two
Tier-one systemTier-one system Tier-two systemTier-two system Upper limit of Tier-two systemUpper limit of Tier-two system
DEMETER of 7 CGCMs SMIP2/HFP of SNU AGCM SMIP2 of SNU AGCM
Investigate seasonal real predictability based on the observed initial condition
and fully coupled GCM
Investigate seasonal real predictability based on the observed initial condition and predicted boundary
condition
Investigate seasonal potential predictability based
on the observed initial condition and observed
boundary condition
4 month x 20 year (1980-1999), 9 ensembles
4 month x 21 year (1979-1999), 6 ensembles
7 month x 21 year (1979-1999), 10 ensembles
7 CGCMs (CERFACS, ECMWF, INGV, LODYC, Meteo-France, Max-Plank Institute, UK
Met Office)
Development of European Multimodel Ensemble system for seasonal-to-interannual
prediction
Description of DEMETER (Tier-one Prediction System)
Development of European Multimodel Ensemble system for seasonal-to-interannual prediction One-tier prediction system using CGCM 9 ensemble members of 7 models 1980-1999 forecast
Institute AGCM Resolution OGCM Resolution
Atmosphere initial
conditionsEnsemble
generation
CERFACS ARPEGE T6331 Levels OPA 8.2 2.0x2.0
31 Levels ERA-40 Windstress and SST perturbations
ECMWF IFS T9540 Levels
HOPE-E 1.4x0.3-1.429 Levels ERA-40 Windstress and SST pe
rturbations
INGV ECHAM-4 T4219 Levels OPA 8.1 2.0x0.5-1.5
31 Levels
CoupledAMIP-typeexperiment
Windstress and SST perturbations
LODYC IFS T9540 Levels OPA 8.2 2.0x2.0
31 Levels ERA-40 Windstress and SST perturbations
Meteo-France ARPEGE T6331 Levels OPA 8.0 182GPx152GP
31 Levels ERA-40 Windstress and SST perturbations
MPI ECHAM-5 T4219 Levels MPI-OM1 2.5x0.5-2.5
23 Levels
Coupled run relaxed to
observed SSTs
Atmosphericconditions from the coupled initialization run (lagged method)
UK Met Office HadAM3 2.5x3.75
19 Levels
GloSea OGCM based on Had
CM3
1.25x0.3-12540 Levels ERA-40 Windstress and SST pe
rturbations
DEMETER CGCM Description
Tier 2 : SNU SST prediction system
3 month lead forecast
Tier 1 : DEMETER
Prediction skill – Correlation with observation of JJA SST
Prediction skill – Correlation with observation of JJA rainfall
Tier 2 : SNU AGCM
Tier 1 : DEMETER
Real predictability: Tier two vs. Tier one Real predictability: Tier two vs. Tier one
Pattern correlation of JJA rainfall anomalies during 1980-1999Pattern correlation of JJA rainfall anomalies during 1980-1999
Western North Pacific region (5-30oN, 110-150oE)
Global domain (60oS-80oN, 0-360oE)
Tier-one: 7 CGCMs average from DEMETER Each CGCMTier-two: SNU AGCM SMIP/HFP with predicted SSTTier-two (upper limit): SNU AGCM SMIP with observed SST
20 yrs mean
0.26
-0.04
20 yrs mean
Real predictability: Tier two vs. Tier one Real predictability: Tier two vs. Tier one
Pattern correlation of JJA rainfall anomalies during 1980-1999Pattern correlation of JJA rainfall anomalies during 1980-1999
Western North Pacific region (5-30oN, 110-150oE)
Global domain (60oS-80oN, 0-360oE)
Tier-one: 7 CGCMs average from DEMETER Each CGCMTier-two: SNU AGCM SMIP/HFP with predicted SSTTier-two (upper limit): SNU AGCM SMIP with observed SST
Real predictability: Tier two vs. tier one Real predictability: Tier two vs. tier one
Pattern correlation of JJA rainfall anomalies during 1980-1999Pattern correlation of JJA rainfall anomalies during 1980-1999
Western North Pacific region (5-30oN, 110-150oE)
Global domain (60oS-80oN, 0-360oE)
Tier-one: 7 CGCMs average from DEMETER Each CGCMTier-two: SNU AGCM SMIP/HFP with predicted SSTTier-two (upper limit): SNU AGCM SMIP with observed SST
20 yrs mean0.380.26
0.28-0.04
20 yrs mean
Real predictability: Tier two vs. tier one Real predictability: Tier two vs. tier one
Pattern correlation of JJA rainfall anomalies during 1980-1999Pattern correlation of JJA rainfall anomalies during 1980-1999
Western North Pacific region (5-30oN, 110-150oE)
Global domain (60oS-80oN, 0-360oE)
Tier-one: 7 CGCMs average from DEMETER Each CGCMTier-two: SNU AGCM SMIP/HFP with predicted SSTTier-two (upper limit): SNU AGCM SMIP with observed SST
20 yrs mean0.380.260.33
0.28-0.04-0.10
20 yrs mean