Documents
Resources
Learning Center
Upload
Plans & pricing Sign in
Sign Out

Jin

VIEWS: 3 PAGES: 45

									 4th International CLIVAR Climate of the 20th Century Workshop
13-15th March 2007, Hadley Centre for Climate Change, Exeter, UK




              The Evolution of
Lead-lag ENSO-Indian Monsoon Relationship
           in GCM Experiments

                        Emilia K. Jin and James L. Kinter III
                      Center for Ocean-Land-Atmosphere studies
                               George Mason University
                         Background and Objectives
  Objectives of this study
Focusing on ENSO-monsoon relationship,
To diagnose the problem in CGCM due to systematic error in ENSO characteristics
To suggest “pacemaker” as an alternative solution to improve the predictability of coupled system
To assess the advantages and shortcomings in “pacemaker” results


 International Climate of the Twentieth Century Project
 Characterize climate variability and predictability of the last ~130 years through analysis of
both observational data and general circulation models, in particular the period since 1949.


 “Pacemaker” Experiments
 The challenge is to design numerical experiments that reproduce the important aspects of
this air-sea coupling while maintaining the flexibility to attempt to simulate the observed
climate of the 20th century.
 “Pacemaker”: tropical Pacific SST is prescribed from observations, but coupled air-sea
feedbacks are maintained in the other ocean basins (e.g. Lau and Nath, 2003).
 Anecdotal evidence indicates that pacemaker experiments reproduce the timing of the forced
response to El Niño and the Southern Oscillation (ENSO), but also much of the co-variability
that is missing when global SST is prescribed.
The Evolution of Lead-lag ENSO-Monsoon Relationship in GCM Experiments


      Influence of CGCM’s Systematic Error On ENSO-Monsoon Predictability
            Influence of model deficiency in the long run on forecast skill
            Systematic errors in ENSO characteristics and forced response


      Improvement through “Pacemaker”
            Simulation of Climatology
            ENSO forced response
            ENSO-monsoon relationship


      Advantage vs. Shortcoming in “Pacemaker”
            Evolution of lead-lag ENSO-Indian monsoon relationship
            Plausible sources of shortcomings
                 Model and Experimental Design

 The NCEP Climate Forecast System (CFS)
                                                Model
                             AGCM                                   OGCM
                           NCEP GFS                                MOM 3.0
  NCEP CFS    (operational Global Forecast System)               1oX1/3 to 1o
                             T62 L64                              40 Levels



 Free long run (Courtesy of K. Pegion)
  52-year simulation
  Analyzing last 50 years (50-yr climatology is subtracted)


 Retrospective forecast
              Lead                                          Initial Condition
              month       Run      Period
                                                        Atm                  Ocean

  NCEP CFS       9        15     1981-2003    NCEP/DOE AMIP R2 GODAS (Behringer et
                                 (23 years)                        al. 2005)
                    Lead-lag Correlation
               (JJMS Extended IMR, NINO3.4)
                                                                              Observed

                                                                              1st month
                                                                              2nd month
                                                                              3rd month
                                                                              4th month
                                                                              5th month
                                                                              6th month
                                                                              7th month




 Extended MR (Indian Monsoon Rainfall Index):
           Total rainfall over 60-100oE, 5-25oN during JJAS
 For retrospective forecasts, reconstructed data with respect to lead time
(monthly forecast composite) is used.
Green solid line denotes 95% significance level
                   Lead-lag Correlation
              (JJMS Extended IMR, NINO3.4)
                                                   Observed
                                                   Long run
                                                   1st month
                                                   2nd month
                                                   3rd month
                                                   4th month
                                                   5th month
                                                   6th month
                                                   7th month




 Purple line: CFS free long run during 52 years
Influence of Systematic Error on Forecast Skill in CFS
                      SEOF 1st mode of SST Anomalies (ENSO mode)

                           1st              5th             9th      Free
Obs.                      month            month           month   long run




                 1                                                     Temporal correlation of PC
                                                                       timeseries with observation
  Correlation




                0.9                                                    Pattern correlation of
                                                                       eigenvector with free long run

                0.8
                                                                    With respect to the increase of lead
                                                                   month, forecast ENSO mode is much
                0.7                                                similar to that of long run, while far
                      1   2    3   4   5     6     7   8     9
                                                                   from the observed feature.
                              Forecast lead month
            ENSO Characteristics in CFS CGCM
                   NINO3.4 Index during 1950-2005


(a) Observation




(b) CFS CGCM (52 year long run)
                       ENSO Characteristics in CFS CGCM
                   Standard Deviation of SST Anomalies over Tropics
                 (a) Observation




                                                             (c) NINO3 region




                                             SST anomalies
Calendar Month




                 (b) CFS long run

                                                                                Observation
                                                                                CFS long run



                                                                 Calendar Month




                         Longitude
   ENSO Characteristics in CFS CGCM
Regression of DJF NINO3.4 Index to SST anomalies
                                (a) Observation




                                (b) CFS long run




                               In CGCM, ENSO SST anomalies
                              show westward penetration with
                              narrow band comparing to the
                              observed.
 Overestimated ENSO forcing in CFS CGCM
         Correlation bet. SST and Latent heat flux
 (a) Observation


                                                     GSSTF ver. 2
                                                     Surface latent
                                                    heat flux during
                                                      1988-2000



 (b) CFS long run


                                                 Positive: Ocean forces the Atmosphere
                                                 Negative: Atmosphere forces the Ocean




 In CGCM, the relationship between SST and latent head flux in the western Pacific
shows the excessive ocean forcing atmosphere. It may be related with too coherent
oceanic response, since the space and time scales of atmospheric internal dynamics
(stochastic forcing) are too coherent (Kirtman and Wu, 2006)
               ENSO-Monsoon in Observation
         Lead-Lag Regressed Map by NINO3.4 Index

Previous JJA                 DJF                   Following JJA
           ENSO-Monsoon in CFS long run
         Lead-Lag Regressed Map by NINO3.4 Index

Previous JJA                 DJF                   Following JJA
The Evolution of Lead-lag ENSO-Monsoon Relationship in GCM Experiments


      Influence of CGCM’s Systematic Error On ENSO-Monsoon Predictability
            Influence of model deficiency in the long run on forecast skill
            Systematic errors in ENSO characteristics and forced response


      Improvement through “Pacemaker”
            Simulation of Climatology
            ENSO forced response
            ENSO-monsoon relationship


      Advantage vs. Shortcoming in “Pacemaker”
            Evolution of lead-lag ENSO-Indian monsoon relationship
            Plausible sources of shortcomings
          “Pacemaker” Experimental Design
In this study, the deep tropical eastern Pacific where coupled ocean-
atmosphere dynamics produces the ENSO interannual variability, is
prescribed by observed SST.


      Pacemaker region


To merge the pacemaker and
   non-pacemaker regions


Outside the pacemaker region


    To handle model drift


   Other forcings (sea ice,
   greenhouse gases, etc)
                       “Pacemaker” Experimental Design
In this study, the deep tropical eastern Pacific where coupled ocean-
atmosphere dynamics produces the ENSO interannual variability, is
prescribed by observed SST.


            Pacemaker region                   165E-290E, 10S-10N




Shaded region denotes that dynamic
 term prevails over thermodynamic
 term in 20-yr NCEP CFS simulation
       dSST
             dynamics  thermod ynamics
        dt

 SST n 1  SST n 1          Fluxsfc
        2 t                  C p H
           “Pacemaker” Experimental Design
In this study, the deep tropical eastern Pacific where coupled ocean-
atmosphere dynamics produces the ENSO interannual variability, is
prescribed by observed SST.




To merge the pacemaker and
                                                   No blending
   non-pacemaker regions

 Global mean SST except pace-maker region




1949                           1950                          1951

        Obs.
        with blending
        without beldning
          “Pacemaker” Experimental Design
In this study, the deep tropical eastern Pacific where coupled ocean-
atmosphere dynamics produces the ENSO interannual variability, is
prescribed by observed SST.


                                                       AGCM
                                                     (NCEP GFS)

                                            Heat fluxes              Blended SST


Outside the pacemaker region                    Slab ocean mixed-layer
                                                     dT   F
                                                              -γTclim
                                                     dt C p H


                       Prescribed mixed-layer depth: Seasonally varying
                       1/3 Smoothed Zonal mean Levitus climatology
                       Except pacemaker region, zonal mean mixed-layer
                       depth of each basin - Pacific, Atlantic, Indian Ocean
                       - has not much differences
          “Pacemaker” Experimental Design
In this study, the deep tropical eastern Pacific where coupled ocean-
atmosphere dynamics produces the ENSO interannual variability, is
prescribed by observed SST.


                                                     AGCM
                                                   (NCEP GFS)

                                           Heat fluxes          Blended SST


Outside the pacemaker region                  Slab ocean mixed-layer
                                                   dT   F
                                                            -γTclim
                                                   dt C p H
    To handle model drift


    Simulated minus observed global mean SST difference   Weak damping of
            with relaxation                                  15W/m2/K
            without relaxation
          “Pacemaker” Experimental Design
In this study, the deep tropical eastern Pacific where coupled ocean-
atmosphere dynamics produces the ENSO interannual variability, is
prescribed by observed SST.


      Pacemaker region                         165E-290E, 10S-10N


To merge the pacemaker and
                                                   No blending
   non-pacemaker regions


Outside the pacemaker region                 Slab ocean mixed-layer


    To handle model drift                  Weak damping of 15W/m2/K


   Other forcings (sea ice,                   Climatological sea ice
   greenhouse gases, etc)                         Constant CO2
                          Model and Experimental Design
 No air-sea interaction          Local air-sea interaction                        Fully coupled system


   Atmosphere                          Atmosphere                                  Atmosphere
  (GFS T62L64)                        (GFS T62L64)                                (GFS T62L64)

                                                             dT   F                          heat flux, wind stress,
                                           heat flux   SST            -γTclim        SST
                                                             dt C p H                       fresh water flux


Observed Climatology           Observed             Slab ocean                         Ocean
                                                 (No dynamics and
  SST        SST                 SST                 advection)
                                                                                  (Full dynamics)

       AGCM                     Mixed layer model + AGCM                                 CGCM
  (1950-2004, 4runs)                (1950-2004, 4runs)                                   (52 yrs)

  Experiment        Extratropics       Tropics          Mixed-layer depth           Period      Member
                 Slab ocean mixed-     Observed      Zonal mean monthly Levitus
                     layer with                                                      55 yr
  Pacemaker       weak damping of     interannual    climatology except Eastern   (1950-2004)       4
                     15W/m2/K             SST          Pacific (reduced as 1/3)
                                       Observed
                                                                                     55 yr
    Control      Climatological SST   interannual              none               (1950-2004)       4
                                          SST

     CGCM                                   MOM3                                     52 yr          1
          JJA Climatology of 55 years (1950-2004)
                      SST                           Rainfall


 Obs.




 Pace




Control




CGCM



    Contour denotes difference from observation.
                     ENSO forcing in Experiments
                   Correlation bet. SST and Latent heat flux


(a) Observation




                                              (c) PACE




(b) CFS long run
             Correlation Map bet. DJF NINO3.4 and SST
            Previous JJA                      DJF   Following JJA

 Obs.




 Pace




Control




CGCM


           Green box denotes pacemaker region.
                 Regressed Map by DJF NINO3.4 Index
                             Rainfall and 850 hPa wind
            Previous JJA                      DJF        Following JJA

 Obs.




 Pace




Control




CGCM


           Green box denotes pacemaker region.
             Indian Monsoon Rainfall Simulations
                        Climatology and Variability
       Climatology of IMR                             Standard Dev. of IMR




 Extended MR (Indian Monsoon Rainfall Index):
          Total rainfall over 60-100oE, 5-25oN during JJAS
                   Lead-lag Correlation
                (JJMS Extended IMR, NINO3.4)




 Observation
 CFS long run
 PACE
 CONTROL


• 26 years during 1979-2006
• Green line denotes 95% significant level
                   Lead-lag Correlation
                (JJMS Extended IMR, NINO3.4)




 Observation
 CFS long run       Ensemble spread of Pace
 PACE               Ensemble spread of Control
 CONTROL


• Shading denotes ensemble spread among 4 members. Note that correlation for
ensemble mean is not the average of correlations for four members.
The Evolution of Lead-lag ENSO-Monsoon Relationship in GCM Experiments


      Influence of CGCM’s Systematic Error On ENSO-Monsoon Predictability
            Influence of model deficiency in the long run on forecast skill
            Systematic errors in ENSO characteristics and forced response


      Improvement through “Pacemaker”
            Simulation of Climatology
            ENSO forced response
            ENSO-monsoon relationship


      Advantage vs. Shortcoming in “Pacemaker”
            Evolution of lead-lag ENSO-Indian monsoon relationship
            Plausible sources of shortcomings
                   Change of Lead-lag Correlation
                 20-year Moving Window during 1950-2004

                              (HadSST and CMAP)




Lag correlation with
respect to 20-yr moving
window during 55 years
   Change of DJF Simultaneous Correlation
        20-year Moving Window during 1950-2004




  Observation
  PACE             Ensemble spread of Pace
  CONTROL          Ensemble spread of Control


• Shading denotes ensemble spread among 4 members. Note that correlation for
ensemble mean is not the average of correlations for four members.
               Indian Monsoon Rainfall Simulations
                         Year-to-year variability




      Observation
      PACE
      CONTROL




 3-year running mean of interannual IMR index                   Period     Cor.

                                                    Pacemaker   1979-2004   0.52
                                                                1979-2004   -0.23
                                                     Control
                                                                1991-2004   -0.52
Change of Regressed Pattern of NINO34 Index
            1950-1974 vs. 1980-2004
      HadSST                  PACE
                                          •Contour denotes
                                            differences of
                                           regressed value:
                                          1980-2004 minus
                                              1950-1974



                                          •Shading denotes
                                           regressed value
                                          during 1950-2004
Change of Regressed Pattern of NINO34 Index
            1950-1974 vs. 1980-2004
      HadSST                  PACE
                                          •Contour denotes
                                            differences of
                                           regressed value:
                                          1980-2004 minus
                                              1950-1974



                                          •Shading denotes
                                           regressed value
                                          during 1950-2004
           Plausible Sources for Recent Shortcoming

The characteristics of recent decadal change is not found in “pacemaker”

    Absence of influence of anthropogenic forcings such as CO2 increase etc.
     Insufficient projection of climate change

    Inadequacies from “pacemaker” experimental design
    1. Role of low-frequency ocean dynamics
    2. Associated atmosphere-ocean coupled mode
    3. Decadal change of monsoon forcing to alter the El Nino
     To supplement this point of view, sensitivity experiments associated with
       decadal change are needed
     For example, change of Q flux with respect to decades can be considered

    Imperfect model
     Wrong atmosphere response
             Annual Mean Global Temperature


      Observation
      PACE




 Even though interannual variability is well matched with observed, “pacemaker
cannot mimic the global warming trend.
                         Summary and Conclusion

 In CFS CGCM, lead-lag ENSO-monsoon relationship is weak and insignificant due to
systematic errors of ENSO and its response.
 In CGCM forecasts, systematic errors of couple models is major factor in limiting
predictability after the influence of initial condition fades away with respect to lead month:
mean error, phase shift, different amplitude, and wrong seasonal cycle, etc.


 To improve the predictability, “pacemaker” experiment is designed and conducted to
reproduce the important aspects of air-sea coupling while maintaining the flexibility to
attempt to simulate the observed climate of the 20th century.
 Surprisingly, “pacemaker” mimics the realistic ENSO-monsoon relationship compared to
other experiments including control (POGA-type) and coupled (CGCM).


 However, the recent change of ENSO-monsoon relationship is missed in “pacemaker”
associated with absence of global warming signal.
 To find out the cause of this discrepancy, supplementary “pacemaker” experiments can be
performed based on this shortcoming.
                          Partial Correlation
          Correlation bet. DJF NINO3.4 and Previous JJA Rainfall
                    Local SST                  NINO 34

 Obs.




 Pace




Control




CGCM
                                     Lead-Lag Relationship
                                   Monthly IMR and NINO3.4 Index
Calendar Month of IMR




                        N34 Lead          N34 Lag    N34 Lead      N34 Lag
   Latent Heat Flux - SST Correlation   Conceptual Model
Observational
Estimates Based on
NASA GFSST2 Data




     Control                                        <HF,SST>
     Coupled Model
     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: Add White Noise to Latent Heat
  Flux
                     Contemporaneous Latent Heat Flux - SST Correlation
Observational
Estimates Based on                                 Increased “Randomness”
NASA GFSST2 Data                                   Coupled Model




     Control
     Coupled Model
                                                    Add White Noise in Space
                                                    and Time to Latent Heat
                                                    Flux in the Western Pacific
                                                    (Ad-Hoc)
Based on What we Know About
Atmosphere Forcing Ocean and
Ocean Forcing Atmosphere,
How Can we Fix the CGCM
Problem in the Central Pacific?
    JJA Partial influence: Local SST vs. Remote forcing



 Partial Correlation
  (Edward, 1979)                (a) Local SST
            R12  R13 R23
 R12,3 
           1  R13 1  R23
                 2      2

 Calculate the partial
effect of local SST and
                                                   (c) Ratio of COA of (a)/(b)
NINO 3.4 SST on the
precipitation anomalies by
removing relationship
between local and NINO3.4
SST
                                (b) NINO 3.4


 COA = COVARIANCE [A,B]
/ σA
    (Kang et al. 2001 JMSJ)
                                                   NINO3.4                  local SST
 To measure an actual
magnitude of quantity of B
related to the reference data
A
 Red denotes the effect of
local SST is larger than
that of remote forcing

								
To top