EGU General Assembly 2020
© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

Multidecadal changes in ENSO properties in the recharge oscillator conceptual model

Lander R. Crespo1,2, Belen Rodriguez-Fonseca3, Irene Polo3, Noel Keenlyside1,2, and Dietmar Dommenget4,5
Lander R. Crespo et al.
  • 1University of Bergen, Geophysical Institute, Norway (
  • 2Bjerknes Centre for Climate Research, Bergen, Norway
  • 3Meteorology and Geophysics Department, Universidad Complutense de Madrid, Madrid, Spain
  • 4School of Earth, Atmosphere and Environment, Monash University, Clayton, VIC 3800, Australia
  • 5ARC Centre of Excellence for Climate Extremes, Australia.

We use a simple conceptual recharge oscillator model for the tropical Pacific to identify multidecadal changes in El Niño-Southern Oscillation (ENSO) statistics and dynamics during the observational record. The model, defined by only two variables, sea surface temperature (SST) and warm water volume (WWV), is fitted to the observations for the period 1901-2010. The variability of ENSO has increased during the 20th century. The model simulates similar changes in variance of SST and WWV. The cross-correlation between SST and WWV also shows significant changes during the observational record. From the 1970s onwards, both observations and model output show that the SST drives WWV anomalies with a lead-time of 10 months and the WWV feedbacks onto the SST with a lead-time of about 8 months. The latter is reminiscent of a recharge-discharge mechanism of the upper ocean heat content. Before the 1970s only the impact of SST on WWV, through implied wind changes, is observed and is reproduced by the model. The periodicity of ENSO has also changed; ENSO has become more frequent changing from a 7-yr periodicity in the beginning of 20th century to a 5-yr periodicity in the recent decades. We find that the full recharge-discharge mechanism of the equatorial upper ocean heat content that characterizes the dynamics of the ReOsc model is only observed from the 1970s onwards and is likely to be a consequence of a stronger observed coupling between WWV and SST and of the leading role of the thermocline feedback. The degrading quality in the observations for earlier periods can also partly explain the decadal changes in the ENSO interactions. We find that the Atlantic Multidecadal Variability and global warming can partly explain the observed and simulated multidecadal changes in ENSO properties.

How to cite: Crespo, L. R., Rodriguez-Fonseca, B., Polo, I., Keenlyside, N., and Dommenget, D.: Multidecadal changes in ENSO properties in the recharge oscillator conceptual model, EGU General Assembly 2020, Online, 4–8 May 2020,, 2020

Comments on the presentation

AC: Author Comment | CC: Community Comment | Report abuse

Presentation version 1 – uploaded on 07 May 2020
  • CC1: Comment on EGU2020-21756, John Bruun, 07 May 2020

    As just mentioned: @ Lander, very interesting work.  You can see some related long term ENSO mode dynamic system properties in our work:

    Bruun, J.T., Allen, J.I. and  Smyth, T.J. (2017), Heartbeat of the Southern Oscillation explains ENSO climatic resonances, J. Geophys. Res.   Oceans, 122, 6746–6772,

    Skákala, J., & Bruun, J.T. (2018), A mechanism for Pacific interdecadal resonances, J. Geophys. Res. Oceans, 123, 6549–6561.

    I think it should be possilble to skill test for the type of dynamic system that is present in the record (observation or simulated). We applied this in the 2017 paper and I think there is an opportunity here to do this in your work with the identification analysis applied to simulation ensembles. It may help us to distinguish ensemble members more precisely through the dynamic class they represent. 

    It would be good to develop this idea further. I'm at

    Best John

  • CC2: Comment on EGU2020-21756, Paul Pukite, 07 May 2020

    Thanks for clearing up how the recharge oscillator model fits into GCMs, in that it is more a simplifying model trained from the output data, so it provides a phenomenological interpretation.

    • AC1: Reply to CC2, Lander R. Crespo, 08 May 2020

      I am glad to hear that my explanation was clarifying. I can try to expand here now with more time. As I said in the chat we train the model with CMIP5 models output data. In practice this means that we fit the recharge oscillator conceptual model to the output data of the CMIP5 models to obtain the parameters of the model (see my slides). Once we have the parameters we can run the model freely and build some statistics with the output variables of the ReOsc mode; SST and thermocline.
      By doing this we can investigate if the recharge-discharge mechanism is well represented by each CMIP model, how is the amplitude of it and if there is a decadal modulation in this mechanism in the models.