1. Mechanisms of the Indian Ocean Dipole influence on following year's El Niño/Southern Oscillation
1) What wind variations matter for El Nino onset?
A westerly wind anomaly and/or a collapse of easterly anomalies? Could wind variation matter as much as its seasonal anomaly?
2) How does IOD generate such wind variation?
Note: this .ppt should be visualised as a diaporama, because of the various appearances (animations).
Takeshi Izumo (1,2), Jérome Vialard (1), Matthieu Lengaigne (1,2), Nicolas Jourdain (3), Hugo Dayan (1), and Iyyappan Suresh (4)
(1) IRD, LOCEAN-IPSL, Paris, France, (2) IISc-NIO-IITM-IRD Joint International Laboratory, NIO, Goa, India, (3) CNRS-LGGE, Saint-Martin d’Hères, France, formerly at ARC, Sydney, Australia, (4) CSIR-NIO, Goa, India

2. Pacific response to a westerly pulse
Typical El Nino windstress
Picaut et al. 1997
=> let's use a realistic linear model (LCS, McCreary 1980, + SST equation) and analyse its impulse response to a Nino-like westerly pulse
Pacific response to a westerly pulse
integration =>
1st peak = initial positive response
2nd peak = classical delayed negative feedback (Schopf and Suarez 1988, Battisti and Hirst 1988)
Izumo et al. 2016

3. Pacific response to wind variations
Linearity => total ocean response = convolution of the wind with the impulse response :
=> ocean response = « causal filter » of τx:
LCS Nino4eqSST causal-filtered τx,cp
integration =>
Dashed lines = approximation as a combination of box filters
=> precise and easy estimate of SST' from past τx
=> For El Nino triggering in spring:
Nino4eq(AMyr1) ≈ τx(FMAyr1) – τx(JASONyr0)/2 = ½ ( τx(FMAyr1) + (τx(FMAyr1) – τx(JASONyr0)) )
=> τx temporal variation between summer-fall and spring matters as much as τx spring anomaly !
Note: related to basin resonances
this equation is able to reproduce SST, SSH and U variations in the central equatorial Pacific well, significantly better than without the delayed feedback (Fig. 5).
=> this feedback is important, about half of the initial response
Formulae valid for τx west Pac.

4. Mechanism of the robust IODyr0 => ENSOyr1 ?
Statistically, in obs. and CMIP:
IODyr0 robust precursor of ENSOyr1, better than ENSO itself, and complementary to WWV in fall (yr0):
ENSO(yr+1) ≈ a WWV – b IOD
Added skill (variance) from IOD:
20-40% in obs., 5-25% in CMIP
Jourdain et al. => not a statistical trap; physical mechanism required, but which one???
(Izumo et al. 2010, 2014, Dayan et al. 2014, Jourdain et al. 2016)
Jourdain et al. 2016
2nd question: What is the physical mechanism behind the influence of IOD on following year's El Nino? Does IOD generate a change of τx between summer-fall and next spring?
FIG. 5. Correlation between DMI in SON (shown by the gray bar) and lagged monthly Niño-3.4 in 21 CMIP5 models. The 90% significance level is indicated by the horizontal dashed lines (t test for 100 degrees of freedom; see caption of Fig. 1).

5. SST/precipitation patterns at the origin of Pacific wind variations
PACLIM exp. setup (7 members)
Dayan et al. 2015
forcing by observed SST
ECHAM AGCM
τx response?
Varying Indian Ocean
Climatological Pacific Ocean
A “pure nIOD”, with its large eastern pole and weaker western pole, forces significant τx,wp easterlies in JASON but not in FMA, leading to a significant FMA-JASON/2 difference in τx,wp.
Main cause: the precipitation seasonal change in the eastern Indian Ocean and maritime continent
Fig. 13 Regression of observed SST (upper, that forces PACLIM AGCM experiment; °C) and precipitation (lower; mm day−1/2στx,wp), plus wind- stress (vectors; N m−2/2στx,wp), on west Pacific τx,wp (averaged over the box region shown in a, as in Fig. 1a) of PACLIM experiment, for the difference FMA–JASON/2 (for both the regressed fields and τx,wp). This figure shows that mainly
the SST/precipitation seasonal change in the eastern Indian Ocean and maritime continent related to the transition from IOD to IOB is at the origin
of this τx,wp difference, i.e. of a strong causal-filtered τx,wp anomaly
Izumo et al. 2016

6. Schematics of mechanism for IOD influence on following year’s El Niño
Summary
1) What wind variations matter for El Nino onset?
=> τx temporal variation between previous summer-fall and spring matters as much as τx spring anomaly !
2) How does IOD generate such wind variation?
IOD eastern pole demise
=> τx(FMAyr1) – τx(JASONyr0)/2 anomaly in west Pacific
=> equatorial wave response in the Pacific
=> significant SST' in central Pacific in spring, modulating ENSO triggering
Schematics of mechanism for IOD influence on following year’s El Niño
Izumo et al. 2010

7. Further reading Thank you
Dayan H, Vialard J, Izumo T, Lengaigne M (2014) Does sea surface temperature outside the tropical Pacific contribute to enhanced ENSO predictability? Climate Dynamics.
Dayan, H., T. Izumo, J. Vialard, M. Lengaigne and S. Masson (2015): Do regions outside the tropical Pacific influence ENSO through atmospheric teleconnections?, Climate Dynamics.
Izumo T, Vialard J, Lengaigne M, de Boyer Montegut C, Behera SK, Luo JJ, Cravatte S, Masson S, Yamagata T (2010) Influence of the state of the Indian Ocean Dipole on the following year’s El Niño. Nature Geoscience.
Izumo T, Lengaigne M, Vialard J, Luo J–J, Yamagata T, Madec G (2014) Influence of Indian Ocean Dipole and Pacific recharge on following year’s El Niño: interdecadal robustness. Climate Dynamics.
Izumo T, J. Vialard, H. Dayan, M. Lengaigne, I. Suresh, (2016) A simple estimation of equatorial Pacific response from windstress to untangle Indian Ocean Dipole and Basin influences on El Niño. Climate Dynamics.
Jourdain N., M. Lengaigne, J. Vialard, T. Izumo and A. SenGupta (2016) Further insights on the influence of the Indian Ocean Dipole on following year’s El Niño-Southern Oscillation in observations and CMIP5 models. Journal of Climate.
Thank you

8. Other supplementary slides
(validation, IOD vs IOBM…)

9. 3rd question: How much predictability skill does the influence of IOD on following year's El Nino bring?
Statistically:
IOD robust precursor of ENSO, complementary to WWV in fall (yr0):
ENSO(yr+1) ~ -IOD+ WWV
Added skill (variance) from IOD:
20-40% in obs., 5-25% in CMIP
FIG. 3. (a) Correlation between Niño-3.4 in NDJ (gray bar) and lagged monthly WWV for 21 CMIP5 models (colors), the CMIP5 multimodel mean (solid black), the SODA reanalysis (long-dashed black), and the NOAA–BMRC observational dataset (short-dashed black). (b) Autocorrelation between Niño-3.4 in NDJ (gray bar) and lagged monthly Niño-3.4 for 21 CMIP5 models (colors), the CMIP5 multimodel mean (solid black), and the multiobservation mean (dashed black). The 90% level of statistical significance for 100 degrees of freedom is indicated by the dashed line
FIG. 5. Correlation between DMI in SON (shown by the gray bar) and lagged monthly Niño-3.4 in 21 CMIP5 models. The 90% significance level is indicated by the horizontal dashed lines (t test for 100 degrees of freedom; see caption of Fig. 1).
FIG. 6. Increase of explained Niño-3.4 variance when a second predictor of ENSO is used in addition to WWV. (a) Increased variance due to DMI (i.e., first predictor is SON WWV and second predictor is SON DMI). (b) Increased variance due to TIO [i.e., first predictor is January–February (JF) WWV and second predictor is JF TIO]. The plot shows the mean value over 105 cross-validations leaving 50% of the samples (randomly chosen) to train the prediction model and keeping 50% for its evaluation. The significance of the increase in variance obtained from this method is indicated within brackets (bold when 90% significant). The date on which the forecast is issued is indicated by the gray bar, and the peak of ENSO in NDJ of year 1 is indicated by the light red bar.
Increase of explained Niño-3.4 variance when IOD is used as a second predictor of ENSO in addition to WWV.

10. The tool: a realistic linear ocean model for interannual variability
The ocean model: LCS, resolving baroclinic modes (McCreary 1980) + SST linear equation, over the Indo-Pacific.
Validation: high correlations of with observations (TAO, AVISO, OSCAR), notably for zonal current and SST in Nino4 (central Pacific), for which the two first baroclinic modes are important, and also for SST and SSH in Nino3 (eastern Pacific).
The AGCM: the ECHAM5 model in its T106 (~1.1º), 31 levels, configuration, realistically simulating the atmospheric response to IOD and IOB (cf. Dayan et al. 2015).
Many sensitivity tests, e.g. varying LCS parameters, closing the Indonesian Throughflow (=> oceanic bridge much weaker (~5-10%) than atmospheric bridge)
=> no significant changes in the results

11. Validation of the LCS combined with a simple linear SST equation
Fig. 2 Validation of the LCS ocean model (green) with in situ TAO data (black) and satellites-related products (red), for the central-west (upper panels) and central-east (lower panels) Pacific. a zonal cur- rent U at 0°, 165°E (cm s−1, 0–20 m, red is for OSCAR product). c SSH at 0°, 140°W (red is for AVISO in cm, black is for TAO thermocline depth z20 in m/1.5). b SST (°C) in Niño4eq region (blue curve if the interannually-varying SST gradient is used, see text). d SST in Niño3.4. The correlations r are between the LCS and the various observational datasets

13. Partial regression in Paclim, showing both IOD and IOB effects, and showing that oceanic bridge exists of course, but is negligeable as compared to atmospheric bridge, confirming most previous studies (e.g. Clarke 1992)
Fig. 11 IOD and IOB respective influences in PACLIM set of experiments on a AGCM west Pacific τx,wp (N m−2/2σIODorIOB), b causal- filtered τx,wp, cf. Eq. (5) and c LCS Niño4eq SST (lower panels) (°C/2σIODorIOB), inferred from the partial regression on −IOD (dark blue) and on −IOB (green; similar method as in Fig. 10). Their sum, i.e. “–IOD–IOB”, is shown in red. In c, the weak oceanic bridge effect of the IOD (cf. Sect. 6.2) is shown in light blue, for comparison to the main atmospheric bridge (e.g. the dark blue line in c) on which the present study focuses. A 3 months running mean is applied to all variables, to reduce intraseasonal noise. Dots on curves indicate sig- nals significant at the 90 % level

14. Fig. 12 Composites (°C) of pure IOD (blue, 2 positive and 4 nega- tive events), pure IOB (green, 4 pos. and 5 neg. events), IOD + IOB cases (red, 4 pos. and 2 neg. events; actual number of members: 7 per event), for DMI (SON) and IOB (JFM) indices (from the observed SST that forced PACLIM), and for the PACLIM AGCM τx,wp response (FMA–JASON/2)×55 [cf. Eq. (6)] and PACLIM LCS SST response in following spring (AMJ). We plot the difference between negative and positive composites (asymmetries between negative and positive events being not statistically significant; not shown). The composite difference shown here corresponds to pure nIOD/nIOB events of 2.4σ amplitude (σIOD and σIOB being of 0.52 and 0.29 °C respectively), hence close to the 2σ convention used in all other fig- ures. For DMI and IOB forcing indices, the square root of the sum of the inter-event variances of DMI and IOB is shown as a thin bar. For the PACLIM ensemble responses, the 90 % interval of confidence (evaluated from a Student t test on the ensembles) is shown as a thick black bar

15. Mechanisms of the Indian Ocean Dipole influence on following year's El Niño/Southern Oscillation
Main SSTA patterns described in literature as likely to influence ENSO through atmospheric teleconnections (Dayan et al. 2015)
Takeshi Izumo (1,2), Jérome Vialard (1), Matthieu Lengaigne (1,2), Nicolas Jourdain (3), Hugo Dayan (1), and Iyyappan Suresh (4)
(1) IRD, LOCEAN-IPSL, Sorbonne Univ. (UPMC, Univ Paris 06)-CNRS-IRD-MNHN, Paris, France, (2) Indo-French Cell for Water Sciences, IISc-NIO-IITM-IRD Joint International Laboratory, NIO, Goa, India, (3) CNRS-LGGE, Saint-Martin d’Hères, France, formerly at ARC, Univ. of New South Wales, Sydney, Australia, (4) CSIR-National Institute of Oceanography, Goa, India
Fig. 2 Main SSTA patterns described in literature as likely to in uence ENSO through atmospheric teleconnections: a the tropical Indian Ocean in fall, the southern Indian, Paci c and Atlantic Oceans in winter, the North Paci c in spring and the tropical Atlantic in summer; b the tropical Indian and the North Atlantic Oceans in spring. The patterns on these gures are obtained as the rst EOF of the SST anomalies in each region at the targeted season