Understanding the double peaked El Niño in coupled GCMs

Abstract

Coupled general circulation models (CGCMs) simulate a diverse range of El Niño–Southern Oscillation behaviors. “Double peaked” El Niño events—where two separate centers of positive sea surface temperature (SST) anomalies evolve concurrently in the eastern and western equatorial Pacific—have been evidenced in Coupled Model Intercomparison Project version 5 CGCMs and are without precedent in observations. The characteristic CGCM double peaked El Niño may be mistaken for a central Pacific warming event in El Niño composites, shifted westwards due to the cold tongue bias. In results from the Australian Community Climate and Earth System Simulator coupled model, we find that the western Pacific warm peak of the double peaked El Niño event emerges due to an excessive westward extension of the climatological cold tongue, displacing the region of strong zonal SST gradients towards the west Pacific. A coincident westward shift in the zonal current anomalies reinforces the western peak in SST anomalies, leading to a zonal separation between the warming effect of zonal advection (in the west Pacific) and that of vertical advection (in the east Pacific). Meridional advection and net surface heat fluxes further drive growth of the western Pacific warm peak. Our results demonstrate that understanding historical CGCM El Niño behaviors is a necessary precursor to interpreting projections of future CGCM El Niño behaviors, such as changes in the frequency of eastern Pacific El Niño events, under global warming scenarios.

Appendices

Appendix 1: The mean state and biases in ACCESS-CM1.3

The mean SST from ACCESS-CM1.3 and SST bias, with respect to the Bureau of Meteorology Research Centre (BMRC) SST reanalyses (Smith 1995) over the period 1980–2004, are illustrated in Fig. 10. ACCESS-CM1.3 is up to $1\,^{\circ }\hbox {C}$ cooler than the reanalysis data in the equatorial Pacific cold tongue region ($180{-}100^{\circ }\hbox {E}$), and up to $2\,^{\circ }\hbox {C}$ warmer east of $100^{\circ }\hbox {W}$ along the coast of South America. ACCESS-CM1.3 displays a warm bias in the South Pacific, in the region of the South Pacific Convergence Zone, and in the tropical North Pacific ($5^{\circ }\hbox {N}$, $160{-}110^{\circ }\hbox {W}$).

Fig. 10

Mean sea surface temperature over the period 1980–2004 (shading) in the a BMRC reanalyses, and b ACCESS-CM1.3. Data are in units of $^{\circ }\hbox {C}$ and the contour interval is $0.5\,^{\circ }\hbox {C}$

The standard deviation of tropical Pacific $SST'$ is indicative of the spatial diversity in ENSO variability (Fig. 2). Variability in the eastern equatorial Pacific in ACCESS-CM1.3 is weaker than in the reanalysis data (the difference in standard deviation is up to $0.6\,^{\circ }\hbox {C}$ at approximately $100^{\circ }\hbox {W}$), including ${<}0.3\,^{\circ }\hbox {C}$ from $160{-}140^{\circ }\hbox {W}$, and slightly stronger (${>}0.2\,^{\circ }\hbox {C}$) west of $180^{\circ }$ longitude in a secondary western peak. Note that the standard deviation of $SST'$ illustrated in Fig. 2 is qualitatively similar to the leading mode of an EOF analysis of ACCESS-CM1.3 $SST'$, which also displays the double peaked pattern of warming and represents 44 % of the $SST'$ variability in ACCESS-CM1.3 (figure not shown).

The annual means of the equatorial surface heat fluxes for ACCESS-CM1.3 are compared with those from the Objectively Analyzed air-sea Fluxes (OAFlux; provided by the Woods Hole Oceanographic Institute (WHOI) OAFlux project, available at http://oaflux.whoi.edu), the TropFlux reanalyses (Kumar et al. 2012), and the Coordinated Ocean-ice Reference Experiments version 2 (CORE-II, which are used to force ACCESS-OM; Griffies et al. 2012) in Fig. 11. The annual mean equatorial longwave radiation and sensible heat flux simulated by ACCESS-CM1.3 are within the range of uncertainty estimated from OAFlux, TropFlux, and CORE-II. Latent heat fluxes in ACCESS-CM1.3 are up to $46\,\hbox {W m}^{-2}$ less than those of the reanalyses, particularly in the eastern equatorial Pacific. Equatorial shortwave radiation values simulated by ACCESS-CM1.3 in boreal winter are up to $38\,\hbox {W m}^{-2}$ different from TropFlux.

Fig. 11

Annual mean of equatorial surface heat flux variables—namely, shortwave, sensible, latent, longwave, and net heat fluxes—from ACCESS-CM1.3 (blue), the CORE-II reanalyses from ACCESS-OM (red), the OAFlux reanalyses (black solid), and the TropFlux reanalyses (black dashed). Data are averaged between $2^{\circ }\hbox {S}$ and $2^{\circ }\hbox {N}$ and are in units of $\hbox {W m}^{-2}$

The mean state of the tropical Pacific MLD in ACCESS-CM1.3 and bias with respect to the UK Met Office (UKMO) subsurface ocean temperature and salinity data (Ingleby and Huddleston 2007) over the period 1980–2005 are compared in Fig. 12. The ACCESS-CM1.3 MLDs are up to 50 m deeper than the UKMO MLDs in bands stretching between $170^{\circ }\hbox {E}$ and $150^{\circ }\hbox {W}$ north and south of the equator.

Fig. 12

Mean mixed layer depth (MLD) over the period 1980–2005 (shading) in the a UK Met Office (UKMO) reanalyses, and b ACCESS-CM1.3. The MLD is defined as the depth at which the density layer $\sigma _t$ deviates from surface values by $0.125\,\hbox {kg m}^{-3}$. Contours show the bias in mean mixed layer depth with respect to the UKMO data. Data are in units of m and the contour interval is 10 m

Appendix 2: Significance of the double peaked El Niño event in ACCESS-CM1.3

Here, we investigate whether the composited double peaked El Niño events are significantly different from the composited eastern Pacific El Niño events. First, the double peaked and eastern Pacific El Niño events from the PiControl simulation of ACCESS-CM1.3 are randomly separated into two groups, groups a and b, and composited. We name these composites $\mu _{x}$ of sample size $n_{x}$, where $x\in \{DP1.3a, DP1.3b, EP1.3a, EP1.3b\}$. We also consider the double peaked El Niño events from the PiControl simulation of ACCESS-CM1.0 and separate them into two composites—$\mu _{DP1.0a}$ and $\mu _{DP1.0b}$—with sample sizes $n_{DP1.0a}$ and $n_{DP1.0b}$, respectively.

The variable for testing the significance of the difference between composites is the Student’s t-distribution:

$$t= \frac{\widehat{\mu _x} - \widehat{\mu _y}}{S\sqrt{\frac{1}{n_x} + \frac{1}{n_y}}}, \hbox {and}$$

(4)

$$S^2= \frac{(n_x - 1)\widehat{\sigma _x^2}+(n_y - 1)\widehat{\sigma _y^2}}{n_x + n_y - 2},$$

(5)

where $n_x + n_y - 2$ is the number of independent observations for the parameter t, and x and y represent the composited El Niño events being tested. The significance value (p value) from each test case is calculated using a two-sided Student’s t-test.

We define a simple test to establish the significance of the El Niño composite events: namely, the double peaked and eastern Pacific El Niño events are significantly different if the following conditions are satisfied during the evolution of the El Niño event (i.e., the first 24 months of the composite):

Test 1 :

the differences between the DP1.3a and EP1.3a composites are greater than the differences between the DP1.3a and DP1.3b composites;

Test 2 :

the differences between the DP1.3b and EP1.3b composites are greater than the differences between the EP1.3a and EP1.3b composites;

Test 3 :

the differences between DP1.3a events from ACCESS-CM1.3 and DP1.0a events from ACCESS-CM1.0 are greater than the differences between the DP1.3a and DP1.3b events from ACCESS-CM1.3; and

Test 4 :

the differences between DP1.3b events from ACCESS-CM1.3 and DP1.0b events from ACCESS-CM1.0 are greater than the differences between DP1.0a and DP1.0b events from ACCESS-CM1.0.

The random sampling is repeated 100 times and median values for the differences between the composites, t, and p across the samples are calculated. The results for tests 1–4 are illustrated in Fig. 13.

Fig. 13

Simple significance testing of $SST'$ composites from randomly selected double peaked and eastern Pacific El Niño events in ACCESS-CM1.3 (DP1.3a, DP1.3b, EP1.3a, and EP1.3b, respectively) and double peaked El Niño events in in ACCESS-CM1.0 (DP1.0a, DP1.0b, respectively) for the 3 years surrounding El Niño events. Data displayed are median t probability density function values calculated from 100 random samples of the test groups a and b. The first column in each row is calculated by subtracting the third column from the second. Differences greater than one standard deviation from the mean are indicated with stippling (‘.’), and differences significant at the 95 % confidence interval with crosses (‘+’). In each case, significance is calculated using a two-sided Student’s t test. The contour interval is 0.1

For test 1, the median difference between DP1.3a and EP1.3a is approximately ${\pm }2$ times greater than the difference between DP1.3a and DP1.3b, which is in the range $[-0.37, 0.19]\,^{\circ }\hbox {C}$ for the 100 samples generated. The differences in DP1.3a and EP1.3a are greater than one standard deviation across the western-central equatorial Pacific during the 12 months prior to the peak of the El Niño event. The greatest differences in the eastern equatorial Pacific occur during the 2 months prior to and 8 months following the peak of the El Niño event. Differences between DP1.3a and DP1.3b across the 100 samples are not statistically significant. A similar result is found for test 2. Even in the PiControl simulations, the sample size of eastern Pacific events in ACCESS-CM1.3 is relatively small – 10 in total – such that the difference between EP1.3a and EP1.3b is likely to be biased by individual events.

The results of tests 3 and 4 illustrate that double peaked events from the ACCESS-CM1.3 model are more similar to each other than to events from ACCESS-CM1.0. Again, the median difference between double peaked events within each model simulation is small (within the range $[-0.22, 0.40]\,^{\circ }\hbox {C}$ for the ACCESS-CM1.0 simulation), while the median differences in double peaked events between the two models are close to ${\pm }2\,^{\circ }\hbox {C}$ during the development of the El Niño event throughout the equatorial Pacific and in the western and eastern Pacific during the decay periods of the El Niño event (the differences are greater than one standard deviation from the mean in each case). These results provide evidence that the composite double peaked and eastern Pacific El Niño events from ACCESS-CM1.3 are sufficiently different to ensure significance in the trends analysis.