By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
 Trends in the seasonal and temporal behaviour of the El Niño–Southern Oscillation over the period 1958–2007 have been assessed using two indices of the phenomenon, NINO3.4 and a non-standardised Southern Oscillation Index (SOI). There is no evidence of trends in the variability or the persistence of the indices, nor in their seasonal patterns. There is a trend towards what might be considered more “El Niño-like” behaviour in the SOI (and more weakly in NINO3.4), but only through the period March–September and not in November–February, the season when El Niño and La Niña events typically peak. The trend in the SOI reflects only a trend in Darwin pressures, with no trend in Tahiti pressures. Apart from this trend, the temporal/seasonal nature of the El Niño–Southern Oscillation has been remarkably consistent through a period of strong global warming.
Vecchi et al.  summarised the long-running debate as to how the El Niño–Southern Oscillation (ENSO) might react to global warming, in the past and in the future. They concluded that theory, models, and observations present divergent views of ENSO's reaction. A similar conclusion was reached by Cane . The focus in most model studies on ENSO and climate change has been on whether the Pacific will tend to a more permanent El Niño state as the world warms due to an enhanced greenhouse effect. Most observational studies have likewise focussed on whether the tropical Pacific has been evolving into a more El Niño-like state, or with a weakened zonal circulation across the equatorial Pacific [e.g., Trenberth and Hoar, 1996, 1997; Vecchi et al., 2006; Solow, 2006; Power and Smith, 2007]. For instance, Vecchi et al.  reported a weakening of the equatorial Pacific pressure gradient since the 1960s, with a sharp drop in the 1970s. A few studies have looked at some other aspects of the climate relating to ENSO. For example, Timmermann et al.  in a model study examined the possibility of an intensification of the annual cycle in the tropical Pacific. Federov and Philander  discussed how changes in the background atmospheric and oceanic state might affect the frequency or periodicity of El Niño. An and Wang  and Zhang et al.  have also discussed possible changes in the frequency of ENSO events. Power and Smith  proposed that the apparent dominance of El Niño during the last few decades was due in part to a change in the background state of the SOI, rather than a change in variability or a shift to more frequent El Niño events alone. Toniazzo et al.  examined how uncertainty in the ENSO response to climate change was related to uncertainty of model parameterisations. Other researchers have examined how the teleconnections between climate variables in many parts of globe, another essential part of ENSO, might vary with time [e.g., Trenberth and Caron, 2000; van Oldenborgh and Burgers, 2005; Power et al., 2006; Kucharski et al., 2007].
 However, there are other ways in which ENSO might change than a trend to either more El Niño-like behaviour or more La Niña-like behaviour, or changes in the frequency of El Niño events, or changes in teleconnections. One possibility is that ENSO might become more or less variable than in the past. Or, the way the phenomenon is phase-locked to the annual cycle may change, perhaps associated with seasonally varying trends in the phenomenon (i.e., a larger trend in one part of the year than at other times of the year). Or, the persistence of the phenomenon across seasons may change, either in the strength of this persistence or in its seasonal variation.
 This paper examines trends in the seasonal and temporal behaviour of ENSO, specifically its phase-locking to the annual cycle, over the past 50 years. As well, trends in ENSO indices have been calculated separately for each month of the year to reveal the seasonality of any trends, rather than the usual approach of examining trends in the annual mean or filtered values of the indices. Examination of these detailed aspects of the ENSO may provide more information to validate the simulations and projections of the phenomenon in global climate models, and to test theories regarding possible changes of ENSO in response to global warming. In particular, this analysis examines the robustness of the relationship between the annual cycle and the ENSO, including its phase-locking to the annual cycle, during a period of strong global warming.
 Two indices of the ENSO, NINO3.4 and the Southern Oscillation Index (SOI), are used here to characterise the ENSO. NINO3.4 is the sea surface temperature averaged across the region 5°S–5°N, 120°W–170°W, expressed as monthly anomalies from the 1971–2000 means. The SOI used here is the non-standardised difference between sea level pressures at Tahiti and Darwin. Most SOI indices in current use are standardized by the mean and standard deviation separately for each month of the year. Such standardization would confuse the examination of the seasonal variation in the variability (since each month would be standardized to equalize their standard deviations – so there would be no seasonal variation in variability), so a non-standardised SOI (actual Tahiti minus Darwin mean sea level pressure) was used here. Data were obtained from http://www.cpc.noaa.gov/data/indices/(accessed 10 April 2008), for the period January 1958 to March 2008. The data were used to determine, for both the SOI and NINO3.4:
 1. The standard deviations for each month of the year, calculated in overlapping 20-yr data windows.
 2. The 3-month serial correlations starting separately for each month of the year, also calculated in overlapping 20-yr data windows.
 3. The 1958–2007 trend calculated separately for each month of the year.
 The standard deviations and serial correlations were calculated using overlapping windows of 20 years to visualize any gradual change in the seasonal pattern or strength of these variables. The variables were also examined in two non-overlapping periods of 25 years. The results from this analysis (not shown) confirmed the results from the overlapping windows. The 50-year trend was calculated with a linear fit to the year, and the strength of the trends in each month is compared using linear correlation coefficients. All correlations were calculated with 50 data pairs. A correlation exceeding 0.235 in magnitude is significant at the 5% level and a correlation exceeding 0.328 in magnitude is significant at the 1% level.
 The standard deviations for each month, for the SOI, are shown in Figure 1a. The variation through the year is similar in all the four 20-year time “slices” (1958–1977; 1968–1987; 1978–1997; 1988–2007) as are the magnitudes in the different months. The variability is largest in the period January–March, drops sharply to a minimum in April–June, and slowly increases towards the end of the calendar year. Figure 1b shows the 3-month serial correlations starting in each month of the year, calculated using data from each of the four time-slices. The 3-month serial correlation is weakest from February (i.e., the correlation between the SOI in February and May), and then gradually increases to a maximum around the end of the year. As is the case with the standard deviations, the seasonal pattern of variation in the 3-month serial correlation, and the magnitude of the correlations, is similar in all the four time slices. The low variability of the SOI in the period April–June reflects the tendency for the SOI to reach minima and maxima associated with the El Niño–Southern Oscillation around the start of the calendar year, and for new events to commence gradually around April–May. The weak serial correlation from February to May reflects the predictability barrier associated with this tendency for El Niño events to commence around this time. The similarity of the pattern and magnitude of the standard deviations and serial correlations in all the four time slices indicates that there has been little if any change in the temporal/seasonal behaviour of the El Niño–Southern Oscillation, as measured by the SOI, over the past 50 years.
 NINO3.4 also exhibits a minimum in its standard deviation around April–May, with a maximum around the start and end of the calendar year (Figure 1c). The seasonal pattern of variability is again very similar in all time slices, although the standard deviations in the earliest (1958–1977) slice are smaller than the later time slices, suggesting fewer or weaker El Niño and La Niña events early in the 50-year period examined here. The 3-month serial correlations for NINO3.4 (Figure 1d) reach a minimum slightly later than is the case for the SOI, with a minimum in March–April (i.e., the weakest 3-month correlations are between March and June or April and July NINO3.4 values). The strength of the serial correlation near its minimum exhibits differences between the time slices, but there is no clear trend with time. In all time slices, the 3-month serial correlations starting in the months July to December are very strong (>0.8) and there is little difference between the various slices. As was the case with the SOI, there is little evidence of any systematic change in the temporal/seasonal behavior of the El Niño–Southern Oscillation over the last 50 years, as measured by NINO3.4. In particular, there is no evidence of a change in the strength of the predictability barrier or a shift in its seasonality. Yu and Kao  also found little decadal variation in the behavior of ocean indices of the El Niño–Southern Oscillation in the central and western Pacific, although they did find strong variations further east.
Figure 2 plots the strength of the trends of the SOI and NINO3.4 (calculated as correlations with the year) for each month of the year. The SOI trends are statistically significant in April–May (and also in August), but are close to zero between November and February. The NINO3.4 trends are weaker than those for the SOI and no month reaches statistical significance, but the weakest NINO3.4 trends are also in the period November–February. From April through to September both NINO3.4 and the SOI exhibit weak trends towards what might be called more “El Niño-like” behaviour (i.e., negative trends in the SOI and positive trends in NINO3.4). The trend (not shown) in sea surface temperatures further west (NINO4; 160°E–150°W) is stronger than for NINO3.4, especially in June–September, while further east (NINO3; 90°W–150°W) the trends are even weaker. All the trends (in NINO3.4 and SOI) change only slightly if the very strong El Niño of 1997 is omitted from the calculation.
 The trend in the SOI arises solely from trends in the western “pole” of the phenomenon, i.e., in Darwin pressure (Figure 2). The trends in Tahiti pressure are very weak throughout the year. The correlation with year of the Darwin pressures averaged over April–September was 0.40, whereas the correlation for Tahiti was 0.02. The correlations with year for pressures averaged over the period November–February, i.e., the peak season for El Niño events, were 0.02 for Darwin and 0.04 for Tahiti. The restriction of the surface pressure trends since 1958 to the western Pacific and the period March–September was confirmed by calculating correlations between the year and surface pressures in the Hadley Centre gridded surface pressure data set (HadSLP2) [Allan and Ansell, 2006] at the Climate Explorer web page (http:://climexp.knmi.nl; accessed 24 April 2008 [van Oldenborgh and Burgers, 2005]). No clear trends in surface pressure since 1958 were found in this data set in the tropical east and central Pacific.
 The strength of the trends in the western and eastern “poles” of the ENSO is illustrated in Figure 3 which shows time series of April–September average values of the SOI as well as Darwin and Tahiti pressures in each year 1958–2007. The trend towards more “El Niño-like” behaviour can be seen (negative trend in SOI). The trend in the Darwin pressures is very similar to that exhibited in the SOI (although opposite in sign), with little trend in Tahiti pressures, although the interannual variations at Tahiti are clearly related to those at Darwin and in the SOI. The absence of trends at Tahiti and the weaker trends in the NINO indices towards the eastern side of the equatorial Pacific (NINO3) contrasts with the trends in Darwin pressures and the sea surface temperatures towards the western edge of the Pacific (NINO4).
 The above results indicate that there has been no substantial modulation of the temporal/seasonal behaviour of the El Niño–Southern Oscillation, as measured by the NINO3.4 and SOI, over the past 50 years, during a period of substantial growth in the atmospheric concentrations of greenhouse gases and of global warming. The variability and persistence of the El Niño–Southern Oscillation, and their seasonal variations including the phase-locking to the annual cycle and the “predictability barrier”, have exhibited considerable robustness through this period. The one clear trend documented here, towards higher pressures at Darwin from March to September, requires an explanation. Why should only one “pole” of the Southern Oscillation exhibit such a trend? And why has this trend occurred only through the time of year when El Niño events are typically developing, and not at the time of year when they typically reach their maximum amplitude? Does this reflect real changes in the El Niño–Southern Oscillation, or another factor impacting only on the western Pacific? And could such a different trend in the two poles of the phenomenon account for the observed changes over time in relationships between the SOI and climate variations in some locations [Nicholls et al., 1996; Nicholls, 2003]?
 This research was supported by the Australian Research Council through Discovery Project DP0877417.