Footprints of IOD and ENSO in the Kenyan coral record



[1] Low-frequency variations in the seasonally phase-locked signal of the Indian Ocean Dipole (IOD) and the Pacific El Niño/Southern Oscillation (ENSO) recorded in 115-year coralδ18O variability from Kenyan coast, the tropical western Indian Ocean, were investigated. A comparison of the monthly coral δ18O corresponding to IOD and ENSO years shows that Kenyan coral distinctly records the East African Short Rain anomaly related to the IOD variability in January, a few months after the Short Rain peak due to oceanographic condition. On the other hand, the ENSO-induced signals do not appear clearly as the positive sea surface temperature (SST) and rainfall anomalies in the monthly coral record. Moreover, annual mean coralδ18O and ENSO show only a weak coherence at the periodicity of 4 to 5 years. These results support the suggestion that the IOD is the dominant climate mode rather than ENSO in the Kenyan coast. The coral records indicate that the negative IOD- like anomalously cold SST condition in the western Indian Ocean precedes the evolution of the Pacific El Niño by one year. The anomalously cold SST condition was prominent in the late 19th century, but weakened in the 20th century. This retreat of the cold SST condition due to warming of the western tropical Indian Ocean may influence the nature of the Pacific ENSO.

1. Introduction

[2] The Pacific El Niño/Southern Oscillation (ENSO) and the Asian monsoon have been considered the major influences on the climate variability in the Indian Ocean [Charles et al., 1997; Torrence and Webster, 1999; Kumar et al., 1999; Cole et al., 2000]. In general, the influence of ENSO on the Indian Ocean has been described as impacts on the SST anomaly through the atmospheric bridge [Liu and Alexander, 2007]. It is known that the El Niño-induced surface heat flux anomaly leads to a basin-wide warming [Klein et al., 1999; Saji et al., 1999; Reason et al., 2000] and this warming activates an atmospheric convection and resultant local rainfall in the surrounding region [Su et al., 2001; Watanabe and Jin, 2002]. The westward propagation of the oceanic Rossby waves generated in the Southern Indian Ocean produces and maintains, for one season longer, the anomalous SST warming in the southwestern Indian Ocean [Alexander et al., 2002; Xie et al., 2002]. Xie et al. [2002]suggested that ENSO-related wind forcing generated these Rossby waves.

[3] The East African rainfall in the coastal region facing the western Indian Ocean was thought to be related to El Niño remotely [Indeje et al., 2000; Camberlin et al., 2001; Jury et al., 2002]. In those studies, however, the two rainfall seasons in the East Africa (“Long Rain” from April to May which links to the southeastern monsoon, and “Short Rain” from October to November which links to the northeastern monsoon) were not separated in the analysis. The Short Rain has been revealed to have large interannual variability with severe flooding impacts in East Africa. In retrospect, this pooling of rainfall into annual increments introduced a serious confusion in the climate research community over the role of El Niño on the interannual rainfall variability.

[4] The Indian Ocean Dipole (IOD) was identified as another dominant climate mode generating climate variability not only in the Indian Ocean but also in the world along with ENSO [Saji et al., 1999; Yamagata et al., 2004]. In the western Indian Ocean, the positive IOD mode is associated with anomalous SST warming and, in particular, heavy East African rainfall anomaly during the peak season of the IOD from September to November. Behera et al. [2005] demonstrated that the IOD influences 79% of the East African Short Rain anomaly. Thus after the discovery of the IOD and the differences in sources of rainfall, both the observational and the model studies modified the view of the influence of ENSO on the tropical western Indian Ocean and East African coast [Schott et al., 2009].

[5] Coral annual bands record the oceanic and climatic conditions, they are useful to reconstruct in situlong-term climate variations. In particular, coral oxygen isotopic ratio (δ18Ocoral) is controlled by both SST [Leder et al., 1996] and δ18O of seawater, which is closely related to sea surface salinity (SSS) resulting from the balance between precipitation/run-off, evaporation, and advection in many parts of the tropics [Gagan et al., 2000; Iijima et al., 2005]. In coral living where seasonal SSS variation is low, coral δ18O is a proxy for SST. Conversely, where the seasonal variation in SST is low, coral δ18O can be a proxy for SSS. In the tropical western Indian Ocean, several studies using coral data showed the long-term SST rise over the last century [Charles et al., 1997; Cole et al., 2000; Nakamura et al., 2009]. Charles et al. [1997] analyzed the monthly coral δ18O for 147 years from the Seychelles, whereas Cole et al. [2000] analyzed the annual coral δ18O for 194 years from Kenyan coast. However, the assessment of interannual ENSO variability and periodicity in these studies using annual-resolved coralδ18O data hinder further studies of seasonal variability in these records.

[6] Nakamura et al. [2009], after deriving the near-monthly coralδ18O signal from Kenyan coast for the recent 115 years, confirmed the long-term SST warming together with the freshening of the sea surface in response to an increase in the local rainfall (Figure S1 in theauxiliary material). They focused on the seasonal signal of the IOD in generating the East African Short Rain anomaly [Kayanne et al., 2006; Nakamura et al., 2009]. They found that, though the cyclic pattern of coral δ18O corresponds to the annual SST cycle, the first half of the low-δ18O peak in every annual band reflects the IOD-related East African Short Rain anomaly, and defined the coralδ18O value corresponding to January in the following year as the coral IOD index [Kayanne et al., 2006] (Figure S2). The time lag between Short Rain peak and coral IOD index is explained by the discharge from Sabaki river reaching the coral site in December by northeast monsoon [Brakel, 1984]. They reconstructed the 115y-coral IOD variability and found changes in the periodicity of the coral IOD through the 20th century [Nakamura et al., 2009]. But the investigation of the influence of ENSO on the Kenyan coral remained unanswered.

[7] Recent climate modeling studies highlight the importance of seasonality in discussing influences of climate modes such as ENSO and IOD and suggests that its exclusion in coral studies of the tropical East African coast may downplay the importance of IOD with respect to ENSO. Besides the Short Rain variability, an understanding of ENSO and IOD connection with the Long Rain variability is needed since the Long Rain is the dominant seasonal signal in East Africa [Hastenrath et al., 1993]. In the present study, we address this issue by carefully examining the seasonality of El Niño signal in the coral δ18O record from the Kenya coast, together with the IOD signal. The expected peak season of the coral record related to Pacific El Niño is the boreal spring, i.e., almost one season (four months lag) after the El Niño maturity [Xie et al., 2002; Yamagata et al., 2004]. Moreover this season corresponds to the East African Long Rain period from April to May. We discuss possible relations among the decrease of the δ18O value, the ENSO-induced basin-wide SST mode and East African Long Rain variation.

2. Materials and Methods

2.1. Study Site and Climate

[8] The coral core (KY16) was obtained in October 2002 from Malindi Marine Park, Kenya (3.2°S, 40.1°E). The SST in Malindi increases to 29.1°C in April and decreases to 25.3°C in August, as averaged over the 1951–2002 period [Rayner et al., 2006]. Precipitation is concentrated during two seasons in East Africa, with Long Rains from April to May and Short Rains from October to November [McClanahan, 1988]. The study site is located 15 km south of the Sabaki River and the discharge in the Short Rain period is brought southward to Malindi by the NE monsoon winds from December to March, while discharge during the Long Rain period is brought northward by the SE monsoon from May to November [Brakel, 1984].

2.2. Analysis for the Seasonal Climate Signals

[9] We reanalyzed the 115 years (1887–2002) near-monthly coralδ18O record (KY16, Figure S1) [Nakamura et al., 2009]. After filtering to remove the long-term trend, the data was arranged with monthly resolution. We classified the 115 years coral annual bands into coral IOD (positive/negative IOD), ENSO (El Niño/La Niña) and neutral years using the coral IOD index and NINO3 index. From the indices, 15 positive and 8 negative coral IOD events [Nakamura et al., 2009], 15 pure El Niño and 12 pure La Niña events, and 11 neutral years were identified (Table S1). For each event, monthly coral δ18O values are analyzed for 3 consecutive years from the previous year (year −1) to the following year (year +1). The years are chosen in such a way that year 0 coincides with the year of IOD/ENSO occurrence. Furthermore, monthly coral δ18O values are averaged for each event, respectively, and lastly the monthly mean δ18O values are arranged and superimposed on each event for the consecutive 3 years.

2.3. Annual Mean Analysis

[10] We averaged the monthly coral δ18O record (KY16) from previous November to October to complete annual resolution as Cole et al. [2000] and took the coherence spectrum between the annual mean coral δ18O and NINO3.4 index.

3. Results and Discussion

[11] The superimposition of monthly mean coral δ18O variability for the consecutive 3 years is shown in Figure 1a. In Coral IOD years the distinct δ18O pattern from November to January in the following year (year 0 to year +1) is confirmed. In particular, the coral δ18O variability in January is much larger than those in ENSO and neutral years (p = 0.01, approximately 0.2‰ freshening (saline) in the positive (negative) IOD year). This analysis validates again the IOD signature in January of our index [Kayanne et al., 2006; Nakamura et al., 2009] (Figure 1a). In the year that the positive IOD and El Niño events occur simultaneously, the lowest δ18O value is found in January year +1. Because the δ18O values in pure El Niño years for the same season are not different from that of the neutral year, we may conclude that the influence of El Niño event on the East African Short Rain anomaly is negligible. Since the basin-wide warming generated by the El Niño increases the SST anomaly off Kenya, we expect more precipitation through a non-linear increase in cumulus activity [Gadgil et al., 1984].

Figure 1.

The superimposition of the seasonal coral δ18O variability for the coral IOD, ENSO, and neutral years. The coral IOD years are identified by the coral IOD index [Nakamura et al., 2009]: the threshold value of the positive (negative) IOD is −0.14 (0.22) ‰ anomaly. ENSO years are defined based on the NINO3 SST index for the period from December to February: the positive (negative) threshold value of 1σ(0.86) determines the El Niño (La Niña) event. When no IOD/ENSO event is found in the consecutive 2 years, those years are considered as neutral years. (a) Pink diamonds and line; pure positive-IOD (n = 11), Pink line and orange diamonds; positive-IOD with simultaneous El Niño (n = 3), Green diamonds and line; pure negative-IOD (n = 7), Orange line; pure El Niño (n = 15), Sky blue line; pure La Niña (n = 12), Brown cross and line; neutral (n = 11). Grey and pink shading represent East African Long and Short Rain period respectively. (b) Same as Figure 1a but only show for pure El Niño, La Niña, and neutral years with standard error bar. (c) Time-series of the coralδ18O value corresponding to March of year 0 in pure El Niño years, just before pure El Niño development (Orange diamonds, n = 12). We excepted the values of the year which negative-IOD occurred in the previous year (n = 3). Orange line represents a regression line. Brown line represents the average of theδ18O value (−5.31‰) corresponding to March of year 0 in neutral years (n = 11).

[12] In order to examine the ENSO influence on the coral δ18O, we examined the δ18O value in the boreal spring of the following year (February–May year +1). This is because the Pacific El Niño signal is expected to appear in corals a few months later. However, the δ18O pattern in El Niño year is almost the same as that in neutral years within a range of standard errors (Figure 1b). The observed SST of the area containing the coral site in May of El Niño year +1 is not different from that of neutral year (Figure S3a), thus the ENSO induced-SST positive anomaly is not shown in the coralδ18O record from Kenya coast. On the other hand, as mentioned this season corresponds to East African Long Rain period and the observed rainfall increases by 100 mm in May of El Niño year +1 (Figure S3b), yet the corresponding low coral δ18O anomaly does not appear in the coral record. During the East African Long Rain period, the coral is exposed to Long Rains on the site while discharge from Sabaki River is advected northeastward by the NE monsoon current [Brakel, 1984]. Thus we expect that the coral δ18O values for the boreal spring reflects SST and SSS (evaporation and advection) rather than the river discharge anomaly contrary to the coral IOD index of boreal fall season. However, SST warming is not observed, and thus, some advection may repress the decrease of the δ18O value resulting from the Long Rainfall. This is puzzling and needs future investigation. Eventually the ENSO-induced signals do not appear clearly as the positive SST and rainfall anomaly in the monthly coral record.

[13] On the assumption that the SST off Kenya is higher in El Niño years, Cole et al. [2000] reported the ENSO signal in the annual coral δ18O variability from Kenya coast. Their study based on a cross-spectral analysis reported coherence between the NINO3.4 SST index and coral record at the 5.5-year periodicity. However, our monthly-resolved analysis does not show the ENSO-induced SST anomaly in the coralδ18O variability. In addition, the coherence between the annual mean coral δ18O (KY16) and NINO3.4 SST index is very weak at 5.5-year periodicity (Figure 2). Following this observation, we conclude that the coherence reported by Cole et al. [2000] is below the confidence level during most part of the analysis period and the IOD is the dominant mode rather than ENSO in the Kenyan coast.

Figure 2.

Coherence spectrum of the coral annual mean record (KY16) and the NINO3.4 SST index. Monthly coral δ18O (KY16) values are averaged from previous November to October. Green line is the 99% significant level.

[14] In contrast, we found values of higher δ18O (up to +0.17‰) one year before the El Niño events (from year −1 to boreal summer year 0) compared with neutral year (p < 0.05, Figure 1b). Although higher East African Long Rain record (up to +200 mm) was observed in Kenya coast just before the occurrence of El Niño (Figure S3b), the coral record shows positive δ18O anomaly, the opposite of what is expected from higher than normal rainfall, and suggests a possible mixture of the signal caused by colder SST and higher SSS concentration resulting from the strong evaporation, advection and coastal upwelling off Kenya coast one year before the Pacific El Niño. Such a cold precondition before pure El Niño based on 3 negative coral IOD events in the total 15 events (20%); colder western pole leads to an El Niño (year 0), is consistent with a recent modeling study [Izumo et al., 2010]. The time series of the δ18O value in March of year 0 just before El Niño shows the highest values in the latter part of the 19th century (Figure 1c), and the possible effect of colder SST in the western Indian Ocean on the ENSO variability seems to be prominent before the 20th century.

[15] Several observational and modeling studies suggest that the anomalous SST distribution in the tropical Indian Ocean may influence the evolution of El Niño [Annamalai et al., 2005]. The cooling condition in the western tropical Indian Ocean as indicated by high values of coral δ18O in the present analysis is expected to influence the evolution of El Niño through intensifying the westerly wind and thus promoting the zonal SST gradient along the equator in the Indian Ocean just like a negative IOD [Izumo et al., 2010]. However, the SST in the western Indian Ocean has recently increased [Alory et al., 2007; Izumo et al., 2008] and it is close to the threshold for the active atmospheric convection [Gadgil et al., 1984]. This recent warming trend in the western side of the Indian Ocean seems to be partly related to the recent global warming [Nakamura et al., 2009]. The coral record from the Kenyan coast suggests that the warming in the tropical western Indian Ocean may reduce the zonal SST gradient and influence the Walker circulation in the Indo-pacific region. This may change the occurrence and the nature of the Pacific El Niño.

4. Conclusion

[16] The Pacific climate variability as typified by the ENSO variability has been considered as a major driver of the Indian Ocean SST. However, the present study shows that the western Indian Ocean SST is mostly independent of ENSO. From the seasonally stratified analysis, we have demonstrated here that the coral δ18O variability from Kenyan coast shows the clear IOD signal in January. But it does not show the Pacific El Niño and La Niña (ENSO) signals quite clearly. On the other hand, the coral δ18O shows the cooler SST condition in the western Indian Ocean one year before the evolution of the Pacific El Niño. This role of the Indian Ocean in the Pacific El Niño development was strong in the late 19th century but has subsequently weakened in the 20th century.


[17] We thank Robert M. Njue, the warden of Malindi Marine Park and Reserve, for permission and support during the coring, and Richard Adera of the Meteorological Department, Kenya, for providing precipitation data. We also thank Joseph Maina and Moses Mwambogo of the Wildlife Conservation Society, Hiroshi Adachi of Geoact Co. Ltd., Fumie Kobayashi of NISSAN Co. Ltd. for logistic support. The research was funded by a grant-in-aid for scientific research from the Ministry of Education, Culture, Sports, Science and Technology, Japan, and by the Japan Marine Science Foundation.

[18] The Editor wishes to thank an anonymous reviewer for assistance evaluating this paper.