Western Indian Ocean marine and terrestrial records of climate variability: a review and new concepts on land–ocean interactions since AD 1660
We examine the relationship between three tropical and two subtropical western Indian Ocean coral oxygen isotope time series to surface air temperatures (SAT) and rainfall over India, tropical East Africa and southeast Africa. We review established relationships, provide new concepts with regard to distinct rainfall seasons, and mean annual temperatures. Tropical corals are coherent with SAT over western India and East Africa at interannual and multidecadal periodicities. The subtropical corals correlate with Southeast African SAT at periodicities of 16–30 years. The relationship between the coral records and land rainfall is more complex. Running correlations suggest varying strength of interannual teleconnections between the tropical coral oxygen isotope records and rainfall over equatorial East Africa. The relationship with rainfall over India changed in the 1970s. The subtropical oxygen isotope records are coherent with South African rainfall at interdecadal periodicities. Paleoclimatological reconstructions of land rainfall and SAT reveal that the inferred relationships generally hold during the last 350 years. Thus, the Indian Ocean corals prove invaluable for investigating land–ocean interactions during past centuries.
KeywordsCoral oxygen isotopes Terrestrial proxy records Indian Ocean Africa India
In light of rapidly changing boundary conditions caused by global warming, decadal to multidecadal climate variability has become a major focus of climate research. In this context, the impact of rising sea surface temperatures (SST) on surface air temperature (SAT) and rainfall over land is of primary importance. Kumar et al. (2004) showed that over last 50 years tropical surface temperatures and precipitation over land and oceans have differing trends. SAT over land increases in phase with the rising SSTs. In contrast, precipitation over land decreases with increasing precipitation over the oceans due to the warmer SSTs (Kumar et al. 2004). This unfortunate combination of increasing surface temperatures and decreasing precipitation over tropical land regions could combine to produce considerable societal consequences. Therefore, changing teleconnection patterns, such as the breakdown of the historical ENSO-Monsoon correlation in recent decades, attracted considerable attention (Kumar et al. 1999). One reason for the very heterogeneous response of precipitation to rapidly warming tropical oceans is the non-linear relationship between SST and rainfall (e.g. Timm et al. 2005; Pfeiffer et al. 2006). In the tropical Indian Ocean, mean SSTs are high, and small changes may have a profound impact on precipitation (Timm et al. 2005).
Examining the relationship between tropical SST and land surface temperatures and precipitation on decadal scales is hindered by the relatively short observational database, particularly in the oceans. Proxy records can potentially extend the oceanic and land climate records far beyond instrumental data. Tree rings, lake sediment records and laminated stalagmites provide annually resolved proxy data for surface temperature and/or precipitation over the African-Asian landmass for several centuries (Fig. 1b; Holmgren et al. 1999; Tyson et al. 2002; Verschuren et al. 2000; Verschuren 2004; Burns et al. 2002; Cook et al. 2003; Deser et al. 2004; Fleitmann et al. 2004; D’Arrigo et al. 2005, 2006; Therell et al. 2006 and references therein). Geochemical parameters in massive coral skeletons from the Indian Ocean are well established in providing monthly to annually resolved time series of changes in SST, salinity and precipitation in the shallow oceans for up to 350 years (Fig. 1b; Charles et al. 1997, 2003; Cole et al. 2000; Cobb et al. 2001; Kuhnert et al. 2000; Zinke et al. 2004a, b, 2005; Pfeiffer et al. 2004, 2006; Timm et al. 2005; Damassa et al. 2006).
This article uses the available Indian Ocean coral δ18O records as proxy time series of SST, and compares their interannual to multidecadal variability to terrestrial time series (instrumental and proxy) (Fig. 1). We will focus on three regions where Indian Ocean SST’s are known to be important for regional temperature and rainfall variability, (1) India, (2) Tropical East Africa and (3) Subtropical South-East Africa (Fig. 1). We aim to bring the observed trends between ocean and land climate as observed by Kumar et al. (2004) into a long-term context. Such a comparison is invaluable in addressing the question to what extend anthropogenic forced warming influenced decadal teleconnections.
The paper is structured as follows: we will initially describe the climate of the Indian Ocean regions studied in this article. We will continue with a description of the instrumental and coral data used, before examining the correlation of the coral records with land temperature and rainfall over the twentieth century. In Sect. 5 we will provide an outlook of how terrestrial and marine paleoclimate records can help to understand past variations in the land-ocean climate interaction.
Climate of the western Indian Ocean and adjacent continents
The climate of the regions surrounding the Indian Ocean is dominated by the Asian monsoon. The monsoon circulation is driven by the pressure gradients resulting from differential heating over the Eurasian continent and the Indian Ocean, modified by the rotation of the Earth (Webster et al. 1998). The region influenced by the Asian monsoon is characterized by a seasonal reversal of surface wind fields and a distinct seasonality of precipitation. During the SW Monsoon (austral winter, June–September) India receives most of its total annual rainfall. During this time, the South Asian continent is very warm and large-scale atmospheric convection occurs over the continent. Over East Africa, the climatological evolution of rainfall is primarily controlled by the movements of the ITCZ from its position north of the equator in austral winter to its most southerly position in austral fall. For this reason, East Africa experiences two distinct rainy seasons, the short rains from October to December (OND) and the long rains from March to May (MAM). OND rainfall experiences larger interannual variability then the MAM rains (Hastenrath and Greischar 1993; Hastenrath and Polzin 2005). MAM rainfall is mainly dominated by tropical Atlantic Ocean variability (Mutai and Ward 2000). January to March (JFM) rainfall has also been in the focus of many studies (e.g. Hugh 2004). This season is considered the mean austral summer rainfall season where rainfall anomalies in East Africa between north and south of 10°S tend to oppose each other (Hugh 2004). The JFM season comprises important parts of rainfall seasons in central-east and southeastern Africa, which are closely linked to Indian Ocean SST. The mean austral summer season for southeast Africa south of 15°S is defined between December and March, when the ITCZ reaches its southernmost position.
The role of the distribution of SST in the Indian Ocean in determining monsoon rainfall variability over India is highly debated (Clark et al. 2000; Vecchi and Harrison 2004). High SST in the Arabian Sea precedes strong monsoon rain, which in turn precedes low SST in austral winter (Cherchi et al. 2007). The cooling in the western tropical Indian Ocean via wind-evaporation feedbacks is the strongest and homogeneously distributed across the western Indian Ocean (20°N–5°S, 50–70°E) for the late Asian monsoon season (September–November, SON). Vecchi and Harrison (2004) identified two distinct regions with strong rainfall anomalies in the India Monsoon domain, Western Ghats and Ganges-Mahanadi Basin, which appear to be independent from each other. This implies different mechanisms that control rainfall variability in the two regions. Western Ghats SAT and rainfall shows the strongest relationship with Arabian Sea SST. On interannual and interdecadal time scales, the monsoon is linked with ENSO. The occurrence of El Nino is usually associated with a weak monsoon and La Nina with a strong monsoon (Webster et al. 1998). However, the relationship between the Asian Monsoon and ENSO is statistically non-stationary and a breakdown of the relationship after 1976 is observed (Kumar et al. 1999).
Rainfall over Eastern Africa is strongly correlated with the two dominant processes of interannual variability affecting the Indian Ocean SSTs, the Indian Ocean Zonal Mode (IOZM) and the El-Niño-Southern Oscillation (ENSO) (Ropelewski and Halpert 1987; Saji et al. 1999; Goddard and Graham 1999; Poccard et al. 2000; Black et al. 2003; Clark et al. 2003). The IOZM is originally defined as the difference in sea surface temperatures (SST) between an eastern (EIO) and western Indian Ocean (WIO) region during boreal fall, it is in its positive mode when WIO-SST are anomalously warmer than EIO-SST. The ENSO originates from the tropical Pacific and is defined as the globally leading interannual coupled atmosphere–ocean process increasing WIO-SST in austral spring and summer (McPhaden et al. 2006). Warm WIO-SST in response to these modes were found to correlate strongly with excess rainfall in the so-called East African short rains (hereafter EASR; from October to December Goddard and Graham 1999).
South Africa Rain is characterized by interdecadal variability between 16 and 20 years, which is observed in several studies of South African summer rainfall variability (Mason 1995; Krueger 1999; Reason et al. 1996; Reason 2002; Reason and Rouault 2002; Tyson et al. 2002; Jury et al. 2004). Warm (cool) SST anomalies in the southwestern Indian Ocean are associated with increased (decreased) rainfall over central and eastern South Africa on seasonal to interdecadal time scales. Several studies concluded that this interdecadal variability is modulated by ENSO-like decadal variability, which influences southern Indian Ocean SST and wind fields on multidecadal time scales (Allan et al. 1995; Reason et al. 1996; Reason and Rouault 2002). Richard et al. (2000) proved that after the late 1960s, the influence of ENSO variability on South Africa rainfall increased, which is associated with a decrease in rainfall over certain areas. Reason and Rouault (2002) postulate that the remote influence of large-scale SST and sea level pressure (SLP) fields in the Indo-Pacific mainly influences the mean position of the tropical temperature troughs, the main rain-producing systems, and therefore, the amount of rainfall received in particular areas across southern Africa. Williams et al. (2006) show evidence for such associations for the recent decade.
Regions and data sets used for the selected regions
2.7–3.3; 5–7; 9–14
3.6; 5–6; 8–11; 18–25
2.2; 3.3; 5–6; 18–25
3.6; 4–6; 8–11; 45
2.7; 3.7; 16–20
3.3; 4–6; 9–11; 16–30
Historical monthly precipitation data are taken from the UEA CRU data product, which cover global land areas and extend over the time period from 1900 to 1998 (Hulme et al. 1998). The gridded data have a resolution of 2.5° × 3.75° longitude. We averaged the precipitation data over the Western Ghats region (7.5–20°N, 72.5–77.5°E), equatorial eastern Africa (5°N–5°S, 30–45°E) and southern Africa (15–25°S, 20–45°E). Western Ghats rainfall was cross-checked with India Monsoon rainfall data per area provided by the Indian Institute of Tropical Meteorology for West Central India centered at 19°N, 78°E including 129 years of data between 1871 and 1999 (Parthasarathy et al. 1991, 1995; Sontakke and Singh, 1996). Both datasets, UEA CRU and IITM products show similar interannual and interdecadal variability (Suppl. Fig. S1).
Table 1 summarizes the instrumental data used in this study. For SAT, we use mean annual values, and for the rainfall, we average the data over the main rain producing seasons in the regions investigated. Major periodicities of SAT and rainfall variability described in the literature are also indicated.
A complete global SST dataset (ERSST) has been constructed by Smith and Reynolds (2003) based on COADS data (Woodruff et al. 1998). The ERSST product provides reconstructed SST anomalies for the global ocean interpolated over a 10° × 10° grid. The monthly analysis begins in 1854, but the signal is dampened before 1880 due to sparse data (Smith and Reynolds 2003). We take the data for grid boxes centered at our various coral locations to groundtruth results obtained with our proxy data.
The coral cores were drilled between 1994 and 2003 from massive colonies of the genus Porites sp. (Porites lobata, Porites lutea, Porites solida), using a commercially available pneumatic drill (Suppl. Inf. Tab 1; Zinke et al. 2005). The average growth rate of the corals sampled varied between 9 and 16 mm per year (Zinke et al. 2005). Thus, the 1 or 2 mm sample spacing provides approximately monthly or bimonthly resolution, respectively. We generate time series of coral δ18O for annual mean and different seasons of the year that represent the main rainfall seasons investigated in this study [(January/February/March (JFM), September/October/November (SON), October/November/December (OND), December/January/February/March (DJFM)]. Details on the analytical procedures and data processing can be found in Zinke et al. (2005).
The coral records assembled here were collected from tropical to subtropical locations within the western Indian Ocean (Fig. 1; Suppl. Inf. Tab. 1). The tropical sites include the Seychelles, Kenya (Malindi; data from Cole et al. 2000) and Mayotte (Comoros) and the subtropical sites include La Réunion and Ifaty (Madagascar) (Fig. 1). From the Seychelles, two coral cores are available: BVB (Charles et al. 1997) and NEP (Pfeiffer and Dullo 2006). Coral data are available for download at the WDC database for paleoclimatology of the National Climatic Data Center (Charles et al. 1997; Cole et al. 2000; Pfeiffer et al. 2004, 2006; Zinke et al. 2004a, b).
Linear correlation coefficients (r) between annual mean coral δ18O and sea surface temperature (ERSST; Smith and Reynolds 2003) for a grid box including the individual coral sites for the time 1950–1995 and 1880–1995
Pfeiffer and Dullo 2006
Pfeiffer and Dullo 2006
Charles et al. 1997
Cole et al. 2000
Zinke et al. 2005
Pfeiffer et al. 2004
For comparison with instrumental SAT, all coral records were normalized by subtracting their mean and dividing by their standard deviation. For comparison with rainfall data, the coral records were normalized and detrended. Our results were crosschecked by using the relevant instrumental SST data from the ERSST data set (Smith and Reynolds 2003).
We compute a tropical coral index from the Kenya (Malindi), Seychelles (MAHE) and Mayotte annual mean δ18O records for the period 1801–1995. The mean series was made by first scaling the individual coral time series so that they had a mean of zero and a variance of one (i.e. standardized to z scores), and then averaging the normalized series. Simple averaging of all the coral records would result in a bias towards the time series with the highest variance. The two Seychelles records were combined into a single composite, as using each record on its own would have resulted in a bias to the Seychelles, which have been sampled twice (Wilson et al. 2006). Between 1847 and 1866, the tropical coral index comprises only the Seychelles and Kenya records, while between 1800 and 1847 the Kenya record was used on its own.
India monsoon region
In this section, we will investigate the relationship between coral proxy data, land temperatures and rainfall over India during the twentieth century. Vecchi and Harrison (2004) established the highest correlation between India monsoon rainfall and tropical Indian Ocean SST for the Western Ghats region only. Therefore, we focus on surface temperatures and rainfall from the Western Ghats region of India, and investigate the relationship between surface temperatures and rainfall with three tropical coral δ18O time series from the western Indian Ocean that are located within the field of the highest correlation established by Vecchi and Harrison (2004), i.e. Kenya (Malindi), Seychelles (MAHE) and Mayotte.
Surface air temperature (Western Ghats)
Asian monsoon rainfall (Western Ghats)
Several authors have explored the relationship between All India Rainfall and coral time series of SST from the western equatorial Indian Ocean (Charles et al. 1997; Cole et al. 2000; Pfeiffer and Dullo 2006). However, a statistically significant relationship could not be observed (Cole et al. 2000; Pfeiffer and Dullo 2006). This study focuses on SAT and rainfall in the Western Ghats region of India, as this region has been shown to be much more sensitive to SST in the western Indian Ocean (Vecchi and Harrison 2004).
The relationship between mean annual Western Ghats SAT and the mean annual average coral record from the western equatorial Indian Ocean is fairly strong and stable over the twentieth century. Warmer SSTs are generally associated with higher SATs. This also holds on long time scales. Coherency is particularly high on interannual time scales and most likely reflects the impact of ENSO and/or the IOZM (Fig. 3).
The relationship between coral δ18O and monsoon rainfall over the Western Ghats region of India is more complex. Correlations are highest for the late Asian monsoon season (SON) when the cooling in the western tropical Indian Ocean via wind-evaporation feedback is strongest and homogeneously distributed across the western Indian Ocean (20°N–5°S, 50–70°E; Vecchi and Harrison 2004). This suggests that the Indian Ocean SST in SON respond passively to the Asian Monsoon in that a stronger monsoon leads to stronger cooling in the tropics. This relationship is very stable throughout the twentieth century, with the exception of the late twentieth century. The strongest correlation with SAT and rainfall is observed with the Seychelles corals due to its position in the center of the cross-equatorial cell of the monsoon circulation. However, the SON time series calculated from both the Seychelles (MAHE) and Mayotte corals are coherent with rainfall in the Western Ghats region on decadal (11–13 years) and interannual (3–7 years) time scales. The interannual frequencies encompass the typical time scales of ENSO variability (e.g. Cole et al. 2000; Cobb et al. 2001; Pfeiffer and Dullo 2006). The tropical Pacific is known to have a strong influence on Indian Ocean SST and Indian monsoon rainfall and the correlation in both depends on the life cycle of ENSO. ENSO impacts on the strength of the Indian monsoon in its early phase, between August and November. El Nino (La Nina) events are typically associated with below-average (above-average) monsoon rainfall (e.g. Webster et al. 1998). The western Indian Ocean also responds to ENSO. ENSO-induced SST anomalies develop in SON and reach their maximum in austral summer (January–April) of the following year, during the mature phase of ENSO. El Nino (La Nina) events are typically associated with warm (cold) SST anomalies. Hence, it is not clear whether the observed coherency between the coral records from the western Indian Ocean and rainfall in the Western Ghats region on interannual and decadal time scales reflects a direct climatic link or mainly results from the impact of ENSO on both regions.
Interestingly, the change of the correlation between the tropical Indian Ocean coral records and Western Ghats rainfall in the late 1970 (Fig. 3a) coincides with the observed break-down of the historical relationship between ENSO and All India rainfall (e.g. Webster et al. 1998; Kumar et al. 1999), which has been attributed either to changes in the mean climate (Webster et al. 1998; Kumar et al. 1999), or stochastic processes (Gershunov et al. 2001). However, at present, it is not clear whether the change in the relationship between the Western Indian Ocean coral record and rainfall in Western Ghats is simply a result of the breakdown of the ENSO-monsoon relationship, as the impact of ENSO on SST in the western Indian Ocean sector has remained stable during the entire twentieth century (e.g. Charles et al. 1997; Pfeiffer and Dullo 2006).
Surface air temperatures
East African rainfall
Throughout the twentieth century, the tropical western Indian Ocean corals (and instrumental SST) follow SAT trends in East Africa (Suppl. Fig. S4). SAT and the coral index indicate periodic temperature variations with periodicities ranging between 30 and 45 years, although this signal is long, relative to the length of the time series investigated here. The 31-year running correlation clearly shows that the interannual relationship between western Indian Ocean SST and East Africa temperatures evolved and strengthened in the course of the twentieth century with the highest correlations in the last 50 years. As instrumental SST shows the same breakdown in the correlation with East African SAT prior to 1940, the observed change in relationship is unlikely to result from poor SST data or proxy data quality issues. We, therefore, speculate that the increase in the correlation between SAT and western Indian Ocean SST reflects the long-term warming trend in the region and possibly the increasing strength of ENSO. The prominent interannual and decadal periodicities found in the period after 1940 are consistent with the main frequencies reported for ENSO (Webster et al. 1998; McPhaden et al. 2006). In addition, the temporal correlation of the Nino3.4 Index with East Africa temperatures mimics the correlation indicated by the corals (Suppl. Fig. S6).
The relationship between the coral indices and East African rainfall is less clear (Suppl. Fig. S5). In general, warmer SSTs/more negative coral δ18O in the western equatorial Indian Ocean are associated with higher rainfall over tropical East Africa, during the short rains (OND) as well as in the mean austral summer season (JFM). However, while instrumental SST from the tropical Western Indian Ocean indicates a stable correlation with East African rainfall over the twentieth century (Figs. 5b, 6b), the coral δ18O time series also indicate periods with a weaker relationship, most notably in the 1930–1940 period (OND, Fig. 5b) and from the 1940–1970s (JFM, Fig. 6b).
The MAHE coral time series shows higher correlations with OND and JFM rainfall in East Africa than the combined MAHE/Mayotte time series. We note that a strong correlation between MAHE and OND precipitation appears in the late 1960s. Although instrumental SST indicates that our results must be viewed with some caution, we speculate that the observed strengthening of the correlation may reflect the warming of the Indian Ocean in the second half of the twentieth century, as rainfall in equatorial East Africa strongly depends on Indian Ocean SSTs (e.g. Latif et al. 1999; Goddard and Graham 1999). Despite the weak linear correlation, the coral records indicate that tropical SST and East African rainfall are coherent on interannual time scales (5–6 years) for OND short rains and JFM mean austral summer rainfall. The decadal coherence is strong for JFM in a frequency band from 9 to 13 years. For OND, coral and rainfall time series indicate decadal (9–11 years) and multidecadal variability (18–30 years), but weaker coherence. These results agree with cross-spectral analysis of OND East Africa rainfall with western Indian Ocean SST (Fig. 5d). The periodicities found in the coral records probably indicate the influence of interannual and decadal ENSO variability on Indian Ocean SST and East African rainfall (e.g. Goddard and Graham 1999; Latif et al. 1999). Recently, Kayanne et al. (2006) found a strong relationship to East Africa rainfall in a 15-year-long monthly resolved coral δ18O record from Malindi (Kenya), especially in Indian Ocean Zonal Mode (IOZM) years. A positive IOZM leads to warmer western Indian Ocean SST and easterly wind anomalies that favor strong precipitation events over tropical East Africa (Saji et al. 1999). Co-occurring IOZM and ENSO events as in 1997/1998 show the strongest response over the western Indian Ocean and eastern Africa (Behera et al. 2005). However, we note that the strong IOZM event which occurred in 1961 (e.g. Saji et al. 1999) is not recorded by the coral indices, although it led to extreme rainfall over East Africa during the OND short rains (Fig. 6a). Therefore, we conclude that longer monthly resolved coral records from offshore Kenya are needed in order to capture better the relationship between seasonal-scale SST anomalies in the western Indian Ocean and rainfall over East Africa.
Surface air temperatures
Southeast Africa rainfall
An important finding is that coral δ18O for DJFM is coherent with south African SAT and with rainfall in the summer rainfall season on interdecadal (16–20 years) time scales. Subtropical southern Indian Ocean SST corroborates these findings. Significant interdecadal variability with periodicities between 16 and 20 years is observed in several studies of South African summer rainfall variability (Mason 1995; Krueger 1999; Reason and Rouault 2002; Tyson et al. 2002; Jury et al. 2004). Warm (cool) SST anomalies in the southwestern Indian Ocean are associated with increased (decreased) rainfall over central and eastern South Africa on seasonal to interdecadal time scales. Several studies concluded that this variability is modulated by interannual ENSO and ENSO-like decadal variability, which influences southern Indian Ocean SST, SLP and wind fields on multidecadal time scales (Allan et al. 1995; Reason et al. 1996; Richard et al. 2000; Reason and Rouault 2002). Zinke et al. (2004a, b) report interdecadal modulations of ENSO teleconnections in the Ifaty-4 coral time series throughout the twentieth century supporting the hypothesis of Indo-Pacific interactions on interdecadal time scales.
On the long-term, our coral data indicate that southern Africa summer rainfall decreased after the 1970s in response to the recent warming of the southwestern Indian Ocean and over southern Africa. This agrees with the observed response of the subtropical African regions to recent global warming (Hoerling et al. 2006).
Summary of coral–climate teleconnections in the twentieth century
The western Indian Ocean tropical and subtropical corals follow trends in SAT in West India, East Africa and South Africa and show coherence on interannual to multidecadal time scales over the twentieth century (Suppl. Fig. S4). The relationship between western Indian Ocean SST and rainfall over adjacent land areas is more complex (Suppl. Fig. S5). We find evidence for strong coherence between ocean temperatures and land rainfall on interannual time scales, most possibly associated with the strong ENSO teleconnection of the Indian Ocean. For the Indian Monsoon region, we observe changing teleconnection patterns on interdecadal time scales probably linked to changes in the background SST through time. Our results are groundtruthed with instrumental data taken from the ERSST dataset (Smith and Reynolds 2003).
Land–ocean interaction from paleoclimatic records covering the past 350 years: an outlook
In this section, we will compare the coral records from the western Indian Ocean with other paleoclimatological records from the monsoon region, such as stalagmites, lake levels, sediments and ice cores. For this comparison, we use the tropical coral index, which extends back until 1801. The tropical coral Index will be compared with paleo-records from India and equatorial East Africa. For South East Africa, we use Ifaty only.
A comparison of various paleo-records is complicated by the fact that each paleoclimatic archive comes with a different temporal resolution ranging from annual to multidecadal and with a different age model uncertainty. Only tree ring chronologies have clear annual growth rings and age models as precise as corals. Nevertheless, we will show that major climatic trends can be traced in various archives, and that most of the relationships established for the twentieth century hold for the past 350 years.
Indian Monsoon region Paleo-records
To date, no temperature reconstruction for the Indian Monsoon region is available, but several temperature and rainfall reconstructions for the Tibetian plateau can be found (e.g. Thompson et al. 2002; Cook et al. 2003; D’Arrigo et al. 2006; Treydte et al. 2006). We selected the Dasuopu ice core, which provides temperature variations over high Asia (Fig. 9b; Davis et al. 2005; Thompson et al. 2002, 2003). Mean annual surface temperature for High Asia in the Dasuopu ice core shows a long-term increase with highest temperatures in the twentieth century. The tropical Indian Ocean coral time series show a long-term warming of tropical Indian Ocean SST largely in-phase with Dasuopu (Fig. 9b). Dasuopu indicates a warming trend of SAT between 1800 and 1900 followed by a short cooling period in the early twentieth century, and then an accelerated warming trend in the second half of the twentieth century. Recently, D’Arrigo et al. (2006) compiled several tree ring temperature reconstructions for Eurasia that are within or at the edge of the Asian monsoon region. They show evidence for anomalous warming and increased precipitation over the Tibetean plateau in the twentieth century. D’Arrigo et al. (2006) proposed that the change in Eurasian land/tropical ocean temperature gradient is coupled to an intensified Asian monsoon over the Tibetian plateau. We observe a strong link between our tropical Indian Ocean corals and SAT over the Tibetian Plateau.
Interestingly, the warm tropical SSTs, as inferred from our tropical coral Index, and temperatures over inner Asia between 1960 and 1990 are associated with lower Monsoon rainfall at Oman, while precipitation increased over the Tibetian Plateau (High Asia) and across inner Asia (D’Arrigo et al. 2006; Treydte et al. 2006 and references therein). These results mark the highly non-linear response of the climate system to anomalous warming in the twentieth century in the Asian Monsoon region.
East Africa Paleo-records
East African lakes are known to be very sensitive to climate variations (Mercier et al. 2002; Bergonzini et al. 2004). Fluctuations in lake level mainly reflect changes in precipitation and evaporation integrated over the lake catchment area. The historical lake level trends in equatorial East Africa show strong similarities (Lake Victoria, Lake Naivasha, Lake Turkana, Lake Bogoria and Lake Baringo; Verschuren et al. 2004). Most lakes show a major highstand between 1880 and 1920, a long decline to a lowstand in the 1950s and recovery thereafter (Verschuren 2004; Cohen et al. 2006). We selected Lake Naivasha, Tanganyika and Lake Victoria representative for historical lake levels in many equatorial East African lakes (Fig. 10a–c). Lake Naivasha shows a different long-term trend, indicating a broader highstand between 1890 and 1940 bracketed by lower lake levels (Fig. 10a; Verschuren 2004). Between 1870 and 2000, Lake Tanganyika and Lake Victoria lake level reconstructions agree, indicating the lowest levels in the first half of the twentieth century and high lake levels between 1870 and 1920 (Alin and Cohen 2003; Stager et al. 2005). Both Lake Victoria and Lake Tanganyika show a sudden jump to higher lake levels in the early 1960s (Birkett et al. 1999; Tate et al. 2004). The tropical Indian Ocean corals broadly follow long-term lake level variations over the nineteenth and twentieth century, although the sudden jump to higher lake levels in the early 1960s has no equivalent in the coral records (Fig. 10d). Therefore, this jump in lake level cannot be explained as a linear response to tropical SST warming. Rainfall station data from East Africa agree with lake level fluctuations on multidecadal time scales, with on average higher precipitation in the early and late twentieth century and a drier period in between.
Southeast Africa paleo-records
For southern Africa, a decadally resolved speleothem δ18O record from Makapansgat Valley (3,000 years) was interpreted to reflect a combination of mean annual temperature and rainfall variability in South Africa (Fig. 11a; Holmgren et al. 1999; Tyson et al. 2002). A lake level record is available from Lake Malawi with annual resolution for the period 1650–1990 (Fig. 11b; Owen et al. 1990; Johnson et al. 2001). A tree ring record from Zimbabwe provides an annually resolved southern Africa summer rainfall reconstruction for November–February for the period 1796–1998 (Therell et al. 2006; Fig. 11c).
The Ifaty-4 coral time series is the most robust subtropical Indian Ocean SST record and shows excellent agreement on the long-term trend with the Cold Air cave speleothem δ18O record from Makapansgat Valley (Cold Air Cave; hereafter CAC) from South Africa (Fig. 11a, d). On multidecadal time scales, we observe an anticorrelation with the CAC speleotheme. This would suggest that warmer SSTs are associated with wetter conditions at CAC (Holmgren et al. 1999). For the Late Maunder Minimum (1675–1715) drier conditions at CAC are associated with cooler SST at Ifaty. The CAC stalagmite shows a dominant 16–20 year periodicity similar to the Ifaty coral. Trends in lake level fluctuations in Lake Malawi show good agreement for most of the 350 year record with the Ifaty coral (Fig. 11b, d; Johnson et al. 2001). Warm subtropical SSTs are associated with higher lake levels and vice versa. The CAC stalagmite co-varies with the Lake Malawi record for most of the 350 years (Fig. 11a, b). This confirms that a large fraction of the stalagmite record is related to the hydrologic conditions over Southeast Africa (Holmgren et al. 1999).
The comparison with the Zimbabwe rainfall reconstruction based on tree ring chronologies shows an anti-correlation on multidecadal time-scales with the annual mean Ifaty coral record (Fig. 11c, d). Cross-spectral analysis indicates that the Zimbabwe rainfall reconstruction is coherent with the Ifaty-coral time series on multidecadal (50–60 year), interdecadal (16–20 years) and interannual frequencies (2–3.5; not shown). This agrees with the dominant time scales of variability observed in instrumental rainfall data from Southeast Africa from the twentieth century and in the Ifaty coral. Most of the interdecadal dry spells observed in the Zimbabwe rainfall reconstruction are associated with warm SST anomalies in the southern subtropical Indian Ocean.
Thus, for the last 350 years, our coral record off Madagascar indicates that the rainfall trends over southern Africa largely follow SST trends in the western Indian Ocean.
In this study, we established the interannual to centennial linkages between three tropical and two subtropical western Indian Ocean coral δ18O time series to land temperatures and rainfall over India, equatorial East Africa and southeast Africa.
The western Indian Ocean tropical and subtropical corals show a strong correlation with land temperatures on interannual to multidecadal time scales. Tropical western Indian Ocean corals show clear decadal linkages at frequencies ranging from 7–13 to 30–45 years with temperatures over western India and East Africa back to 1800. The subtropical corals follow southern Africa SAT trends over more than 300 years, and are coherent with surface air temperatures on interdecadal frequencies ranging between 16 and 20 years.
The relationship between western Indian Ocean SST and rainfall over adjacent land areas is more complex. Running correlation analysis suggests varying strength of the interannual relationship. We find evidence for changing teleconnection patterns between SST/land temperatures and rainfall on interdecadal time scales, possibly associated with changes in the background temperatures through time. The most prominent change occurred in the late twentieth century, when teleconnection patterns changed dramatically. Nevertheless, three tropical Indian Ocean corals are coherent with equatorial East African rainfall on interannual (5–6 years) and decadal (9–13 years) frequencies, the main modes of interannual (ENSO) and decadal (ENSO-like) climate variability in the tropical Indo-Pacific. The subtropical corals indicate strongest coherence with southern African rainfall on interannual and interdecadal frequencies ranging between 2–3, 5–6 and 16–20 years, respectively.
Overall, this study reveals the great potential of tropical and subtropical Indian Ocean corals for reconstructing the complex relationship between tropical surface air temperatures and precipitation over land and oceans during the twentieth century and potentially beyond. This will provide a framework for a detailed analysis of decadal to multidecadal interactions between land and ocean at time scales well beyond the instrumental record.
Comparison with paleoclimatological reconstructions of either rainfall or temperature over land areas reveals that most relationships established for the twentieth century, also hold for the last 350 years. Thus, we believe that the network of high-resolution proxy records from the Indian Ocean and the surrounding land mass will prove invaluable for a better understanding of land-ocean interactions over several centuries.
We would like to thank the TESTREEF-group (contract Nr. EV5 V-CT 94-0447) for the drilling of the coral cores between 1994 and 1996. We especially thank B. Thomassin for his enthusiastic support. We thank DAF/SPEM Mayotte, Université de La Réunion and Centre d’Océanologie Marseille for supporting our field campaigns and the CITES permits. We thank H. Erlenkeuser (Leibniz Laboratory, University Kiel) and M. Joachimski (University Erlangen) for carrying out part of the oxygen isotope analysis. This research was supported by the German Science Foundation (Leibniz Award to Prof. Wolf-Christian Dullo and SFB 460, poject B1), the Dutch NWO-ALW grant (awarded to Jens Zinke, grant number 854.00.034) and the European Community’s Human Potential Programme under contract HPRN-CT-2002-00221 (STOPFEN). O. Timm was supported by the Japan Agency for Marine-Earth Science and Technology (JAMSTEC) through its sponsorship of the International Pacific Research Center. This is IPRC contribution number 547.
This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.
- Brohan P, Kennedy JJ, Harris I, Tett SFB, Jones PD (2006) Uncertainty estimates in regional and global observed temperature changes: a new dataset from 1850. J Geophys Res 111. doi:10.1029/2005JD006548
- Burns SJ, Fleitmann D, Mudelsee M, Neff U, Matter A, Mangini A (2002) A 780 yr annually-resolved record of Indian Ocean monsoon precipitation from a speleothem from south Oman. J Geophys Res 107. doi:10.1029/2001JD001281
- Damassa TD, Cole JE, Barnett HR, Ault TR, McClanahan TR (2006) Enhanced multidecadal climate variability in the seventeenth century from coral isotope records in the western Indian Ocean. Paleoceanography 21. doi:10.1029/2005PA001217
- D’Arrigo R, Wilson R, Li J (2006) Increased Eurasian-tropical temperature amplitude difference in recent centuries: Implications for the Asian Monsoon. Geophys Res Lett 33. doi:10.1029/2006GL027507
- Davis ME, Thompson LG, Yao T, Wang N (2005) Forcing of the Asian monsoon on the Tibetian Plateau: evidence from high-resolution ice core and tropical coral records. J Geophys Res 110. doi:10.1029/2004JD004933
- Hastenrath S, Polzin D (2005) Mechanisms of climate anomalies in the equatorial Indian Ocean. J Geophys Res 110. doi:10.1029/2004JD004981
- Jury MR, Enfield DB, Mélice J-L (2002) Tropical monsoons around Africa: Stability of El Nino-Southern Oscillation associations and links with continental climate. J Geophys Res 107. doi:10.1029/2000JC000507
- Kayanne H, Iijima H, Nakamura N, McClanahan T, Behera S, Yamagata T (2006) Indian Ocean Dipole index recorded in Kenyan coral annual density bands. Geophys Res Lett 33. doi:10.1029/2006GL027168
- Parthasarathy B, Munot AA, Kothawale DR (1991) Indian summer monsoon rainfall indices: 1871–1990. Meteorol Mag 121:174–186Google Scholar
- Parthasarathy B, Munot AA, Kothawale DR (1995) Monthly and seasonal rainfall series for All-India homogeneous regions and meteorological subdivisions: 1871–1994. Contributions from Indian Institute of Tropical Meterorology, Research Report RP-065Google Scholar
- Pfeiffer M, Timm O, Dullo W-C (2004) Oceanic forcing of interannual and multidecadal climate variability in the southwestern Indian Ocean: evidence from a 160 year coral isotopic record (La Reunion, 50°E, 21°S). Paleoceanography 19. doi:10.1029/2003PA000964
- Reason CJC, Rouault M (2002) ENSO-like decadal variability and South African rainfall. Geophys Res Lett 29. doi:10.1029/2002GL014663
- Saji HH, Goswami BN, Vinayachandran PN, Yamagata T (1999) A dipole mode in the tropical Indian Ocean. Nature 401:360–363Google Scholar
- Therell MD, Stahle DW, Ries LP, Shugart HH (2006) Tree-ring reconstructed rainfall variability in Zimbabwe. Clim Dyn 26. doi:10.1007/s00382-005-0108-2
- Timm O, Pfeiffer M, Dullo W-C (2005) Non-stationary ENSO-precipitation teleconnection over the equatorial Indian Ocean documented in a coral from the Chagos Archipelago. Geophys Res Lett 32. doi:10.1029/2004GL021738
- Vecchi GA, Harrison DE (2004) Interannual Indian Rainfall variability and Indian Ocean sea surface temperature anomlaies. In: Wang C, Xie SP, Carton JA (eds) Earth climate: the Ocean-Atmosphere Interaction. AGU Geophysical Monograph Series 147, pp 247–259Google Scholar
- Verschuren D (2004) Decadal and century-scale climate variability in tropical Africa during the past 2000 years. In: Battarbee RW (ed) Past Climate Variability through Europe and Africa. Kluwer, DordrechtGoogle Scholar
- Williams CJR, Kniveton DR, Layberry R (2006) Climatic and oceanic associations with daily rainfall extremes over southern Africa. Int J Climatol 20. doi:10.1002/joc.1376
- Wilson R, Tudhope S, Brohan P, Osborn T, Briffa K, Tett S (2006) 250 years of reconstructed and modeled tripical temperatures. J Geophys Res 111. doi:10.1029/2005JC003188
- Woodruff SD, Jenne RL, Slutz RJ, Steurer PM (1998) A comprehensive Ocean-Atmosphere data set. Bull Am Meteorol Soc 68:1239–1248. doi:10.1175/1520-0477(1987)068<1239:ACOADS>2.0.CO;2Google Scholar
- Zinke J, von Storch H, Mueller B, Zorita E, Rein B, Mieding B, A - Eisenhauer- KIHZ-Consortium et al (2004a) Evidence for the climate during the Late Maunder Minimum from proxy data and model simulations available within KIHZ. In: von Storch H, Raschke E, Floeser G (eds) The climate in historical times—towards a synthesis of Holocene proxy data and climate models. Springer, BerlinGoogle Scholar