3.1. Development of Surface Hydrography
 The G. ruber/G. sacculifer δ18O records of Site 709C from the tropical western Indian Ocean [Shackleton and Hall, 1990], of Site 763A from the subtropical eastern Indian Ocean, and of Site 214 from the tropical eastern Indian Ocean (Figure 4a) [Karas et al., 2009] show similar long-term trends with decreasing values from the early Pliocene until ∼3.5 Ma, changing toward more positive values during mid-Pliocene global cooling [Ravelo et al., 2004]. However, absolute δ18O values differ between cores. δ18O values at Site 709C are commonly more positive than those at sites 214 and 763A, possibly indicating higher salinities there. Since ∼3.5 Ma, δ18O values at Site 763A start to deviate showing at times more positive values compared to Site 214. Foraminiferal Mg/Ca implies similar SSTMg/Ca during the early Pliocene at all sites with a long-term warming trend of ∼2°C toward the mid-Pliocene at ∼4–3.6 Ma (Figures 4b and 5a). This observation is in accordance with Brierley et al. , who showed that during the early Pliocene the tropical warm pool was expanded with a reduced temperature gradient between the equator and the subtropics. During the mid-Pliocene, SSTMg/Ca from the tropical western Indian Ocean Site 709C remained stable at ∼26°C (Figure 5a) and broadly resemble those from tropical eastern Indian Ocean Site 214 [Karas et al., 2009]. The long-term decrease of ∼1°C, which is evident at sites 214 and 806 [Wara et al., 2005; Karas et al., 2009], is not seen at Site 709C implying that the tropical western Indian Ocean might have been less influenced by mid-Pliocene global cooling [Ravelo et al., 2004]. From ∼3.3 Ma onward, SSTMg/Ca at Leeuwin Current Site 763A became significantly cooler by 2–3°C than at tropical sites 214 and 709C from the present-day Indian Ocean Warm Pool. This gradient in SSTMg/Ca is comparable to modern conditions with an annual SST difference of ∼2°C between sites [Locarnini et al., 2006], implying that present-day SST conditions were already reached during the mid-Pliocene.
 Within the critical time period at 3.5–3 Ma, when we register a different hydrographic development at the ocean surface between sites 214 and 763A, there is evidence for distinct tectonic induced changes in the subsurface level (∼300–450 m water depth) G. crassaformis Mg/Ca derived temperatures and δ18Oseawater (Figures 1, 5a, 5b, 5d) [Karas et al., 2009] at tropical eastern Indian Ocean Site 214. The observation of more saline subsurface conditions during ∼3.3–3.1 Ma at Site 214 (Figure 5d) was interpreted as a consequence of the tectonic reorganization of the Indonesian Gateway [Karas et al., 2009]. Either the contribution of cooler and fresher ITF waters to this site was reduced and replaced by warmer and saltier tropical Indian Ocean waters or a switch back to more warm and saline South Pacific source waters occurred [Karas et al., 2009]. After ∼2.95 Ma, the change in ITF subsurface waters from a South to a dominant North Pacific source finalized [Karas et al., 2009] indicated by fresher and cooler conditions at the subsurface level at Site 214 (Figures 5b and 5d).
3.2. Pliocene Changes in the Leeuwin Current and Indian Ocean Polar Heat Transport
 It is important to note, however, that at Site 763A the surface ocean definitely cooled before the switch of ITF source waters finalized at ∼2.95 Ma. Surface cooling began at ∼3.5 Ma, and after ∼3.3 Ma the impact of warm tropical waters diminished compared to the tropical eastern and western Indian Ocean sites 214 and 709C (Figure 5a). At the same time, subsurface waters at Site 214 became more saline in line with a reduction of the ITF (Figure 5d) [Karas et al., 2009]. Both, the more saline conditions at the subsurface level at Site 214 and the cooler surface conditions at Site 763A suggest a significant reduction of the ITF at 3.3–3.1 Ma. As today the Leeuwin Current is controlled mainly by the ITF [e.g., Feng et al., 2003], we suspect that it was clearly reduced during the mid-Pliocene when the ITF declined due to the new tectonic setting in the Indonesian Gateway [Cane and Molnar, 2001; Gaina and Müller, 2007]. We presuppose in this respect that the Leeuwin Current was present also during the mid-Pliocene indicated by G. sacculifer δ18Oseawater records from sites 214 and 763A, which are very similar and imply a common ITF source water (Figure 5d).
 In fact, various modeling studies pointed out the effects of the mid-Pliocene ITF reduction [Hirst and Godfrey, 1993; Godfrey, 1996; Lee et al., 2002]. With an entirely closed Indonesian Gateway, these models generate scenarios quite similar to our reconstructions. Our observed SSTMg/Ca pattern in the Indian Ocean suggests a weaker Leeuwin Current being ∼2°C cooler, while SSTMg/Ca at the tropical eastern and western Indian Ocean sites 214 and 709C remain rather stable. In this respect, a cooling of surface ITF waters during this time interval causing the SSTMg/Ca decline in the Leeuwin Current area seems rather unlikely as the SSTMg/Ca at Site 214 (and at Site 806) [Wara et al., 2005] are hardly changing (Figure 5a) [Karas et al., 2009]. We therefore consider the reduction in ITF and not just a cooling as causative for the cooling at Leewin Current Site 763A. This notion is further supported by tectonic reconstructions of the ITF region, which propose a shoaling of the Indonesian Gateway with the emergence of small islands like Timor [Cane and Molnar, 2001; Gaina and Müller, 2007; Kuhnt et al., 2004], and in consequence, a restricted throughflow [Kuhnt et al., 2004].
 The tectonic reorganization might indeed have reduced the surface throughflow volume since ∼3.3 Ma, whereas the subsurface ITF after ∼3.1 Ma again started to cool and freshen due to the switch to North Pacific source waters [Karas et al., 2009]. This switch might have supported the surface layer cooling of Site 763A through mixing processes of the cold and fresh subsurface waters from the Indonesian region [Hirst and Godfrey, 1993; Song and Gordon, 2004]. In contrast, changes in the monsoon systems (Indian and Asian monsoon), driving oceanographic changes in that area today and possibly on glacial-interglacial timescales [Gordon et al., 2003; Xu et al., 2008], are unlikely of having cooled the (sub)surface during the mid-Pliocene time period ∼3.5–3 Ma, when Indonesian surface and subsurface flow changed. Significant changes in the monsoon systems and in South China SST clearly appeared after 3 Ma [Gupta and Thomas, 2003; Jia et al., 2008].
 Our findings of a reduced Leeuwin Current are consistent with marine and terrestrial palynological studies off northwestern Australia and from southwestern Australia, respectively [Martin and McMinn, 1994; Dodson and Macphail, 2004]. These studies suggest the expansion of aridity around 3 Ma in the southwest, the disappearance of rain forest and the development of shrub/grasslands in northwestern Australia. Indeed cooler SST in that area through a reduced Leeuwin Current would have caused a significant reduction in precipitation in the coastal areas of western and southwestern Australia [Feng et al., 2003].
 Apart from climatic effects on western Australia, we observe from 3.5 to 3 Ma a similar surface layer cooling at Site 1084 in the Benguela upwelling system [Marlow et al., 2000] as we registered at Site 763A (Figure 5a). Both temperature records show an almost identical development until 2.4 Ma, when alkenone-derived SST from Site 1084 further drops significantly. The good match of both SST records until late Pliocene times supports our notion that the restriction of the Indonesian Gateway possibly contributed to the cooling of the Benguela upwelling system [Karas et al., 2009]. Such cooling most likely resulted not only from the proposed cooling of subsurface waters in the tropical eastern Indian Ocean [Karas et al., 2009], but also from the reduced surface throughflow of Indonesian waters. Support comes from modeling studies [Hirst and Godfrey, 1993; Godfrey, 1996], which suggest both a weaker Leeuwin Current when the ITF is closed, and a weaker Agulhas Current resulting in a significant surface cooling of ∼1°C of the Agulhas outflow region close to Site 1084. In consequence, the poleward heat flux in the Indian Ocean would have been considerably reduced [Hirst and Godfrey, 1993; Gordon, 2005] explaining the enhanced meridional temperature gradient in the Indian Ocean and the cooling of the Benguela upwelling system during a time of global warmth with reduced meridional temperature gradients [Brierley et al., 2009]. After ∼2.4 Ma, the significant cooler alkenone SST at Site 1084 compared to the SSTMg/Ca at Site 763A are most likely related to the intensification of the trade winds and marked cooling of the Southern Ocean, which initiated the modern-like Benguela upwelling system (Figure 5a) [Etourneau et al., 2009]. At the same time, SSTMg/Ca at Site 763A remained relatively warm caused by a still flowing Leeuwin Current. Even though it was cooler and/or reduced, it prevented strong coastal upwelling, which would likely have developed without the presence of the Leeuwin current [Smith et al., 1991; Morrow et al., 2003].