High Arabian Sea productivity conditions during MIS 13 – odd monsoon event or intensiﬁed overturning circulation at the end of the Mid-Pleistocene transition?

. Marine isotope stage (MIS) 13 ( ∼ 500 000 years ago) has been recognized as atypical in many paleoclimate records and, in particular, it has been connected to an exceptionally strong summer monsoon throughout the Northern Hemisphere. Here, we present a multi-proxy study of a sediment core taken from the Murray Ridge at an intermediate water depth in the northern Arabian Sea that covers the last 750 000 years. Our results indicate that primary productivity conditions were anomalously high during MIS 13 in the Arabian Sea and led to extreme carbonate dissolution and glauconitization in the deep-sea sediments. These ob-servations could be explained by increased wind driven upwelling of nutrient-rich deep waters and, hence, by the occurrence of an exceptionally strong summer monsoon event during MIS 13, as it was suggested in earlier studies. How-ever, ice core records from Antarctica demonstrate that atmospheric methane concentrations, which are linked to the ex-tent of tropical wetlands, were relatively low during this period. This constitutes a strong argument against an extremely enhanced global monsoon circulation during MIS 13 which, moreover, is in contrast with results of transient climate modelling experiments. As


Introduction
The Mid-Pleistocene transition (MPT) characterises a fundamental change in the climate state which allowed ice sheets to expand and evolve from a dominant 41-kyr (obliquity) to a quasi ∼100-kyr rhythm (Clark et al., 2006;Lisiecki and Raymo, 2005;Raymo and Nisancioglu, 2003;Raymo et al., 2006;Shackleton and Opdyke, 1976). The end of the MPT between ca. 600 and 500 ka is described by a series of events (Schmieder et al., 2000). First, the transition between MIS 14 and 13 (i.e. termination T VI ) is the least pronounced termination of the past 640 ka. Ice volume has increased insignificantly during MIS 14, compared to the other late Pleistocene glacial periods. A record from Lake Baikal indicates, for instance, that mountain glaciations were reduced in central Eurasia from 580 to 380 kyrs ago (Prokopenko et al., 2002). In particular, the record documents a continuous forestation, suggesting that mild winter conditions prevailed with relatively little snow cover.
MIS 13 is, on the other hand, exceptional. It marks an extreme δ 13 C max associated with a major reorganization in the carbon reservoir of the global ocean (Wang et al., 2003). Several peculiarities occured in the ocean during this time, such as thick laminated layers of the giant diatom Ethmodiscus rex in the Atlantic Ocean (Romero and Schmieder, 2006). Also, the climate changed dramatically during this period with high terrigenous influx at Ceara Rise (Harris et al., 1997), indicating heavy precipitation in the Amazon Basin, or the exceptional thick soil horizon S5 found at the Chinese loess plateau (CLP) (Guo et al., 2009;Sun et al., 2006b). Moreover, extreme African and Indian monsoon intensity, inferred from the occurrence of the anomalous sapropel Sa in the Mediterranean and a peak in planktic oxygen isotope records from the equatorial Indian Ocean (Bassinot et al., 1994a;Rossignol-Strick et al., 1998), is commonly linked to this event (Guo et al., 2009;Yin and Guo, 2008).  Furthermore, the transition between MIS 14 and 13 coincides with the onset of the Mid-Brunhes dissolution interval (MBDI), which lasts until ∼280 ka (Barker et al., 2006;Bassinot et al., 1994b;Droxler et al., 1988). This period of extensive dissolution in the deep sea is probably not related to enhanced greenhouse gas forcing, since Antarctic ice core data and foraminiferal boron isotopes generally indicate low atmospheric pCO 2 levels, even within interglacial periods during this time (Hönisch et al., 2009;Petit, 1999). An alternative explanation for the MBDI invokes an increase in lowlatitude shelf carbonate production (Droxler et al., 1997). To add to that, it has been suggested that pelagic carbonate production increased globally due to the proliferation of the coccolithophore Gephyrocapsa (Bollmann et al., 1998), thereby, causing widespread dissolution in the deep sea (Barker et al., 2006). The most severe dissolution occured during MIS 11, which followed on from the so-called Mid-Brunhes event at ca. 430 ka (i.e. termination T V ), representing the largestamplitude change in δ 18 O of the global ocean over the past 6 million years (Wang et al., 2003).

Sediment core MD04-2881
The sedimentary sequence of the Murray Ridge provides an excellent archive of past primary productivity and Indian summer monsoon intensity (Pourmand et al., 2004;Reichart et al., 1997Reichart et al., , 1998Reichart et al., , 2004Schulz et al., 1998). IMAGES Core MD04-2881 was recovered on 14 October 2004, from a water depth of 2387 m at the Murray Ridge (22 • 12 .5 N -63 • 05 .5 E) (Fig. 1). The sediment consists of homogeneous, dark brownish to olive greenish to light greenish/yellowish grey hemipelagic mud. The upper 34 m of the core have been sub-sampled in 10 cm resolution. XRF and magnetic susceptibility scans have been performed in 1 cm resolution.

Analytical methods
An Avaatech XRF core scanner at the Royal Netherlands Institute of Sea Research (NIOZ, Texel, Netherlands) has been used to measure the bulk elemental composition of the sediment core in high-resolution. The split core surface was cleaned and covered with a 4 µm thin SPEXCer-tiPrep Ultralene foil to avoid contamination and prevent desiccation. Each section was scanned four times at 0.1 milliamps (mA)/5 kilovolts (kV) (no filter), 0.15 mA and 10 kV (no filter), 0.5 mA and 30 kV (Pd-thick filter) and 1 mA/50 kV (Cu-filter). A 1 cm 2 area of the core surface was irradiated with X-rays using 30 s count time (120 s for the 50 kV setup). For further technical details on the XRF scanning technique, see (Richter et al., 2006).
Reliability of XRF scanning counts has been tested by comparing it to a lower-resolution sample set (10 cm) for XRF measurements on discrete samples. 3-5 g of freezedried sediment was thoroughly ground. Residual moisture, organic matter and carbonates were removed using a Leco TGA (Thermo-Gravimetric Analysis), 600 mg of the residue was mixed with 6 g flux (consisting of 66% lithium tetraborate, Li 2 B 4 O 7 and 34% lithium metaborate, LiBO2) and 0.500 ml of a 30% lithium iodide solution and fused to glass beads. Glass beads were measured using an ARL9400 X-ray fluorescence spectrometer. Analytical precision, as checked by parallel analysis of international reference material and in-house standards, is better than 2% for Al, Ti better than 3% for Ba.
In general, XRF scanning is less suited for light elements (Richter et al., 2006;Tjallingii et al., 2007). When comparing the elemental scanning counts for Al with the absolute measurements on discrete samples, we find a low correlation (r 2 =0.38). This low correlation coefficient implies that normalization to Aluminum (Al), which is commonly done for elemental data, will lead to large uncertainties for the XRF scanning dataset. We, therefore, rely only on the raw counts for Barium (Ba), Calcium (Ca), Strontium (Sr), the sum of the terrestrial elements and Bromine (Br). A comparison between depth profile of Ba scanning-counts with the Ba/Al profile derived from conventional XRF measurements on discrete samples shows a perfect match between the two (Fig. 3d). This perfect match is why we conclude that closed-sum issues did not influence our record, in this particular case.
Magnetic susceptibility of discrete samples was measured on a Kappabridge KLY-2. Susceptibility was divided by the sample's dry weight, giving the mass magnetic susceptibility [m 3 /kg].
Stable isotope ratios were measured on the benthic foraminifera Uvigerina peregrina (single specimen, size fraction 150-600µm) and the planktic foraminifera Neogloboquadrina dutertrei (∼20 specimen, 300-350 µm) and Globigerinoides ruber (∼50 specimen, 212-300 µm). A single specimen of the benthic foraminifera and aliquots of the homogenized G. ruber samples were loaded into individual reaction vessels and each sample reacted with three drops of H 3 PO 4 (specific gravity = 1.92) using a Finnigan MAT Kiel III carbonate preparation device at Utrecht University. Long-term analytical precision was estimated to be ±0.07 for δ 18 O and ±0.03 for δ 13 C by measuring eleven standards (international NBS-19 and in house NAXOS) with each set of 38 samples. The Neogloboquadrina samples were analyzed in an ISOCARB common bath carbonate preparation device linked on-line to VG SIRA24 mass spectrometer also at Utrecht University. Isotope Values were calibrated to the PeeDeeBelemnite (PDB) scale. Analytical precision was determined by replicate analyses and by comparison to the international (IAEA-CO1) and in-house carbonate standard (NAXOS). Replicate analyses showed standard deviations of ±0.06 and ±0.1 for δ 13 C and δ 18 O, respectively.
Size-normalized weights of the planktic foraminiferal species G. ruber were measured to estimate the amount of carbonate dissolution. These measurements were done on the same relative narrow size fraction (212-300 µm) used for stable isotope analysis. The shells were weighed using a microbalance (precision 0.1 µg) and the mean weight is taken to represent that population.
Total numbers of the deep-dwelling planktic foraminiferal species Globorotalia truncatulinoides and Globoratalia crassaformis were counted on splits of the 150-600 µm size fractions from the wet, sieved freeze-dried sediment. The counts are expressed as number per gram dry sediment. Certain intervals of the core are characterised by high abundances of "green grains", which were counted on the same sample splits and are expressed as number per gram dry sediment.

Chronology
Age constraints are based on correlating the benthic δ 18 O U. peregrina record to the LR04 benthic oxygen isotope stack (Lisiecki and Raymo, 2005) (Fig. 2). This correlation shows that MD04-2881 covers the past ∼750 000 years, although the oldest ∼100 000 years are less well confined. The amplitude variations in the δ 18 O U. peregrina record are comparable to the global benthic stack, except for the interval below ∼600 ka, which shows only minor variations. The planktic δ 18 O records from N. dutertrei and G. ruber largely confirm the benthic isotope chronology. We do not find exceptionally light isotope values in any of the two planktic records during MIS 13, thereby questioning a monsoon related basin-wide flooding event in the northern Indian Ocean during MIS 13 (Rossignol-Strick et al., 1998). On the other hand, one could argue that also today most of the large river runoff from India is directed towards the Bay of Bengal and, therefore, the local salinity in the northern Arabian Sea was potentially less affected by an extreme increase in monsoon feed river discharge in the past. Similar to the U. peregrina record, a dampened δ 18 O signal is found in the record of N. dutertrei beyond ∼650 ka. The resulting age model indicates that interglacial periods are characterised by lower sedimentation rates compared to glacial periods. Sedimentation rate is, in particular, low during MIS 5 which may even suffer from a hiatus.
The reason for the dampened isotopic signal in the lower part of the core has not yet been solved, but it is well known  (Lisiecki and Raymo, 2005). that the benthic isotope signal in the Arabian Sea has been altered by OMZ variability through changes in carbonate ion concentrations and supralysoclinal dissolution (Schmiedl and Mackensen, 2006). Furthermore, changes in Arabian Sea intermediate water masses between glacial and interglacial periods potentially influence the isotope signal (Jung et al., 2001;Zahn et al., 1991), although it is not clear why this would affect both benthic and planktic δ 18 O records. Perhaps an increased diagenetic alteration of the isotopic signal with depth may have played a critical role. Clearly, the magnetic susceptibility record of MD04-2881 shows a decreasing down-core trend with flat values below ∼650 ka (Fig. 2), indicating the diagenetic removal of the magnetic properties in the sediment by the decomposition of organic matter and associated changes in the redox conditions of the pore waters within this interval (Reichart et al., 1997).

OMZ intensity and productivity changes
Marine organic carbon (MOC) content of Murray Ridge sediment cores has previously been used as productivity and/or OMZ intensity proxy (Reichart et al., 1998). It has recently been shown that the Br counts from XRF scanning enabled a fast and robust procedure to estimate the MOC content of the sediment (Ziegler et al., 2008). The Br record of MD04-2881 indicates that maximum MOC contents occur during glacial periods, whereas the lowest values coincide with glacial terminations (Fig. 3). These minimum values are accompanied by peak occurrences of G. crassaformis and G. truncatulinoides (Fig. 3) Ti/Al (discrete samples) . 3. Proxy records from MD04-2881 versus the global benthic isotope stack LR04 (black stippled line) (Lisiecki and Raymo, 2005). Ziegler, 2009). Similar to the ice rafted debris layers in the North Atlantic, peak occurrences of the Globorotalids usually do not last for more than a few thousand years and their abundances always return to very low baseline values before rising again. It has been suggested that their occurrences are indicative for periods of intensified winter mixing due to extreme cold winter monsoons, resulting in a breakdown of the OMZ (Reichart et al., 1998). Others argued that evidence for the required salinity and/or sea surface temperature changes in such a mechanism are missing and that the winter mixing theory is, therefore, hypothetical (Schulte et al., 1999). These authors linked a break-down of the OMZ instead to processes in the global oceanic circulation. The interval from 470 to 570 ka is remarkable, as it is the longest interval in the record where no G. crassaformis or G. truncatulinoides specimen occur. Amongst others, Reichart et al. (1997Reichart et al. ( , 1998 showed that the MOC content of the Murray Ridge records co-varies with other upwelling productivity indicators (e.g. Globigerina bulloides abundances and Ba/Al). Ba, for instance, has been successfully applied as proxy for primary productivity (Dehairs et al., 1980;Gingele et al., 1999;Jacot Des Combes et al., 1999;Shimmield and Mowbray, 1991). Barite crystals precipitate in microenvironments within decaying organic matter (Dehairs et al., 1980). One problem in the interpretation of Ba as productivity indicator lies in the distinction of biogenic and detrital Ba. Normalization with Al is, therefore, commonly used to assess the detrital Ba component (e.g. Gingele et al., 1999). The relative contribution of detrital Ba appears to be small at the Murray Ridge (Schenau et al., 2001), so that the Ba records we obtained from MD04-2881 by XRF scanning and discrete sampling will primarily reflect changes in productivity. Note that we will primarily use the raw counts for Barium in our discussion, because they are highly correlated with the Ba/Al ratios derived from the discrete samples of the last 462 ka (Fig. 3).
Evidently, the Ba record co-varies with the benthic oxygen isotope record, indicating highest primary productivity conditions during interglacial periods as was previously found (Shimmield, 1992). This implies that the maximum MOC contents during glacial periods, at the depth of our studied core, are most likely related to other processes than increased productivity conditions only, as has been suggested for other Arabian Sea MOC records (Clemens and Prell, 2003;Murray and Prell, 1992;Schmiedl and Leuschner, 2005).
A comparison of sediment cores from different water depths at the Murray Ridge indicated that relatively shallow cores from within the modern OMZ contain the highest MOC contents during interglacial periods and that they vary inphase with other productivity proxies, while the deeper sites (i.e. well below the present-day OMZ) contain the highest MOC contents during glacial periods (Ziegler, 2009). This suggests that the oxygen content, of the bottom water at the core depth, and thereby the extension of the OMZ, is an important factor in controlling the depth dependent preservation of organic matter. Primary productivity is a second factor, which becomes dominant in records that are constantly within the OMZ. Higher sedimentation rates during glacial periods would have further facilitated the preservation of organic carbon (Clemens and Prell, 2003), but this process cannot explain the differences in MOC content between various water depths. On this basis, we may conclude that the Br enrichments during glacial periods in MD04-2881 coincide with an extreme downward extension of the OMZ. In turn, the relative low Ba concentrations within the MOC maxima during glacial periods could be due to early diagenetic processes. Arabian Sea sediments that are deposited well within the modern OMZ are characterised by high C org /Ba bio ratios, because of a lower preservation of Barite upon deposition through sulfate-reducing conditions (Schenau et al., 2001).

Dissolution and dilution processes
Bulk elemental concentrations of Ca and Sr versus the sum of Al, Si, Ti, Fe and K reflect the input and preservation of biogenic carbonate versus the relative input of terrestrial material (Fig. 3). Because of its elevated location, the site is shielded from the input of turbidities and fan sedimentation of the Indus. The terrestrial material is, therefore, most likely eaolian (Reichart et al., 1997). Changes in the Ti/Al ratio of the sediments from the Murray Ridge have been applied in former studies as indicators for grain size and, thus, wind speed, since Titanium is concentrated in heavy minerals in the coarser size fraction (Reichart et al., 1997). The Ti/Al record of MD04-2881 (derived from conventional XRF measurements on discrete samples, not from XRF scanning) shows a close relationship with glacial-interglacial variability (Fig. 3) as was previously found for the Oman Margin, with higher Ti/Al values corresponding to an increased coarse-grained lithogenic flux into the Arabian Sea during dry glacial periods (Clemens et al., 1996). The total concentration of terrestrial elements in MD04-2881 shows, however, no clear glacial-interglacial variability. Several interglacial periods are even characterised by increased terrestrial element concentrations. This suggests that the bulk variations in terrestrial elements are dominated by the production and preservation of biogenic carbonate rather than by dilution.
Increased Ca and Sr contents and lower contents of terrestrial elements characterise the MBDI from 280 to 480 ka, with the exception of MIS 11 (Fig. 3). Similar to MD04-2881, this carbonate plateau has been found in other Indian Ocean cores and was related to long-term eccentricity-driven cycles in the production of coccolithopores (Rickaby et al., 2007). Extreme minimum Ca and Sr contents coincide with MIS 5 and 13. These interglacial periods are characterised by the lowest sedimentation rates and, hence, point to periods of severe carbonate dissolution (Fig. 3).
Calcite dissolution may occur above the lysocline when the metabolic release of CO 2 during organic matter remineralization leads to carbonate under-saturation in the pore waters (Adler et al., 2001;Jahnke et al., 1994). This supralysocline dissolution process typically occurs below the OMZ in the Arabian Sea, where a high flux of organic material is accompanied by oxygen availability (Klöcker et al., 2007;Schulte and Bard, 2003;Tachikawa et al., 2008). The water depth of the studied core at around 2400 m was apparently strongly influenced by supralysoclinal dissolution during interglacial periods, when productivity conditions were significantly enhanced.
Size normalized weights (SNW) of planktic foraminifera have been used as an indicator for surface (Barker and Elderfield, 2002) and bottom water carbonate ion concentration [CO 2− 3 ] (Broecker and Clark, 2001;Lohmann, 1995). The SNW of G. ruber shows a good correlation with the Ba record, but also with the extensive OMZ intensities during the glacial periods (Fig. 3). This suggests that the SNW records may represent an even better picture of productivity variations in the Arabian Sea than the Ba record, which could have been altered during extended OMZ conditions. Anomalous low SNW values are found during MIS 13. Due to the complete dissolution of foraminifers during MIS 5, no SNW data could be obtained from this interval.
Furthermore, MIS 5 and 13 are characterised by large numbers of light green to dark green grains in the sand size fraction (Fig. 3). Green grains commonly occur at the edges of oxygen-minimum zones and are composed of authigenic minerals, most commonly Glauconite (Kelly and Webb, 1999;Mullins et al., 1985). They often form within granular substrates such as faecal pellets or foraminiferal chambers. Glauconite forms at or near the sediment surface and requires low sedimentation rates, so that enough time is available for biological alteration of detrital clay minerals (Worden and Morad, 2003). The process of glauconization is often associated with relatively shallow water depths (<1000 m). The core depth of 2347 m is, to our knowledge, one of the deepest water depth where in-situ Glauconite formation has been found yet (see also Wiewiora et al., 2001).

Intensity of the Indian-Asian monsoon
The atmospheric methane record from Antarctic ice cores largely reflects the strength of tropical monsoon with a secondary input from boreal sources (Loulergue et al., 2008;Ruddiman and Raymo, 2003). Widespread wetlands, during periods of increased summer monsoon precipitation, are an important source of methane production when organic material decays under reducing conditions. Therefore, the atmospheric methane record provides important constraints for the interpretation of productivity changes and associated supralysoclinal dissolution intervals in our studied core from the Arabian Sea in terms of monsoon variability.
Currently, the longest methane record is derived from EPICA Dome C, which covers the last 800 000 years (Fig. 4). Changes in methane concentrations are dominated by the ∼100-kyr glacial rhythm superimposed on the 23-kyr precession component (Loulergue et al., 2008;Spahni et al., 2005). The strong imprint of the precession cycle is consistent with the outcome of climate model experiments, which indicate that tropical monsoons respond primarily to changes in Northern Hemisphere summer insolation on orbital timescales (Kutzbach, 1981). The link between monsoon variations and methane concentrations is supported by East Asian summer monsoon records from Chinese speleothem records, which show the same precession phase for maximum summer monsoon intensity (Wang et al., 2008). Recently, we carried out a transient simulation with the intermediate complexity model CLIMBER-2 that included both insolation and ice volume variations (Weber and Tuenter, 2010;Ziegler, 2009). Indeed, this simulation reveals that the intensity of Indian-Asian summer monsoon precipitation responds to both forcing parameters, in accordance with the Antarctic methane record over the past 650 kyr (Fig. 4). However, the methane record shows much stronger 100 000 year glacial-interglacial component, which is probably introduced by methane contribution from boreal wetlands (Loulergue et al., 2008).
Overall, the variations in Ba and SNW records of MD04-2881 and, thus, productivity changes in the Arabian Sea and associated changes in the carbonate ion concentration of the water, share features with the methane record and model simulation (Fig. 4). However, a detailed comparison of the two records shows an almost anti-phase relationship at the precession scale. A further marked difference, is the anomalous high productivity peak and carbonate dissolution event associated with MIS 13. During this time, methane concentrations are lower than in every other interglacial period of the last 500 000 years (Fig. 4). Also from a modelling perspective, the extreme summer monsoon maximum in MIS 13 is unexpected, because (1) benthic isotope records indicate that MIS 13 is a relatively cool interglacial (Lisiecki and Raymo, 2005), with remnant ice sheets in the Northern Hemisphere, and (2) Northern Hemisphere summer insolation maxima are not particularly strong in this period, although the earth's eccentricity was at a maximum around 500 ka (Laskar et al., 1993).
We note that high productivity conditions in the Arabian Sea during MIS 13 linked to enhanced summer monsoon activity would to some extend match with earlier interpretations. The anomalous sapropel (Sa) in the Mediterranean at 528-525 ka and a synchronous peak in planktic oxygen isotope records from the equatorial Indian Ocean have been interpreted as indicators of an unusually heavy monsoon event over Africa and Asia at the start of MIS 13 (Bassinot et al., 1994a;Rossignol-Strick et al., 1998) more recently, the timing of the Sa sapropel was evaluated by Lourens (2004), showing that it occurs within MIS 14, ∼20 000 years earlier as originally proposed, thus, questioning the correlation with the isotope excursion in the equatorial Indian Ocean. In addition to that, the Eastern Mediterranean planktic oxygen isotope records presented by Lourens (2004) indicate no extreme freshwater signal in connection with the sapropel Sa. In the following, we argue, based on the evidence from the methane record, that MIS 13 was most likely not characterised by an extreme, global summer monsoon event. This line of reasoning is further substantiated by new results from the Sanbao Cave speleothems. The extended cave record shows no anomalous isotope signature during MIS 13, arguing against abnormally high rates of precipitation during MIS 13 (H. Cheng, personal communication, 2009). We also note that the equatorial Indian Ocean isotope peak is a relatively short-lived event which contrasts the Arabian Sea productivity maximum, which appears to cover the whole MIS 13. This might indicate that different mechanisms are responsible for the observed events. As a consequence of our argumentation here, the equatorial Indian Ocean oxygen isotope excursion in MIS 13 requires a new explanation. Future research on new, long sedimentary records from the Bay of Bengal will provide additional information, which is necessary to solve this open question.

Inferences from the Chinese loess plateau
The Chinese loess plateau (CLP) is considered another important climate archive for the reconstruction of the Asian summer and winter monsoon as far back as 22 million years ago (Ding et al., 1995;Guo et al., 2002;Kukla et al., 1988;Porter and An, 1995). The winter monsoon transports dust from the Asian inlands to the CLP, while the summer monsoon brings precipitation (Porter and An, 1995). Successive loess and soil layers are, therefore, interpreted as alternating periods of strengthened winter (cold and dry) and summer monsoon (wet and warm), respectively. Recently it has been suggested that it is actually the breakdown of the Siberian High during spring that produces windstorms and associated dust deposition (Roe, 2009). Most proxies that have been used to unravel the history of the loess sequence (e.g. magnetic susceptibility) reflect the degree of chemical weathering and, thus, soil formation (Liu and Ding, 1998). Many loess records are dominated by glacial-interglacial variability superimposed by millennial scale events, which correlate to Heinrich events (Ding et al., 1995;Liu and Ding, 1998;Porter and An, 1995).
The Ba and, to a lesser degree, SNW records of MD04-2881 show a high similarity with a magnetic susceptibility stack from the CLP (Clemens et al., 2008). In contrast to the Antarctic methane record and model simulation, the exceptional high productivity conditions reached during MIS 13 coincided with an exceptional thick soil horizon S5 in some loess records of the central CLP (Guo et al., 2000;Sun et al., 2006b), and with an extreme event in a monsoon record from the Tibetan plateau (Chen et al., 1999). There are, however, noticeable regional differences in the expression of the S5 soil horizon (Sun et al., 2006a). While records from the central CLP expose a thick well-developed soil horizon, the S5 is hardly detected in the northwestern area. It was suggested that maximum intensities of summer monsoon precipitation did not reach this region until MIS 11 (Sun et al., 2006a). The latter observation is in much better agreement with the Antarctic methane record, which shows that methane concentrations were significantly lower during MIS 13 than during the interglacial periods after the Mid-Brunhes event, MBE, at ∼430 ka (Loulergue et al., 2008;Spahni et al., 2005).
Another major difference between the loess records of the central and northwestern site of the CLP is that in the central region soil occurrences are determined by glacialinterglacial variability, while they exhibit a strong precession imprint in the northwest (Sun et al., 2006a). The latter observation is not only in good agreement with the Antarctic methane record, but also with the Indian-Asian summer monsoon reconstructions derived from the Chinese speleothem oxygen isotope records of the Sanbao and Hulu caves, which indicate primarily 23-kyr precession cycles over the last 225 000 years (Wang et al., 2008). Similar to the loess records, the speleothem-derived monsoon record is overprinted by rapid events, which occur synchronously with climate variations in the North Atlantic region (Wang et al., 2005b;Wang et al., 2001).

Cause of the extensive productivity conditions during MIS13
Comparison of the Chinese loess records with temperature records from Antarctica have led to the suggestion that the climates of both hemispheres are unusually asymmetric during MIS 13 (Guo et al., 2009). Accordingly, Northern Hemisphere mean annual temperatures, evidenced by extreme soil formation in the Loess Plateau record, weakest Asian winter monsoon and lowest Asian dust and iron fluxes, were much warmer than at the Southern Hemisphere, because the global oxygen isotope record is characterised by relatively positive values (Guo et al., 2009). Moreover, the Deuterium (δD) record of the EPICA Dome C ice core showed relatively cold interglacial temperatures during MIS 13, indicating that at least Antarctic temperatures were cold with respect to the successive interglacial periods (Jouzel et al., 2007). On the other hand, data from a glaciomarine sedimentary sequence from the West Antarctic continental margin suggest that the interval spanning MIS 15-13 was one single, prolonged interglacial period, which potentially experienced a collapse of the West Antarctic Ice sheet (Hillenbrand et al., 2009). Warm Northern Hemisphere annual temperatures are consistent with the continuous forestation and inferred reduced mountain glaciations in central Eurasia throughout MIS 15 to 11 (Prokopenko et al., 2002). Tree growth is particularly sensitive to wintertime climate. Therefore, this period was probably characterised by mild winters, with relatively little snow cover. Such mild winter conditions would explain the absence of G. crassaformis or G. truncatulinoides in our Arabian Sea record in this interval. In addition, the higher winter temperatures may explain the thick soil horizon S5 in the central CLP. First it may facilitate pedogenesis through enhanced chemical weathering, and secondly a less intense winter monsoon may lead to a reduction of dust flux to the loess sites. As an alternative explanation from a modelling study, it was suggested that a precipitation maximum during MIS 13 could have occurred because of a reinforcement of the summer monsoon by an intermediate sized Eurasian ice-sheet . Such a scenario, however, does not explain the regional differences between the loess records and absence of a distinct monsoon event in the EPICA methane record during MIS 13. We, therefore, suggest that the anomalous climate patterns observed worldwide during MIS 13 are not primarily linked to changes in the intensity of the monsoon, but reflect an important turnover in the Atlantic circulation.
During the interim state of the MPT, the formation of North Atlantic deep water (NADW) was decreased and deep waters were influenced by a large Southern Hemisphere component (Raymo et al., 1997;Schmieder et al., 2000). Around T VI , a series of events occurred in the South Atlantic, which point to a significant increase in NADW formation during that time (Gingele and Schmieder, 2001;Romero and Schmieder, 2006;Schmieder et al., 2000): (1) A very high production of NADW has been inferred from globally distributed benthic carbon isotope records (Raymo et al., 1997). (2) During MIS 13 an extreme δ 13 C max occurs, which has been interpreted as a major reorganization in the carbon reservoir of the global ocean (Wang et al., 2001).
(3) A certain group of benthic foraminifera became extinct Kawagata et al., 2006). (4) An increased poleward heat transport in the Atlantic Ocean has been evidenced by pollen records offshore Greenland (de Vernal and Hillaire-Marcel, 2008). These records suggest that the size of the Greenland ice-sheet was much more reduced than today, even though the benthic isotope record indicates a larger global ice volume during MIS 13.
A modelling study showed that increased NADW formation affects primary productivity and OMZ intensity in the Arabian Sea through increased nutrient availability on millennial time scales (Schmittner et al., 2007). In a separate study, we argued that the orbitally-induced primary productivity changes in the Arabian Sea are also very sensitive to the global ocean circulation rather than only summer monsoon intensity, therefore, causing a much longer precession phaselag (Ziegler, 2009). Similarly, we propose that the productivity peak and associated anomalous dissolution event during MIS 13 relates to increased Atlantic overturning circulation www.clim-past.net/6/63/2010/ Clim. Past, 6, 63-76, 2010 around T VI . At the same time, increased heat transport to high northern latitudes might have caused the exceptionally mild winter conditions in Eurasia. Denton et al. (2005) suggested that the winter climate was much more sensitive to past changes in Atlantic meridional overturning, due to seaice related feedback mechanisms. Accordingly, intensified AMO may have resulted in mild winter conditions, facilitating soil formation on the central CLP. This implies that both Arabian Sea productivity and CLP soil formation was effectively decoupled from Asian summer monsoon intensity during MIS 13.

Conclusions
A high-resolution multi-proxy record from the north-eastern Arabian Sea of the past 750 ka reveals productivity changes, which oscillate primarily in concert with the ∼100 kyr glacial-interglacial rhythm. Highest productivity peaks are associated with interglacial periods. In contrast, the base of the OMZ deepens during glacial periods, suggesting that intermediate water ventilation played an important role. Termination T VI differs from the other major late Pleistocene terminations (T I−V and T VII ) by the absence of a strong winter monsoon-related event in the Arabian Sea. During MIS 13, primary productivity conditions were anomalously high and led to extreme carbonate dissolution and glauconitization in the deep-sea sediments. An intensive Atlantic overturning circulation during this time may have triggered mild winter conditions found in large parts of the Northern Hemisphere and, thereby, weakened the Asian winter monsoon. In turn, enhanced NADW production during T VI may have increased the supply of nutrients to the Arabian Sea, thereby, setting the stage for the anomalously high productivity conditions and the carbonate dissolution event during MIS 13. The presented interpretation constitutes an alternative view on MIS 13, which has been linked to an extreme boreal summer monsoon event in earlier studies. Future research, especially on long sedimentary records from the Bay of Bengal will potentially provide crucial information, which is necessary to finally answer the isotope stage 13 monsoon question.