Kalimantan hydroclimate variability since the last glacial period

Changes in convective activity and hydroclimate over Northeastern Kalimantan are key features to understand glacial to interglacial climate evolution in the center of the West Pacific Warm Pool during the Late Pleistocene to Holocene. We use high-resolution X-ray fluorescence (XRF) scanner-derived elemental ratios in sediment Core SO217-18522 (1º 24.106’ N, 119º 4.701’E, 975 m water depth) recovered from the northern Makassar Strait to reconstruct changes in precipitation-related weathering and erosion over Northeastern Kalimantan over the last 50 kyr. Enhanced seasonality of rainfall and an extended dry season during Heinrich Stadials (HS4 to HS1) and the Younger Dryas (YD) suggest weakening of the tropical convection associated with a southward shift of the tropical rain belt and the annual mean position of the Intertropical Tropical Convection Zone. Increasing sediment discharge and intensification of convective activity occurred during the early to mid-Holocene during an interval of high Northern Hemisphere insolation, elevated atmospheric pCO2 levels and global warming. Our reconstructions in comparison with regional terrestrial and marine records highlight the high spatial variability of Kalimantan hydroclimate on millennial to glacial-interglacial timescales.


Introduction
The Indonesian seas are situated in the climatological center of deep atmospheric convection associated with the ascending branch of the Walker Circulation that leads to biannual precipitation maxima in the tropics (Aldrian and Susanto 2003;Hendon 2003;Sprintall et al. 2014;Cahyarini et al. 2021). Several paleoclimate studies that investigated the variability of deep atmospheric convection over Kalimantan since the last glacial (Partin et al. 2007(Partin et al. , 2015Carolin et al. 2013Carolin et al. , 2016Hendrizan et al. 2017;Schröder et al. 2018;Zhang et al. 2018;Krause et al. 2019) suggested a southward shift of the Intertropical Convergence Zone (ITCZ) during intervals of Northern Hemisphere cooling. In contrast, proxy records of the vegetation response to precipitation changes in Kalimantan since the last glacial period indicated a regionally highly variable hydroclimate (Wurster et al. 2017). However, reconstructions of past precipitation patterns may be affected by several biases. For instance, speleothem precipitation records may be influenced by changes in the isotopic composition of moisture sources. Precipitation estimates using paleosalinity reconstruction-based riverine discharge signals in marine sediment cores are affected by intense surface currents within the Indonesian seas. These paleosalinity-based riverine discharge signals are additionally influenced by vigorous vertical mixing of upper ocean waters by tides, monsoonal winds and the effect of rapid deglacial sea-level rise Zhang et al. 2018). The low spatial coverage of high-resolution marine salinity records is an additional limitation on their 1 3 potential for accurate reconstructions of regional precipitation variations.
Heterogeneous vegetation changes on the maritime continent between close canopy vegetation on the Sundaland landmass, including Kalimantan, and open canopy vegetation in separate exposed landmasses (Sulawesi and Palawan) may have occurred during the last glacial period but remain elusive (Wurster et al. 2017(Wurster et al. , 2019. In addition, the carbon isotope composition of cave guano suggests that dense rainforest persisted continuously over at least the last 15 ka in Northeastern Kalimantan (Wurster et al. 2019). However, a Southern Hemisphere eastern Kalimantan site suggested that significant drying occurred between 7.7 and 6.3 ka (Wurster et al. 2017). This evidence contradicts the general paradigm that sea-level rise and increased Holocene insolation led to an increase in Indonesian monsoonal precipitation and points to more regional variability, with a decrease in monsoonal rainfall in the mid-Holocene at Southern Hemisphere sites. Here, we attempt to resolve some of these contradictions using the elemental composition of terrigenous riverine discharge derived from X-ray fluorescence (XRF) core scanning to reconstruct hydroclimate variability in Northeastern Kalimantan in comparison to published data from Makassar Strait sites located in the Southern Hemisphere (Hendrizan et al. 2017;Schröder et al. 2018).
X-ray fluorescence scanner records of marine sediment cores are commonly used to reconstruct sediment discharge and evaluate changes in continental hydroclimate (Lianwen et al. 2002;Chen et al. 2006;Weltje and Tjallingii 2008;Kujau et al. 2010;Hu et al. 2013;Nace et al. 2014;Clift et al. 2014;Fraser et al. 2014;Kuhnt et al. 2015;Lo Giudice Cappelli et al. 2016;Hendrizan et al. 2017;Gebregiorgis et al. 2020). Marine sedimentary archives preserve climate proxies, including provenance-related variations in the elemental composition of terrigenous discharge (Log(K/ Ti)) (Kujau et al. 2010;Hendrizan et al. 2017), chemical weathering in the source area (Log(K/Al) and Log(K/ Rb)) (Hu et al. 2013;Clift et al. 2014;Lo et al. 2017), and amount of riverine terrigenous sediment runoff (Log(Ti/Ca), Log(Fe/Ca), Log((Fe + Ti + Al + Si + K)/Ca), Log(K/Ca) and Log(Al + K/Ca)) (Fraser et al. 2014;Nace et al. 2014;Kuhnt et al. 2015;Lo Giudice Cappelli et al. 2016;Gebregiorgis et al. 2020). However, the elemental composition of the terrigenous discharge derived from XRF core scanning as a proxy for hydroclimate variability in the catchment area of larger rivers is still under-exploited in the Indonesian seas. For example, the log-normalized ratios of (Ti/Ca), (Fe/Ca), and (K/Ca) were used to evaluate changes in river sediment runoff related to monsoonal rainfall in southern Indonesia and Australia (Kuhnt et al. 2015;Ardi et al. 2020) and precipitation changes at the northern edge of the Indonesian Seas particularly in Mindanao (Fraser et al. 2014). However, these river runoff proxies are not appropriate in shallower intermediate water sites in the Makassar Strait due to the influence of intense bottom water currents of the Indonesian Throughflow (ITF) that remove large parts of the finegrained clay component of the fluvial discharge (Hendrizan et al. 2017). These proxies also require near-constant biogenic carbonate accumulation rates and may be biased in proximal tropical settings with high variability in marine carbonate production and accumulation under the influence of local paleogeographic change and global sea-level variations (Aiello et al. 2019;Bova et al. 2020).
Here we present a high-resolution XRF scanner record of Core SO217-18522 from the Makassar Strait and compare it with other proxies in Kalimantan that document convection changes over the last 50 kyr ( Fig. 1; Table 1). For this study, we use the logarithmic ratios of potassium to titanium (Log(K/Ti)), titanium to aluminum (Log(Ti/Al)), potassium to aluminum (Log(K/Al)), and potassium to rubidium (Log(K/Rb)), which are driven by the intensity of chemical weathering, erosion and transport in the catchment area of the Berau River ( Fig. 1). We relate fluctuations in these elemental ratios to the variability of potassium-rich clay (illite) input from mountainous regions of the elevated part of the catchment (Kujau et al. 2010;Hu et al. 2013;Hendrizan et al. 2017;Lim et al. 2019), and thus to spatial and temporal changes in rainfall amount and seasonality associated with changing precipitation and tropical convection patterns in Northeastern Kalimantan.  -18522 (1º24.1060' N, 119º4. 7010'E, 975 m water depth) off the Berau Delta in Kalimantan and location of marine and terrestrial records referred to in this study. Cores SO217-18517 (Hendrizan et al. 2017) and 91GGC (Dubois et al. 2014) are indicated with yellow circles; Bau Bau and Gomantong Caves (Wurster et al. 2017), Niah Cave (Wurster et al. 2010), and Gunung Buda Cave (Partin et al. 2007(Partin et al. , 2015Carolin et al. 2013) are indicated with magenta circles. Gray shaded areas represent exposed shelves during sea-level lowstand at the Last Glacial Maximum (LGM) (~ 120 m isobaths). Map created with GeoMapApp (http:// www. geoma papp. org) based on Global Multi-Resolution Topography synthesis (Ryan et al. 2009)

Geological setting and sediment source
The Berau River catchment is located in the Berau subbasin, a part of the Tarakan Basin in East Kalimantan (Fig. 2a). The Berau sub-basin is bounded by Mesozoic and older rocks of the Semporna High in the northern region, Mesozoic and Eocene Melange complexes of the Kuching High in the western, Mangkalihat High in the southern, and the Makassar Trough of the Sulawesi Sea in the eastern part (Fig. 2a). The sub-basin is geologically subdivided into the Lati syncline and Rantaupanjang anticline with NNW-SSE fold axes (Suwarna and Hermanto 2007). The sub-basin was formed by extension and subsidence during the Middle-Late Eocene, which created wrench faults and resulted in the appearance of major NW-SE oriented structures (Subroto et al. 2005;Sudradjat and Hamdani 2018). Economically important lithologies filled the Berau sub-basin from the Cretaceous Telen and Benggara Formations, followed by shales with tuffs of the Eocene-Oligocene Sembakung Formation, Late Oligocene-Early Miocene carbonate sequences of the Selior and Tabalar Formation, and Plio-Pleistocene fluvio-deltaic and shallow marine deposit of Domaring and Sajau Formation (Krisnabudhi et al. 2020).   Situmorang and Burhan 1995;Djamal et al. 1995;Sukardi et al. 1995;Supriatna and Abidin 1995;Heryanto and Abidin 1995) 1 3 The Berau sub-basin is mainly covered by Neogene-Quaternary sediments in the eastern area and Mesozoic-Paleogene in the western region (Krisnabudhi et al. 2020). Detailed geological mapping in the Berau catchment area were carried out by Djamal et al. (1995), Heryanto and Abidin (1995), Situmorang and Burhan (1995), Sukardi et al. (1995), Supriatna and Abidin (1995). The geology of the Berau catchment in the eastern area ( Fig. 2b) comprises young Quaternary river deposits and Tertiary shallow to deep marine sediments, including limestone, deltaic deposits, glauconitic sandstone with abundant coral and molluscs, as well as volcanic deposits in the middle part of the catchment (Fig. 2b). In the western area, the geology of the catchment (Fig. 2b) consists of Cretaceous/Tertiary intrusive-oceanic crust deposits, including volcanic deposits, biotite-hornblende granite, andesite and diorite intrusion, quartzite, feldspathic, and micaceous sandstones of marine origin. The most important sources of the terrigenous load in the Berau River are Cenozoic marine biogenic shallow-water limestone deposits of the Selior and Tabalar Formations (Krisnabudhi et al. 2020). These carbonate-rich sediments are considered the primary sedimentary parent rocks for the Ultisols cover in the Berau district that is easily eroded by the river and transported to the Sulawesi Sea (Buschman et al. 2012). This large component of re-deposited carbonates in the Berau River runoff renders commonly used calcium-normalization of terrigenous sediment runoff problematic since this normalization is based on the assumption that the Ca-component of the marine sediment is entirely or primarily derived from marine biogenic carbonate.
The Berau River in northern Kalimantan is situated in the heart of the West Pacific Warm Pool (WPWP) in an area with sustained year-round precipitation. Today, the convective activity above the Berau River catchment area is related to the seasonal migration of the ITCZ, which passes over the area twice during its swing between a southerly position during December to February ( Fig. 3a and c) and a northerly position during June to August ( Fig. 3b and d). Precipitation ranges from 250 to 350 mm/month at the core site, with two peaks in May, when the ITCZ shifts to the north, and in November, when the ITCZ moves to the south (Fig. 3e). Kalimantan also experiences interannual precipitation changes during El Niño/ La Niña phases (Partin et al. 2007(Partin et al. , 2015. During El Niño events, warm SST in the central and eastern Pacific drive convection eastward, leading to drier conditions in northeastern and northern Kalimantan. During La Niña events, the convective activity strengthens, and the precipitation increases in northeastern and northern Kalimantan (Cobb et al. 2003).
The Berau River is located between two other large rivers (Kayan and Mahakam Rivers) that drain northern and central Kalimantan in an eastward direction into the Sulawesi Sea (Fig. 2b). The Berau River's discharge averages 605 m 3 s −1 at Gunung Tabur (Buschman et al. 2009(Buschman et al. , 2012. A plume of Berau River discharge flows to the Makassar Strait up to 15 km over the shelf, but it may also reach as far as the barrier reef, about 30 km to the north, during the wet season (Renema 2006). The river is influenced by a mixed tidal regime, predominantly semidiurnal, with a tidal range of about 1 m during neap tide and 2.5 m during a spring tide (Buschman et al. 2012). In addition, river transport is influenced by seasonally changing wind patterns (Tarya et al. 2015). During the southeast monsoon, from July to September, the wind speed increases to an average of 4 ms −1 , increasing northeastward freshwater transport (Fig. 3b, d). During the northwest monsoon (November to February), the wind is dominantly from a northerly direction, further enhancing the southward river transport into the shelf area (Fig. 3a, c).
The catchment area of the Berau River consists of ca. 100 km 2 rainforest vegetation, sustained by high rainfall of about 2900 mm year −1 (Damste 2016). The suspended matter load in the river is about 2 Mt year −1 and the sediment yield from the Berau catchment averages over 6.5 weeks is 140 t km −2 year −1 (Buschman et al. 2012). The distribution of clay mineral assemblages in the Berau River and its surroundings show strong similarities to that of the Mahakam River. The clay assemblages in the Berau River are dominated by high kaolinite content of > 50% associated with illite and smectite (< 20%) (Liu et al. 2012). Despite the similar climate-forced chemical weathering, the Mahakam River has, however, a lower kaolinite content (> 40%) than the Berau River, which is probably due to the more extensive area cover surrounding the Mahakam River (Liu et al. 2012).
Besides sediment derived from the Berau River delta, the core location of Core SO217-18522 is also influenced by lateral advection of ITF water masses (Fig. 2a). The ITF acts in northern Makassar Strait as a western boundary current along the Kalimantan continental slope (Susanto and Gordon 2005;Brackenridge et al. 2020). The surface (upper 50 m) flow is primarily wind-driven, with an average velocity of 0.5 m/s (Gordon et al. 2008). Velocities increase with depth, reaching a maximum of 1 m/s within the thermocline between 100 and 150 m. (Mayer and Damm 2012;Susanto et al. 2012). The flow and velocity of the ITF vary seasonally in response to monsoonal forcing. In particular, the ITF velocity is weaker/stronger during the northwest/southeast monsoon Sprintall et al. 2019). The ITF also controls sedimentation by winnowing and lateral advection of fine-grained sediment. For example, the fine-grained prodelta of the Berau delta system is deflected to the south by currents with maximum velocities of up to 80 cm/s at the delta front (Brackenridge et al. 2020). The Berau delta system was also strongly influenced by deglacial sea-level changes. During the sea-level lowstand of the Last Glacial Maximum (LGM), the paleo-Berau river sediment discharge occurred close to the location of Core SO217-18522, where terrigenous sediment was directly deposited (Supp. Fig. 3). The stepwise sea-level rise during the last deglaciation dramatically altered the topography of the Berau delta and the location of the Berau river mouth. It also affected the offshore current circulation by opening new ocean passages and connections to the South China Sea (SCS). Sea level rise started at 19.6 ka with a massive pulse of 10 m over ~ 800 years (Hanebuth et al. 2000(Hanebuth et al. , 2009Schröder et al. 2018). Afterwards, the sea-level rise slowed down (Heinrich Stadial 1/HS1), but abruptly increased at 14.5 ka during Meltwater Pulse 1a (MWP 1a), flooding a large part of the Paternoster Platform and Sunda Shelf (Hanebuth et al. 2009;Schröder et al. 2018). In addition, the narrow connection between the SCS and the Sulawesi Sea through the Sibutu passage widened during MWP 1a . The Sunda Shelf became fully flooded, and the southern connection to the SCS opened during Meltwater Pulse 1b (MWP 1b) at 9.5 ka, resulting in a regional circulation pattern similar to the present day (Hanebuth et al. 2000(Hanebuth et al. , 2009Xu et al. 2008;Schröder et al. 2018).

Materials and methods
Marine sediment core SO217-18522 was recovered with a split piston corer during R/V Sonne Cruise 217 Makassar-Java (MAJA) in 2011 . Core SO217-18522 is 12.33 m long and composed of homogenous dark gray/green fine-grained clay (Supp. Fig. 1a). The core is located offshore the delta of the Berau River, which is discharging runoff and terrigenous sediment into the Sulawesi Sea (Renema 2006). The regional hydrography is additionally influenced by surface and intermediate water circulation within the main ITF pathway .
XRF core scanning was carried out on the split core surface of the archive halves with the second generation Avaatech core scanner at the Institute of Geosciences, Christian-Albrechts-University, Kiel. The split core surface was covered with 4 µm thickness Ultralene plastic film to avoid contamination of the XRF measurement unit and desiccation of the sediment. Scanning was performed at 1 cm interval on an oriented, flat, and smooth surface with a slit size and a step size of 10 mm (X-ray voltage: 10 kV, 30 kV, and 50 kV, count time: 10 s for 10 kV and 20 s for 30 kV and 50 kV). Raw X-ray spectra were converted into area counts per second (area_cps) using the iterative least-square software package WIN_AXIL from Canberra Eurisys and a core-specific model. A total of 17 elements (Al, Si, P, S, Cl, K, Ca, Ti, Mn, Fe, Cu, Zn, Br, Rb, Sr, Zr, and Ba) were analyzed. Logarithmic ratios of elemental counts are used to avoid method biases related to method-specific effects (e.g., matrix effect, rock density, pore volume) (Weltje and Tjallingii 2008).
We used Log(K/Ti) and Log(K/Al) as a proxy of fluviatile clay discharge. Log (K/Ti) and Log(K/Al) values increase with the amount of clay-rich sediments characterized by high illite content from physical weathering and erosion in mountain catchment areas (Tian et al. 2011;Gebregiorgis et al. 2020). Log(K/Al) is commonly used as a proxy of chemical weathering (Ruxton 1968), following the assumption that mobile water elements such as K are leached by weathering the rock, rising the proportion of water-immobile elements such as Al and Rb in the resulting sediment (Hu et al. 2013). However, rapid erosion in a monsoonal, strongly seasonal climate regime tends to increase K/Al ratios in the discharged sediment (Tian et al. 2011). A proxy for chemical weathering and precipitation can be identified by K/Rb ratio, chemical weathering and precipitation increases the K/Rb ratio due to the smaller radius and greater bonding force of K than Rb (Lo et al. 2017). We also applied Log(Ti/Al) as a proxy of riverine terrigenous input, based on the assumption that Al is easily transported seaward in fine-grained clay, whereas Ti is deposited close to the river mouth (Lim et al. 2019). We used the Log normalized Zr/Rb to estimate changes in grain density (related to grain size changes). Higher Log(Zr/Rb) is due to enriched Zr in heavy minerals (zircon) and associated with a coarser-grained fraction, whereas Rb is enriched in clay minerals and characteristic of fine grain sizes (Dypvik and Harris 2001;Hendrizan et al. 2017;Kylander et al. 2011;Schneider et al. 1997).

Chronology
We used 11 published AMS 14 C dates between 0 and 30 ka  to establish the age model of Core SO217-18522. In addition, the δ 18 O minimum associated with Antarctic warm event A1 provided a further control point to tie our age model of the benthic δ 18 O record of Core SO217-18522 to the Antarctic oxygen isotope temperature chronology (EPICA 2006) (Supp. Fig. 1b). Sedimentation rates in Core SO217-18522 range from 30 to 52 cm/kyr between 40 and 25 ka, from 30 to 40 cm/kyr between 25 and 17.7 ka, and from 17 to 31 cm/kyr between 17.7 ka and present Fig. 4a; Supp. Fig. 1a). The highest sedimentation rates of more than 50 cm/kyr occur within the LGM (25-27 ka), with a marked decrease to ~ 30 cm/kyr associated with Heinrich Stadial 2 (HS2, 25-23 ka). The high glacial values finally decrease to less than 20 cm/kyr during HS1 (18-15 ka), followed by a marked rise to ~ 28 cm/kyr during the Bølling-Allerød (B-A), the YD and the earliest Holocene (15-10 ka). Sedimentation rates below 20 cm/kyr are low during the early Holocene (10-3 ka) and only increase in the late Holocene, which may be in part artificially due to stretching of the soft sediment in the upper part of the core by the piston-coring technique.

X-ray fluorescence (XRF) scanner-derived elemental records
The logarithmic ratio of XRF scanner-derived Ca (cps) to the sum of the terrestrial elements K, Ti, Al, Fe and Si, abbreviated as Log(Ca/Terr), shows marked fluctuations during the last glacial period compared to the interglacial period (Fig. 4b). Log(Ca/Terr) varies between -0.4 and 0.2 between 50 and 18 ka, then exhibits a sharp increase from -0.05 to 0.45 during the deglaciation. Log(Ca/Terr) decreases continuously from 0.5 to 0.1 between 8 and 0 ka. Log(K/Ti), Log(K/Al), Log (K/Rb), and Log(Ti/Al) fluctuate from 0.3 to 0.35, 0.6 to 0.75, 1.7 to 2, and 0.25 to 0.5, respectively, during the glacial period (Figs. 4c-d, 6e-f). A sharp rise in Log(K/Ti), Log (K/Al), and Log(Ti/Al) during the glacial/ interglacial transition to the Holocene is interrupted by slight decreases in Log(K/Ti), Log(K/Al), and Log(Ti/Al) during HS1 and the YD (Figs. 4c-d, 6f).
We identify five distinct stages in the evolution of carbonate and terrigenous elements in Core SO217-18522 during Northern Hemisphere cooling (Figs. 4, 6). (1) A slight decline in carbonate concentration relative to the terrigenous input indicated by Log(Ca/Terr), which started at ~ 40 ka and lasted until ~ 38 ka, marks the initial phase of HS4. The decrease in Log(Ca/Terr) is associated with a reduction of Log(K/Ti) and increases in Log(K/Al) and Log(Ti/Al). Further declines in Log(Ca/Terr) occurred (2) between 30 and 27.8 ka, (3) between 25.8 and 24 ka, (4) between 18 and 15 ka, and (5) between 12.9 and 11.5 ka. These drops in Log(Ca/Terr) together with decreases in Log(K/Ti), Log(K/ Al), and Log(Ti/Al) reveal declines in carbonate concentrations associated with Northern Hemisphere cooling episodes during HS3, HS2, HS1, and the YD.
The Log(Ca/Terr), Log(K/Ti), Log(K/Al), and Log(Ti/ Al) curves exhibit similar patterns from the last glacial maximum (LGM) until the mid-Holocene, but differ markedly in the mid-late Holocene (Figs. 4, 6). Log(Ca/Terr), Log(K/Ti), Log(K/Al), and Log(Ti/Al) display high-amplitude fluctuations from the LGM to the end of Termination 1. During HS1 and the YD, Log(Ca/Terr), Log(K/Ti), Log(K/ Al), and Log(Ti/Al) decrease from -0.2 to 0.2, from 0.35 to 0.3, from 0.8 to 0.6, and from 0.3 to 0.5, respectively. Log(Ca/Terr) increases continuously between 11 and 6 ka from − 0.1 to 0.5 (Fig. 4b), followed by a decreasing trend after 6 ka with values between 0.5 and 0.1. A continuous increase in Log(K/Ti) occurs from 11 until 4 ka with values between 0.27 and 0.5 (Fig. 4c) followed by a decrease from 4 to 0 ka. Log(Ti/Al) and Log(K/Al) increase from 11 to 0 ka with values between 0.7 and 1.2 and between 0.3 and 0.6, respectively (Figs. 4d, 6f).

Role of carbonate dilution
It is commonly assumed that in a two-component system of marine biogenic carbonate and terrigenous supply within the discharge area of tropical rivers the rates of carbonate sedimentation are more or less constant, while terrigenous runoff is strongly fluctuating. This high variability is driven by changes in monsoonal precipitation, erosion and sediment discharge along the adjacent continental margin. Bulk sedimentation rates are expected to vary with the terrigenous input, increasing with rising terrigenous discharge. However, sedimentation rates in Core 18522 are high during HS3, the LGM, the B/A, the YD, the earliest Holocene and the Common Era (CE) (Fig. 4a), when (except for the CE) the proportion of terrigenous sediment was relatively low or decreasing. Moreover, sedimentation rates remained overall low during HS2 and HS1, when the terrigenous input was relatively high and variable. This unusual pattern suggests that changes in carbonate accumulation were affected by the discharge of eroded carbonate from exposed fossil carbonate platforms into the Berau River delta during the last 50 kyr (Fig. 4b, Supp. Fig. 2). The high variability of carbonate input at the location of Core SO217-18522 is also apparent, when comparing XRF scanner Ca areal counts and the sum of terrigenous element areal counts (Supp. Fig. 7), which highlights the high variability of Ca counts compared to terrigenous element counts. Two main factors may have contributed to the high variability of carbonate accumulation in the Berau River discharge area: (1) eroded material from local carbonate sources of closely adjacent reefs and carbonate platforms contributed to the biogenic marine carbonate accumulation and (2) variability of the sediment transport to the core location during the last glacial and deglaciation due to changes in sea level, topography of the Berau Delta and ITF current activity (Hendrizan et al. 2017).
The variability of marine carbonate production in the vicinity of Core SO217-18522 may have influenced Log(Ca/ Terr), as modern carbonate production in warm shallow marine platform areas along the east coast of Kalimantan includes widespread and prolific patch reefs, atolls, fringing and barrier reefs, extensive carbonate platforms and mixed carbonate/clastic shelves (MacKinnon et al. 1996;Wilson and Evans 2002). Slope and deeper water environments receive re-deposited carbonates, particularly from the northern margin of the Mangkalihat Peninsula (Wilson and Evans 2002). However, the production of pelagic carbonate and re-deposition of fine-grained carbonate material from the carbonate platforms in this area is highly variable and was likely influenced by changing climate and sea level during the last deglaciation, particularly during the B/A and the early Holocene. However, the surface geology of the Berau River catchment is dominated by limestone deposits that are a significant contributor to the composition of the regional soil cover. Following erosion by the river and its tributaries, carbonates form the dominant an important part of the sediment load that is discharged into the Sulawesi Sea (Buschman et al. 2012). This large significant component of re-deposited carbonates in the Berau River runoff renders the commonly used calcium-normalization of terrigenous sediment runoff discharge problematic, since calcium-normalization assumes that the Ca-component of the marine sediment is exclusively derived from marine biogenic carbonate. To overcome this issue, we used Log(K/Ti), Log(K/ Al), Log(Ti/Al), and Log(K/Rb) to evaluate the temporal variability of potassium-rich and coarse-grained sediment discharge from the Berau River in Core SO217-18522.

Influence of winnowing by bottom currents
The grain size of the transported sediment load reflects the riverine current speed and transport efficiency, which are directly related to precipitation intensity. If sediment composition is mainly determined by the riverine discharge, grain size and the proportion of terrigenous input should be highly correlated. However, decreasing Log(Zr/Rb) in Core SO217-18522 indicates decreasing grain size during Northern Hemisphere cooling episodes (HS4-HS1, and the YD) (Fig. 4e). These decreases in grain size are associated with a low carbonate content (Log(Ca/(Terr)) and a high contribution of terrigenous clay. We therefore suggest that winnowing through bottom currents also affected the ratio of terrigenous elements to Ca in the Makassar Strait. Similar variations at Core SO217-18517 in the Makassar Strait during the YD were interpreted as reduced winnowing due to the weakening of ITF (Hendrizan et al. 2017). Today, a strong ITF flow transports clay minerals and other fine-grained particles from Kalimantan southward along the Makassar Strait (Eisma et al. 1989;Gingele et al. 2001;Kuhnt et al. 2004;Hendrizan et al. 2017), which may resuspend illite and kaolinite from the < 2 μm fraction as the primary source of K and Al. Thus, the decreased Log(Zr/ Rb) (Fig. 4e) and increased terrigenous input Log(K/Ca), Log(Fe/Ca), Log(Ti/Ca), and Log(Al/Ca) (Supp. Fig. 2) at Site SO217-18522 may indicate a slowdown of the deep thermocline ITF in the Sulawesi Sea during HS4-1 and the YD. This weaker thermocline flow would reduce winnowing and resuspension in the vicinity of Core SO217-18522, leading to a local increase in the terrigenous clay component of the sediment (Supp. Fig. 2).

Using K/Ti, K/Al and K/Rb as tracers of illite-dominated monsoonal clay input from the Berau river catchment
We use the discharge of potassium-rich (illite-rich) clays as an indicator of seasonal (monsoonal) precipitation following previous studies of monsoonal terrigenous sediment discharge in tropical regions (Tian et al. 2011;Hu et al. 2013;Clift et al. 2014;Hendrizan et al. 2017). This interpretation is based on the assumption that increases in monsoonal precipitation would cause enhanced erosion in the mountainous parts of the Berau River catchment. The correlation coefficient (r 2 = 0.96) of K and Ti in Core SO217-18522 is higher than the r 2 of other elemental ratios (K vs. Al, Ti vs. Al, K vs. Fe, K vs. Si, and Zr vs. Rb) (Fig. 5a) and is also higher than the correlation coefficient of K and Ti in Core SO217-18517 (r 2 = 0.92) offshore Mahakam Delta (Hendrizan et al. 2017). Higher K values indicate higher illite contribution to the clay mineral assemblages. Hendrizan et al. (2017) suggested that increased monsoonal rainfall in mountainous elevated areas results in increased content of illite and Ti-rich minerals in the terrigenous discharge. Enhanced K and Ti ratio also identified in the East Asian monsoon region indicates intensified erosion during summer monsoon in the SCS (Wei et al. 2003). We consider here a similar mechanism of K and Ti enrichment through changes in erosional intensity and provenance of mountainous elevated area. Therefore, the erosional processes related to increase in monsoonal rainfall could concentrate more K and Ti in the erosional products.
The clay mineral assemblages in the Mahakam and Berau Rivers (Liu et al. 2012) show similarities to that of the Pearl River and Mekong discharges in the northern and southern SCS, supporting that the erosional efficiency of monsoonal rainfall results in increased illite concentrations in riverine discharge (Hu et al. 2013). This interpretation is supported by the variability of Log(Ti/Al), Log(K/Al) (Figs. 4d, 6f) which shows similar trends as that of Log(K/Ti) over the last 1 3 50 ka in Core SO217-18522 (Fig. 4c). Overall, the comparison to monsoonal discharge patterns in the SCS indicates that variabilities of Log(K/Ti), Log(K/Al), and Log(Ti/Al) are widely applicable indicators of monsoonal erosion in the catchment and sediment discharge areas related to the seasonality and amount of precipitation.
The K/Rb ratio has been used as a proxy for increased chemical weathering and precipitation (Lo et al. 2017). In Core SO217-18522, Log(K/Rb) increases following a prolonged period rainfall between 25 and 11 ka, interrupted by an aridification spell associated with the YD (6e). However, the variability of Log(K/Rb) exhibits differences to Log(K/ Ti), Log(Ti/Al) and Log(K/Al) during the Holocene period (Figs. 4c-d, 6f). Log(K/Al) is related to the abundance of K-rich illite, whereas Al occurs as free oxide and/or substituted form in weathering-resistant clay minerals. Thus, Log(K/Ti), Log(K/Al), and Log(Ti/Al) are not only related to the amount of precipitation, but are also indicators of seasonality ("monsoonality"), the erosional power of rainfall and the increased efficiency of sediment transport in Fig. 5 Correlation of XRF scanner terrigenous elemental counts in Core SO217-18522. a Correlation between K and Ti (r 2 = 0.96); b Correlation between K and Al (r 2 = 0.95); c Correlation between K and Si (r 2 = 0.92); d Coefficient between K and Fe (r 2 = 0.86); e Correlation between Ti and Al (r 2 = 0.94); f Correlation between Zr and Rb (r 2 = 0.51) seasonally swollen rivers. In contrast, Log(K/Rb) is a proxy for chemical weathering, comparable to the chemical index of alteration in sediments, which relates to climate humidity and annual average of rainfall intensity.

Variability in glacial convective precipitation
Our records provide evidence of millennial-scale hydroclimate variability in the Sulawesi Sea over the last glacial period. The decreased sediment discharge during Heinrich Stadials is in agreement with records of lower precipitation from calcite δ 18 O in Northwestern Kalimantan cave records from (Carolin et al. 2013 ; Fig. 6c). The reduced rainfall is also reflected by increased sea surface salinity ; Fig. 6b). We suggest that millennial-scale hydroclimate variations in Kalimantan were driven by high-latitude climate change in the Northern Hemisphere (Fig. 6a), particularly the weakening of the Atlantic Overturning Thermohaline Circulation (AMOC) in the North Atlantic region (Partin et al. 2007;Carolin et al. 2013). This mechanism drove a southward migration of the ITCZ, which resulted in decreased precipitation over the Northern Hemisphere tropics, including Kalimantan (Chiang and Bitz 2005;Zhang and Delworth 2005;Carolin et al. 2013;Hendrizan et al. 2017).
The weakening of the AMOC is reflected by the thermohaline flow in South East Asian tropical regions, where it strongly influences the strength and vertical structure of the ITF . We attribute the decline in ITF intensity during HS4-HS1 and the YD, as indicated by lower Log(Zr/Rb) (Fig. 4e), to weakening of the AMOC and the inter-ocean flow through the Makassar Strait (Hendrizan et al. 2017). These millennial-scale changes in circulation may have contributed to increases in the ratio of terrigenous elements to Ca and lower Ca counts during HS4, HS1 and the YD in Core SO217-18522 (Fig. 4b, Supp. Fig. 6a).
However, no reduced Ca to terrigenous ratios are recorded for HS3 and HS2, despite the consistent supply of terrestrial sediment from the Berau River, as indicated by elevated sedimentation rates (Fig. 4a). During HS4, an increase in Log (Ti/Al) and Log(K/Al) (Figs. 4d, 6f) indicates enhanced erosion in the Berau River catchment, resulting in increased riverine clay input at the location of Core SO217-18522. This interpretation of enhanced erosion and increased riverine clay input during HS4 is supported by an increase in Log(Zr + Rb + Ti)/(K + Al) (Supp. Fig. 2e), as Zr, Fe, and Ti preferentially occur in heavy minerals and indicate higher erosional and transport capacity of rivers, whereas K and Al occur in background river-transported fine-grained clay. Overall, XRF scanning records and sedimentation rates (Figs. 4, 6) reflect a decrease in sediment runoff, erosion, and riverine clay input during Heinrich Stadials with the possible exception of HS4, supporting speleothem evidence of aridification during Heinrich Stadials in the northern Kalimantan region (Carolin et al. 2013).

Increased seasonality in Northeast Kalimantan during HS1 and the YD
The highest concentrations of terrigenous elements in Core SO217-18522 occur in sediments deposited during HS1 and the YD (Supp. Fig. 2). However, sediment provenance, the proximity of the coastline and the Berau River mouth, carbonate production and re-sedimentation, as well as the resuspension/winnowing of clay in marine sediments along the main ITF pathway, need to be considered when evaluating hydroclimate proxy records. Thus, it is crucial to assess  , c Speleothem record from Gunung Buda, Northern Kalimantan (Partin et al. 2007(Partin et al. , 2015Carolin et al. 2013); d Boreal summer insolation 2°N during June-August (black curve) from Laskar et al. (2004); e Log(K/Rb) at 1 cm interval (blue curve); f Log(K/Al) at 1 cm interval (black curve); g Composite of CO 2 concentration in the Antarctic ice core (Indermühle et al. 2000;Monnin et al. 2004). Green shadings indicate Heinrich Stadials (HS4-HS1) and the Younger Dryas (YD) whether the decreases in Log(K/Ti), Log(Ti/Al), Log(K/Rb), and Log(K/Al) in Core SO217-18522 during HS1 and the YD (Figs. 4c-d, 6e-f) reflect decreased sediment erosion in the catchment of Berau River, consistent with dry conditions in Kalimantan and Sulawesi (Partin et al. 2007;Hendrizan et al. 2017;Wicaksono et al. 2017;Zhang et al. 2018) during HS1 and the YD and/or a more proximal position of the coring site relative to the coastline and river mouth (Supp. Fig. 3).
The increase in Log(Ti/Al) during the B/A is paralleled by a rise of Log(K/Ti) and Log(K/Al), interpreted as enhanced riverine clay input due to higher rainfall erosion in the catchment of Berau River. Today, the freshwater plume of the Berau River, which consists of a ~ 2 m thick surface water layer, responds to changes in wind forcing with enhanced southward transport further into the shelf area during the wet monsoon (Tarya et al. 2015). Southward surface flow transports sedimentary organic matterrich clay and fine-grained particles, derived from soil erosion in the Berau River, to the shelf area (Booij et al. 2012;Damste 2016). The minima in Log(Ti/Al) (Fig. 4d) and the decreases in Log(K/Ti) (Fig. 4c) and Log(K/Al) (Fig. 6f) indicate reduced riverine terrigenous input during HS1 and YD, suggesting reduced monsoonal rainfall over the region. Our results support previous interpretations of a southward shift of the ITCZ causing dry conditions in northern Kalimantan and wetter conditions in southern Indonesia during HS1 and the YD (Zhang and Delworth 2005;Kuhnt et al. 2015;Hendrizan et al. 2017;Zhang et al. 2018).
However, the δ 13 C fatty acid (CFA) in marine sediments and Guano, used as a proxy for vegetation change on Kalimantan, yielded contrasting results for HS1 and the YD (Fig. 7). A decrease of δ 13 CFA and C 4 herb pollen in northern Kalimantan indicated reduced grass vegetation and aridity related to intensified precipitation during HS1 and the YD (Dubois et al. 2014, Fig. 7b). In addition, studies based on Guano δ 13 C insect cuticles indicated reduced C4 biomass in the Kalimantan cave region due to an increase in monsoonal rainfall over these two periods (Wurster et al. 2019(Wurster et al. , 2017(Wurster et al. , 2010Fig. 7f). We do not equate less seasonality to dryer conditions-conditions can be very wet with low seasonality (i.e. when looking at the present regional precipitation records with elevated rainfall during all seasons). Seasonality can also increase when dry seasons (or dry years during El Niño years) occur. We think "increased seasonality" may be an explanation for conflicting results of precipitation reconstructions across HS1 and the YD in Kalimantan region. The occurrence of wet conditions in northern Kalimantan (Fig. 7) due to increased seasonality could be related to differences in local insolation at latitudes 3°N and 3°S (Supp. Figs. 4,5), which were translated via vegetation feedbacks and SST changes in the WPWP (Wurster et al. 2019). Moreover, increased Northern Hemisphere solar radiation from June to August (Supp. Figs. 4,5) may have led to increased regional precipitation in northern regions of Kalimantan.

Increased precipitation during the early to mid-Holocene
The paleoclimate record in Core SO217-18522 indicates a common trend of increasing Log(K/Ti) between ~ 10 and ~ 6 ka (Fig. 7e), which is in agreement with the  (Partin et al. 2007(Partin et al. , 2015, blue curve represents δ 18 O corrected and black curve indicates δ 18 O uncorrected ; d δ 18 O sw in Sulawesi Sea Core SO217-18522 ; green curve); δ 18 O sw in Core SO217-18517 off the Mahakam Delta (Hendrizan et al. 2017; red curve); e Normalized ratio of Log(K/Ti) measured at 1 cm interval in Core SO217-18522; f Green and black curve represent Guano δ 13 C at Gomantong and Bau-Bau, East Kalimantan (Wurster et al. 2017), red curve shows Guano δ 13 C at Niah Cave, Northern Kalimantan (Wurster et al. 2010). Blue shadings indicate Heinrich Stadial 1 (HS1) and the Younger Dryas (YD), green shading marks the Bölling/Alleröd (B/A) and Mid-Holocene decreasing δ 18 O trend in the Gunung Buda stalagmite records, indicating increasing precipitation from the early to mid-Holocene (Partin et al. 2007(Partin et al. , 2015Fig. 7c). These increasingly wet conditions led to intensified sediment erosion (Log(K/Ti)) and riverine terrigenous input (Log(Ti/ Al)) from the Berau River catchment to the Sulawesi Sea (Figs. 4c -d, 7e). Increased rainfall over northern Kalimantan has been related to intensification of the East Asian monsoon (Hunt and Premathilake 2012) and to seasonal northward shifts in the ITCZ position, which intensified convection and drove higher SST (Partin et al. 2007).
The δ 18 O sw record from Core SO217-18522 also shows a decrease during the early to mid-Holocene (Fig. 7d) consistent with increased precipitation, sediment erosion, and riverine terrigenous input into the Sulawesi Sea. In contrast, a recent study based on δ 13 C of insect cuticles from the Guano record at Bau-Bau, East Kalimantan, shows higher δ 13 C values, reflecting the increased abundance of grass vegetation due to reduced precipitation during the early-mid Holocene (Wurster et al. 2017;Fig. 7f). We suggest that these differences may be interpreted as regional changes due to the northward shift of the ITCZ, which is supported by records offshore of the Mahakam Delta and south of Java indicating dry conditions south of Java and East Kalimantan during the early to mid-Holocene (Mohtadi et al. 2011;Hendrizan et al. 2017).
Northern Hemisphere summer insolation associated with elevated pCO 2, and global warming have been considered as the primary forcing of hydroclimate changes in Kalimantan during the early to mid-Holocene. Increased precipitation at 10-6 ka in Northeastern Kalimantan has been related to global climate change during the Holocene Climate Optimum, which coincided with maximum Northern Hemisphere summer insolation (Fig. 6d). The records from Core SO217-18522 show good agreement with PMIP models, reflecting the northward movement of rain belts from the tropics to the sub-tropics at 6 ka (Oh and Shin 2016). Core SO217-18522 is located close to the southernmost boundary of the Northern Hemisphere tropical rain belt, and maximum precipitation would have occurred at this location during the early to mid-Holocene. In contrast, dry conditions prevailed at Southern Hemisphere eastern Kalimantan sites, as indicated by δ 18 O sw values offshore the Mahakam River (Hendrizan et al. 2017) and higher values of δ 13 C in insect cuticles at Bau-Bau (Wurster et al. 2010(Wurster et al. , 2017 (Fig. 1).

Conclusion
Hydroclimate proxies (Log(K/Ti), Log(K/Al), Log(K/Rb), and Log(Ti/Al)) indicate drying in Northeastern Kalimantan during Heinrich Stadials. Decreases in Log(K/Ti), Log(K/ Al), Log(K/Rb) and Log(Ti/Al) reflect reduced sediment erosion, chemical weathering, and riverine clay input from the Berau River due to reduced precipitation over Northeastern Kalimantan. These records are consistent over the entire region of Kalimantan, suggesting that the main driver of aridity in the Berau River catchment during HS4 to HS1 and the YD was a southward shift of the ITCZ. However, the heterogeneity of Holocene climate records underlines the complexity of climate patterns in Kalimantan. During the early to mid-Holocene (10-6 ka), the hydroclimate proxies of Core SO217-18522 indicate intensification of precipitation in Northeastern Kalimantan whereas south of the equator Southeastern Kalimantan remained dry. This regional contrast in precipitation was maintained during the early to mid-Holocene by increased convective activity in the northern part of Kalimantan during intensified Northern Hemisphere precessional insolation. Comparison of our results with regional terrestrial and marine reconstructions highlights the high spatial variability of Kalimantan hydroclimate over the past 50 kyr.
Acknowledgements The authors acknowledge the invaluable help of the R/V Sonne captain and crew during the SO-217 MAJA cruise. We also thank Nils Andersen, Matthias Hüls, and A.Z. Rakowski for stable isotope and AMS 14 C measurements. We are grateful to Dieter Garbe-Schönberg, and Ulrike Westernstroer for helping with the X-Ray Fluorescence Scanner and to Dörte Mikschl for technical assistance during the laboratory preparation. This study was funded by the German Bundesministerium für Bildung und Forschung (BMBF, grant SO217 MAJA, 03G0217A) and the Deutsche Forschungsgemeinschaft (DFG, KU649/32-1). We acknowledge Beasiswa Saintek-BRIN 2019 for support of doctoral research at Institut Teknologi Bandung and Kemendikbudristek for support of research in the framework of 2021 Doctoral Dissertation Research (Contract No: 2/E1/KP.PTNBH/2021 to Nining Sari Ningsih).
Author contributions All authors are main contributors to the article. MH conceived and wrote the initial draft of the paper; WK, AH, SYC and NSN discussed the results, and modified the manuscript.
Funding Open Access funding enabled and organized by Projekt DEAL.
Data availability Data presented in this paper are archived at https:// www.pangaea.de.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.