Last Deglaciation—Holocene Australian-Indonesian Monsoon Rainfall Changes O ﬀ Southwest Sumba, Indonesia

: Previous studies suggested the multi-millennial scale changes of Australian-Indonesian monsoon (AIM) rainfall, but little is known about their mechanism. Here, AIM rainfall changes since the Last Deglaciation (~18 ka BP) are inferred from geochemical elemental ratios (terrigenous input) and palynological proxies (pollen and spores). Pollen and spores indicate drier Last Deglaciation (before ~11 ka BP) and wetter Holocene climates (after ~11 ka BP). Terrigenous input proxies infer three drier periods (i.e., before ~17, ~15–13.5, and 7–3 ka BP) and three wetter periods (i.e., ~17–15, ~13.5–7, and after ~3 ka BP) which represent the Australian-Indonesian summer monsoon (AISM) rainfall changes. Pollen and spores were highly responsive to temperature changes and showed less sensitivity to rainfall changes due to their wider source area, indicating their incompatibility as rainfall proxy. During the Last Deglaciation, AISM rainfall responded to high latitude climatic events related to the latitudinal shifts of the austral summer ITCZ. Sea level rise, solar activity, and orbitally-induced insolation were most likely the primary driver of AISM rainfall changes during the Holocene, but the driving mechanisms behind the latitudinal shifts of the austral summer ITCZ during this period are not yet understood.


Introduction
Despite its importance to the livelihood of people in the densely populated Southern Indonesia region, Australian-Indonesian monsoon (AIM) rainfall is still poorly understood [1,2]. A study on AIM past changes is significant to generate robust analogs as the basis to predict and model AIM rainfall future changes [3,4]. Previous studies from both marine and non-marine proxies in the AIM region (Southern Indonesia and Northern Australia) suggested millennial -multi-millennial scale changes of AIM during the Last Glacial Maximum (LGM)-Holocene [1,[5][6][7][8][9][10]. Based on modern conditions as analog, drier (wetter), or lower (higher) rainfall conditions characterizes the Australian-Indonesian winter (summer) monsoon intensification [2,11]. In general, drier condition was inferred at Last Glacial a more complete record of past rainfall to clarify the primary drivers of AIM changes since the Last Deglaciation (~18 ka BP).

Present Climate
The climate of the Indonesian Archipelago is dominated by the contrast condition between AISM and Australian-Indonesian winter monsoon (AIWM) [41]. A moisture-rich summer monsoon occurring during November-March induces the wet season, while the drier winter monsoon delivers the dry season during April-September. This bi-annual monsoon switch is induced by the seasonal latitudinal shift of the Intertropical Convergence Zone (ITCZ) [41,42] (Figure 1). The peak of AISM coincides with the SH (austral) summer, while during the NH (boreal) summer, AIWM reaches its peak [43,44]. The rainfall differences between the wet and dry season are most prominent in Southern Indonesia, from Southern Sumatra to the Lesser Sunda Islands, indicating more monsoon influence on this region [27,45]. In this region, the monthly rainfall reaches its peak in January, while August is the month with the lowest rainfall [27]. Based on the higher (lower) rainfall during AISM (AIWM) in modern Southern Indonesia, it is assumed that the past increases (decreases) in AIM rainfall were related to AISM (AIWM) enhancing [1,2,6]. The southward shift of the ITCZ during the austral summer (boreal winter) is induced by the northern hemisphere (NH) cold surges that blow through the South China Sea (SCS) [46] (Figure 1). The different dominant winds during AISM (northwest winds) and AIWM (southeast winds) affect the wind-driven surface currents in the Indonesian seas and Eastern Indian Ocean (i.e., Karimata Strait Throughflow, Equatorial Counter Current, and South Java Current) [47][48][49] (Figure 1). The bi-annual switch of AIM also affects the Indonesian Throughflow (ITF), part of warm water backflow of thermohaline circulation through changes in the direction of Karimata Strait Throughflow (KSTF) [47,50,51] ( Figure 1). The KSTF brings the low salinity surface waters from SCS and other seas it goes through (i.e., Karimata Strait, Java Sea) to the Flores Sea [48] during AISM. The presence of low salinity surface waters in the Flores Sea impedes the ITF's surface flow, as a result, it enhanced the ITF's intermediate flow [48,52,53]. The distribution and preservation of land-derived pollen and spores in the marine sediments were most likely influenced by the ITF and the bi-annual switch on the direction of winds and surface currents [36- 38,54].  [8,41,42,[46][47][48]50,55]. Black arrows: dominant wind direction during AIWM (solid) and AISM (dashed). Light blue dashed two-way arrows: wind-driven surface currents (i.e., ECC, SJC, and  [8,41,42,[46][47][48]50,55]. Black arrows: dominant wind direction during AIWM (solid) and AISM (dashed). Light blue dashed two-way arrows: wind-driven surface currents (i.e., ECC, SJC, and KSTF). Dark blue arrows: ITF. Purple arrows: NH cold surges. Red shading: boreal summer ITCZ. Yellow shading: austral summer ITCZ. The red rectangle indicates the Sumba region.

Methodology
Core ST08 was retrieved from off southwest of Sumba Island at the depth of 2966 m below sea level (mbsl) as part of the 2016 Widya Nusantara Expedition (E-WIN) using Baruna Jaya VIII research vessel (R.V.) operated by Research Center for Oceanography, Indonesian Institute of Sciences (LIPI). The sampling location was at 10.19495 • S and 118.77246 • E or approximately 80 km southwest of Sumba Island (Figure 2). This deep marine sediment core has a length of 235 cm and consists of foraminifera-bearing gray silt with a few mollusk fragments [56]. Five radiocarbon-dated samples and foraminiferal zonation revealed the sediment deposition period of core ST08, which was at Late Pleistocene-Holocene [56,57].
The samples for XRF were collected at 2 cm interval (117 sub-samples were taken) to ascertain that their measurement was carried out at the same depth level as other proxies (i.e., foraminifera, palynology). The samples were covered by plastic wrap (polypropylene) to avoid contamination (from the desiccation of sediments) before being measured by Thermo NITON XL3t 500, a Field Point X-Ray Fluorescence (FP-XRF) device at The Chemistry of Geological Resources Laboratory, Research Center for Geotechnology of LIPI, Bandung, Indonesia. The elemental data were measured in parts per million (ppm). Titanium (Ti) and Potassium (K) used as proxies to infer the AIM rainfall after being normalized by Calcium (Ca). Ti represents the constituent minerals of mafic and ultramafic rocks on the surrounding islands [2,58,59], K could be originated from the weathering of micas found on the exposed land outcrops [60], and Ca is mainly derived from the carbonate test and skeleton of marine organisms [61]. The increase in Ti and K indicates the enhancement of riverine input to the ocean, which could be triggered by the enhancement of AIM rainfall [1,2]. As the single element data obtained from XRF has limited capabilities [30,32,34], natural logarithmic (ln) values of elemental ratios are utilized to infer the past rainfall changes instead. Ca was used as the denominator for Ti and K, hence the past changes of AIM rainfall were inferred from ln Ti/Ca and ln K/Ca proxies instead. The use of Ca as a denominator needs to be validated, due to the presence of carbonate formations on Sumba Island [62,63], which indicates the presence of land-originated Ca. Validation is performed through the linear correlation of Ti vs Ca, which produces low correlation values ( Figure 3). This indicates that Ca is generally derived from marine biogenic processes, hence its use as a denominator for Ti and K may remove the biogenic components from the elemental record [59,64,65].
Pollen and spores distribution data were obtained from the pollen and spores distribution chart of core ST08 [66]. The analysis was performed on 58 sub-samples (collected in 4-8 cm interval and weighed~6.5 g), and their preparation was done using acetolysis method [67] in The Chemistry of Geological Resources Laboratory, Research Center for Geotechnology of LIPI, Bandung. A total of 150 specimens of pollen and spores were counted and determined for each sample to sufficiently represent the vegetation diversity [68]. The determination and grouping of pollen and spores were conducted based on Pollen and fern spores recorded in recent and Late Holocene marine sediments from the Indian Ocean and Java Sea (APSA) [69], and the Plant Zonation [70]. The abundance of Poaceae, grassland pollen, lowland/peatland pollen, and Pteridophyta were employed to infer the past changes of AIM rainfall. The higher (lower) abundance of Poaceae and grassland pollen indicate drier (wetter) climate or lower (higher) AIM rainfall while the higher abundance of lowland/peatland pollen and Pteridophyta indicate the opposite condition [37][38][39]71].

Palynological Proxies (Pollen and Spores)
The records of pollen and spore proxies show less fluctuation compared to the terrigenous input proxies. The abundances of grassland pollen and non-arboreal pollen (NAP) were higher before~11 ka BP (Last Deglaciation) and declined afterward. Poaceae show a similar trend, but with the addition of a slight decrease in abundance~16.5-15 ka BP. On the other hand, the abundances of lowland/peatland pollen, mangrove pollen, montane pollen, and arboreal pollen (AP) showed the opposite trend, lower during the Last Deglaciation and increased afterward. The abundance of Pteridophyta was also lower in the Last Deglaciation and increased since~9 ka BP ( Figure 6).  (Table S1) and relevant foraminifera data from previous studies on core ST08 (total foraminifera, Neogloboquadrina spp. abundance, and foraminiferal zones) [56,57] (Table S2). Red dashed lines terminated by gray circles indicate the stratigraphic position of AMS 14 C samples. Black dashed line terminated by downwards arrow with tip leftwards indicates the first occurrence (FO) of Gl. fimbriata.

Discussion
Results of this study show that the record of terrigenous input proxies is more fluctuating than the record of pollen and spore proxies, which indicate their higher sensitivity to the changes in rainfall. Pollen and spore proxies infer drier condition (lower rainfall) at Last Deglaciation and wetter condition (higher rainfall) at Holocene. Terrigenous input proxies infer three drier periods (i.e., before~17 and 15-13.5, and~7-3 ka), with a more significant rainfall decrease before~17 ka BP. Three wetter periods (i.e.,~17-15,~13.5-7, and after~3 ka BP) are also inferred from these proxies (Figure 7).

Palynological Proxies (Pollen and Spores)
The records of pollen and spore proxies show less fluctuation compared to the terrigenous input proxies. The abundances of grassland pollen and non-arboreal pollen (NAP) were higher before ~11 ka BP (Last Deglaciation) and declined afterward. Poaceae show a similar trend, but with the addition of a slight decrease in abundance ~16.5-15 ka BP. On the other hand, the abundances of lowland/peatland pollen, mangrove pollen, montane pollen, and arboreal pollen (AP) showed the opposite trend, lower during the Last Deglaciation and increased afterward. The abundance of Pteridophyta was also lower in the Last Deglaciation and increased since ~9 ka BP ( Figure 6).
Pollen proxies (i.e., grassland pollen, lowland/peatland pollen, and Poaceae) show similar trends to the oxygen isotope (δ 18 O) records from the Antarctic [81] and Greenland ice cores [82] that reflect changes in temperature (Figure 8). This indicated that pollen proxies were also sensitive to temperature changes, thus climate during the Last Deglaciation (Holocene) might be characterized by cooler (warmer), and drier (wetter) conditions [71,[86][87][88] (Figure 8). Effect of the Last Deglaciation-Holocene temperature changes in the Sumba region was also confirmed by the higher sea surface salinity (SSS) during the Last Deglaciation on the southwestern Savu Sea (near ITF pathway) [72] (Figure 8). The higher (lower) salinity was most likely linked to the weakening (strengthening) of AMOC during the cooler (warmer) Last Deglaciation (Holocene) [72,89]. The weakening of AMOC slowed down the global thermohaline circulation [90,91] and changed the nature (e.g., salinity) of its constituent flows, including ITF [53,92,93]. Pollen proxies did not detect the Last Deglaciation climate events, except Heinrich Stadial 1 (HS1), which were indicated by a decrease in Poaceae abundance (~16.5-15 ka BP) (Figure 8). On the other hand, the increase in Pteridophyta abundance~9 ka BP was most likely linked to sea-level rise-triggered rainfall increase. The increase in rainfall~9 ka BP was also indicated by δ 18 O record from the Liang Luar cave (Flores) [16] and was supported by the high sea level in this period [83][84][85] (Figure 8). The rise of sea level submerged the Indonesian continental shelf, triggering rainfall enhancement through increased moisture supply [6,7]. The lower sensitivity of pollen and spore might be associated with their more extensive source area, which most likely encompassed the Lesser Sunda Islands and Northwestern Australia. Their Lesser Sunda Islands origins were indicated by the occurrence of Podocarpus and Dacrycarpus, typical montane taxa of this region [37,94] (Figure 8). During the Last Deglaciation, the exposed Northwestern Australia shelf could be the additional source area for Poaceae pollen grains, which resulted in its high abundance as its grains easily transported by the winds [36, 38,66,95]. The occurrence of Phyllocladus, a typical montane taxon of Australia and New Zealand [66,96], also supports this suggestion ( Figure 8). As there were pieces of evidence for the spatial variation of AIM rainfall [5,13,18,28], the changes in AIM rainfall would affect the vegetation (through the changes in drought threshold) not at the same time in each location, so they weakened the pollen and spores-derived signals of monsoon [3,28]. This indicates that pollen and spore proxies are not ideal to reconstruct the past rainfall in this study.
Terrigenous input proxies infer the variability of riverine runoff related to the changes of rainfall during AISM (AISM rainfall) [1,2,8]. During the Last Deglaciation, the rainfall changes were most likely connected to the high latitude climate events, i.e., early stage of Deglaciation (ED) (beforẽ 17 ka BP), HS1 (~17-15 ka BP), ACR (~15-13.5 ka BP), and Younger Dryas (~13.5-11 ka BP) (Figure 7). Wetter periods coincided with the NH cold events (i.e., HS1 and YD) while drier periods indicated NH warm events (ED) and southern hemisphere (SH) cold events (ACR). [19]. Higher rainfall during HS1 and YD might be induced by the southward shift of the austral summer ITCZ while lower rainfall during ED and ACR were linked to the northward shift of the austral summer ITCZ [8,11] (Figure 9). The NH cooling (during HS1 and YD) enhanced the boreal winter cold surges, which pushed the ITCZ southward [8]. This mechanism also resulted in a southward shift of the boreal summer ITCZ, indicated by the lower EASM rainfall [97,98]. The lower rainfall during ACR could be explained by its co-occurrence with B-A [99] hence the northward shift of austral summer ITCZ was most likely linked to the NH warming (during ED and B-A) [8,19]. The Last Deglaciation rainfall records of ST08 were similar to δ 18 O records from Bali Gown Cave (Northwestern Australia) [14] (Figure 7), indicating the southern boundary of the austral summer ITCZ during wetter periods (HS1 and YD) [8] (Figure 9). On the other hand, ∆ 18 O (δ 18 OGs.ruber−δ 18 OG.bulloides) records (which indicated AIWM changes) from off south Java [1] and terrigenous input proxies (ln Ti/Ca) from off southwest Java [10] showed the opposite rainfall changes during the ACR (Figure 7). This indicated that the records from Java region responded to B-A instead, so they hinted a considerable influence of NH cross-equatorial moisture transport and the southern boundary of the austral summer ITCZ [8] (Figure 9).
The YD wetter period continued until Early Holocene (EH). The abrupt rainfall increase in EH, which was inferred in off southwest Java [10], not detected on ST08. This could be linked to the relatively constant terrigenous input in off southwest Sumba due to the considerable distance from the recently drowned-Sunda Shelf, as opposed to off southwest Java [10]. Δ 18 Ο records on off south Java, which changes closely followed boreal summer insolation, increased during YD-EH [1] ( Figure  7). This indicated the simultaneous increase of AISM and AIWM, but the effect of the strengthening AIWM and lower austral summer insolation (which should result in AISM weakening) was most likely distressed by the enhanced moisture supply related to abrupt sea-level rise [1,7,10] (Figure 7).
During Mid-Holocene, the rainfall records of ST08 were similar to the ln Ti/Ca records from off northwest Sumba, which inferred drier (wetter) Mid (Late) Holocene [2] (Figure 7). Lower rainfall during the Mid Holocene (MH) (~7-3 ka BP) was most likely linked to the decrease in solar activity (hence lower sunspot numbers) [2,20], which suppressed the effect of increasing austral summer insolation [2,21]. During the Late Holocene (LH) (after ~3 ka BP), an increase in austral summer insolation and solar activity resulted in the enhancement of rainfall [2,20,21]. The southern boundary of the austral summer ITCZ during MH was most likely located around its position during ED and ACR [8,100,101] and shifted southward during LH to around its position during HS1, YD, and EH [2,8,100] (Figure 9), but their relation to solar activity and orbitally-induced austral summer insolation is still not understood [5]. The carbon isotope composition of the C30 n-alkanoic fatty acids (δ 13 CFA) records from the southwestern Savu Sea [28] showed contradictive rainfall records ( Figure  7). This contradiction might be related to the differences in climate signals recorded on terrigenous input and δ 13 CFA proxies. δ 13 CFA reflects the dry season (AIWM) water stress connected to the amount  [72], (e). δ 18 O record of Antarctic (EDML) ice core [81], (f). δ 18 O record of Greenland (GISP2) ice core [82], (g). Reconstructed relative sea level from δ 18 O of Red sea benthic foraminifera [83][84][85], (i). δ 18 O record of stalagmites from Liang Luar cave (Flores) [16]. Data are plotted against the mean ages obtained from the age model. Black curves indicate smoothed values (exponential smoothing, df: damping factor) ( Table S5).
The YD wetter period continued until Early Holocene (EH). The abrupt rainfall increase in EH, which was inferred in off southwest Java [10], not detected on ST08. This could be linked to the relatively constant terrigenous input in off southwest Sumba due to the considerable distance from the recently drowned-Sunda Shelf, as opposed to off southwest Java [10]. ∆ 18 O records on off south Java, which changes closely followed boreal summer insolation, increased during YD-EH [1] (Figure 7). This indicated the simultaneous increase of AISM and AIWM, but the effect of the strengthening AIWM and lower austral summer insolation (which should result in AISM weakening) was most likely distressed by the enhanced moisture supply related to abrupt sea-level rise [1,7,10] (Figure 7).
During Mid-Holocene, the rainfall records of ST08 were similar to the ln Ti/Ca records from off northwest Sumba, which inferred drier (wetter) Mid (Late) Holocene [2] (Figure 7). Lower rainfall during the Mid Holocene (MH) (~7-3 ka BP) was most likely linked to the decrease in solar activity (hence lower sunspot numbers) [2,20], which suppressed the effect of increasing austral summer insolation [2,21]. During the Late Holocene (LH) (after~3 ka BP), an increase in austral summer insolation and solar activity resulted in the enhancement of rainfall [2,20,21]. The southern boundary of the austral summer ITCZ during MH was most likely located around its position during ED and ACR [8,100,101] and shifted southward during LH to around its position during HS1, YD, and EH [2,8,100] (Figure 9), but their relation to solar activity and orbitally-induced austral summer insolation is still not understood [5]. The carbon isotope composition of the C 30 n-alkanoic fatty acids (δ 13 C FA ) records from the southwestern Savu Sea [28] showed contradictive rainfall records (Figure 7). This contradiction might be related to the differences in climate signals recorded on terrigenous input and δ 13 C FA proxies. δ 13 C FA reflects the dry season (AIWM) water stress connected to the amount of rainfall during AIWM (AIWM rainfall) [28]. We suggest a joint analysis of terrigenous input and δ 13 C FA proxies from the same site in future studies to reconstruct the past changes of both AISM and AIWM rainfall, so more robust AIM rainfall records are produced.

Conclusions
Proxies analyzed in this study (i.e., terrigenous input and pollen and spores) show different sensitivity to AIM rainfall changes since the Last Deglaciation (~18 ka BP). Pollen and spores indicate drier climate during the Last Deglaciation and wetter climate during the Holocene. Terrigenous input proxies are more sensitive as they can detect three drier periods (i.e., before ~17, ~15-13.5, and 7-3 ka BP) and three wetter periods (i.e., ~17-15, ~13.5-7, and after ~3 ka BP) which represent the AISM rainfall changes. The lower sensitivity of pollen and spore proxies was related to their wider source area (the Lesser Sunda Islands and Northwestern Australia). This might be the reason of the absence of high latitude climatic event signals (i.e., ED, HS1, ACR, and YD) in nearly all pollen and spore proxies (except Poaceae) despite their high responsiveness to temperature changes. During the Last Deglaciation, AISM rainfall changes indicated a strong response to high latitude climatic events. The higher rainfall periods were linked to the southward shift of the austral summer ITCZ during the NH cold events (i.e., HS1 and YD) while the northward shift of the austral summer ITCZ resulted in lower AISM rainfall during the NH warm event (ED) and SH cold event (ACR, coeval with B-A). AISM rainfall changes during the Holocene were related to the sea-level rise (EH), solar activity (MH and LH), and orbitally-induced austral summer insolation (LH). Differences in recorded climate signals caused the contradiction between our AISM rainfall records and the rainfall records from the southwestern Savu Sea (which inferred AIWM rainfall). A joint analysis of terrigenous input and δ 13 CFA proxies from the same site is suggested in future studies to develop more robust AIM rainfall records.

Conclusions
Proxies analyzed in this study (i.e., terrigenous input and pollen and spores) show different sensitivity to AIM rainfall changes since the Last Deglaciation (~18 ka BP). Pollen and spores indicate drier climate during the Last Deglaciation and wetter climate during the Holocene. Terrigenous input proxies are more sensitive as they can detect three drier periods (i.e., before~17,~15-13.5, and 7-3 ka BP) and three wetter periods (i.e.,~17-15,~13.5-7, and after~3 ka BP) which represent the AISM rainfall changes. The lower sensitivity of pollen and spore proxies was related to their wider source area (the Lesser Sunda Islands and Northwestern Australia). This might be the reason of the absence of high latitude climatic event signals (i.e., ED, HS1, ACR, and YD) in nearly all pollen and spore proxies (except Poaceae) despite their high responsiveness to temperature changes. During the Last Deglaciation, AISM rainfall changes indicated a strong response to high latitude climatic events. The higher rainfall periods were linked to the southward shift of the austral summer ITCZ during the NH cold events (i.e., HS1 and YD) while the northward shift of the austral summer ITCZ resulted in lower AISM rainfall during the NH warm event (ED) and SH cold event (ACR, coeval with B-A). AISM rainfall changes during the Holocene were related to the sea-level rise (EH), solar activity (MH and LH), and orbitally-induced austral summer insolation (LH). Differences in recorded climate signals caused the contradiction between our AISM rainfall records and the rainfall records from the southwestern Savu Sea (which inferred AIWM rainfall). A joint analysis of terrigenous input and δ 13 C FA proxies from the same site is suggested in future studies to develop more robust AIM rainfall records.