Holocene hydroclimate reconstruction based on pollen, XRF, and grain-size analysis and its implications for past societies of the Korean Peninsula

The dynamics of the East Asian Summer Monsoon (EASM) and their link to past societies during the Holocene are topics of growing interest. In this study, we present analyses of a ca. 6,000-year pollen record, as well as X-ray fluorescence (XRF) and grain-size data from the STP18-03 core sampled from Miryang in the Korean Peninsula, which 15 spans ca. 8.3–2.3 ka BP. In-phase relationships of these proxies revealed an imprint of the Holocene Climate Optimum (HCO) during the early to mid-Holocene and subsequent drying toward the late Holocene in accordance with decreasing solar insolation. At centennial timescales, our study indicates wet conditions during ca. 8.3–7.5, 7.1–6.4, 6.0–4.8, and 3.6– 2.8 ka BP, and a drier climate during ca. 7.5–7.1, 6.4–6.0, and 4.8–3.6 ka BP. Notably, our finding for ca. 6.4–6.0 ka BP contributes further evidence of a drying event in the Korean Peninsula during this period. We suggest that the Pacific Ocean 20 played a role in the underlying mechanism of hydroclimate change in the region. A strong Kuroshio Current (KC) and longterm El Niño–Southern Oscillation (ENSO)-like variability in the Western Tropical Pacific (WTP) were closely linked to the influence of the EASM over the Korean Peninsula. In particular, dry phases during ca. 4.8–3.6 and 2.8–2.3 ka BP, which were synchronous with a more active ENSO, closely corresponded to lower population levels according to a summed probability distribution (SPD) of archaeological records assembled in the Korean Peninsula. This finding implies that past 25 human societies of Korea were highly vulnerable to climate deterioration caused by precipitation deficits.

Holocene. For example, although many paleoclimate studies have focused on mainland China as a key region representing East Asia, site-specific and varying imprints of the EASM within the region are being increasingly addressed between the interior and coastal parts of China (Chen et al., 2015b;Zhou et al., 2016;Zhu et al., 2017) as well as between China and Korea . In the Korean Peninsula, the EASM has been suggested to be primarily driven by the Pacific 35 Ocean, in processes such as the Kuroshio Current (KC) flowing from the Western Tropical Pacific (WTP) and the El Niño-Southern Oscillation (ENSO) in the equatorial Pacific (Lim and Fujiki, 2011;Park et al., 2019;Park et al., 2016;Lee et al., 2020). Therefore, more high-resolution studies are needed to examine EASM dynamics closely in relation to oceanic forcing.
However, despite increasing research efforts, high-resolution studies of the EASM in the Korean Peninsula remain scarce.
Although pollen analysis has been the most common approach to reconstruct past precipitation changes (Lim and Fujiki, 40 2011;Park et al., 2019;Lee et al., 2020;Park et al., 2016), this methodology contains physical limitations in enhancing the temporal resolution between samples. In contrast, speleothem-based Holocene EASM reconstructions, which permit finer resolution, have been confined to the last few thousand years in the Korean Peninsula and have not defined a clear link with the Pacific Ocean (Jo et al., 2017;Hong et al., 2012;Yu et al., 2016). Additionally, oxygen isotope values in speleothems sometimes do not purely reflect rainfall signals Maher and Thompson, 2012;Caley et al., 2014;Lachniet, 45 2009). Therefore, there is a significant need to assemble multiple types of proxies and maintain high temporal resolution for more accurate analysis of the link between the EASM and oceanic forcing.
Another topic of rising interest is the relationship between climate change and human societies. This issue has been addressed by many studies worldwide over the past few decades, in regions including Europe (Tallavaara et al., 2015;50 Büntgen et al., 2011), Greenland (D'Andrea et al., 2011, America (Munoz et al., 2010), Mesopotamia (Weiss et al., 1993;Carolin et al., 2019), India (Dixit et al., 2018), Egypt (Manning et al., 2017), and China Xie et al., 2013;Wang et al., 2014). Most of these studies have supported a dominant climatic role in the vicissitudes of past societies, whereas others have recently reported human resilience to environmental stresses (Flohr et al., 2016;Blockley et al., 2018).
Thus, the relationship between climate and human societies is not always simplistically determined and may be case-specific. 55 Even within sub-continents of China-e.g., Northwest, North, and South China, the Qinghai Tibetan Plateau-and Japan, human activity patterns during the Holocene appear to be distinct Crema et al., 2016). It remains unclear whether this discrepancy reflects local differences in climate imprints or differences in societal responses to climate. Therefore, to analyze socio-environmental relationships in a particular area accurately, a direct comparison of continuous, local, and high-resolution paleoclimate proxy data with human activity indicators is crucial. However, in the Korean 60 Peninsula, this work remains in the nascent stage Park et al., 2019). The topic of climate-society relationships in Korea often depends on individual historical records of natural disasters, leading to assumptive storytelling about their potential impact on contemporary societal unrest (Cho, 2009). For older periods with scarce disaster records, qualitative and broad climate interpretations tend to be expressed as "warm/cold" or "wet/dry", often in comparison with https://doi.org/10.5194/cp-2020-98 Preprint. Discussion started: 7 August 2020 c Author(s) 2020. CC BY 4.0 License. temporal and spatial patterns of archaeological findings (Ahn and Hwang, 2015), due to a lack of reliable high-resolution 65 climate data.
Therefore, in the present study, we performed a multi-proxy analysis of Holocene hydroclimate change in the Miryang region of the Korean Peninsula from ca. 8.3 to 2.3 ka BP. We combined pollen, X-ray fluorescence (XRF), and grain-size data to reconstruct the EASM history over the Korean Peninsula. The high temporal resolution of our XRF analysis allowed 70 us to analyze the paleoclimate dataset at decadal to annual scales. We examined the role of the Pacific Ocean as an underlying mechanism for EASM history in the Korean Peninsula, particularly in terms of ENSO variance and heat and moisture supply along the KC. We also identified a synchronous change between hydroclimate and past population in the Korean Peninsula and explored the processes through which past human societies may have responded to climate impacts during the Holocene. 75

Regional Setting
Our coring site (35°26'18.84" N, 128°46'41.26" E; 4.64 m a.s.l.) is located in a floodplain of the Miryang River in the southeastern part of the Korean Peninsula ( Fig. 1a-c). The Miryang River is a tributary of the Nakdong River, which eventually flows out to the Korea Strait, between the Korean Peninsula and the Japanese Archipelago. The backdrop of the coring site is a group of small mountains several hundred meters in elevation, including Jongnamsan (663 m) and 80 Palbongsan (391 m) (Fig. 1c). The Korean Peninsula is situated amid the East Asian monsoonal system, and experiences seasonal differences in air pressure due to its location between the Eurasia continent and the Pacific Ocean (An, 2000) (Fig.   1a). In summer, hot and humid southeasterly wind flows into the Korean Peninsula under the influence of the warm KC, whereas in winter, cold and dry northwesterly wind from the Siberian High dominates. According to a 30-year (1981-2010 AD) climate record from a meteorological observation station 6 km north of the coring site, the mean annual temperature is 85 13.3 ℃, with a January minimum of 0.0 ℃ and August maximum of 25.8 ℃. The mean annual precipitation is 1229.4 mm; precipitation is lowest in December (16.4 mm) and highest in July (269.5 mm) (Fig. 1d). A total precipitation of 829.5 mm during June-September, JJAS) accounts for 67.5 % of the total annual precipitation, indicating significant influence of the EASM on the hydroclimate of the region (Korea Meteorological Administration, 2020).

Core materials and dating
In April 2018, the 20-m STP18-03 core was collected in 1-m sections from a floodplain of the Miryang River (Fig. 1). Depth 100 zones of the uppermost 0-1.25 m and lowermost 14-20 m were excluded from all analyses as the former were regarded to have been affected by human activities and the latter consisted mainly of gravel. We sent a total of 16 samples to the Korea Institute of Geoscience and Mineral Resources (KIGAM), Republic of Korea (Table 1) for age dating; eight were measured using the optically stimulated luminescence (OSL) dating method. These samples were treated with Na4P2O7, HCl, H2O2, and H2SiF6 to extract quartz with a diameter of 4-11 μm. Then, OSL signals were measured using a TL-DA-20 reader (Risø 105 DTU, Roskilde, Denmark) equipped with a blue light-emitting diode (LED; 470 ± 20 nm) stimulation source. Plant fragments from the other eight samples were used for radiocarbon dating by accelerator mass spectrometry. Based on the results, we constructed an age model using the bacon ver. 2.3 R package (Blaauw and Christen, 2011) with the IntCal13 calibration dataset (Reimer et al., 2013) (Fig. 2b). Two radiocarbon dates from depths of 795 cm and 1032 cm were omitted due to their anomalous ages in relation to those of other samples. 110

Proxy Analyses
We performed palynological analysis of a total of 137 samples at intervals ranging from 1 to 70 cm (1-6 cm for sections from 395 to 1010 cm and 9-70 cm for the remaining sections). The average temporal resolution was 41.4 years, with minimum and maximum values of 12 and 348 years, respectively. For sample preparation, we followed the standard protocol of Faegri et al. (1989) including HCl (10 %), KOH (10 %), HF (40 %), and acetolysis. KOH treatment was repeated twice to 115 remove organic matter completely. For highly humic samples in the range of 501-528 cm, the KOH procedure was repeated up to three times. For each sample, one lycopodium tablet containing 177,745 spores was added, and at least 300 pollen grains and spores were counted on each slide using a Leica microscope at 400 × magnification. For palynomorph identification, we referred to a pollen atlas from Lake Suigetsu, Japan (Demske et al., 2013). The percentage of pollen was calculated for each taxon relative to the total sum of non-aquatic pollen grains and spores in the sample. The result was 120 visualized using the Tilia software ver. 2.0.41 (Grimm, 1991), and stratigraphically constrained cluster analysis was conducted using CONISS (Grimm, 1987) based on non-aquatic taxa.
Grain-size analysis was performed using a Mastersizer 2000 laser diffraction particle-size analyzer (Malvern Instruments, UK) at KIGAM. Approximately 300 mg of each sample was collected at 10-cm intervals from the 205-1390 cm section. 125 These subsamples were treated with H2O2 (35 %) and HCl (1 N) to remove organic matter and carbonates. Grain sizes of < 4 https://doi.org/10.5194/cp-2020-98 Preprint. Discussion started: 7 August 2020 c Author(s) 2020. CC BY 4.0 License. μm, 4-63 μm, and > 63 μm were classified as clay, silt, and sand, respectively. XRF analysis was also conducted at KIGAM using an XRF core scanner (Avaatech B.V.; Alkmaar, Netherlands), which extracts elemental concentration data in a nearly continuous manner (Croudace et al., 2006;Löwemark et al., 2011). XRF signals were measured from split core surfaces from depths of 12.5-1293.5 cm, with settings of 10/50 kV and 0.25/1.0 mA, and a sampling time of 30 s. In total, 2,088 130 values were collected at a resolution of 0.5 cm. However, we did not perform paleoenvironmental interpretations on data at depths above 365 cm, because these sections correspond to periods later than ca. 2.3 ka BP, for which there is evidence of agriculture in the Miryang region (Yoon et al., 2005) Cross-spectral analysis on the proxy data was conducted using the REDFIT-X software ver.

Chronology
The chronology of the STP18-03 core contained a record of ca. 6420 years, from the mid-to late Holocene, representing the period from ca. 8340 cal yr BP (1280 cm depth) to ca. 1920 cal yr BP (350 cm depth) (Fig. 2b). The sedimentation rate was 0.38 cm per year for the 790-1280-cm segment and 0.09 cm per year for the 380-790-cm segment (average, 0.14 cm per year). Sedimentation was continuous throughout the core, except for the 901-905 and 1100-1140-cm segments, perhaps due 145 to on-site technical problems while coring or disturbance by underground water.

Zone 1 (790-1390 cm)
This zone consists mainly of dark brown clay, although sand is frequently observed in multiple layers (Fig. 2a). Sand percentages fluctuate significantly between 0.5 % and 79.1 % (Fig. 2c). The amount of titanium (Ti) also varies, with a large 150 amplitude generally in the opposite direction to that of sand content (Fig. 2d). The tree pollen percentage is generally low at 1145-1280 cm, with a minimum value of 77.2 % (Fig. 2e). Artemisia (mugwort), Poaceae (wild grass), and fern species comprise 20.6 % of the pollen composition (Fig. 2f); this trend is reversed at 790-1090 cm, where the proportion of arboreal pollen remains high and stable at 87.1-96.4 %, and that of Artemisia and Poaceae pollen together with fern spores remains low, at an average of 8.7 %. The largest proportion of arboreal pollen is constituted by Quercus (oak) (Fig. 3); other 155 broadleaf tree genera include Alnus (alder), Fraxinus (ash), and Ulmus (elm), which reach the highest proportion in this zone https://doi.org/10.5194/cp-2020-98 Preprint. Discussion started: 7 August 2020 c Author(s) 2020. CC BY 4.0 License. of the core. Notably, at a depth of 950 cm, a sudden increase in Alnus pollen to 58.0 % (Fig. 3) coincides with abrupt shifts in sand and Ti content ( Fig. 2c and d). This abnormal value implies a sudden local disturbance event that perturbed the preexisting vegetation and gave rise to pioneer species with high environmental tolerance (McVean, 1953;Weng et al., 2004).

Zone 2 (400-790 cm) 160
From 790 to 400 cm, the clay content gradually increases as depth decreases, and the colour changes from light brown to grey (Fig. 2a). Sand percentages and Ti content stabilize, changing in tandem with the pollen data ( Fig. 2c-e). Overall, lower sand content is concurrent with a lower proportion of tree pollen and higher Ti values, and vice versa. Between 590 and 635 cm, where the sand percentage temporarily decreases to 28.8%, the Ti content increases sharply as the proportion of tree pollen drops to as low as 83.1 % due to a decline in the main arboreal genera such as Quercus, Pinus, Alnus, Fraxinus, and 165 Ulmus (Fig. 3). A similar trend is also observed between 420 and 500 cm, where the proportions of sand and arboreal pollen drop to below 10 % and 80 %, respectively. Upland herbs such as Artemisia and Poaceae occupy this relatively open area, whereas Pteridium ferns rise to a maximum value in this zone. This trend is in contrast with the 500-590 and 635-790 cm sections, in which the proportions of sand and tree pollen remain high as Ti XRF values are low. Notably, a high proportion of tree pollen at 500-590 cm is led by a climax in Quercus pollen (Fig. 3). 170

Zone 3 (210-400 cm)
This part of the core is characterized by brown clayey silt ( Fig. 2a) with low sand content (Fig. 2c) and high Ti values (Fig.   2d). Pollen deposition is interrupted after an explosive increase in upland herbs (mainly Artemisia and Poaceae), Cyperaceae (sedges), and Polypodiales undiff. ferns at the beginning of this zone (Figs. 2e, f and 3). Considering the near absence of sand content in this zone (Fig. 2c), this disruption may have been caused by a cessation of water supply from the river to the 175 floodplain due to either climate drying or river route alteration, and the subsequent exposure of the site to air.

Role of our proxy data as paleoclimate indicators
At millennial timescales, our Ti, pollen, and sand content data are consistent with a declining trend of summer solar insolation in the Northern Hemisphere from the mid-to late Holocene (Berger and Loutre, 1991) (Fig. 4a-e). The gradual 180 decrease in the proportion of arboreal pollen (Fig. 4c) reflects cooling and drying associated with a southward migration of the Intertropical Convergence Zone (ITCZ) induced by orbital forcing (Berger and Loutre, 1991;Haug et al., 2001). Our sand proportion data also follow this trend, as fluvial sand discharge by the Miryang River would have weakened due to less precipitation in the late Holocene relative to earlier periods (Fig. 4e). Ti XRF values, in the opposite direction, change in tandem with these two proxies, such that the signal generally increases towards the late Holocene (Fig. 4b). In many studies, 185 Ti has been used as an indicator of terrestrial erosion, although its paleoenvironmental interpretation may vary according to https://doi.org/10.5194/cp-2020-98 Preprint. Discussion started: 7 August 2020 c Author(s) 2020. CC BY 4.0 License. regional context (Sun et al., 2008;Bakke et al., 2009). In the present study, considering its synchronicity with arboreal pollen proportion ( Fig. 4b and c), we interpret Ti as reflecting hydroclimate change in the Korean Peninsula. During wet periods, more tree growth (mainly oak and pine trees, Fig. 3) in nearby hills would have suppressed soil erosion via the anchoring effect of roots, leading to lower Ti XRF values. However, as climate became drier towards the late Holocene, tree 190 replacement by herbs and ferns would have weakened this effect, resulting in greater Ti erosion.
This attribute of Ti data as a climate proxy in relation to vegetation change is further supported by cross-spectral analysis (Ó lafsdóttir et al., 2016) (Fig. 5). The analysis of Ti and arboreal pollen data implies high coherency at frequencies of 518, 148, 127, and 104 years (Fig. 5a). A ~500-year frequency is widely detected in East Asian monsoonal regions and has been 195 attributed to oceanic influences such as thermohaline circulation (THC) or the ENSO, possibly modulated by solar activity (Roth and Reijmer, 2005;Dima and Lohmann, 2009;Xu et al., 2014;Xu et al., 2019;Stebich et al., 2015;Park et al., 2019). This link is corroborated by additional analysis with the WTP SST record from the MD98-2176 core  and hematite-stained grains (HSG) in the North Atlantic (Bond et al., 2001), which commonly show the ~500-year frequency observed in our Ti values and in the tree pollen data (Fig. 5b-c and e-f). Smaller frequencies of ~150, ~130, and ~100 years 200 are likely attributable to a solar origin (Scuderi, 1993;Roth and Reijmer, 2005); these signals have been interpreted in terms of solar modulation on the EASM strength at Qinghai Lake, central China (Ji et al., 2005) and Jeju Island , south of the Korean Peninsula ( Fig. 1a and b). Our additional analysis of total solar irradiation (TSI) data ) contributes further evidence of a solar contribution to these cycles ( Fig. 5d and g), although ~130-and ~100-year frequencies fail to reach statistically significant levels between tree proportion and TSI (Fig. 5g), possibly due to the lower 205 temporal resolution of pollen analysis relative to XRF scanning.

Climate change in the Korean Peninsula during the Holocene
From ca. 8.3 to 5.4 ka BP, highly sustained proportions of arboreal pollen (Fig. 4c) likely reflect the influence of the Holocene Climate Optimum (HCO) Wanner et al., 2008;Zhou et al., 2016), which resulted in warmer summers in the Korean Peninsula, possibly increasing annual average temperatures by 1-2 ℃ compared with the pre-210 industrial period (Renssen et al., 2012). The climate was generally warm and humid, influenced by the northward advance of the EASM (Yang et al., 2015). However, one exception is an abrupt drop in the proportion of tree pollen at ca. 8.2 ka BP, when pollen from herb taxa including mugwort (Artemisia) and wild grass (Poaceae) suddenly increased (Fig. 3), reflecting the "8.2 ka event", an abrupt global cooling phenomenon (Alley et al., 1997;Veski et al., 2004;Cheng et al., 2009), which corroborates our previous reports of this event in the Korean Peninsula (Park et al., 2018;Park et al., 2019). 215 At centennial timescales, several periods of wet and dry climate alternate throughout the mid-to late Holocene. Our Ti, pollen, and sand data indicate relatively wet climates for ca. 8.3-7.5, 7.1-6.4, 6.0-4.8, and 3.6-2.8 ka BP, and drier conditions for ca. 7.5-7.1, 6.4-6.0, and 4.8-3.6 ka BP (Fig. 4b-e). A pronounced feature of this multi-centennial-scale https://doi.org/10.5194/cp-2020-98 Preprint. Discussion started: 7 August 2020 c Author(s) 2020. CC BY 4.0 License. environmental change is a close connection with SST records from the Pacific Ocean Stott et al., 2004) 220 ( Fig. 4f and g). Periods of warm and wet climate are accompanied by high SST records obtained for the Okinawa Trough ( Fig. 4f), which is located directly on the path of the KC (Fig. 1a), which is an important contributor to EASM strength in coastal East Asia, where it is a major transport mechanism of heat and moisture from the WTP (Jian et al., 2000;Zhou et al., 2009;Lim and Fujiki, 2011;Hu et al., 2015;Park et al., 2016;Constantine et al., 2019;Lee et al., 2020). During ca. 8.3-7.5, 7.1-6.4, 6.0-4.8, and 3.6-2.8 ka BP, greater heat and water vapor supply along the KC would have enabled active 225 atmospheric convection and stronger EASM influence over the Korean Peninsula, whereas the opposite would have occurred during ca. 7.5-7.1, 6.4-6.0, and 4.8-3.6 ka BP.
Among these periods, a sign of drying and/or cooling around 6.4-6.0 ka BP (Fig. 4b-e) is consistent with previous findings at Lake Pomaeho in the central Korean Peninsula   (Fig. 1b). This drying signal may have been 230 muted in our previous study of Gwangyang   (Fig. 1b, GY-1) due to lower temporal resolution (~80 years) around the sedimentary section. However, evidence of climate deterioration during 6.4-6.0 ka BP is not consistent across different regions of East Asia. For example, Daihai Lake  and Gonghai Lake

in North
China and Dongge Cave in South China (Wang et al., 2005) (Fig. 1a) record abrupt shifts toward less precipitation. However, in Lake Xiaolongwan Xu et al., 2019) and Lake Sihailongwan  in Northeast China 235 (Fig. 1a), this signal is not clear. Although this discrepancy may reflect regionally different climate patterns, we cannot rule out the possibility of inherent bias in the proxy-based reconstructions. For example, although the DA stalagmite in Dongge Cave (Wang et al., 2005) (Fig. 1) recorded significant drying around 6.4-6.0 ka BP, the D4 stalagmite (Dykoski et al., 2005), which was obtained from the same cave, does not exhibit clear changes. The reliability of EASM precipitation signals among Chinese stalagmites has been questioned due to mixed effects of the Indian Summer Monsoon (Fig. 1a, ISM) and/or 240 hydrologic processes affecting oxygen isotope values (Maher and Thompson, 2012;Chen et al., 2015a;Caley et al., 2014).
Likewise, in pollen records, source area and/or overestimation issues inherent in palynological methodology (Seppä and Bennett, 2003) can lead to potential bias in blurring climate signals. In this study, we suspect that a methodological issue explains the smaller amplitude of the changes in pollen proportions during ca. 7.5-7.1 ka BP, whereas the other sedimentary proxies, XRF and sand percentage data exhibit clearer phase shifts with the Pacific Ocean ( Fig. 4b-g). Similarly, in pollen 245 records from Daihai Lake  and Gonghai Lake , drying signals during ca. 7.5-7.1 ka BP are less evident than those during ca. 6.4-6.0 ka BP. Considering this complexity among different sites and time periods, imprints of EASM weakening during ca. 7.5-7.1 and 6.4-6.0 ka BP are not yet clearly explicable at the regional scale in East Asia and therefore require further research with abundant high-resolution data across different study sites.

250
Our proxy data also indicate a pronounced drying trend in the Korean Peninsula during ca. 4.8-3.6 ka BP (Fig. 4b-e), which is consistent with global findings of significantly decreased precipitation and/or temperature around this period (Bond et al., 1997;Wang et al., 2005;Wanner et al., 2011;Bond et al., 2001). In East Asia, this climate impact may have been amplified https://doi.org/10.5194/cp-2020-98 Preprint. Discussion started: 7 August 2020 c Author(s) 2020. CC BY 4.0 License. by long-term ENSO-like variance, which strengthened from the mid-Holocene Conroy et al., 2008;Donders et al., 2008). The ENSO possibly exerts a dampening effect on EASM intensity by affecting low-to mid-latitude 255 atmospheric patterns Feng and Hu, 2014;Hu et al., 2015), evidence of which has been found in coastal East Asian regions, particularly during El Niño-like phases (Chen et al., 2015b;Park et al., 2017;Park et al., 2016). Active ENSO-like forcing would frequently have pushed warm seawater pools in the western Pacific farther to the east (Timmermann et al., 2018), consequently lowering WTP SST values (Fig. 4g) and weakening the KC (Hu et al., 2015) (Fig.   4f). 260

Response of past societies to mid-to late Holocene climate change
From ca. 6 ka BP to 2.3 ka BP, the Ti value of the STP18-03 core closely follows the summed probability distribution (SPD) plots of radiocarbon dates collected from archaeological samples in the Korean Peninsula (Oh et al., 2017) (Fig. 6a). The SPD method is increasingly used as a proxy for past population levels Ahn and Hwang, 2015;Tallavaara et al., 2015;Crema et al., 2016;Bevan et al., 2017;Oh et al., 2017;Xu et al., 2019) by assembling radiocarbon age 265 calculations from archaeological findings (Gamble et al., 2005;Surovell et al., 2009). In our data, Ti content gradually decreased and remained low until ca. 4.8 ka BP, recovered high values by ca 3.6 ka BP, and then diminished again before increasing at ca. 2.8 ka BP (Fig. 6b). This trend is coherent with, but opposite to, changes in the SPD data, which indicate larger populations during ca. 6.0-4.8 ka BP and ca. 3.6-2.8 ka BP, and lower populations during ca. 4.8-3.6 ka BP and after ca. 2.8 ka BP (Fig. 6a). Three abrupt transition points around 4.8, 4.2, and 4.0 ka BP are found in both datasets, validating 270 their link ( Fig. 6a and b). Together with our previous research Park et al., 2019), this robust synchronicity between Ti data and archaeological records contributes to accumulating evidence that past societies of the Korean Peninsula responded strongly and with great sensitivity to climate change.
Given its synchronicity with arboreal pollen data, we used the Ti XRF signal as a proxy of climate change and identified two 275 periods with a relatively warm and wet climate, when past populations increased: ca. 6-4.8 ka BP and ca. 3.6-2.8 ka BP (Fig.   6). This link is accompanied by high SSTs in the Okinawa Trough (Fig. 4f). Notably, the former period corresponds to the latter part of the HCO, characterized by a warm and wet climate during the mid-Holocene both globally (Wanner et al., 2008;Renssen et al., 2012;Renssen et al., 2009) and in the Korean Peninsula . During this period, favorable climate conditions would have enabled successful hunting and gathering, with sufficiently abundant food resources to sustain 280 population growth . Besides hunting and gathering, evidence of foxtail, broomcorn, and legume cultivation has been found at Korean Middle-Late Chulmun (Neolithic) sites for as early as ca. 5.5 ka BP (Lee, 2011).
Although the degree to which farming was an important food source for the Chulmun people remains unclear, it is likely that a warm and humid climate provided better conditions for successful agriculture Xu et al., 2019).
Moreover, the latter period of ca. 3.6-2.8 ka BP was a time of explosive increase in Bronze Age settlements in the Korean 285 Peninsula, beginning as early as ca. 3.9 ka BP (Kim and Bae, 2010). Similarly, Lee (2011) suggested a major transition from https://doi.org/10.5194/cp-2020-98 Preprint. Discussion started: 7 August 2020 c Author(s) 2020. CC BY 4.0 License.
Chulmun to Mulmun (Bronze) culture in the Korean Peninsula at ca. 3.4 ka BP, with clear evidence of intensive agriculture, including domesticated plants such as rice. The impact of temporary climate deterioration around 3.2 ka BP (Fig. 6b-d), which may reflect enhanced ENSO activity   (Fig. 6e) and/or the 3.2-ka event (Kaniewski et al., 2017), was not large enough in the Korean Peninsula to interrupt the increasing population trend during this cultural boom (Fig. 6a). 290 During ca. 4.8-3.6 ka BP, high Ti values and a low proportion of arboreal pollen indicate a cool and dry climate in the Korean Peninsula when human activity declined (Fig. 6). Evidence of significant cooling and drying events around this period (Bond et al., 1997;Wang et al., 2005;Wanner et al., 2011;Bond et al., 2001) and their potential impacts on the shrinkage or unrest of past societies (DeMenocal, 2001) have been widely reported from various sites worldwide including 295 Mesopotamia (Weiss et al., 1993;Carolin et al., 2019), India (Dixit et al., 2018;Dixit et al., 2014), China Li et al., 2018;Xiao et al., 2018;Xu et al., 2019), Japan (Kajita et al., 2020;Kawahata et al., 2009), and Korea (Constantine et al., 2019;Park et al., 2019). In these periods, climate drying would have increased dietary stress by hindering successful hunting, gathering, millet cultivation and even livestock domestication (Roffet-Salque et al., 2018;Kawahata et al., 2009).
Abrupt and synchronous changes in Ti and pollen data together with the decline in archaeological SPD values at ca. 4.8, 4.2, 300 and 4.0 ka BP imply a significant impact of sudden climate deterioration on Korean prehistoric societies (Fig. 6). These changes were likely triggered by the onset of an active ENSO at ca. 4.8 ka BP (Fig. 6e) and modulated by lower SSTs in the Okinawa Trough and WTP until ca. 3.6 ka BP (Fig. 6f).
Our Ti XRF data suggest synchronicity of climate deterioration with a decline in population in the Korean Peninsula from ca. 305 2.8 ka BP until ca. 2.3 ka BP (Fig. 6a-d). During this period, SSTs in the Okinawa Trough and WTP decreased ( Fig. 4f and g), and ENSO activity around 2.7 ka BP (Fig. 6e) may have amplified the climate impact. Although no palynological record is available for this period ( Fig. 3 and 6c), our previous study of Gwangyang   (Fig. 1b, GY-1) supports this socio-environmental link (Fig. 6d). In the Korean Peninsula, most Bronze Age pottery disappears during this period, as observed for Misa-ri, Garak-dong, and Heunam-ri-type pottery around 2.9-2.8 ka BP and Yeoksam-dong, Songguk-ri, and 310 Geomdan-ri-type pottery during 2.8-2.4 ka BP (Lee, 2017;Ahn and Hwang, 2015). In this sense, it is likely that decreasing SPD values after ca. 2.8 ka BP (Fig. 6a) primarily reflect a collapse of Bronze Age culture in Korea, and that this change was influenced by climate drying and/or cooling at that time ( Fig. 6b-d). As during ca. 4.8-3.6 ka BP, the population decline after ca. 2.8 ka BP was a widespread phenomenon that has been detected in Korea and at many sites worldwide including mainland China , Turkey (Woodbridge et al., 2019), and Britain and Ireland (Bevan et al., 2017). 315 Therefore, climate impact on human societies during this period should be understood within a global context, possibly in connection with a cooling trend after the Bond event 2 (Bond et al., 1997;Wanner et al., 2011).

Conclusion
Our multi-proxy analysis of pollen, XRF, and grain-size data was used to reconstruct the EASM history of the Korean Peninsula from ca. 8.3 to 2.3 ka BP. We identified a concurrent change between hydroclimate and past human activity. The 320 Holocene climate in Korea was sensitively modulated by the strength of the KC and ENSO-like variance. Wet conditions prevailed during ca. 8.3-7.5, 7.1-6.4, 6.0-4.8, and 3.6-2.8 ka BP, when SSTs in the western Pacific Ocean were sufficiently high to enhance EASM strength. However, during ca. 7.5-7.1, 6.4-6.0, and 4.8-3.6 ka BP, climate became drier due to KC weakening and an increase in ENSO-like variance that likely dampened the EASM. Although regional imprints of climate deterioration during ca. 6.4-6.0 ka BP remain unclear in East Asia, the findings of our study contribute to our knowledge of 325 this climate event in the Korean Peninsula. The reconstructed hydroclimate change was also synchronous with past population levels inferred by the SPD of archaeological remains. Past societies flourished amid favorable climate conditions during ca. 6-4.8 ka BP and ca. 3.6-2.8 ka BP, but suffered from precipitation deficits during ca. 4.8-3.6 ka BP. This finding is consistent with multiple findings of contemporary collapses of past civilizations worldwide and confirms that this global socio-environmental link was present in the Korean Peninsula. Nevertheless, it should be noted that the relationship between 330 climate and past societies is not always straightforward. Further research is required to elaborate our understanding of its dynamics.

Data availability
Data used in this study are available from Jinheum Park (jinheum94@snu.ac.kr)

Competing interests
The authors declare that they have no conflict of interest.