A new Dead Sea pollen record reveals the last glacial paleoenvironment of the southern Levant

The southern Levant is a key region for studying vegetation developments in relation to climate dy- namics and hominin migration processes in the past due to the sensitivity of the vegetation to climate variations and the long history of different anthropogenic occupation phases. However, paleoenvir- onmental conditions in the southern Levant during the Late Pleistocene were still insuf ﬁ ciently understood. Therefore, we investigated the vegetation and ﬁ re history of the Dead Sea region during the last glacial period. We present a new palynological study conducted on sediments of Lake Lisan, the last glacial precursor of the Dead Sea. The sediments were recovered from the center of the modern Dead Sea within an ICDP campaign. The palynological results suggest that Irano-Turanian steppe and Saharo-Arabian desert vegetation prevailed in the Dead Sea region during the investigated period (ca. 88,000 e 14,000 years BP). Nevertheless, Mediterranean woodland elements signi ﬁ cantly contributed to the vegetation composition, suggesting moderate amounts of available water for plants. The early last glacial was characterized by dynamic climate conditions with pronounced dry phases and high but unstable ﬁ re activity. Anatomically modern humans entered the southern Levant during a climatically stable phase (late MIS 4 e MIS 3) with diverse habitats, constant moisture availability, and low ﬁ re activity. MIS 2 was the coldest phase of the investigated timeframe, causing changes in woodland composition and a widespread occurrence of steppe. We used a biome modeling approach to assess regional vegetation patterns under changing climate conditions and to evaluate different climate scenarios for the last glacial Levant. The study provides new insights into the environmental responses of the Dead Sea region to climate variations through time. It contributes towards our understanding of the paleoenvironmental conditions in the southern Levant, which functioned as an important corridor for human migration processes.


Introduction
The study of the paleoclimate and its influences on environments is essential not only to understand current and future climate changes (Masson-Delmotte et al., 2013) but also to reconstruct the history of mankind. The Levant in the southeastern Mediterranean region is a possible meeting point of anatomically modern humans (AMH) and Neanderthals, where gene flow between the two hominins may have occurred (Kuhlwilm et al., 2016). The Levantine fossil record provides evidence for the first migration of AMH out of Africa (Hershkovitz et al., 2018) and for later hominin migration processes towards Eurasia (Mellars, 2011). A renewed migration of AMH out of Africa, probably leading to the occupation of Europe, took place during the last glacial. Earliest fossil evidence dated to ca. 55 ka BP (kilo years before present; all radiocarbon dates in this paper have been calibrated) originates from Manot Cave, Israel (Hershkovitz et al., 2015). In addition, the extinction of Neanderthals occurred during the last glacial, namely between 45 and 30 ka BP, leading to ongoing discussions about the causes (Shea, 2008). Among others, climatic causes have been postulated for AMH migration and Neanderthal extinction processes (e.g., Shea, 2008;Müller et al., 2011).
However, there are ongoing discussions on the hydroclimatic conditions in the southern Levant during the last glacial. While the majority of climate records from the Eastern Mediterranean agree on cold and arid conditions during the last glacial (e.g., Fleitmann et al., 2009;Langgut et al., 2011;Müller et al., 2011;Pickarski et al., 2015), most studies from the Dead Sea region suggest more humid conditions compared to today. Indications for an increased glacial wetness were provided by various studies dealing, for instance, with lake-level reconstructions, geochemical compositions of sediments, and speleothem activities. Lake-level reconstructions of Lake Lisan, the precursor of the Dead Sea, indicate major lake-level high stands during the last glacial of up to ca. 240 m above typical Holocene levels (e.g., Bartov et al., 2002;Torfstein et al., 2013b). During the highest stands, Lake Lisan even merged with the Sea of Galilee (Bartov et al., 2003;Hazan et al., 2005;Torfstein et al., 2013b), being separated today by more than 100 km. The sediments that deposited during the occurrence of Lake Lisan are mainly comprised of aragonite-detritus laminae (e.g., Begin et al., 1974;Katz et al., 1977;Machlus et al., 2000). Aragonite formation requires an increased input of freshwater to the lake to provide bicarbonate to the Ca-chloride brine (Stein et al., 1997;Barkan et al., 2001). Therefore, the high lake levels and aragonite deposition might indicate increased precipitation rates during the last glacial (Stein et al., 1997;Torfstein et al., 2013b). Increased glacial wetness is also suggested by the deposition of speleothems in areas that were too dry for speleothem growth during the Holocene (Vaks et al., 2003(Vaks et al., , 2006Bar-Matthews et al., 2017).
Despite its outstanding role in terms of archeology and paleoclimate, little was known about the paleovegetational conditions in the southern Levant during the last glacial. The study of fossil pollen and other palynomorphs contained in sediments enables the reconstruction of past environments, particularly the paleovegetation and paleoclimate (Faegri and Iversen, 1989). However, a detailed palynological study based on an independent chronology was still missing for the last glacial Dead Sea region.
Previous palynological studies from the Levant were either based on marine or terrestrial sediments. Marine studies were conducted in the Levantine Basin and suggest cold and dry glacial conditions (Cheddadi and Rossignol-Strick, 1995;Langgut et al., 2011). However, the pollen assemblages reflect a huge pollen source area given the basin size and were influenced by African vegetation due to the Nile outflow (Langgut, 2018). A terrestrial pollen record encompassing the whole last glacial was obtained from Yammouneh Basin, Lebanon (Gasse et al., , 2015. This study also indicates warm and wet interglacials and cold and dry glacials. However, given the high altitude of the Yammouneh Basin, the vegetation might have been influenced by water deficiencies due to water storage as ice or frozen soils . The same orographic climate effect might have affected the Birkat Ram area on the Golan Heights, where palynological results for the last 30 ka suggest a similar vegetation pattern (Schiebel, 2013). Strikingly, a last glacial sequence from the Sea of Galilee, located in the Jordan Valley below sea level, also indicates less water availability for plants compared to the Holocene (Miebach et al., 2017). Pollen records from the Hula Basin, northern Israel, might also encompass large parts or even the whole last glacial period (e.g., Horowitz, 1979;Weinstein-Evron, 1983). However, problems with radiocarbon ( 14 C) dating (e.g., Meadows, 2005;van Zeist et al., 2009) and UraniumeThorium (UeTh) dating (Weinstein-Evron et al., 2001) of sediment cores from Hula Basin make a convincing correlation to other records difficult. Major chronological uncertainties also occur in a sediment core from the Ghab Valley, northwestern Syria. The pollen sequence suggested the spread of Mediterranean forests during marine isotope stages (MIS) 3 and 2 (Niklewski and van Zeist, 1970). However, Rossignol-Strick (1995) revised the chronology to a late glacial and Holocene age. Horowitz (1992) presented a vegetation model for the Dead Sea region during the Late Quaternary based on several low-resolution pollen sequences. The correlation to climate records was based on few age determinations and the following hypothesis: during times corresponding to even numbered MIS, woodland vegetation predominated (e.g., MIS 4 and 2), while periods corresponding to odd numbered MIS were characterized by the dominance of steppe vegetation (e.g., MIS 3 and 1) and/or desert vegetation (e.g., MIS 5 and 1). Therefore, further knowledge is required about the vegetation history of the Levantine lowlands based on a terrestrial high-resolution pollen record with a robust chronology.
The study of microscopic charcoal can provide valuable insights into the history of fire activity and the relationship between the paleovegetation, anthropogenic activities, and fire regimes (Whitlock and Larsen, 2001). However, almost none of the last glacial palynological studies from the Levant investigated microcharcoal in addition to pollen. Exceptions are studies from the Ghab Valley (Yasuda et al., 2000) and the Hula Basin (Turner et al., 2010) dating to the late glacial and Holocene. However, large uncertainties in the radiocarbon chronologies of the sediment cores from the Ghab Valley and the Hula Basin (e.g., Meadows, 2005;van Zeist et al., 2009) make the timing of events and the correlation to other records speculative. Still, these records provide comparisons between changes in fire activity and vegetation because charcoal and pollen originated from the same sediment sequences. The fire history during earlier times of the last glacial and particularly at the Dead Sea remained unknown.
Here, we present a new palynological study conducted on sediments of the last glacial Lake Lisan (Lisan Formation). The investigated sediments were recovered in the framework of the International Continental Scientific Drilling Program (ICDP) from the central Dead Sea (Stein et al., 2011a, b) and have an independent chronology based on 14 C and U-Th dating (Neugebauer et al., 2014;Torfstein et al., 2015;Kitagawa et al., 2017). The palynological study provides new and detailed insights into the last glacial vegetation and fire history of the southern Levant in relation to climate changes. We test previous hypotheses on the paleovegetation in the Dead Sea region, namely the coincidence of woodland expansion and retraction with marine isotope stages (Horowitz, 1992). We evaluate climate scenarios for the last glacial Levant regarding their influence on the biome distribution using a biome-climate transfer function and the accordance with the new palynological dataset. In addition, we draw conclusions about the paleoenvironmental setting for occupation phases of AMH and Neanderthals.

Dead Sea
The Dead Sea is situated in the southern Levant bordering Israel, Jordan, and the Palestinian Territory West Bank (Fig. 1). It is a terminal lake and is primarily fed by the perennial Jordan River but also by groundwater and several streams, which are mainly ephemeral, i.e., they experience occasional flashfloods. The total drainage area comprises 42,200 km 2 (Greenbaum et al., 2006). The Dead Sea occupies the lowest continental depression on Earth (currently ca. 430 m below mean sea level (m bmsl)). With a surface area of about 760 km 2 (76 km in NeS direction, up to 17 km in WeE direction), it is the largest lake in the region (Litt et al., 2012). A sill separates a northern deep basin with a water depth of about 300 m from a shallower southern basin. While the water depth of the northern basin has been steadily declining during the last decades due to human impact, the southern basin is nowadays occupied by evaporation ponds (Greenbaum et al., 2006). The Dead Sea is a hypersaline water body with a salinity of ca. 27.5% (ca. 340 g/l), i.e., the salinity is multiple times higher compared to seawater (Gavrieli and Stein, 2006).
The Dead Sea is located at the Dead Sea Transform, a tectonic boundary between the Sinai and Arabian plates. The Dead Sea Basin is the largest and oldest of several pull-apart basins, which originated along the transform during the plate motion process (Garfunkel, 2014). Since its formation in the early Miocene, the Dead Sea Basin subsided continuously and acted as a major sediment trap (Garfunkel, 1997). During the late Neogene, the valley was filled with water coming from the Mediterranean Sea and forming the marine Sedom Lagoon (Stein, 2014). After the disconnection of the Sedom Lagoon from the Mediterranean Sea, the Dead Sea Basin was occupied by a series of lakes. One of them was Lake Lisan, which occurred during the last glacial. While chronostratigraphic analyses at the shore indicate an age of ca. 70 to 15 ka BP for the duration of Lake Lisan (e.g., Torfstein et al., 2013a), the new ICDP core drilled at the deepest part of the Dead Sea suggest that the  Greenbaum et al. (2006). transition to Lake Lisan occurred ca. 15e20 ka earlier (Neugebauer et al., 2016).

Climate and vegetation
The southern Levant is a transition zone of different climate regimes (Fig. 2). The northern part is characterized by a Mediterranean climate with hot, dry summers and mild, wet winters. The precipitation is mainly brought by Mediterranean mid-latitude cyclones, e.g. Cyprus Lows Goldreich, 2003). The southern part is occupied by a dry, subtropical desert (Kushnir et al., 2017). Here, precipitation arrives mainly as flash floods by the tropical Active Red Sea Trough (Dayan and Morin, 2006). While precipitation rates generally decrease towards the south and with lower elevations, temperatures generally decrease towards the north and with higher elevations. Seasonality increases towards the east (Goldreich, 2003). The Dead Sea itself lies in a hyperarid area with 50e100 mm mean annual precipitation (Greenbaum et al., 2006).
The natural vegetation of the southern Levant is primarily shaped by the precipitation distribution but also by temperatures and soils (Fig. 2;Zohary, 1962Zohary, , 1982Danin and Plitmann, 1987;Danin, 1992). The northern part is characterized by the Mediterranean biome with arboreal climax communities. Mediterranean woodland reaches southward to the Judean Mountains and along the upper slopes of the Rift Valley east of the Dead Sea. The area receives 350e1200 mm mean annual precipitation (Zohary, 1962). Common trees are deciduous Quercus (mostly Q. ithaburensis and Q. boissieri; Q. libani at high elevations), evergreen Quercus (Q. calliprinos), Pistacia spp., Ziziphus spp., Rhamnus spp., Ceratonia siliqua, Phillyrea latifolia, Styrax officinalis, Arbutus andrachne, and several Rosaceae species. They are accompanied by conifers such as Pinus halepensis and Juniperus spp. (Baruch, 1986 and references therein;Danin, 1992). Irano-Turanian steppe occupies areas with ca. 100e350 mm mean annual precipitation. It is characterized by herb and dwarf-shrub communities dominated by Artemisia herbaalba. Saharo-Arabian desert vegetation occurs in the southern part, where the mean annual precipitation falls below 100 mm. It is a vegetation type with sparse plant cover and low diversity. Important representatives of the Saharo-Arabian vegetation are Amaranthaceae. Sudanian vegetation occupies tropical oases of the Jordan Valley. Mainly trees and shrubs such as Maerua crassifolia, Acacia tortilis, Balanites aegyptiaca, and Ziziphus spina-christi compose this vegetation type (Zohary, 1962).

Drilling campaign and stratigraphy
The Dead Sea Deep Drilling Project (DSDDP) under the auspices of the ICDP was intended to gain the first long, continuous, and high-resolution sediment sequence from the Dead Sea. The drilling campaign took place in 2010/2011 (Stein et al., 2011a, b). We analyzed sediment samples from the deepest borehole (core 5017-1-A, site 5017-1, N 31 30 0 28.98 00 , E 35 28 0 15.60 00 ) with a total drilled length of 455.34 m. It is located at the center of the northern basin ( Fig. 1). In this study, the sediment depth of the analyzed samples encompasses 199.07e92.35 m. The investigated interval belongs to the Lisan Formation (Neugebauer et al., 2014(Neugebauer et al., , 2016. 76% of the Lisan Formation of the 5017-1-A core are mass transport deposits (MTDs), showing disturbed, slumped, and homogenous sediment sections. The remaining sediments are composed of laminated alternating aragonite and detritus (aad facies) with some gypsum laminae (Kagan et al., 2018). We only sampled the latter facies, resulting in sampling gaps of different length.

Chronology
The current chronology of the investigated sediment sequence is based on a linear interpolation of published 14 C and UeTh dates (Table 1). Twelve 14 C dates were conducted from terrestrial plant remains of core 5017-1-A (Neugebauer et al., 2014;Kitagawa et al., 2017). We calibrated the 14 C dates using the calibration dataset IntCal13 (Reimer et al., 2013) within the software OxCal 4.2 (Ramsey, 2009). UeTh dating was performed on four samples of primary aragonite from core 5017-1-A (for more details see Torfstein et al., 2015). In addition, Torfstein et al. (2015) correlated a massive gypsum deposit in the 5017-1-A core to UeTh dated counterparts in exposed margins of the Dead Sea Basin (Torfstein et al., 2013a). To account for the high number of MTDs in the Lisan Formation, we constructed an event-free age-depth model following Kagan et al. (2018). MTDs larger than 50 cm (Fig. 5) were excluded from the model. Debrites (homogenites, turbidites, and breccias) were completely removed, whereas slumps were considered to include some original laminae. Thus, only 2/3 of the slump thickness was excluded. According to the age-depth model, the investigated sediment section of this study encompasses ca. 88e14 ka BP.

Palynological analyses
Pollen preparation of 203 sediment samples with a sample volume of mostly 4e6 cm 3 was processed following a standard protocol described by Faegri and Iversen (1989). The chemical treatment included 10% hydrochloric acid (HCl) for 10 min, 40% hydrofluoric acid (HF) for 48 h, 10% hot HCl for 10 min, glacial acetic acid (C 2 H 4 O 2 ), hot acetolysis with 1 part concentrated sulfuric acid (H 2 SO 4 ) and 9 parts concentrated acetic anhydride (C 4 H 6 O 3 ) for a maximum of 3 min, and C 2 H 4 O 2 . Sieving and ultrasonic sieving were carried out to remove particles coarser than 200 mm and particles finer than 10 mm, respectively. Lycopodium tablets with a known number of spores were added to each sample to calculate pollen, non-pollen palynomorph, and micro-charcoal concentrations (Stockmarr, 1971). Samples were preserved in glycerol and were stained with safranin.
Palynomorphs were identified with a Zeiss Axio Lab.A1 light microscope with the help of palynomorph atlases and keys (Reille, 1995(Reille, , 1998(Reille, , 1999Beug, 2004) as well as the pollen reference collection of the Institute of Geosciences and Meteorology, University of Bonn. At least 500 terrestrial pollen grains were counted in each sample. Obligate aquatic plants were not included in the terrestrial pollen sum to exclude local taxa growing in the lake (Moore et al., 1991). Furthermore, destroyed and unknown pollen were excluded from the terrestrial pollen sum, which was used to calculate percentages of the pollen assemblage. Quercus pollen was grouped into two morphotypes. Evergreen Quercus pollen was separated from deciduous Quercus pollen according to the morphological features of the evergreen Q. ilex type (Beug, 2004). The pollen types were named after evergreen Q. calliprinos and deciduous Q. ithaburensis, respectively e today's most common species in the region e following the nomenclature rules by Birks (1973). Some pollen types were grouped to higher taxonomic levels in the summary pollen diagrams. Likewise, Alnus, Fraxinus excelsior type, Platanus orientalis, Salix, Tamarix, Ulmus/Zelkova, and Vitis were grouped to riverine trees and shrubs following van Zeist et al. (2009). Pollen count rations of Artemisia to Amaranthaceae (A/ A) and Quercus ithaburensis type to Amaranthaceae (Q/A) were calculated to evaluate their use as climate indicators (El-Moslimany, 1990;Zhao et al., 2012). Charcoal particles were divided into two size fractions with diameters of 25e100 mm and 100e200 mm. If the size fraction is not stated hereafter, the sum of both fractions is given.
A stratigraphically constrained cluster analysis using a square root transformation was performed by CONISS (Grimm, 1987) to aid pollen zonation. All taxa with more than 1% of the terrestrial pollen sum and the sum of trees and shrubs were used for the analysis. Pollen diagrams were prepared with the software Tilia, Version 2.0.41 ( © 1991e2015 Eric C. Grimm).

Biome modeling
The (Bayesian) biome model (Sch€ olzel, 2006;Litt et al., 2012; Table 1 Radiocarbon ( 14 C) dates of terrestrial plant remains and UraniumeThorium (UeTh) dates of Dead Sea core 5017-1-A with sediment depths. The median of calibrated ages is given for 14 C dates. AMS: accelerator mass spectroscopy. Ohlwein and Wahl, 2012) is generally used for probabilistic reconstructions of past climate states given a biome distribution. It includes the probability for the occurrence of each biome given the corresponding pollen spectra, the biome-climate transfer function (also called likelihood function), and a prior probability distribution. In this study, spatial variations of the biomes in the Levant e the Mediterranean, Irano-Turanian, and Saharo-Arabian biome e were modeled using climate information. Therefore, we estimated  (Meusel et al., 1965;Thoma, 2017) used for biome modeling, b) original biome distribution in the Levant, c) modeled probability for the occurrence of biomes in the surrounding of Lake Lisan under changed climate parameters (the analyzed area is marked by the box in b), P A ¼ annual precipitation sum, T W/S ¼ mean summer and winter temperature. only the biome-climate transfer function in the biome model e the probability for the occurrence of a biome given a certain climate state. This probability was determined by using a quadratic discriminant analysis (QDA), where the probability for each biome is dependent on the other two biomes and where the probabilities are normalized to one. We used the CRU version 4.01 dataset comprising monthly terrestrial climate parameters for the period from 1901 to 2016. This global dataset was created by interpolating quality checked station data on a regular 0.5 Â 0.5 latitude-longitude grid (Harris et al., 2014). We used three climate parameters, namely annual precipitation, winter temperature (average December, January, February values), and summer temperature (average June, July, August values). The recent biome distribution is based on vegetation maps by Meusel et al. (1965) that were digitized on the same grid ( Fig. 3a and Sch€ olzel et al., 2002;Litt et al., 2012;Thoma, 2017).
Biome modeling was performed using the freely available R software (R Core Team, 2016). Further details of the model, including the mathematical description, were given by Ohlwein and Wahl (2012) and Litt et al. (2012).

Biome modeling
We performed biome modeling to explore the spatial implications of Levantine biomes in response to climate fluctuations. A series of plotted scenarios (Fig. 3c) draw clues about the climatic sensitivity of biomes in the surrounding of Lake Lisan (Fig. 3b). The model reveals major biome shifts in response to precipitation and temperature changes: I) The Mediterranean biome spreads with increasing precipitation, but only a temperature reduction of more than 4 C causes a decline of the Mediterranean biome; II) Decreasing temperatures trigger a spread of the Irano-Turanian biome, while precipitation changes do not necessary alter the Irano-Turanian biome distribution; III) The Saharo-Arabian biome shrinks under colder conditions and with increasing precipitation.
The model enables us to explore the spatial distribution of biomes under various climate scenarios for the last glacial period though does not account for changing boundary conditions such as atmospheric CO 2 and insolation variations. Given the existing pollen data of the region, those scenarios can be evaluated. Several studies have estimated the glacial temperature in the southern Levant based on proxy data or climate models. For MIS 3, those temperature estimations range between a reduction of ca. 2e5 C with a mean of ca. 3 C (~2 C (Stockhecke et al., 2016) . Despite ongoing discussions on glacial precipitation in the southern Levant (e.g., Torfstein et al., 2013b), few studies are available that quantify the amount of precipitation for the last glacial. Enzel et al. (2008) and Vaks et al. (2006) suggest a doubling of annual precipitation based on rates of aragonite deposition in Lake Lisan and speleothem growth in the northern Negev. Rohling (2013) and Barkan et al. (2001) even assume a 5-fold and 6-fold precipitation rate compared to today, respectively. In contrast, Stockhecke et al. (2016) suggest a reduction of annual precipitation for the last glacial Levant, i.e., ca. 15% during MIS 3 and ca. 30% during MIS 2 (Stockhecke et al., 2016: Fig. 12).
In addition, various aspects of seasonality changes were discussed for the last glacial climate. Such are, for example, an increased precipitation seasonality (Prentice et al., 1992), a decreased precipitation seasonality (Orland et al., 2012), an increased frequency of floods (Ben Dor et al., 2018), and a seasonal moisture deficit due to increased snowfall (Robinson et al., 2006). While we assess total precipitation changes in the biome model, the current biome model does not account for seasonality changes. Still, we discuss the influence of glacial seasonality changes on plants in section 4.3.4.
The modeled biome distribution given the different climate scenarios (Fig. 4bee) illustrates the variability of recent hypotheses in the literature. All vegetation patterns differ considerably from today's modeled biome distribution with unchanged climate parameters (Fig. 4a). The reduction of temperature particularly favors the Irano-Turanian biome, which replaces the Saharo-Arabian biome over a wide area. In some areas such as the Yammouneh Basin, the temperature decline triggers a limitation of the Mediterranean biome due to orographic effects such as frozen soils and snowfall (cf. Develle et al., 2011). Precipitation changes have major impacts on the Mediterranean biome. While an increased precipitation facilitates the spread of the Mediterranean biome, particularly southeast of the Dead Sea, precipitation declines reduce the probability of the Mediterranean biome occurrence.
The biome model outputs are discussed regarding their coincidence with regional pollen data in sections 4.3.3 and 4.3.4.

Pollen zonation
The results of the palynological investigation are summarized in Table 2 and Figs. 5 and 6. Five pollen assemblage zones (PAZs) were defined according to the cluster analysis (Fig. 5). The Roman numeral II follows the hierarchical classification of pollen assemblage superzones after Tzedakis (1994) for a better overview of future synthesis pollen records from the whole Dead Sea profile 5017-1 (see also Chen and Litt, 2018

. General remarks and interpretations
The vegetation in the Dead Sea region was generally dominated by open vegetation during the last glacial period (ca. 88e14 ka BP; Figs. 5 and 6). Large pollen amounts of Amaranthaceae (amaranth family including the former goosefoot family Chenopodiaceae), Artemisia, and Poaceae (grasses) indicate the widespread occurrence of herb and dwarf-shrub communities. Amaranthaceae are the main representatives of the Saharo-Arabian desert biome, which nowadays receives very low annual precipitation values of 100 mm or less (Zohary, 1962;Litt et al., 2012). The plant family contains many drought-adapted and salinity-tolerant species (Rossignol-Strick, 1995;van Zeist et al., 2009). Artemisia is the dominant plant of today's Irano-Turanian steppe biome (Zohary, 1962) and tolerates aridity though less extreme than Amaranthaceae (Rossignol-Strick, 1995). Artemisia is adapted to somewhat higher precipitation rates of ca. 100e350 mm (Zohary, 1962). Modern studies showed that Artemisia pollen increases and Amaranthaceae pollen decreases with decreasing aridity. Therefore, A/A ratios can be used as a moisture indicator, particularly in primary non-forested areas (El-Moslimany, 1990;Zhao et al., 2012). The expansion of Irano-Turanian steppe and suppression of Saharo-Arabian desert under reduced temperatures, as shown by the biome model (Figs. 3 and 4), implies that increased A/A ratios can also be indicative of reduced temperatures. Poaceae are also associated with the Irano-Turanian steppe biome (Litt et al., 2012), particularly humid steppes (van Zeist and Bottema, 2009). Yet, its various species also admix into a range of vegetation types (Danin, 1992. Poaceae comprise of a wild pollen type and a Cerealia Table 2 Pollen assemblage zones (PAZs) with depths, ages, mean spatial and temporal resolution of samples, main components of pollen assemblages (arboreal pollen (AP) and nonarboreal pollen (NAP) with mean percentages), pollen concentration (PC), and definition of lower boundaries (LB   6. Pollen diagram of Dead Sea core 5017-1-A plotted against age with selected taxa, pollen concentrations, ratios of Artemisia to Amaranthaceae and Quercus ithaburensis type to Amaranthaceae with dots marking dry events, charcoal concentrations, and marine isotope stages (MIS) after Lisiecki and Raymo (2005). The chronology is based on a linear interpolation of calibrated radiocarbon dates (triangles), UraniumeThorium dates (squares), and a correlated UraniumeThorium date (star). pollen type, which can morphologically be distinguished. While the wild type refer to the majority of wild grasses, the Cerealia type contain mainly domesticated cereals and their ancestors, which are native to this region (Beug, 2004).
Trees and shrubs never dominated the Dead Sea region during the investigated period. Nevertheless, they contributed substantially to the pollen composition (averagely 25.5%). Quercus ithaburensis type (deciduous oak) was the most abundant arboreal pollen type, followed by Juniperus type (juniper and/or Mediterranean cypress), Quercus calliprinos type (evergreen oak), and Pistacia (pistachio). While deciduous and evergreen oaks are usually well represented to overrepresented (van Zeist et al., 1975;Rossignol-Strick, 1995), Pistacia and Juniperus type are usually underrepresented in the pollen precipitation (Rossignol-Strick, 1995;van Zeist et al., 2009). All of these trees and shrubs represent the Mediterranean biome (Baruch, 1986;Litt et al., 2012), which occurs nowadays in the most humid areas of the southern Levant (Zohary, 1962). Modern Levantine vegetation studies indicated that a reduction of arboreal vegetation coincides with a decline in moisture (Zohary, 1962;Kadmon and Danin, 1999). Likewise, the Mediterranean biome retreats under reduced precipitation rates in the biome model (Figs. 3 and 4). This relation has also been described and discussed by Litt et al. (2012) based on botanical-climatological transfer functions by using a Holocene Dead Sea pollen record. Therefore, the amount of arboreal pollen is strongly related to changes in available moisture for plants in the past, i.e., effective moisture, which depends mainly on precipitation and evapotranspiration (the sum of evaporation and plant transpiration). The Q/A ratio e directly comparing the most abundant components of the Mediterranean and Saharo-Arabian biomes e also mainly reflects moisture changes. In addition, temperature reductions can limit the Mediterranean biome distribution, particularly at high altitudes (Figs. 3 and 4). Still, other factors such as seasonality, local habitats, and insolation must be considered for evaluating vegetation changes through time.
Pollen concentrations generally indicate vegetation density and pollen productivity in the lake catchment. However, Dead Sea samples that contain gypsum show low pollen concentrations, as indicated in Fig. 5, because gypsum laminae have higher sedimentation rates than alternating aragonite-detritus laminae (Stein, 2001;Torfstein et al., 2015). Still, the overall decreasing trend in pollen concentrations along the record suggests a stepwise vegetation density reduction, particularly at the MIS 5/4 transition and at ca. 35 ka BP.
The analysis of charred particles in sediments is one of the primary methods to reconstruct fire activity, i.e., fire frequency and/ or fire intensity, in the past. The comparison with pollen assemblages allowed insights into the relationship between fire, vegetation, and climate (e.g., Swain, 1973;Daniau et al., 2010;Vanni ere et al., 2011). Charcoal particles above 100 mm in size are not transported far from fire sources. Thus, they indicate local fires. Smaller charcoal particles are able to travel longer distances and therefore reflect regional fires (Whitlock and Larsen, 2001). Charcoal concentrations in the investigated Dead Sea sediments vary from 0 to 7457 particles/cm 3 (1476 particles/cm 3 on average). Still, the highest charcoal concentrations are several times lower than during the last interglacial (Chen and Litt, 2018). Charcoal particles <100 mm constitute 84.6% of the charcoal sum and indicate the prevalence of regional fires.
Different fire regimes are generally generated by atmospheric conditions, ignition agents, and the availability of consumable resources, i.e., vegetation (Moritz et al., 2010). According to Whitlock et al. (2010), the highest fire activity is usually related to grassland and savanna biomes. Daniau et al. (2010) reviewed changes in fire regimes during the last glacial on a global scale and concluded that they were primarily related to changes in plant productivity. Also previous studies from the Mediterranean and Near East that investigated the fire history during the last glacial connected enhanced fire activity to higher arboreal pollen percentages and an increased terrestrial biomass caused by higher temperatures and increased moisture (e.g., Daniau et al., 2007;Turner et al., 2010;Pickarski et al., 2015). In contrast, the results from the Dead Sea suggest that forest density, grassland occurrence, and thus availability of fuel was not the primary trigger for changes in fire activity because there is neither a significant linear correlation between charcoal concentrations and arboreal pollen percentages nor with Poaceae percentages. Hence, other climate variations are possible factors that primarily influenced the Dead Sea charcoal record. Such climate variations could be, for instance, longer/more intensive summer droughts (Vanni ere et al., 2008 (umbellifers). Moderate pollen frequencies of Quercus ithaburensis type and small pollen amounts of Pistacia, Juniperus type, and Quercus calliprinos type indicate the occurrence of Mediterranean woodland elements. Trees and shrubs did probably not occur in a closed forest belt but were patchily distributed in habitats with locally more available moisture. They most likely formed a mosaic with Irano-Turanian steppe components.
The charcoal record indicates a rapidly fluctuating fire frequency with an increasing trend (Figs. 6 and 7). Highest charcoal concentrations of the investigated period appear in this PAZ. The charcoal record supports unstable environmental conditions with a high but fluctuating fire activity.
Fluctuations in the vegetation indicate millennial-scale climate oscillations with four phases of reduced available moisture expressed by low A/A and low Q/A ratios (Figs. 6 and 7). They strikingly coincide with the deposition of gypsum (Figs. 5 and 7;Neugebauer et al., 2014). Neugebauer et al. (2016) correlated dry phases during the early last glacial derived from micro-facies analyses from the Dead Sea core 5017-1-A to cold phases in the North Atlantic, coinciding with stadial conditions (mostly cold and dry) in other Mediterranean records. Following this interpretation, the detected dry phases might correlate to Greenland stadials ( Fig. 7; Rasmussen et al., 2014), i.e., pollen would show a Dansgaard-Oeschger signature. However, the current chronology does not allow a convincing correlation to single climate events of other paleorecords. Thus, the connection between rapid vegetation changes in the Dead Sea region and high latitude climatic conditions during this phase remains speculative.

The environment during late MIS 4 to MIS 3 and implications for modern humans
PAZ II3 (168.05e130.99 m; 62.6e34.7 ka BP) corresponds to late MIS 4 and early/middle MIS 3. The onset of PAZ II3 marks the strongest shift in the vegetation during the investigated time, as indicated by the cluster analysis (Fig. 5). Artemisia pollen percentages increase while the pollen frequencies of a range of other herbs and dwarf shrubs, namely Amaranthaceae, Poaceae, Tubuliflorae, Plantago, Liguliflorae, Brassicaceae, and Rumex, decline. The change in non-arboreal pollen is also indicated by increased A/A values (Fig. 6). Moreover, the onset of PAZ II3 displays an increase of arboreal pollen. Overall highest mean percentages of Quercus ithaburensis type occur in PAZ II3. The pollen composition mirrors a spread of Artemisia steppe and Mediterranean woodland components dominated by deciduous oak, implying an increase of available moisture for plants compared to earlier phases. PAZ II2b (130.99e114.56 m; 34.7e30.6 ka BP) corresponds to late MIS 3. Its vegetation composition and abundance of single taxa largely resemble PAZ II3. However, Artemisia percentages further increase in PAZ II2b. The abundance of Artemisia is not only associated with an increase of effective moisture but can also be indicative for reduced temperatures. The complete absence of thermophilous Olea europaea (olive tree) in PAZ II2b e although never represented by high numbers in the pollen record e supports the latter explanation. A reduction of temperatures since ca. 35 ka BP coincides with the study by Ayalon et al. (2013) suggesting conditions too  (Gasse et al., , 2015, B) AP and Artemisia/Amaranthaceae ratios with dashed trend lines and dots marking dry events from Dead Sea (this study), C) charcoal concentrations in particles/cm 3 with dashed trend lines from Dead Sea (this study), D) gypsum deposits in Dead Sea core 5017-1-A (Neugebauer et al., 2014), E) Dead Sea lake-level reconstructions (Waldmann et al., 2009;Torfstein et al., 2013b) with today's lake level (dashed line), F) speleothem growth periods from Mizpe Shelagim Cave, Mt. Hermon (above; Ayalon et al., 2013) and several caves west of the Dead Sea (below; Vaks et al., 2003;Lisker et al., 2010), G) south Levantine hominin record (Shea, 2008;Hershkovitz et al., 2015), H) precipitation anomaly for the Levant expressed as percental deviation from total mean (today has a value of ca. 18%) with average in dark grey (Stockhecke et al., 2016), I) Greenland ice core isotope record with numbered Dansgaard-Oeschger events above  and numbered Heinrich events below , and I) June and December insolation for the 30 th parallel north (Berger and Loutre, 1991). Marine isotope stages (MIS) refer to Lisiecki and Raymo (2005). cold for speleothem growth at Mt. Hermon, northern Israel, between ca. 35 and 16 ka BP (Fig. 7). During late MIS 4 and MIS 3, both factors e relatively high moisture availability and cool conditions e probably played an important role for the environment.
The charcoal record supports an environmental shift around 63e62 ka BP. The fire activity became low and stayed low until the late glacial. A possible trigger could be weaker/shorter summer droughts due to cooler and/or moister summer conditions.
Our observation of high effective moisture during late MIS 4 and MIS 3 is supported by the deposition of speleothems in areas that were too dry for speleothem growth during the Holocene, namely the rain shadow semi-desert and the northern Negev ( Fig. 7; Vaks et al., 2003Vaks et al., , 2006Lisker et al., 2010;Bar-Matthews et al., 2017), and the continuous deposition of aad laminae in the Dead Sea core 5017-1-A (lithology in Fig. 5; Kagan et al., 2018) indicating a positive freshwater balance (e.g., Stein et al., 1997). Therefore, we cannot support the hypothesis that MIS 3 was a dry phase, as suggested by several authors who correlated warmer phases (e.g., Holocene, MIS 3) at the North Atlantic with drier phases in the southern Levant (e.g., Bartov et al., 2003;Haase-Schramm et al., 2004;Stein et al., 2010). Likewise, our palynological results contradict a previous vegetation model (Horowitz, 1992) that suggested the reduction of woodland components in the Dead Sea region during MIS 3. This discrepancy might be ascribed to the nature of MIS 3. Although defined as a separate isotope stage in the marine realm, the benthic d 18 O values defining MIS 3 are much higher than during full interglacial stages such as MIS 5e and 1, and therefore they are rather similar to the full glacial MIS 4 and 2 (Lisiecki and Raymo, 2005). In addition, MIS 3 is not expressed as an interglacial as defined by Jessen and Milthers (1928) for terrestrial records.
To test which climate might have caused the described vegetation pattern with large proportions of steppe though still allowing enough moisture for Mediterranean woodland, we modeled the biome distribution in the Levant for MIS 3 given two different climate scenarios in the literature: a temperature reduction of 3 C (see section 4.1) with I) a precipitation decrease of 15% following Stockhecke et al. (2016) (Fig. 4b) and II) a precipitation increase of 100% following Enzel et al. (2008) and Vaks et al. (2006) (Fig. 4c). Both model outputs fit very well with vegetation reconstructions for the Yammouneh Basin ( Fig. 7; Gasse et al., 2011Gasse et al., , 2015, indicating a sparse herbaceous flora in nowadays forested areas of Lebanon. However, the biome model reveals much different vegetation patterns for the Dead Sea region. Scenario I suggests a wellmixed balance of the three biome types with an increased probability for the Irano-Turanian biome compared to today. In contrast, scenario II suggests a strong expansion of the Mediterranean biome, causing lower probabilities of the other two biomes. The pollen signal from the Dead Sea (Fig. 6) supports the first scenario due to high proportions of Irano-Turanian taxa and moderate amounts of Mediterranean taxa. The model indicates that a small reduction in precipitation together with reduced temperatures does still allow sufficient moisture availability for the growth of Mediterranean taxa in the Dead Sea region.
The comparison of the Dead Sea region to northern latitudes reveals a contrasting pattern during late MIS 4 and MIS 3. Greenland and Europe were influenced by frequent and rapid climate fluctuations (Fig. 7; Rasmussen et al., 2014), causing rapid expansion and contraction of arboreal vegetation over southern Europe (e.g., Fletcher et al., 2010). In comparison to that, the Dead Sea region saw more stable environmental conditions with very low fire activity, with steady woodland occurrence (lowest AP standard deviation of the record), and without pronounced dry phases resulting in constant moisture availability. Still, climate tracers such as the pollen ratios A/A and Q/A indicate some variations in the pollen assemblage that might refer to Dansgaard-Oeschger oscillations, as discussed above.
AMH occupied the southern Levant during this stable climate phase after an absence of several thousand years. Earlier dispersals from Africa into the southern Levant were dated to 194e177 ka BP based on fossils from the Misliya Cave, northern Israel (Hershkovitz et al., 2018), and to ca. 130e90 ka BP based on human remains from the Skhul Cave and the Qafzeh Cave, northern Israel (e.g., Valladas et al., 1988;Mercier et al., 1993;Grün et al., 2005). The fossils from the Misliya Cave provide the earliest evidence for AMH outside Africa (Hershkovitz et al., 2018). The oldest fossils from the investigated last glacial period date to 54.7 ± 5.5 ka BP and were excavated in the Manot Cave in northern Israel (Hershkovitz et al., 2015). Hershkovitz et al. (2015) suggested that those people could be closely related to the first AMH who eventually dispersed into Europe. Earliest evidence for the colonization of Europe by AMH dates to 45e43 ka BP (Benazzi et al., 2011).
AMH might have benefited from a humid phase between ca. 56 and 44 ka BP when crossing the nowadays arid regions between northeastern Africa and the southern Levant . After entering the southern Levant, however, AMH lived in a refugial area with favorable environmental conditions on longer time scales according to our study. The vegetation formed a mosaic of Irano-Turanian steppe with herbs and dwarf shrubs, open Saharo-Arabian desert vegetation, and Mediterranean woodland elements. This diverse landscape offered a large variety of habitats for animals and humans. No pronounced dry phases occurred, and water was constantly available in the ecosystem. Temperatures were certainly lower than today, but the continuous occurrence of frost-sensitive Pistacia indicates mild winters, at least at lower altitudes (Rossignol-Strick, 1995). The fire activity was constantly very low, suggesting weak/short summer droughts.
In contrast, Neanderthals had already occupied the southern Levant since the late MIS 5. Neanderthal remains from the Amud, Kebara, and Geula Caves were dated to ca. 81e42 ka BP (Shea, 2008 and references therein). Thus, they already lived in the southern Levant during a climatically dynamic phase with pronounced dry phases, higher fire activity, and intensified resource uncertainty.
To conclude, Neanderthals tolerated a wide spectrum of environmental conditions in the glacial Dead Sea region, and diverse and rather stable environmental conditions since ca. 63 ka BP provided great potential for the residence of AMH.  (Rossignol-Strick, 1995). Together with the high abundance of Artemisia and the complete absence of Olea europaea, these changes in the pollen assemblage indicate the coolest phase of the investigated timeframe. This conclusion fits well with temperature calculations inferred from the Soreq Cave speleothems, central Israel (McGarry et al., 2004;Affek et al., 2008) and alkenone-based sea surface temperatures for the southeastern Levantine Basin (Almogi-Labin et al., 2009). Low pollen concentrations suggest a sparser vegetation compared to previous phases.
The constantly low fire frequency, as indicated by low charcoal concentrations, coincides with observations by Orland et al. (2012). They suggested a decreased seasonal rainfall gradient prior to 15 ka BP inferred from the Soreq Cave speleothems. After 15 ka BP and particularly during the Holocene, the climatic conditions were characterized by higher seasonality with distinct wet and dry seasons.
The comparison of pollen records from the Yammouneh Basin in Lebanon (Gasse et al., , 2015, the Birkat Ram maar lake on the Golan Heights (Schiebel, 2013), the Sea of Galilee in northern Israel (Miebach et al., 2017;Schiebel and Litt, 2018), and the Dead Sea (Litt et al., 2012; this study) allows an assessment on vegetation and climate pattern in the Levant during MIS 2. The Yammouneh Basin and the Birkat Ram are located in regions that are nowadays characterized by Mediterranean climate with a climax of dense Mediterranean woodland. The Sea of Galilee lies at the southern edge of the Mediterranean biome bordering the Irano-Turanian steppe. The Dead Sea is located in a hyperarid area, where Saharo-Arabian desert vegetation prevails. The surrounding mountains are covered by Irano-Turanian steppe vegetation and Mediterranean woodland. Strong gradients in precipitation, temperature, and the vegetation distribution between these localities occur nowadays ( Fig. 2; Zohary, 1962) and occurred during the Holocene (Fig. 8).
In contrast, Mediterranean forests were considerably reduced in the vicinity of the Yammouneh Basin, the Birkat Ram, and the Sea of Galilee during MIS 2. Arboreal components were more abundant in the Dead Sea region compared to the Holocene. However, a continuous and strong human impact, reducing the amount of trees and shrubs, has to be considered for the Holocene (Miller, 1991;Rollefson and K€ ohler-Rollefson, 1992), also given that mean arboreal percentages for the last interglacial optimum were more than twice as high as for the Holocene (Litt et al., 2012;Chen and Litt, 2018). Dwarf shrubs, grasses, and other herbs dominated the glacial vegetation in the whole study area, and there was no continuous and dense vegetation belt of the Mediterranean biome in the northern parts. Thermophilous trees were probably patchily distributed at moister habitats, particularly in the Jordan Rift Valley, where arboreal pollen percentages are somewhat higher than in the mountainous areas of the north.
The comparison of the described vegetation pattern with modeled biome distributions (Fig. 4) enables us to evaluate two different climate scenarios: a temperature reduction of 6 C (see section 4.1) with I) a precipitation decrease of 30% following Stockhecke et al. (2016) (Fig. 4d) and II) a precipitation increase of 100% following Enzel et al. (2008) and Vaks et al. (2006) (Fig. 4e). Both model outputs fit well with the pollen data from Yammouneh and Birkat Ram, suggesting the dominance of steppe and reduction of forest during MIS 2. However, the vegetation patterns differ greatly in the Dead Sea region: While scenario I triggers a similar distribution of the Mediterranean biome with lower probabilities of occurrence, scenario II suggests a spread of Mediterranean woodland into regions close to the Dead Sea that are nowadays arid. While scenario I slightly underestimates the proportion of Mediterranean woodland compared to the Dead Sea pollen data, scenario II clearly overestimates the distribution of the Mediterranean biome. Further studies are needed to estimate precipitation rates for MIS 2, which were probably slightly reduced but by not as much as 30%. In addition, precise temperature estimations, particularly for interior regions such as the Jordan Rift Valley, are needed because temperature also modifies the biome distribution.
Slightly lowered precipitation rates during MIS 2 could have still allowed relatively high available moisture for plants and a positive freshwater balance in the Dead Sea region. Reduced temperatures ( Fig. 7; Rasmussen et al., 2014), low summer insolation (Figs. 7 and 8; Berger and Loutre, 1991), and reduced catchment wide evaporation (due to lower temperatures and higher relative humidity; Bar-Matthews et al., 2017) would cause higher effective moisture compared to recent conditions. The total evapotranspiration would have been additionally lowered because of the low woodland cover in the northern mountain range, leading to reduced plant transpiration. This can have a significant impact on the water budget (Schiller et al., 2002(Schiller et al., , 2010Ungar et al., 2013). A lower basin wide evaporation and lower plant transpiration would have still allowed a positive freshwater balance in the lakes, leading to the increased lake levels even under reduced precipitation rates (Stockhecke et al., 2016). In addition, plant cover could have been shaped by available water in their habitats, while additional precipitation was stored as snow on high mountains (Robinson et al., 2006) and released as flash floods. The frequency of flash floods was considerably increased during rising Lake Lisan levels (Ben Dor et al., 2018). Plants could probably not sufficiently use the water brought by rapid snowmelts in springs and rapid drains during flash floods. Moreover, plant growth could have been limited by reduced atmospheric CO 2 levels during the last glacial, as previous studies suggested (e.g., Cowling and Sykes, 1999;Prentice et al., 2017). According to experimental studies, low CO 2 can result in decreased plant fitness and increased water use, especially of C3 plants (Gerhart and Ward, 2010 and references therein). Therefore, a reduction of forests and a shift from C3 plant to C4 plant dominance were ascribed to low CO 2 levels in other regions (e.g., Levis et al., 1999;Harrison and Prentice, 2003). However, a correlation of woodland retreat and a spread of C4 plants (e.g., many Amaranthaceae species) with gradually decreasing atmospheric CO 2 levels during the last glacial (Petit et al., 1999) cannot be detected in the Dead Sea pollen record. Therefore, the impact of CO 2 on the Levantine vegetation composition seems to be limited. Still, decreasing CO 2 levels could explain the gradually shrinking pollen concentration that indicates a declining vegetation density.

Preliminary insights into the late glacial environment
PAZ II1 (95.99e92.35 m; 15.4e14.2 ka BP) corresponds to part of the late glacial. The phase begins with a pronounced peak of Amaranthaceae and a small peak of Asteraceae (Tubuliflorae and Liguliflorae). Simultaneously, most other taxa drop. Thereafter, the composition and abundance of taxa resembles PAZ II4 with a diverse herbaceous flora and low amounts of Artemisia. It is probable that the initial peaks of Amaranthaceae and Asteraceae reflect a local spread of pioneer plants, engendering a temporal overrepresentation in the pollen assemblage and a statistical suppression of other taxa. A spread of local plants is supported by a simultaneous major lake-level drop of Lake Lisan from ca. 260e465 m bmsl, one of the lowest stands during the Late Quaternary ( Fig. 7; Stein et al., 2010;Torfstein et al., 2013b). The exposed shores could have been vegetated by saline-tolerant pioneer communities including species of Amaranthaceae and Asteraceae, as modern observations suggest (Aloni et al., 1997). Due to the wind-pollination of Amaranthaceae, local stands of this family are particularly overrepresented in the pollen diagram. A similar peak was observed after a strong lake-level decline at the Sea of Galilee at ca. 24e23 ka BP (Miebach et al., 2017), and the same phenomenon probably played a role at the Dead Sea during the transition to the last interglacial period (Chen and Litt, 2018). Still, a short dry period, as suggested by the deposition of gypsum in the Dead Sea core 5017-1-A (Figs. 5 and 7; Torfstein et al., 2013b;Neugebauer et al., 2014), and the strong lake-level reduction of Lake Lisan ( Fig. 7; Torfstein et al., 2013b) most likely additionally triggered the spread of herbs and dwarf shrubs.
Frost-sensitive Pistacia and Olea europaea occur consistently again after a virtual absence of many millennia. This indicates a return to higher temperatures at least during winters (van Zeist et al., 1975;Rossignol-Strick, 1995). The reduction of Artemisia and Juniperus type, which were common during the cold MIS 2 including the Last Glacial Maximum, also imply higher temperatures compared to MIS 2.
Since ca. 16.2 ka, the fire activity increased again. A probable cause could be a higher seasonality with distinct wet and dry seasons. This explanation is in line with the study by Orland et al. (2012), suggesting an increased seasonality since the late glacial inferred from the Soreq Cave speleothems.

Summary and conclusions
The 5017-1 profile obtained by the DSDDP is the longest continuous sediment record from the Dead Sea Basin, enabling the detailed reconstruction of the southern Levantine environmental history (Neugebauer et al., 2014). The detailed vegetation history of the southern Levant in particular was still insufficient understood given the scarceness of long continuous lacustrine sediment sequences and major chronological uncertainties in many pollen records (e.g., Rossignol-Strick, 1995;Weinstein-Evron et al., 2001;Meadows, 2005). Here, we analyzed the pollen and microscopic charcoal assemblage of the Lisan Formation of the Dead Sea core 5017-1-A, spanning ca. 88e14 ka BP. Moreover, we performed biome modeling to assess shifts in biome distribution in response to climate variations. The biome modeling results illustrate the climate sensitivity of the regional vegetation and help to evaluate climate scenarios for the last glacial Levant.
The pollen record indicates a mixture of Irano-Turanian steppe, Saharo-Arabian desert vegetation, and Mediterranean woodland elements. Although changes in the pollen composition might not Fig. 8. Comparison of a north-to-south gradient in the Levant. Arboreal pollen (AP) with mean percentages shown as dashed lines from A) Yammouneh, Lebanon (Gasse et al., , 2015, B) Birkat Ram, Golan Heights (Schiebel, 2013; discontinuous sedimentation might occur between ca. 10 and 17 ka BP), C) Sea of Galilee, northern Israel (MIS 1: Schiebel and Litt, 2018;MIS 2: Miebach et al., 2017), and D) Dead Sea, Israel/Palestine/Jordan (MIS 1: Litt et al., 2012; MIS 2: this study). E) June and December insolation for the 30 th parallel north (Berger and Loutre, 1991). F) Lake-level reconstructions of the Sea of Galilee (Hazan et al., 2005), the Dead Sea, and its precursor Lake Lisan (Torfstein et al., 2013b). Marine isotope stages (MIS) refer to Lisiecki and Raymo (2005). provide clear evidence for variations in the precipitation amount, it indicates the availability of water for plants, i.e., the effective moisture. Decreased pollen ratios of AP/NAP, A/A, and Q/A indicate low effective moisture probably pointing to a Dansgaard-Oeschger signature. Hence, four dry phases occurred during the early last glacial (MIS 5b/a and early MIS 4), which coincide with the deposition of gypsum in Lake Lisan. A high and dynamic fire activity took place.
An increased proportion of Irano-Turanian steppe vegetation and Mediterranean woodland elements suggests consistently high effective moisture during late MIS 4, MIS 3, and MIS 2, when fire activity was continuously low. Biome modeling suggests that no precipitation increase is needed for such amounts of effective moisture. Lower insolation, reduced catchment wide evapotranspiration, and low temperatures were probably sufficient for a positive water balance. MIS 2 was the coolest period of the investigated timeframe, as indicated by a change in arboreal taxa.
The residence of anatomically modern humans in the southern Levant during the last glacial was supported by stable environmental conditions with relatively high effective moisture and low fire activity. In contrast, Neanderthals had already lived in a dynamic ecosystem with changing water availability and higher fire activity, thus tolerating a wide spectrum of environmental conditions.
The comparison of Levantine pollen records along a north-tosouth gradient indicates that, at least during MIS 2, there was no gradient of available water for plants comparable to the Holocene and today. An overall more similar, open vegetation occurred, and there was no dense Mediterranean woodland belt in the north. Scattered stands of thermophilous woodland components e particularly found at lower altitudes such as the Jordan Rift Valley e formed a heterogeneous landscape.
Major environmental changes occurred during the late glacial. However, further high-resolution analyses based on a robust chronology are needed to reveal the detailed vegetation and fire history.
The new palynological record contributes towards our understanding of the influence of long-term and short-term climate oscillations on the environment. New insights into the Levantine vegetation history help to reconstruct and evaluate the regional paleoclimate, which also carries implications for recent and future climate changes. Furthermore, the detailed knowledge of the environmental setting is essential to reveal the relationship between environmental developments and anthropogenic processes in the past.

Data availability
The palynological dataset related to this article is available on the PANGAEA database (https://doi.pangaea.de/10.1594/PANGAEA. 900564).