Abrupt climate and vegetation variability of eastern Anatolia during the last glacial

Detailed analyses of the Lake Van pollen, Ca /K ratio, and stable oxygen isotope record allow the identification of millennial-scale vegetation and environmental changes in eastern Anatolia throughout the last glacial (∼ 111.5–11.7 ka BP). The climate of the last glacial was cold and dry, indicated by low arboreal pollen (AP) levels. The driest and coldest period corresponds to Marine Isotope Stage (MIS) 2 (∼ 28–14.5 ka BP), which was dominated by highest values of xerophytic steppe vegetation. Our high-resolution multi-proxy record shows rapid expansions and contractions of tree populations that reflect variability in temperature and moisture availability. These rapid vegetation and environmental changes can be related to the stadial-interstadial pattern of Dansgaard–Oeschger (DO) events as recorded in the Greenland ice cores. Periods of reduced moisture availability were characterized by enhanced occurrence of xerophytic species and high terrigenous input from the Lake Van catchment area. Furthermore, the comparison with the marine realm reveals that the complex atmosphere–ocean interaction can be explained by the strength and position of the westerlies, which are responsible for the supply of humidity in eastern Anatolia. Influenced by the diverse topography of the Lake Van catchment, more pronounced DO interstadials (e.g., DO 19, 17–16, 14, 12 and 8) show the strongest expansion of temperate species within the last glacial. However, Heinrich events (HE), characterized by highest concentrations of ice-rafted debris (IRD) in marine sediments, cannot be separated from other DO stadials based on the vegetation composition in eastern Anatolia. In addition, this work is a first attempt to establish a continuous microscopic charcoal record for the last glacial in the Near East. It documents an immediate response to millennial-scale climate and environmental variability and enables us to shed light on the history of fire activity during the last glacial.


Introduction
The last glacial inception was marked by the expansion of continental ice sheets and substantial changes in oceanographic conditions in the North Atlantic as well as in atmospheric temperature and moisture balance in the Northern Hemisphere (e.g., Blunier and Brook, 2001;Cacho et al., 2000Cacho et al., , 1999Chapman and Shackleton, 1999;Sánchez Goñi et al., 2002;Svensson et al., 2008Svensson et al., , 2006Wolff et al., 2010). Between Marine Isotope Stages (MIS) 5d and 2, the climatic conditions were characterized by numerous abrupt millennial-scale oscillations, known as Dansgaard-Oeschger events (DO; Dansgaard et al., 1993). These are most prominently documented in Greenland ice cores and exhibit an abrupt warming (Greenland interstadials; GI), followed by a gradual cooling and a final rapid temperature drop towards a cold Greenland stadial (GS; e.g., NGRIP members, 2004;Rasmussen et al., 2014;Svensson et al., 2008;Wolff et al., 2010). About 25 such stadial to interstadial transitions, varying in amplitude from 5 to 16 • C, are defined in the NGRIP record during the last glacial period (NGRIP members, 2004;Rasmussen et al., 2014;Wolff et al., 2010). Although the climatic and environmental impacts of the DO cycles have been intensively studied during the last decades, the mechanism behind them is still under debate (e.g., Cacho et al., 2000Cacho et al., , 1999Rasmussen et al., N. Pickarski et al.: Abrupt climate and vegetation variability of eastern Anatolia 2014; Wolff et al., 2010). The main process proposed as a cause for the recurring pattern is freshwater being forced from ice-sheets that affected the extent of the sea ice, ocean heat transport, and Atlantic Meridional Overturning Circulation (AMOC; Bond and Lotti, 1995;Cacho et al., 2000Cacho et al., , 1999Chapman and Shackleton, 1999;Hemming, 2004;Hodell et al., 2008;McManus et al., 1999;Rasmussen et al., 2014;Wolff et al., 2010). The most extreme cold intervals are Heinrich events (HE; Bond et al., 1993, 1992, Heinrich, 1988, characterized by reduced sea surface temperatures (SST; Cacho et al., 1999) with highest concentrations of ice-rafted debris (IRD) in marine sediments due to massive iceberg discharges, which mainly originated from the Laurentide ice sheet (Alvarez-Solas and Ramstein, 2011).
Since the first long lacustrine sediment sequences were recovered at Lake Van in summer 2010 Litt et al., 2012), numerous high-resolution data have been gathered, providing insight into short-term changes in past climatic and environmental conditions in eastern Anatolia (e.g., XRF measurements described by Kwiecien et al., 2014; total organic carbon content (TOC), Stockhecke et al., 2014a). The sensitivity of this region was already welldocumented by previous palynological data sets covering the Late Glacial and the Holocene periods (Litt et al., 2009;Wick et al., 2003). First pollen results of the new ∼ 219 m long composite profile encompassing the last 600 ka, have been described based on lower temporal resolution (between ∼ 900 and 3800 years) by . A detailed highresolution pollen analysis (between ∼ 100 and 800 years) for the last interglacial (131.2-111.5 ka BP) is documented in Pickarski et al. (2015).
Here we present new biotic data (pollen, microscopic charcoal remains) and combine them with already available Lake Van abiotic proxies (stable oxygen isotope and element measurements) of the last glacial period (∼ 111.5-11.7 ka BP). Special focus is given to the centennial-to millennial-scale climate variability, as known from Greenland, and the regional response of vegetation to abrupt paleoenvironmental changes in eastern Anatolia. After the examination of climate and vegetation changes on a local level, we compare our results to selected global reference archives. Furthermore, we provide the first continuous sedimentary microscopic charcoal record from Lake Van to give insights into the coupling and feedback between fire activity and major changes in climate, vegetation, and fuel amount during the last glacial.

Regional setting
Lake Van (38.6 • N,42.8 • E) is a deep terminal alkaline lake (3574 km 2 ; max. depth > 450 m) situated on the eastern Anatolian high plateau at 1647 m above sea level (a.s.l., Fig. 1). It is the largest soda lake in the world (Degens and Kurtman, 1978), which is partly fed by numerous small rivers around the basin. In the south, the lake is surrounded by the Bitlis Massif reaching altitudes of more than 3500 m a.s.l. Two large active stratovolcanoes, Nemrut (2948 m a.s.l.) and Süphan (4058 m a.s.l.), border the lake to the west and north (Fig. 1b).
The present-day climate of eastern Anatolia is controlled by seasonal changes in the position and strength of the following atmospheric components: (a) the mid-latitude westerlies, (b) the sub-tropical high-pressure system, and (c) the Siberian high-pressure system (Akçar and Schlüchter, 2005;Türkeş, 1996). The regional climate at Lake Van is continental with warm and dry summers (mean temperature > 20 • C; Turkish State Meteorological Service) and cold and wet winters, marked by regular frosts. The minimum average temperature of the coldest month is far below 0 • C (−7.9 • C in Van; see Fig. 1b for the location). Total rainfall varies from 385 mm a −1 (Van) to 816 mm a −1 (Tatvan) and peaks in the winter months (October to February), with a second rainfall maximum during spring (March to May). Higher elevations of the west-east oriented mountain ranges along the Bitlis Massif ( Fig. 1b) are affected by the strength and position of the "Cyprus cyclones" from the Mediterranean Sea with precipitation values up to 1200 mm a −1 in Bitlis (Turkish State Meteorological Service; . The modern distribution of vegetation at Lake Van is closely related to rough orography and spatial rainfall variability. The southward slopes of the Bitlis Massif are covered by the Kurdo-Zagrosian oak steppe-forest (Quercetea brantii), which extends from the Taurus Mountains (east-central Turkey) via the Bitlis complex (SW shore of Lake Van) to the Zagros Mountains (SW Iran;Zohary, 1973). It consists of several oak species, which are accompanied by Pistacia atlantica, P. khinjuk, Acer monsplessulanum, Juniperus oxycedrus, Pyrus syriaca, Crataegus spp., Prunus and Amygdalus spp. (Frey and Kürschner, 1989). In the rain-shadow, where rainfall decreases drastically, the north-eastern part of the lake drainage is covered by Irano-Turanian steppe vegetation, dominated by Artemisia fragrans, steppe forbs and grasses (Zohary, 1973).

Material and methods
Sedimentary record "AR" (Ahlat Ridge; 38.667 • N, 42.669 • E; 357 m water depth) was collected on a bathymetric ridge in the northern part of the Tatvan Basin ( Fig. 1b) during the ICDP (International Continental Scientific Drilling Program) project PALEOVAN in summer 2010 Litt et al., 2012). Here we present data of the uppermost 3.87-41.72 m of the event-corrected composite record (mcblf-nE; depth scale, which excludes volcanic ash layers and mass flow deposits; Stockhecke et al., 2014a), representing the time span from 9.48-111.39 ka BP.

Chronology
The chronology of the Lake Van sedimentary sequence is based on independent proxy records, e.g., high-resolution XRF measurements (Kwiecien et al., 2014), total organic carbon (TOC; Stockhecke et al., 2014a), and pollen data , which were used for the construction of the agedepth model as published in Stockhecke et al. (2014a). By adding radiometric dating techniques, the Lake Van chronol-ogy was correlated by using "age control points", derived from visual synchronization with the GICC05-based NGRIP isotope data for this relevant interval (0-116 ka; NGRIP members, 2004; Rasmussen et al., 2006;Svensson et al., 2008;Wolff et al., 2010). Additionally, a correlation with three 40 Ar / 39 Ar dated onshore tephra layers was implemented in the age-depth model of the composite profile, i.e., the Nemrut Formation (NF) at 32.70 ± 2.55 ka BP, the Halepkalesi Pumice (HP-10) fallout at 61.60 ± 2.55 ka BP as well as the Incekaya-Dibekli Tephra at ∼ 80 ka BP (Stockhecke et al., 2014b;Sumita and Schmincke, 2013). Within the last glacial period, the paleomagnetic Laschamp excursion at ∼ 41 ka BP (Vigliotti et al., 2014) could be identified in the core sequence. Here we want to stress that, among data sets used for visual correlation, pollen data published in  show the glacial-interglacial changes with the largest signal amplitude. However, the age-depth model of Stockhecke et al. (2014b) is based on tuning with the NGRIP event stratigraphy. The correlation points of the Lake Van sedimentary record have been mainly defined by abiotic proxies (i.e., TOC) caused by a higher time resolution of this data set in comparison to the pollen samples available during that time. Even if we present a high-resolution pollen record in this paper, leads and lags between different biotic and abiotic proxies related to climate events have to be taken into account.
To make these pollen counts statistically representative, a minimum of ∼ 500 identified pollen grains per sub-sample were counted for the calculation of terrestrial pollen percentages (100 %), composed of arboreal pollen (AP) and nonarboreal pollen (NAP). Spores of green algae (e.g., Pediastrum boryanum spp., P. simplex, P. kawraiskyi), dinoflagellate cysts, pollen grains of aquatic taxa and damaged pollen grains were excluded from the terrestrial pollen sum and pollen concentration. Percent calculation, cluster analysis to define pollen assemblage zones (PAZ), and construction of the pollen and charcoal diagram ( Fig. 2) was carried out by using TILIA program; version 1.7.16 ( © 1991-2011 Eric C. Grimm).

Stable isotope analysis
Lake Van carbonates consist of a mixture of calcite and aragonite precipitated in surface water. We selected 200 subsamples at the same stratigraphic level which was used for the pollen analysis (20 cm sampling resolution). The freezedried and ground sediment samples were analyzed at the University of Kiel using a Finnigan GasBenchII with carbonate option coupled to a DELTAplusXL IRMS. The isotopic composition is given relative to the VPDB standard in the conventional δ-notation and was calibrated against two international reference standards (NBS19 and NBS18). The standard deviation for reference analyses was 0.06 ‰ for the stable isotope signature.

Profiling measurements
Profiling measurements of the complete Lake Van sedimentary record (Ahlat Ridge site) are published and described in detail by Kwiecien et al. (2014) as well as in the high-resolution study of the last interglacial by Pickarski et al. (2015).
The sediment cores were scanned with an AVAATECH XRF Core Scanner III at MARUM (Bremen). Intensities of major elements (e.g., Ca, K) on sediment cores were collected every 2 cm down-core over a 1 cm 2 area. The row XRF measurements were processed by the Iterative Least square software (WIN AXIL) package from Canberra Eurisys, providing intensity data in total counts (t c ). The Ca / K ratio presented in this paper is unitless.

Palynology
The palynological data of the last glacial are presented in Fig. 2. This sequence can be divided into four pollen assemblage superzones (PAS IIa, IIb, IIIa, IIIb) following the criteria described in Tzedakis (1994 and references therein), which were applied in  for the low-resolution 600 ka long Lake Van pollen record. The PAS IIa and IIb can be further subdivided into six pollen assemblage zones (PAZ; Fig. 2) based on changes in the AP / NAP ratio and changes in the relative frequency of individual taxa. Main characteristics of each pollen zone and sub-zone as well as criteria for defining the lower boundaries are given in Table 1.

δ 18 O bulk and XRF
The oxygen isotopic composition of bulk sediments (δ 18 O bulk ) reflects regional climate changes and local temperature variability. However, the interpretation of δ 18 O bulk is complex as it can be influenced by a number of climate variables, such as air and water temperature, seasonality of precipitation, moisture source and precipitation-to-evaporation ratio. According to Litt et al. (2012Litt et al. ( , 2009) and Lemcke and Sturm (1997), the δ 18 O bulk values of carbonates at Lake Van are primarily controlled by evaporation processes. Furthermore, Kwiecien et al. (2014) and Pickarski et al. (2015) mention that changes in seasonal rainfall have a significant effect on lake water isotope values. During the last glacial, the oxygen isotope signature of carbonates was apparently heavier during interstadials and lighter during stadials (Fig. 3b).
The presented Ca / K record displays a ratio between authigenic carbonate precipitation and siliciclastic material input from the drainage (Kwiecien et al., 2014). According to Kwiecien et al. (2014), the Ca / K ratio shows higher values ascribed to higher amounts of authigenic carbonate during warmer periods (interstadials / interglacial) and lower values related to increasing detrital input during stadials/glacial (Fig. 3c).

Long-term vegetation dynamics at Lake Van
Variations in the orbital configuration of the Earth are responsible for changes in the climate system from one state to another; on millennial timescales, for glacial-interglacial cy-  cles (Berger, 1978;Berger et al., 2007). However, higher frequency oscillations (e.g., Dansgaard-Oeschger events;Dansgaard et al., 1993) are superimposed on the long-term orbitally driven climate dynamics. These abrupt changes of the climate system are not directly driven by orbital forcing, but can be interpreted as transitions between two states of the inter-hemispheric Atlantic Ocean circulation driven by largescale thermal and salinity gradients (e.g., Bond and Lotti, 1995;Cacho et al., 2000Cacho et al., , 1999Chapman and Shackleton, 1999;Hemming, 2004;Hodell et al., 2008;McManus et al., 1999;Rasmussen et al., 2014;Wolff et al., 2010). In particular, changes in the oceanic circulation affected regional and local atmospheric circulation patterns, for example, the strength and position of the westerlies in the Northern Hemisphere, which are responsible for the moisture supply in east-ern Anatolia (Akçar and Schlüchter, 2005;Roberts et al., 2008). According to Jessen and Milthers (1928) and , an interstadial stage is an interval of temporary improved climate within a glacial phase, which has been either too short to permit full expansion of thermophilous trees and/or too cold or dry to reach the climate optimum of an interglacial period in the same region. In comparison, stadial stages correspond to cold / dry intervals marked not only by global but also by local ice re-advances (Lowe and Walker, 1984).
Below, we will discuss only the most pronounced interstadials (e.g., MIS 5c and 5a) and Dansgaard-Oeschger interstadials (AP > 10 %; e.g., DO 19,14,12,8 and 1), here identified on the basis of the Lake Van pollen record (see also . All other "warm / wet" phases with lower expansion of temperate trees (AP < 10 %) are not explicitly mentioned in the following section.

MIS 5e-5a
The last interglacial at Lake Van (MIS 5e; 131.2-111.5 ka BP; Pickarski et al., 2015) is characterized by an oak steppe-forest during the climate optimum (129.1-124.1 ka BP), while coniferous species (e.g., Pinus) dominated the late interglacial period between 124.1 and 111.5 ka BP at the minimum peak in summer insolation (Pickarski et al., 2015;Fig. 3a, e). The expansion of drytolerant and/or cold-adapted Pinus (probably Pinus nigra) going along with a reduction of warm-temperate species (e.g., deciduous oak) demonstrates cooler temperatures and summer-dry conditions during the late interglacial period (Pickarski et al., 2015). In this regard,  argued that positive summer temperature anomalies and negative winter temperature anomalies during the late interglacial lead to a strong continentality, more precisely, to strong temperature variations with still low moisture availability in eastern Anatolia. MIS 5d is marked by a significant expansion of steppic herbaceous plants at the expense of wooded landscape, which signals a considerable climate deterioration from ∼ 111.5-107.8 ka BP (Herning stadial; AP < 20 %; Fig. 3e). The regional climate at Lake Van was characterized by a strong seasonal moisture deficiency (δ 18 O bulk values > 2 ‰, Fig. 3b). Cold and/or dry climatic conditions with marked seasonality in precipitation during the stadial were responsible for poor soil development and enhanced erosion of regional material as recorded by the Ca / K ratio (Fig. 3c).
An abrupt shift in most proxies (pollen data, δ 18 O bulk , Ca / K ratio) displays the onset of two pronounced interstadials at 107.8-87.2 ka BP (Brørup interstadial; MIS 5c) and at 84.9-77.5 ka BP (Odderade interstadial; MIS 5a; Fig. 3). The rapid decrease in steppic herbaceous elements (e.g., Chenopodiaceae) and the slow increase of summer-green oaks suggest that climate in eastern Anatolia became progressively warmer and wetter during these interstadials. Similar to the late interglacial stage (124.1-111.5 ka BP; Pickarski et al., 2015), the Brørup interstadial recorded a slight climatic amelioration that continued with the predominance of cold-and/or summer-dry adapted conifers (Pinus > 60 %, Fig. 3e; Pinus concentration > 20 000 grains cm −3 , Fig. 2). Changes in seasonal rainfall inferred from depleted δ 18 O bulk values (up to −2 ‰) to more positive oxygen isotope signatures (> 1 ‰) and generally lower winter temperatures have a decisive negative impact on moisture-requiring thermophilous plants such as deciduous oaks (< 10 %, Table 1; Fig. 3e). Therefore, the likely occurrence of deciduous Quercus in higher altitudes, for instance at southern slopes of the Bitlis Massif, would be caused by increasing orography-related precipitation amounts ). An open oak steppe-forest, which was predomi-nant during the MIS 5e at Lake Van, did not become dominant at any time during the last glacial ( Fig. 3e; Pickarski et al., 2015).
The new high-resolution microscopic charcoal data show that fire frequency had an immediate response to climate variability (Fig. 3d). According to previous high-resolution pollen studies by Wick at al. (2003) and Pickarski et al. (2015), rising global temperature and increased moisture availability leads to higher vegetation productivity (e.g., higher vegetation density due to the spread of warmtemperate grasslands) that correlates to considerably more fuel for burning during interstadials (> 1000 charcoal particles cm −3 ; Fig. 3d). During stadials, lower microscopic charcoal concentrations (between 300 and 850 particles cm −3 ) characterizes an open dry desert-steppe landscape with low vegetation density (e.g., MIS 5b; Daniau et al., 2010;Sadori et al., 2015;Vanniere et al., 2011;Fig. 3d).
After a short-term climatic deterioration between ∼ 87.3 to 84.9 ka BP (MIS 5b; Rederstall stadial, AP < 10 %) characterized by a similar expansion of steppic herbaceous plants as documented for MIS 5d (Herning stadial), the spread of deciduous oaks defined the beginning of the MIS 5a. Environmental conditions of the Odderade interstadial (MIS 5a; ∼ 85-77 ka BP) are difficult to resolve due to the eruption of the Incekaya-Dibekli volcano at ∼ 80 ka BP (Sumita and Schmincke, 2013). The fragmentary documentation of the vegetation signal, primarily due to the respective admixture of pyroclastic material (Stockhecke et al., 2014b), complicates the study of vegetation and climate evolution at Lake Van. Nevertheless, the briefly rising AP level (AP > 10 %) consisting of deciduous Quercus, Betula and the sporadic occurrence of Pistacia cf. atlantica, points to short-term favorable climatic conditions and an increased moisture availability at the beginning of MIS 5a (Figs. 2, 3e). Relatively high fire intensity (charcoal concentration up to 2000 particles cm −3 ; Fig. 3d) and lower detrital input (Fig. 3c) support an advancing vegetation cover due to warmer climatic conditions. However, Pinus, which was dominant during MIS 5e and MIS 5c, is no longer growing in the vicinity of the lake (less than 4 %; Pinus concentration < 200 grains cm −3 , Figs. 2, 3e). The shift from depleted oxygen isotope signature (−1.90 ‰) to more positive values (1.81 ‰, Fig. 3b) confirms the reduction in precipitation and/or increasing evaporation throughout the MIS 5a.
In general, the abrupt variability of temperate AP from Lake Van and δ 18 O NGRIP values are more or less synchronous (Fig. 4). Leads and lags between the proxy records, illustrated in detail in Fig. 5, are difficult to assess due to their heterogeneous resolution. In any case, we cannot expect a perfect matching between biotic and abiotic proxies related to climate events due to their different response time. In addition, the lack of correspondence between the pollen signal and the timing of some DO events could also be explained by uncertainties in the current age-depth model (see Stockhecke et al., 2014a). Still, as expected from various eastern Mediterranean pollen records, the Lake Van pollen record documents that temperate taxa tend to reach their maxima after the onset of a warming phase and, therefore, lag behind the Ca / K increase, which responds immediately to climate changes (Fig. 5).
The longest and most pronounced variability of tree populations in the Lake Van pollen record is shown during the MIS 3 (∼ 61-28 ka BP), suggesting a rapid alternation of warmer / wetter interstadials and cooler / drier stadials. High-amplitude variations in the Ca / K ratio of Lake Van sediments indicate changes in erosion of regional material, due to unstable environmental conditions in the catchment area. Larger interstadials such as DO 19,14,12,and 8 are indicated by peaks in the Ca / K ratio, which are concurrent with AP maxima and increased regional fire frequency (Fig. 5). As described above, this millennial-scale variability in the proxy record was indirectly modulated by orbitaldriven changes and variations in the atmospheric circulation of the Northern Hemisphere (e.g., Cacho et al., 2000Cacho et al., , 1999Chapman and Shackleton, 1999;Hemming, 2004;Hodell et al., 2008;McManus et al., 1999;Rasmussen et al., 2014;Wolff et al., 2010). Consequently, the vegetation and environmental conditions at Lake Van respond to abrupt shifts in temperature and moisture availability related to the position and strength of the westerlies, which brought mild and humid conditions from the North Atlantic into the eastern Mediterranean region (Akçar and Schlüchter, 2005;Allen et al., 1999;Fletcher et al., 2010;Müller et al., 2011;Roberts et al., 2008). The Lake Van δ 18 O bulk signature supports the suggestion of favorable environment by the receipt of isotopically depleted meltwater supply and/or increased precipitation between ∼ 57 and 54 ka BP (δ 18 O bulk from −1.21 up to ∼ 1 ‰; Fig. 3b). Furthermore, we propose that the interval of constantly heavier δ 18 O bulk values during DO 14 to 12 reflect higher evaporation at Lake Van.
The most pronounced DO stadials, i.e., the stadials preceding DO 17,12,8,4, and DO 1 refer to Heinrich events (HE) 6, HE 5, HE 4, HE 3, and HE 1 (Fig., 4a; Bond and Lotti, 1995;Bond et al., 1993;Heinrich, 1988). The reduction of oceanic heat transport (weakening or shut down of the AMOC; Böhm et al., 2014;Bond and Lotti, 1995;Bond et al., 1993;Broecker, 1994;Cacho et al., 2000Cacho et al., , 1999Chapman and Shackleton, 1999) led to significantly cooler Mediterranean SST's and climatic conditions in the Mediterranean region (e.g., Allen et al., 1999;Fletcher et al., 2010;Müller et al., 2011;Sánchez Goñi et al., 2002;Tzedakis, 2005;Tzedakis et al., 2004). According to Kwiecien et al. (2009), a decreasing atmosphere-sea-surface thermal gradient of the Mediterranean Sea would have caused a reduction in the frequency and strength of storm tracks, which were responsible for an intensifying aridity in the eastern Mediterranean region. However, since tree populations were already limited at Lake Van during the last glacial (< 10 % AP), the pollen signal is relatively insensitive to severe climatic deterioration such as Heinrich events. Hence, both types of stadials, Heinrich events and DO stadials, lead to a similar reduction of tree taxa (mainly deciduous Quercus), and therefore cannot be clearly distinguished from an average cooling event (Fig. 4b). An exception might be the HE 5 (∼ 49 ka BP; Fig. 4a). A collapse of AP taxa from 10 % to less than 2 % (Fig. 4b) and a short-term detrital supply (Fig. 3c) show that the climate of HE 5 was as cold and dry as the glacial maximum of MIS 4 between 70 and 60 ka BP.
Between ∼ 28 and 14 ka BP (MIS 2), very low AP percentages (<10 %; Fig. 3e) and a decreasing fire frequency (Fig. 3d) without significant fluctuations underline considerably stable climate conditions but pronounced regional cooling and aridity in comparison to MIS 3. It led to an overall reduction in terrestrial biomass production and thus to a decrease in fuel availability for burning in the catchment  TM  T5  pM  p5  fM  f5  4M  45  5M  55  6M  65  7M  75   TM  T5  pM  p5  fM  f5  4M  45  5M  55  6M  65  7M  75  AgeVykaVBP%   ya%VSVδ T8 OVNGRIP  HE  DO   T  T   p  p   f  4  5  6  f 4  5 6 7 8  TM TT Tp Tf  T4  T6  T8  T9 yb%VSVLakeVVanVythisVstudy% APVys%VxVTM APVys%V area. Furthermore, the presence of Juniperus, indicative for an unstable soil cover, enhanced minerogenic input into the lake (low Ca / K; Fig. 3c) and a drop in δ 18 O bulk values (up to −1.6 ‰; Fig. 3b) support the assumption of wide open plains around Lake Van. Insolation changes became the major driver (e.g., low summer insolation, Fig. 3a) and leads to a cooling of global climatic conditions, which results in a maximum ice extent of the Northern Hemisphere during the last glacial maximum (LGM; Cacho et al, 2000Cacho et al, , 1999Chapman and Shackleton, 1999;NGRIP members, 2004;Rasmussen et al., 2014;Sánchez Goñi and Harrison, 2010;Wolff et al., 2010).
After the LGM (at 21 ± 2 ka; Tzedakis, 2007 and references therein), high summer insolation (Fig. 3a), decreasing global ice volume (Wolff et al., 2010), the resumption of the westerly activity, enhanced precipitation (depleted δ 18 O bulk values; Fig. 3b), and locally rising temperatures in eastern Anatolia promoted an expansion of an oak steppe-forest at Lake Van (DO 1 at ∼ 14 ka BP; synonymous with the Bølling-Allerød warm period; Figs. 3e, 4). However, the Late Glacial-Holocene transition (Termination I) was interrupted by the impact of the Younger Dryas (YD) climate reversal between ∼ 12.8-11.6 ka BP (Litt et al., 2009;Wick et al., 2003). This dramatic event is characterized by the δ 18 O bulk peak (4.46 ‰; Fig. 3b Fig. 4e) in the Tenaghi Philippon area (Müller et al., 2011). The marine pollen record from the south-eastern Levantine Basin (Fig. 4c) also points to increasing AP percentages during ∼ 56.0 and 44.5 ka BP (Langgut et al., 2011). However, there are some differences between the pollen records. More developed forests exist around the central and eastern Mediterranean Sea and the Black Sea compared to the Near East. Despite the intensive aridification in eastern Anatolia during glacial, the vegetation composition of Lake Van and the Levantine Basin differs from the terrestrial Mediterranean pollen records. Firstly, drought-sensitive taxa such as Ulmus, Carpinus betulus and Fagus were frequently present in Italy, e.g., at Lago Grande di Monticchio, Valle di Castiglione, Stracciacappa and Lagaccione (Allen et al., 1999;Follieri et al., 1998) even during stadials. Secondly, the high-resolution Tenaghi Philippon and Ioannina sequences (Müller et al., 2011;Tzedakis et al., 2002) show that thermophilous trees (e.g., deciduous Quercus) increased rapidly during each interstadial without migrational lags (Fig. 4e). It suggests that these sites were better suited in sustaining refugial temperate tree population due to the effects of orographic precipitation (e.g., Müller et al., 2011;Tzedakis et al., 2002). Thirdly, both the diversity and the low amplitude of variations in temperate tree taxa of the eastern Mediterranean pollen records (Lake Van and Levantine Basin; Fig. 4b, c) indicates greater distances and/or slow migration rates from refugia during the glacial interval. Such areas might be found in the south and south-east Black Sea Mountains (Euxinian vegetation) and the Caucasus mountains (Hyrcanian vegetation), which receive increased atmospheric moisture and higher orographic precipitation from the Black Sea (Bottema, 1986;Leroy and Arpe, 2007;Shumilovskikh et al., 2012). Especially the Black Sea region was characterized by mean winter temperatures close to or above 0 • C during the last glacial (Shumilovskikh et al., 2014).
Moreover, the different vegetation development in the Mediterranean region demonstrates a west to east vegetation gradient from an open temperate forest (including evergreen species) in the central (Allen et al., 1999;Follieri et al., 1998) and eastern Mediterranean (Müller et al., 2011) to a semiarid grassland in the Near East during DO interstadials. In eastern Anatolia, the availability of precipitation is the limiting factor for the establishment of an open oak steppe-forest. This moisture gradient reflects an increasing continental affect and a decreasing influence of changes in the atmospheric circulation of the Northern Hemisphere.

Conclusions
1. By using a range of paleoenvironmental proxies, we were able to detect subtle climate changes even during generally cold and arid glacial phases. This study illustrates the great potential of Lake Van as an archive of eastern Mediterranean climate and environment over the entire Quaternary.
2. Our new palynological results show the climatic teleconnection between Lake Van and the Atlantic Ocean. It reflects the complex underlying drivers of high frequency regional climate and environmental variability caused by seasonal insolation changes, ice sheet dynamics, and the ocean circulation in the North Atlantic.
3. The comparison with central and eastern Mediterranean and Black Sea pollen sequences reveals a west to east gradient of decreasing moisture in the Mediterranean region due to an increasing continental effect and a reduction of the atmospheric impact of the North Atlantic Ocean.
4. Dansgaard-Oeschger cycles are clearly recognized in the Lake Van pollen record and by other abiotic proxies. Interstadials are characterized by the spread of temperate vegetation (e.g., deciduous Quercus) suggesting regional moisture availability. Stadials can be recognized by the dominance of steppe elements (e.g., Artemisia, Chenopodiaceae) pointing to cold temperature and an increasing aridity.
5. Heinrich events cannot clearly be distinguished from an average cooling event (stadials). They show a similar impact on the vegetation and the environment.
6. The supply of detrital material seems to respond directly to changes in the vegetation composition, e.g., the terrestrial supply is low (high authigenic carbonate precipitation) when the vegetation cover in the catchment area is dense. In contrast, an open landscape favors a physical erosion including local detrital and dust input.
7. Moreover, the fire frequency at Lake Van indicates an immediate link to climate changes. The fire frequency is high when global temperatures and regional moisture levels increase providing a higher terrestrial biomass production.

Data availability
The complete pollen data set is available on the PAN-GAEA database (http://doi.pangaea.de/10.1594/PANGAEA. 853814), and the microscopic charcoal record can be found on the Global Charcoal Database (GCD; www.paleofire.org).