Introduction

In the eastern Mediterranean region, interactions between weather systems and the local topography allow for steep gradients in temperature and wetness. These gradients result in a complex mosaic of biomes, rich biodiversity, and a particular sensitivity to changes in climate, such as during the transition from the last glacial to the present (Holocene) interglacial. This transition occurred in multiple stages, the most pronounced being Heinrich Event 1 (contemporary with the Oldest Dryas stadial), the Bølling–Allerød interstadial, and the Younger Dryas stadial (Comes and Kadereit 1998; Allen 2003; Goldreich 2003; Médail and Meyers 2004; Hewitt 2011; Kushnir et al. 2017; Robins et al. 2023). Reconstructions of the timing and nature of millennial-scale climate variability within glacial terminations, especially from low and mid-latitude terrestrial environments far from polar ice sheets and coastlines, provide valuable insight into how global climate forcing is expressed on regional and local scales (Moulin et al. 2022; Vadsaria et al. 2022).

In the Southern Levant, numerous paleoenvironmental studies have reconstructed the environmental conditions of the last deglacial transition though they often paint conflicting pictures of environmental and hydroclimate change during the late Pleistocene. While early studies focused on interpreting sensors such as shoreline positions, carbonate deposition, lake levels, and vegetation patterns as records of mean annual precipitation or temperature (Horowitz 1973; Stiller and Hutchinson 1980; Baruch and Bottema 1991), more recent studies recognize that the mechanisms by which these sensors respond to environmental forcing are more nuanced. Here, many of the environmental shifts previously ascribed exclusively to changes in mean annual precipitation or temperature are now recognized to be possible via changes in, for example, precipitation seasonality or wind regime (Enzel et al. 2008; Roskin et al. 2011; Vardi et al. 2018; Miebach et al. 2019; Ludwig and Hochman 2022; Robins et al. 2023). For this reason, studies utilizing diverse methods to reconstruct different parts of the Late Pleistocene Southern Levant climate system have become increasingly important in clarifying our understanding of the millennial and centennial scale changes of the most recent glacial termination, a time period that served as the backdrop for profound changes in human lifeways, including the emergence of sedentism and the development of agriculture (Goring-Morris and Belfer-Cohen 1997; Tchernov 1997; Grosman and Belfer-Cohen 2002; Kushnir et al. 2017; Belfer-Cohen and Goring-Morris 2020).

Within the archeological record of the eastern Mediterranean region, the transition from the Last Glacial Maximum (“LGM”) to the Holocene Interglacial occurred approximately 25/23—11.7 cal ka BP (calibrated millenia before present, where the present refers to the year 1950) and corresponds to the period known as the Epipaleolithic. Over the course of the Epipaleolithic, a shift can be seen in the region from nomadic cultural entities to semi-sedentary and finally sedentary communities. Cultural phases in the Mediterranean climatic zone of northern Israel include the Masraqan, Kebaran, and Nizzanan (Early Epipaleolithic), Geometric Kebaran (Middle Epipaleolithic), and Natufian (Late Epipaleolithic cultures). These cultures became increasingly sedentary, transitioning from bands of a few dozen mobile hunter-gatherers to small hamlets of up to one hundred individuals. Settlement facilitated the development of increasingly complex methods to modify local environments to encourage growth of plant and animal species of economic interest (Belfer-Cohen and Goring-Morris 2011; Zeder 2011), although hunter-gatherers in the Levant may have begun to significantly impact the landscapes in which they lived as early as the LGM (Ramsey et al. 2015).

While the impact that late Pleistocene climate changes had on human lifeways is debated, environmental reconstructions of the last deglaciation in the Levant provide necessary context for understanding the dramatic changes in human lifeways and subsistence strategies of the time. Paleoclimate-proxy records produced from sediment cores and speleothems from the Eastern Mediterranean region provide valuable overviews of the regional air temperature, precipitation histories, wind regimes and water availability. However, local conditions at archeological sites are best inferred from sediments collected from the sites themselves or their immediate surroundings, as the region’s geomorphology is often the most important control of local (hydro)climatic conditions (Cornuault et al. 2016; Vardi et al. 2018; Bar-Matthews et al. 2019; Miebach et al. 2019; Orland et al. 2019; Langgut et al. 2021; Matthews et al. 2021). Paleoenvironmental reconstructions for Levantine Epipaleolithic sites not only shed light on the environmental conditions experienced by the sites’ prehistoric inhabitants, but when combined with robust chronological frameworks, they also sharpen our understanding of the timing of primary cultural and environmental developments and of regional climate change patterns, since reliable chronologies are crucial for the comparison of conditions between contemporary sites.

Here we present a comprehensive chronostratigraphic model and proxy paleoenvironmental data for Jordan River Dureijat (“JRD”), an open-air excavation in northern Israel and one of the only sites in the region to document the entire Epipaleolithic period. Furthermore, the resolution of this sedimentary record makes it unique in its ability to record centennial-scale changes in the environment and hydroclimate of the Southern Levant during this time. Although JRD was first described more than 20 years ago, until now no comprehensive examination of the sediment geochemistry, grain size or depositional environment has been attempted. Early descriptions of the site describe JRD as a lakeshore environment where the archeological artifacts were deposited subaerially. However, these interpretations rely on limited observations of only the site’s uppermost sediments (Marder et al. 2015). In this study, we describe JRD’s depositional environments and establish its chronostratigraphic framework. Specifically, we address the following questions: What was the nature of the waterbody at JRD during the late Pleistocene? At which times was the JRD site utilized by the various cultural entities? Finally, we compare and discuss our inferences within the context of regional studies to improve the current understanding of Levantine hydroclimate conditions during the late glacial, including the importance of northern precipitation sources to the hydrological balance of Lake Lisan and the impact of changing wetness regimes on local populations.

Study area

The Hula Valley

The Hula Valley is a pull-apart basin located along the Dead Sea Transform in northern Israel (Fig. 1; Garfunkel 1981; Heimann 1990; Rybakov et al. 2003). With an area of approximately 150 km2 and a valley floor elevation of 60–70 m above sea level (masl), the Hula Valley is both the smallest and highest of the Jordan Valley basins within Israel, as well as the northernmost. The basin is bordered by the Naftali Mountains (Cretaceous sedimentary rocks, 500–900 masl) to the west and the Golan Plateau (Pleistocene volcanics, 400–1200 masl) to the east. The Hermon range (up to 2800 masl) constitutes the northeastern limit of the valley, while in the north and northwest the Hula Valley is contiguous with the Hasbani River Valley. The southern boundary of the Hula Valley is defined by the uplifted Korazim block (basalt, 300 masl; Fleischer 1968; Horowitz 1973; Heimann 1990; Belitzky 2002; Heimann et al. 2009).

Fig. 1
figure 1

A Location of the JRD excavation [33° 1′ 19.7″ N, 35° 37′ 42.1″ E] and other locations mentioned in the text, labelled as follows: (1) Sediment Core SL112 (2) Sediment Core MD84-632 (3) Soreq Cave (4) Masada M1 section (5) northwest Negev dunefield. B Simplified geologic map of the Hula Valley and surrounding area. C Map of the historical Hula Lake area showing the lake surface configuration from 1935 (mean water level 70 masl). Contour interval is 25 m. Geologic and regional maps (panels A and B) are redrawn after Horowitz (1973). Hula Lake area topography and hydrology (panel C) from the Survey of Palestine 1:20,000 map sheets 20–27 and 21–27 (1941) and 1:100,000 map sheet 3 (1947)

Full size image

The area of the Hula Valley catchment is 1470 km2, with precipitation falling in the surrounding mountain ranges entering the valley primarily as perennial fluvial flow via the Dan, Banias and Hasbani rivers. Together, these are the major tributaries that form the uppermost course of the upper Jordan River, which in turn along the lower Jordan River, drains to the Dead Sea. Additionally, four aquifers discharge a total of more than 300 million cubic meters of groundwater annually to the Hula Valley by way of springs located along the valley’s margins and in the Hermon Range (Babad et al. 2019; Fig. 1c). This water naturally collects in the Hula Basin, where it has historically formed a shallow freshwater lake (Lake Hula). The valley is drained by the Jordan River as it flows through the Jordan River Gorge en route to the Sea of Galilee. In the 1950s this outflow was widened and deepened in order to drain Lake Hula. Extensive construction related to both the drainage of the valley, the channelization of the Jordan River as well as subsequent agricultural/light industry activity, has significantly transformed the nature of the valley. However, two small nature reserves (Agmon HaHula and the Hula Nature Reserve; Fig. 1b) have reconstructed environments similar to those of the pre-drainage valley, including shallow waterbodies surrounded by wetlands (Zohary and Orshansky 1947; Hambright and Zohary 1998; Inbar 2002; Payne 2012). At the time of its drainage, the Hula Lake was 5.3 km long and 4.4 km wide and occupied the southern third of the valley, while the central third of the valley was dominated by extensive wetlands and crossed by a network of streams. In the first half of the twentieth century, the surface area of the lake fluctuated between 12 km2 in the summer and 14 km2 in the winter, corresponding to an approximately 1.5-m rise in lake level during the wet season (Dimentman et al. 1992; Hambright and Zohary 1998).

The Plio-Pleistocene infill of the Hula Basin comprises interfingered alluvial sediments, conglomerates, lacustrine marls, and basalt flows, and is estimated to be up to 4600 m thick at its depocenter, now the location of Kibbutz Hulata (Rybakov et al. 2003; Fig. 1c). The Pliocene and Pleistocene sediments have been described as lacustrine and fluvial deposits with intercalated basalt flows, indicating that lakes have occupied this basin at least intermittently since the Neogene (Picard 1963; Horowitz 1973; Kafri et al. 1983). Some researchers suggest that the absolute elevation of the lake surface has remained unchanged, with the sediment-accumulation rate reflecting only subsidence and compaction (Kafri et al. 1983). Others propose that the lake volume increased and decreased in response to regional environmental changes coinciding with global climate cycles (Horowitz 1971, 1973; Stiller and Hutchinson 1980; Weinstein-Evron 1983). The position of the shorelines, the distribution of sedimentary environments and facies, and the elevation of the lake surface is probably controlled by a combination of climate and tectonics (Kafri and Lang 1987).

The modern climate of the Hula Valley is temperate with dry, hot summers (Köppen classification Csa). At Kfar Blum, located 17 km northwest of JRD and at a similar elevation, the Israeli Meteorological Service has measured average daily temperatures of 11.7 °C and 28.0 °C for January and July, respectively (calculated for the years 1995–2009). Modern precipitation is mainly sourced from the Eastern Mediterranean (Goldreich 2003; Kushnir et al. 2017). Mean annual precipitation is 507 mm, with 94% of the precipitation falling November–April (Israeli Meteorological Service; calculated for the hydrological years September 1990–2019). Typical vegetation of the Mediterranean environment at this elevation includes Quercus ithaburensis—Pistacia atlantica forests and batha landscapes characterized by grasses and shrubs (e.g., jujubes, styrax, and tamarisk) growing in either alluvial, terra rossa or rendzina soils (Zohary and Orshansky 1947; Horowitz 1979; Zohary 1982; Weinstein-Evron 1983; Langgut et al. 2021). During the wettest months of the year, JRD is fully flooded by the Jordan River. Every spring, as the river recedes, a dense carpet of vegetation grows over the site, the roots of which penetrate the uppermost sediments of the investigated sequence.

The JRD site

The JRD site is located near the southern margin of the Hula Valley, on the east bank of the Jordan River at an elevation of approximately 60 masl (Fig. 1). Archeologists previously studied this stretch of the Jordan River during the excavations of two older Paleolithic sites located along its banks: the Acheulian site of Gesher Benot Ya’aqov (400 m south of JRD; Goren-Inbar and Belitzkey 1989) and the Mousterian site Nahal Mahanayeem Outlet (600 m to the north; Sharon et al. 2008).

At JRD, stone tools dating to the Epipaleolithic were first discovered in 1999, embedded in piles of sediment removed during the course of routine river-dredging operations (Sharon et al. 2002a, b). Site surveys and preliminary excavations were carried out in 2002 and 2014 (Marder et al. 2015), followed by comprehensive excavations each summer from 2015 to 2021, with a hiatus related to the SARS‑CoV‑2 pandemic during the summer of 2020 (Sharon et al. 2020). Excavation at JRD exposed a rich assemblage of bone and stone tools, of which fishing equipment is among the most significant finds (Sharon et al. 2020; Pedergnana et al. 2021). This site documents almost the entire Epipaleolithic of the Levant and, in particular, three of its most prominent lithic traditions: the Masraqan, Natufian, and Geometric Kebaran (Sharon et al. 2020).

The JRD site comprises multiple open-air excavation areas, of which the largest and most extensively excavated is known as Area B (Fig. 2; c. 36 m2). Here, the preservation of stone and bone artifacts document repeated visits to this site throughout the terminal Pleistocene (Sharon et al. 2020; Pedergnana et al. 2021). The archeological stratigraphy of the site has been described previously and includes seven artifact-bearing layers (Sharon et al. 2020). Many of these layers can be traced throughout the site, while some are present in only a limited spatial extent (Sharon et al. 2020). Due to the water-logged nature of the site, botanical and microfaunal remains including pollen, bones and plant macrofossils are well-preserved in the site’s sediments (Sharon et al. 2020; Pedergnana et al. 2021; Langgut et al. 2021).

Fig. 2
figure 2

A Photo of Jordan River Dureijat Excavation Area B, highlighted in gray. Dotted lines demarcate the grid used in the excavation of the site, where the sides of each square measure one meter. The location of the sampled section along the site’s east wall is highlighted in dark gray. The direction of the flow of the Jordan River is shown using a hollow arrow. B Drone photo of the JRD excavation. C Stratigraphy of the east wall of JRD excavation Area B showing beds 7–13

Full size image

Materials and methods

Stratigraphic logging of the JRD site

On the exposed walls of JRD excavation Area B, several distinct stratigraphic units are clearly visible, including all seven of the cultural horizons described in Sharon et al. (2020). The exposed east wall of the excavation area was carefully logged at the conclusions of the 2016 and 2017 excavation seasons (Fig. 2c). The sedimentogic units are described and defined in this study on the basis of changes in the sediments visible at outcrop scale, including changes in the type and concentration of shells in the sediments and boundaries visible between the beds. This differs from the method used to define the site’s archeological stratigraphy, described in Sharon et al. (2020).

Sediment sampling

A 2.2-m thick stratigraphic section located along the east wall of excavation Area B (termed the “East Wall Primary Section” (EWPS)) was sampled continuously at a vertical resolution of 1 cm, resulting in the acquisition of 228 contiguous samples. All were stored in plastic bags and dried overnight at 35 °C. Aliquots of each sample were then removed and used for the grain-size and geochemical analyses (performed at Lanzhou University, China) and paleontological analyses (University of Iceland).

Grain-size analysis

Particle-size distributions of each of the 228 samples were measured using a Malvern Mastersizer 2000 coupled to a Hydro2000G wet dispersion unit. Measurements were made in manual mode while the stirrer was operating at 700 RPM, the pump at 1750 RPM, and the inline ultrasound operating at 80% of maximum power. Prior to measurement, each c. 0.3 g sample was treated with 10% hydrogen peroxide and 12% hydrochloric acid to remove organic matter and carbonate material, respectively. These reactions were performed while the samples were positioned on a 220 °C hotplate and stirred continuously. Immediately prior to measurement, 10 mL 25 g L−1 sodium hexametaphosphate solution was added to each sample to break apart any aggregates that remained.

The model “General purpose-normal sensitivity” was used to transform the intensity of the scattered energy detected at each sensor to spherical volume. This model, which utilizes the principles of the Mie theory of light scattering, relies on user-assumed refractive and absorptive indices of the sample material, which we have defined as 1.52 and 0.1, respectively.

The curve-fitting tool CFLab (Wu et al. 2020) has been used to decompose the sediment-grain-size distributions. This MATLAB application performs curve-fitting on particle-size distributions using Weibull Probability Distribution Functions to determine the shapes and relative contributions of multiple subpopulations.

Total organic carbon measurements

Total organic carbon (TOC) was measured using a modification of the methods presented in Schollenberger (1927) and Walkley and Black (1934). In this method, crushed samples of known weight were placed in test tubes, treated with equal parts 0.8000 N potassium dichromate solution and concentrated sulfuric acid and heated in an oil heater to 170 °C for 5 min. The TOC in each sample is removed via the reaction with sulfuric acid and potassium dichromate to create carbon dioxide, which is released to the atmosphere. The amount of potassium dichromate consumed in this reaction is then quantified by titration with 0.2 N ferrous sulfate solution.

Carbonate-content determination

To determine the calcium carbonate (CaCO3) content of each sample 50–80 mg of pulverized, homogenized material was reacted with 5 ml 10% HCl in a sealed Büchner flask, where the HCl was injected via a syringe connected to the hose barb and the top of the flask was sealed with a rubber stopper attached to a pressure gauge. The reaction of carbonate in the samples with the hydrochloric acid produces carbon dioxide gas. The resulting pressure increase was then converted into the carbonate concentration of the sediment sample.

Paleontological analysis

For samples from Unit II, approximately 30 g of material from samples every eighth centimeter were weighed and soaked in a weak hydrogen-peroxide solution overnight before being sieved over 63, 125, 250, and 1000 µm screens using tap water. The sieve residues were then dried overnight at 60 °C and examined under a low power stereomicroscope to record plant and animal remains such as charcoal, seeds, gyrogonites, ostracods, fish teeth, bones, and foraminifera. Additionally, 2–4 g of each sample from Unit II were dried, treated with peroxide, and sieved using a 250 µm screen; these samples were examined for the presence of ostracod valves. Additional samples from Unit I were treated and examined in the same ways but at a lower resolution.

Radiocarbon dating

We chose to date charcoal from terrestrial plants growing locally in the catchment to avoid reservoir effects resulting from the large amount of older carbonate in the catchment. The charcoal fragments were examined and identified to genus level using a stereomicroscope or characterized by measuring the δ13C of the material.

In total, 24 charcoal specimens were dated via accelerated mass spectrometry radiocarbon dating. Of these, fourteen come from the east wall of Area B, while nine were recovered from other locations within Area B. Sixteen samples were submitted to the Poznan Radiocarbon Laboratory and measured at the AMS Laboratory of Adam Mickiewicz University and eight samples were dated at Beta Analytics in Miami, USA. Prior to measurement, all samples were crushed and washed with hydrochloric acid and sodium hydroxide to remove carbonates and organic contaminants, respectively.

Establishment of the chronostratigraphic model

In the creation of the chronostratigraphic model of JRD’s Area B (Fig. 3), we used outlier and depositional modeling tools in OxCal v. 4.4 (Bronk Ramsey 2008, 2009a). Due to the uneven thicknesses of the layers over the excavation area and their lack of internal structure, we chose to include only radiocarbon data from the exposed east wall of the site. All charcoal included in the chronostratigraphic model comes either from the EWPS or from laterally equivalent sediments less than 30 cm away. Outliers were identified and removed with OxCal‘s General Outlier Model using a student’s t-distribution and an outlier prior probability of 20% (Bronk Ramsey 2009b). Of the twelve dates recovered from Unit II, two were identified as outliers and excluded from the depositional model: Beta-457488 and Beta-457486 (posterior probabilities of 73% and 92%, respectively). The depositional model uses the calibration curve IntCal20 (Reimer et al. 2020) to produce a likelihood distribution for the calibrated age of each sediment sample. Here, a Poisson sequence was utilized to model sediment deposition against depth and we suggested a prior of one depositional event per centimeter and allowed for two orders of magnitude variation in each direction (K0 = 1; Log10(k/k0) = − 2 to + 2). For the deepest 12 cm of the sequence, linear interpolation was used to estimate the ages of samples stratigraphically lower than the oldest date. This interpolation uses the average sedimentation rate calculated for the 20 cm of material immediately overlying the oldest date, equal to 43 cm ka–1 [2325 years per meter] and is indicated by a dashed line in the chronostratigraphic model shown in Fig. 3.

Fig. 3
figure 3

Stratigraphy and chronology of the JRD site. The age-depth relationship of the sediments of the East Wall Primary Section is shown in black with uncertainty in grey; solid black lines indicate depositional modeling results from OxCal; dashed line indicates linear interpolation. Calibrated radiocarbon dates from the vicinity of the East Wall Primary Section are shown along with their 2-sigma ranges: grey circles show dates rejected on the basis of OxCal outlier modeling; black circles are dates included in the depositional model. Lithostratigraphic units and beds as defined in this study. Archeological units and cultural horizons as defined in Sharon et al. (2020)

Full size image

Results

Stratigraphy

The JRD outcrop is composed of two lithostratigraphic units: a lower unit (Unit II) composed of dark-colored, fine-grained mud which includes the cultural horizons of the archeological site, and an upper, structureless unit of lighter-colored, coarser material (Unit I). The boundary between these two units is found at approximately 57.8 masl on the east wall of the excavation area and is sharp and uneven (Fig. 4).

Fig. 4
figure 4

Results of the geochemical and grain-size analyses (three-point running means) plotted against the depth of the sedimentary samples. Cultural horizons as they are defined in Sharon et al. (2020) are labeled on the log to the left; lithologic beds are labeled to the right. Ostracod valves (not shown) were found in all samples; hollow circles in the microfossil column indicate studied samples for which no other microfossils were found

Full size image

Unit II comprises 13 distinct beds of very dark grey sediment (Munsell color 2.5Y 3/1, wet measurement) distinguishable primarily by changes in the concentration of clam (genus Unio) and gastropod (genus Melanopsis) shells. Odd-numbered beds contain few to no mollusk remains while each even-numbered bed contains up to 30% whole shells by volume. All beds are structureless and—where mollusk remains are present—the shells are embedded within a muddy matrix and do not form clast-supported structures or layers, such as shell mounds or lag deposits. Within Unit II, many of the even-numbered beds are correlated with the archeological horizons of the JRD site (Fig. 4). Contacts between the beds of Unit II are sharp and well-defined.

In the overlying Unit I, the sediments are visibly coarser and lighter in color (dark olive brown: Munsell color 2.5Y 3/3, wet measurement). These sediments contain fragmented shells and roots are seen intruding from above (Fig. 4). Unit I comprises a single structureless bed.

Sedimentology

The JRD sediments are low in organic carbon and calcium carbonate content and comprise mostly silts and sandy silts (Fig. 4). In lithostratigraphic Unit II, TOC concentrations vary between 1 and 11% (average = 2.8%; Table 1). Total organic carbon content is generally highest in beds 3, 5, 7 and 9 and reaches local minima in beds 2, 4, 6, 8, 10 and 12 (Fig. 4). Calcium carbonate concentration is between 13 and 37% in all samples, with an average of 21% (Table 1). Of the 228 samples analyzed from Unit II, nearly all are comprised of silt (n = 104) or clayey silt (n = 105). Approximately 40% of the resulting grain-size-frequency curves are unimodal, of which all are leptokurtic (Graphic Kurtosis > 1.1) with the mode occurring between 5 and 22 µm (Fig. 5a–d). Weibull function curve fitting shows, however, that these unimodal particle-size distributions most closely resemble the summed distributions of three independent populations. Most of the polymodal samples contain either two (37%) or three (20%) modes. However, curve fitting shows these distributions are commonly comprising between three and five subpopulations. The major mode of each polymodal sample is typically also between 5 and 22 µm, with the most common modes occurring at 7–10 µm, 13–18 µm, and 54–64 µm.

Table 1 Summary of sedimentological data
Full size table
Fig. 5
figure 5

Representative particle-size distributions for sediments from Unit II decomposed into Weibull subpopulations usinig CFLab (Panels A–D). Fine component (purple) has a mode < 1 μm and is interpreted as autochthonous (bio)chemical sediment. Overlapping, intermediate-sized subpopulations (modes between 4 and 90 μm; shown in green and blue) represent various hydraulic inputs to the lake (suspended load). Coarse subpopulations (orange; 300–400 μm) are interpreted as hydraulic sedimentation via saltation. Panel E shows the locations of potential major hydraulic sediment sources (fluvial inputs) of sediment to the basin in relation to the sampling site under larger surface area (blue outline) and reduced surface area (purple outline) conditions (not to scale)

Full size image

The average concentration of sand in the Unit II samples is 9%, with the least sandy beds found at the base of the sequence (Table 1; Fig. 4). While every Unit II sample contains some sand, most of it in the very fine sand fraction, much of this sand comes from the coarse tails of populations with major modes in the silt fraction. Samples with minor modes in the fine or medium sand fractions are present in most JRD beds, and approximately half of the analyzed curves include a mode greater than 200 µm. However, modal values for these populations typically fall well below 1%. Additionally, many samples include a clay mode located between 0.8 and 2 µm (typical modal values between 1 and 2%).

The sediment samples recovered from Unit I contain, on average, more organic carbon and less calcium carbonate than the samples of Unit II (Table 1). All of the 26 samples analyzed from lithostratigraphic Unit I belong to the silt-sand-clay textural classification, all include modes within the fine or medium sand fraction and most (n = 15) include a mode between 0.7 and 0.9 μm. All but one sample is multimodal, with the most common modes being located at 4–12 μm, 20–60 μm, and either at 112–225 μm or 350–500 μm.

Microfossils

In Unit II, plant and animal remains smaller than 1 mm are present in all samples, with the most common types of microfossils being ostracod valves, small gastropod shells, gastropod opercula, and charcoal (Fig. 4). Stonewort gyrogonites are common in lithostratigraphic Unit II’s even-numbered beds but nearly absent from the odd-numbered beds. The teeth and bones of small, aquatic amphibians and fish are also common, especially in samples near the top of Unit II (Fig. 4).

Table 2 Results of the radiocarbon dating. Depths given in the second column are centimeters beneath the contact between lithological units I and II along the east wall of excavation Area B
Full size table

In Unit I, small gastropod shells—most belonging to the genus Bithynia—and gastropod opercula are the most common microfossils, followed by benthic foraminifera and stonewort gyrogonites. A small number of teeth and ostracod valves (on average, 2.4 valves per gram) are also found in these sediments.

Radiocarbon ages and chronostratigraphic model

Of the 23 charcoal samples analyzed from Unit II, 22 fall within the period 20.5–11.5 cal ka BP, while one postdates 1950 AD (Table 2). The chronostratigraphic model (Fig. 3) indicates that Unit II was emplaced from approximately 21.1 cal ka BP and terminating at approximately 11.3 cal ka BP. The chronostratigraphic model also reveals a near total cessation in sediment accumulation between 18 and 17 cal ka BP that was not observable in the site stratigraphy. At this time, accumulation is modelled to have dropped from approximately 27 cm ka−1 to between 2.0 and 2.5 cm ka−1 for a few hundred years before recovering. Two additional significant changes in the sediment-accumulation rate occurred: one at c. 19.6 cal ka BP, from approximately 43 to 27 cm ka−1, and another at c. 17 cal ka BP, from approximately 27 to 23 cm ka−1. The single charcoal sample analyzed from Unit I yielded a conventional radiocarbon date of 173 ± 19 14C a BP (Beta-457485), suggesting Unit I was emplaced at a much later date than Unit II (Table 2). Modeled ages of the Unit II beds (Table 3) indicate that each bed represents between three hundred and two thousand years of sediment accumulation.

Table 3 Modeled ages of the beds correlated with the site’s stratigraphy, presented together with their cultural affinities (after Sharon et al. 2020)
Full size table

Discussion

The JRD excavation comprises two sedimentary units—a lower unit accumulated between 21.1 ± 0.2 and 11.3 ± 0.2 cal ka BP (Lithostratigraphic Unit II) unconformably overlain by modern sediments (Lithostratigraphic Unit I). The accumulation of Unit II sediments likely continued after 11.3 cal ka BP but sediments were probably removed erosionally prior to the emplacement of Unit I, either by the flow of the Jordan River or agricultural/infrastructure development in the valley. The single radiocarbon date recovered from Unit I sediments provides a calibrated age of post-290 cal a BP. This can be interpreted as approximately the earliest these sediments could have been emplaced as this charcoal is likely older and re-deposited here. The modern sediments that constitute Unit 1 are not discussed further in this study.

Implications of grain-size-frequency distributions

Particle-size distributions of the Unit II sediments are mostly multimodal, suggesting more than one source for the JRD sediments. However, when the particle-size distributions are unmixed using the Weibull curve fitting tool of Wu et al. (2020) even the unimodal distributions are revealed to be the sum of multiple subpopulations (Fig. 5a–d). The presence of multiple overlapping silt and very fine sand populations is interpreted as deposition of suspended sediment via multiple hydraulic inputs. Changes in the seasonal discharges/suspended sediment loads of the rivers entering the lake may also result in discrete modes in the particle-size distributions (Sun et al. 2002). Here, the dominant sediment sources are suggested to be silty sediments entering the lake via the suspended sediment load of rivers such as the Upper Jordan River and the Dishon and Mahanayim streams (Fig. 5a–d, blue and green subpopulations). Direct aeolian input to the lake is presumed to be minimal due to the distance between the study site and potential dust sources. Furthermore, the high flux of fluvial sediments to the basin is expected to overprint the potential dust signal.

Populations with modes in the fine-to-coarse sand grain-size fraction are suggested to represent saltation, potentially generated by heavy storm events (Fig. 5a–d, orange subpopulation). This material was probably transported to the lake via streams entering the lake from the east, closest to JRD. Changes in the volume of the coarse population and also its mode are suggested to track changes in the distance between JRD and the side stream inflows, with large amounts of coarse sediments present when the paleoshoreline and stream inlets are closest to JRD (Fig. 5e). In distributions where this coarser population is absent we suggest that the site was located far enough from the paleoshoreline, or at least from the location of the stream inlets, to not have been influenced by these smaller streams, or that no such heavy storm events occurred.

Inferences based on sediment geochemistry

Total organic carbon concentrations in the even-numbered beds are typically low (Fig. 4). We relate this to the rapid reuptake of organic carbon by organisms in a shallow and biologically productive environment. High or rising TOC concentration immediately after the deposition of each even-numbered bed is probably indicative of conditions for the preservation of organic carbon becoming more favorable as local water level rose and plant material was rapidly buried. However, there is no evidence that the bottom water was anoxic due to the abundance of ostracod remains throughout the sequence (Fig. 4).

Carbonate present in the sediments likely originated from a combination of autochthonous and allochthonous sources, as carbonate rocks are present throughout the catchment (Fig. 1b) and mollusk and crustacean remains are present throughout the stratigraphic sequence (Björgvinsson 2017; Valdimarsson 2017). The amount of carbonate present in each sediment sample is not correlated with the amount of shell material visible in the beds at outcrop scale. However, some correlation does exist between the samples’ calcium carbonate concentration and the proportion of fine grains in the samples: samples with the highest percentage of grains < 4 µm tend to contain the least carbonate (Spearman’s rho = − 0.26) while siltier samples tend to contain more carbonate (Spearman’s rho = 0.24). This correlation reflects the changing water level at the site, where mollusks and crustaceans, especially ostracods, are more abundant in the shore-proximal environments where silt deposition dominated and the diversity of microhabitats is higher.

Significance of mollusk remains

At JRD, the beds of Unit II were delineated on the basis of shells visible on the exposed walls of the excavation area during logging of the primary section (Fig. 2c). Though many species of mollusks are found within the JRD sediments (Marder et al. 2015; Björgvinsson 2017), only the largest species are visible at the JRD EWPS outcrop: These are primarily freshwater bivalves belonging to the family Unionidae and Melanopsis gastropods. Other shells found abundantly in the JRD sediments include gastropods belonging to the genera Theodoxus and Bithynia and bivalves belonging to the genera Corbicula and Pisidium.

Freshwater mussels belonging to the family Unionidae are commonly found in the shallow waters of lakes and riverbanks throughout northern Israel (Milstein et al. 2012). The mussels at JRD have previously been ascribed to the species Unio terminalis (Marder et al. 2015). However, it is possible that some or all belong to one or several undescribed (extinct) species, or to species no longer found in this region (Sharon et al. 2020). The taxonomy of these bivalves in the Eastern Mediterranean region is undergoing rapid revision (Araujo et al. 2018; Lopes-Lima et al. 2021), and it is therefore beyond the scope of this paper to attempt to identify these mussels to species level.

Melanopsis are commonly found in the freshwater bodies (lakes, rivers, streams and springs) of the southern Levant, where they typically live attached to rocks or emergent vegetation and are not normally found living in deep-water or turbid environments (Tchernov 1973; Heller et al. 1999; Heller and Sival 2001; Marder et al. 2015; Bandel 2020). During the spring and summer of 2017, the authors observed these snails living within a few centimeters of the water surface, either attached to the stalks of reeds and other emergent vegetation or on rocks in many environments: along the shores of the Sea of Galilee as well as in springs and along riverbanks around central and northern Israel. Rarely did we observe individual snails living on fine-grained sediments or in water deeper than 20 cm.

At JRD, unionid shells are present in all beds of lithostratigraphic Unit II but are most abundant in the even-numbered beds (corresponding to cultural horizons 5 through 3-A; Fig. 4). Melanopsis shells, however, are found only in the even-numbered beds, where their abundance implies the presence of a suitable substrate. As no stones large enough to support communities of Melanopsis close to the air–water interface are recorded at JRD we assume that the snails here were supported by plants such as emergent vegetation. This observation supports the hypothesis that sediments of even-numbered beds were deposited in a shore-proximal environment and are representative of local water-level minima while the odd-numbered beds were deposited in deeper water, farther from the paleoshoreline, in an environment too deep to support emergent vegetation and large populations of Unionidae.

As the mollusk remains at JRD are supported by the mud matrix and unsorted, we suggest that the shells are either in situ or were deposited very close to where the animals lived. No winnowing of the fine sediments is observed, suggesting that the depositional environment was low energy, and it is unlikely that the shells were transported via downslope movement, waves or currents. This is consistent with the observations of Björgvinsson (2017), who noted that the gastropod remains are well-preserved and thus unlikely to have been transported far. Although gastropods belonging to the family Viviparidae have previously been observed in sediments from JRD—calling into question the timing of the extirpation of these snails from the Hula Valley—no Viviparidae are observed in the sediments from JRD investigated here, supporting the interpretation published in Marder et al. (2015) that these shells were washed out of a nearby Middle Pleistocene outcrop.

Depositional environments of the archeological site

The uniform nature of the JRD Lithostratigraphic Unit II beds and the fine nature of the sediments, together with the absence of bedforms or sedimentary features typically associated with moving water (e.g., ripples, lag deposits or cross beds), indicate that Unit II was deposited in a stable, low-energy lentic environment as opposed to a beach or lakeshore. Good preservation of pollen and plant remains and an absence of infilled mud cracks, soil or paleosol horizons further suggest that this site never underwent significant subaerial exposure. There are no visible unconformities, and our chronostratigraphic model shows that the rate of sediment accumulation remained largely constant. Furthermore, the scattered nature of the archeological finds, together with frequent but random distribution of charcoal but without evidence of distinct hearths, supports the interpretation that the entire JRD sedimentary sequence accumulated subaqueously. Typical wetland or stream-channel features are also not seen at JRD, and TOC are significantly lower in the JRD sediments than in older Hula Valley freshwater deposits previously analyzed and interpreted as wetland deposits (Kafri and Lang 1987). It is therefore suggested that the sediments of JRD were deposited in the calm, shallow water of a nearshore lake environment and that Unit II represents a continuous record of the site’s paleoenvironmental conditions.

In contrast to JRD, the Pleistocene sedimentary infill of the Hula Basin to the north of JRD contains both (interfingered) lacustrine and wetland sediments as well as fluvial and alluvial sediments related to the runoff of water and transport of sediments from the surrounding highlands (Horowitz 1971, 1973). At JRD, the sedimentology, stratigraphy, and microfossils indicate that the entire archeological site is composed of lacustrine sediments deposited in a lake that existed here continuously from at least 21.1 ± 0.2 to at least 11.3 ± 0.2 cal ka BP. The sediments accumulated at JRD have primarily been transported to the site via fluvial processes originating from various sources (rivers and streams draining to the valley from the north, east and west; Fig. 1c), most likely including some aeolian input as well.

The types, conditions, and abundances of shells present within the sediments—in combination with the sediment–grain–size distributions, carbonate content and organic matter content—indicate changes in the water depth at JRD and in the distance from JRD to the paleo-shoreline. Following the LGM, Paleolake Hula appears to have repeatedly expanded and contracted in response to changes in the regional hydroclimate regime of the surroundings. The deposition of lake and wetland sediments at the nearby archeological site of Nahal Mahanayeem Outlet (dated between c. 74 and c. 44 ka; Kalbe et al. 2014) and in a core drilled between the Nahal Mahanayeem Outlet and Jordan River Dureijat sites (c. 90 ka; Eyþórsdóttir 2019) suggests the lake underwent repeated expansions and contractions throughout the Late Pleistocene. Here, the even-numbered beds represent the emplacement of shallow-water, shore-proximal deposits as indicated by coarser grains, more abundant macrofossils, the presence of remains of emergent vegetation, and higher calcium carbonate and lower TOC concentrations. In contrast, the odd-numbered beds, which contain finer-grained sediments, preserve more organic carbon, contain less calcium carbonate, and contain fewer mollusk and crustacean remains, represent sedimentation in an environment up to approximately a couple meters deeper, farther from the paleo-shoreline.

The lacustrine sediments at JRD indicate that the site was located within the margins of the late Pleistocene-early Holocene Paleolake Hula. At the time of its utilization during the Epipaleolithic, the inhabitants of this region likely waded into the lake to fish or collect other resources such as plants or shellfish, leaving behind artifacts such as stone line weights and net sinkers and bone-fish hooks at the location of the archeological site (Sharon et al. 2020). These artifacts are found exclusively in the site’s archeological horizons, which correspond to the even-numbered beds and Bed 13 and represent near-shore, shallow-water conditions (Fig. 3). The high concentration of the artifacts in these beds suggests that fishing activity was most common in shallow-water areas near the paleoshoreline, as opposed to in deeper water. At JRD, fishing hooks are found primarily in beds 10 and 12 (Pedergnana et al. 2021). If these fish hooks were lost during use, they may indicate that water depths at this time did not exceed the average waist height of adult humans living in this area during the Epipaleolithic. Using the reconstructed Natufian body heights presented in Hershkovitz and Gopher (2008), we speculate that water depths greater than c. 1 m (approximately waist-height) would have been too deep for hook and line fishing and therefore consider this a rough estimate of the maximum water depth during the deposition on these sediments during at least part of the year.

Lake level and environmental history of Paleolake Hula at JRD

As a hydrologically-open lake located in a broad valley with gently-sloping sides the subtle lake-level changes reconstructed from the JRD sediments can reflect potentially large changes in the water balance of the valley. Here, we relate the lake-level changes inferred from JRD’s sedimentary record with other environmental studies from the Levant.

Beds 1–3: LGM lake highstand (21.1 ± 0.2 to 17.2 ± 0.1 cal ka BP)

Beds 1 and 3, located below 56.5 masl, are the thickest beds at JRD and contain both the site’s finest sediments and the fewest shells. The sediments of these beds accumulated at a slightly higher sediment-accumulation rate than the others, indicating prolonged sedimentation in a stable, low-energy environment distal to the paleoshoreline. We interpret this as the highest stand of Paleolake Hula recorded at JRD. The consistent transport of fine sediment to the comparatively deep lake is suggested to represent a wet environment where enhanced riverine discharge and/or overland-water flow transported the fine grains deep into the basin. However, as these are the oldest sediments preserved at the site we are not able to compare them with sediments deposited earlier in the glacial, during which discharge to the lake may have been even larger. The absence of coarse grains in Beds 1 and 3 supports the interpretation that the study site was far from any inflow to the lake during this time, and particle-size distributions for Beds 1–3 suggest hydraulic transport of suspended silt-sized grains to be the dominant sediment source during this time. Based on our chronology, this highest late Pleistocene level of Paleolake Hula occurred from 21.1 ± 0.2 to 17.2 ± 0.1 cal ka BP, with a minor regression during the time period 19.0 ± 0.1 to 18.4 ± 0.1 cal ka BP (Bed 2/Horizon 5). In a study of pollen and palynomorphs from the JRD sediments, Langgut et al. (2021) reconstructed abundant dwarf-shrub and steppe vegetation during the late glacial (Pollen Zone 1a; JRD Beds 1 and 2) and concluded that the area surrounding JRD was up to 5 °C colder and received 60 mm less annual precipitation than today. Significantly suppressed evaporation throughout the year but particularly during LGM summers has been suggested as responsible for maintaining high lake levels in the waterbodies along the Dead Sea Transform during this time (Bartov et al. 2002; Ludwig and Hochman 2022), and may have been a contributing factor to the high lake levels reconstructed here for Paleolake Hula.

The emplacement of the shell-rich sediments which comprise Bed 2 is suggested to represent a slight, temporary drop in water level immediately following the LGM (Fig. 6) lasting, per our age model, approximately six hundred years. A small amount of sand in these samples also supports the interpretation that the study site was nearer to the paleoshoreline at this time. Stone tools recovered from this level consist primarily of scrapers, burins, and bladelets and have been assigned to an Early Epipaleolithic cultural entity (Sharon et al. 2020). The age of Bed 2 modeled here is consistent with the later Early Epipaleolithic and beginning of the Middle Epipaleolithic in the chronological scheme of the terminal Pleistocene cultural entities presented in Belfer-Cohen and Goring-Morris (2020). During this time, the inhabitants of the Hula Valley likely experienced cool conditions with abundant freshwater despite slightly reduced precipitation relative to the present (Langgut et al. 2021; Rice et al. 2023).

Fig. 6
figure 6

Paleoenvironmental records from the Eastern Mediterranean region 21–11 cal ka BP. A CaCO3, B TOC, and C grain-size data from this study. D Palynological data from Langgut et al. (2021). E Soreq Cave stable oxygen isotopes from Bar-Matthews et al. (2003a, 2003b) and lake-level curve for Lake Lisan from Torfstein et al. (2013b). F 30°N July and December insolation curves from Berger and Loutre (1991). Reconstructed Eastern Mediterranean Sea surface temperatures from Essallami et al. (2007) sediment core MD84-632

Full size image

Lake-level drawdown at 17.2 ± 0.1 cal ka BP

The chronostratigraphic modeling suggests a temporary drop in the sediment-accumulation rate occurred sometime between 18 and 17 cal ka BP, approximately at the location of the boundary between Bed 3 and Bed 4. During this period of suppressed sediment accumulation the sediment-accumulation rate dropped from c. 27 cm ka–1 to less than 3 cm ka–1 for several hundred years before rising again to c. 27 cm ka–1. While no visible changes in the outcrop-scale stratigraphy indicate a change in the sedimentation regime or any sort of hiatus, from the particle-size analysis a major change in the depositional environment is inferred here. In sediments deposited above c. 56.5 masl, a significant shift in the nature of the sediments and microfossils suggests changes in both the local environment and the size/shape of the lake. The sudden increase in sand deposition reflects the transition to a more shore-proximal depositional environment, and points to a significant lowering of the lake level at c. 56.5 masl. Although the chronostratigraphic model is not sensitive enough to determine exactly when, this rapid drawdown of the lake most likely happened between 17.6 and 17.1 cal ka BP based on our sensitivity testing. Hereafter, we refer to the timing of the drawdown as occurring at 17.2 ± 0.1 cal ka BP as reconstructed in our initial modeling efforts.

In sediments deposited after 17.2 ± 0.1 cal ka BP (Beds 4–13), the alternating deposition of shell-rich and shell-poor beds reflects changing water depth in a generally smaller and shallower lake, and increased deposition of coarse sediments is related to the proximity of the site to the paleoshoreline. Here, the deposition of JRD’s even-numbered beds occurred at times when lake level was shallowest and the paleo-shoreline was nearest to the location of JRD. This is consistent with the interpretation of the lake environment presented in Rice et al. (2023), whose reconstruction of Paleolake Hula characterizes the lake as large and hydrologically open during the deposition of Bed 1 (Horizon 6) and Bed 2 (Horizon 5), and smaller and more poorly-buffered during the deposition of Bed 6 (Horizon 4-A) and Bed 8 (Horizon 3-C).

Beds 4–6 (17.2 ± 0.1 to 16.0 ± 0.1 cal ka BP): Heinrich Event 1 lowstand

Beds 4 and 6, corresponding roughly to the Geometric Kebaran horizons 4-A, 4-B, and 4-C, are some of the site’s sandiest beds. Decreased distance from the shoreline to the sampling site is likely responsible both for the sand-sized modes in these sediments as well as for the abundance of foraminifera (Fig. 4), as the foraminiferal tests are likely allochthonous and sourced from the Cretaceous to Eocene marine limestones of the Galilee Mountains as observed elsewhere in Pleistocene lakes of the Dead Sea Transform (Arnardóttir 2018) (Fig. 1). Alternatively, coarse grains in these beds (subpopulation mode c. 250 μm) may reflect dust deposition from proximal or local sources.

The sediments of Bed 5 represent a shoreline transgression which lasted approximately 200 years. Beds 5 and 6 correspond to the pollen zone 2 of Langgut et al. (2021), who documented a profound increase in deciduous trees and woodland vegetation during this period together with a decrease in precipitation. Their findings suggest a local environment that was both warmer and drier than the LGM, but still with sufficient amount of water available to support a dense community of plants including oak and Pistacia trees. The competency of streams entering the lake at this location was also high enough to facilitate the transport of relatively large silt and sand grains to the study area at least intermittently. Frequent storms during this time are also inferred from the sub-annual carbonate stable isotope records of Melanopsis shells presented in Rice et al. (2023).

The deposition of Beds 4–6 occurred synchronously with the deposition of gypsum at Masada (Section M1, dated 17.1 to 15.5 ± 0.5 ka; Torfstein et al. 2013a, b) and corresponds to the lake-level minimum reconstructed for Lake Lisan at c. 17 cal ka BP at both the location of the modern Dead Sea and Sea of Galilee (Torfstein et al. 2013a; Lev et al. 2019; Fig. 6). It has been hypothesized that the low lake level at this time was related to reduced precipitation during Heinrich Event 1 (Bartov et al. 2003). However, low lake levels may also be the result of increased evaporation: the mobilization of vegetated sand dunes in the northwestern Negev dunefield during Heinrich Event 1 (Roskin et al. 2011; Robins et al. 2022) and high fluxes of Saharan dust in the Levantine Basin sediment core SL 112 (Hamann et al. 2008) both point to increased windiness during this time. Although particle-size distributions from JRD Beds 4–6 do not show silt consistent with the size and sorting of North African dust (c. 40 um, Beuscher et al. 2017), this same increase in windiness may also be responsible for low water levels here via enhanced evaporation. Regardless of their cause, the contemporaneous water-level drops reconstructed for these waterbodies support the observation that they responded to changes in the regional hydroclimate in similar manners (Nadel 2017; Lev et al. 2019).

Artifacts attributed to Middle Epipaleolithic Geometric Kebaran hunter-fisher gatherers including burins, scrapers, and limestone net sinkers are found in JRD Beds 4 and 6 (Sharon et al. 2020). The timing of the deposition of these artifacts—between 17.2 ± 0.1 and 16.0 ± 0.1 cal ka BP—is consistent with the reconstructed timing of the Middle Epipaleolithic, which lasted from c. 18.5 to 15.0 cal ka BP (Belfer-Cohen and Goring-Morris 2020).

Bed 7: Late Oldest Dryas highstand (16.0 ± 0.1 to 14.8 ± 0.1 cal ka BP)

A lake-level highstand lasting approximately twelve hundred years reflects wet conditions in the catchment following Heinrich Event 1. A period of higher sand influx compared to the other deeper-water facies at JRD together with larger sediment mean grain sizes indicate an increase in the competency of overland flow and/or rivers entering the lake. This is in agreement with the reconstruction of Langgut et al. (2021), who inferred low temperatures and steeply rising annual precipitation from the pollen record (Pollen Zone 3), based on an increase in dwarf-shrub steppe vegetation pollen and the preservation of very little arboreal pollen. This highstand reconstructed for Paleolake Hula occurred at the same time as Lisan Event 1, the most recent in a series of high lake-level periods reconstructed for Lake Lisan (Torfstein et al. 2013a, b). Bed 7 also corresponds to a period of decreasing oxygen isotope values in Soreq Cave speleothems indicating a shift to wetter conditions in central Israel at this time (Bar-Matthews et al. 2003b; Fig. 6e).

Beds 8–10: lake-level change during the Bølling–Allerød and Older Dryas (14.8 ± 0.1 to 13.3 ± 0.1 cal ka BP)

Beds 8 (Cultural Horizon 3-C: 14.8 ± 0.2 to 14.1 ± 0.2 cal ka BP) and 10 (Horizon 3-B: 13.5 ± 0.1 to 13.3 ± 0.1 cal ka BP) correspond to the Natufian cultural horizons 3-C and 3-B, respectively, and are suggested to represent low water level during the warm (Bølling, Bed 8; and Allerød, Bed 10) oscillations of the Bølling–Allerød Interstadial. They are separated by a short-lived period of higher lake level, suggested to represent peak stadial conditions during the Older Dryas (Bed 9: 14.1 ± 0.2 to 13.5 ± 0.1 cal ka BP).

In the pollen record of Langgut et al. (2021), the sediments of Beds 8 and 9 correspond to Pollen Zone 4, characterized by low levels of Artemisia pollen, high arboreal pollen, and a significant shift in the aquatic plant assemblage, including the replacement of watermilfoils (Myriophyllum) and pondweeds (Potamogeton) by Pediastrum algae. In a previous study of pollen from Paleolake Hula, Van Zeist et al. (2009) proposed that the growth of Pediastrum was encouraged by enhanced transport of organic matter to the lake via increased overland-water flow to a deeper water environment. At JRD, Pediastrum occurs at all levels above 57 masl (Beds 8–13). However, the abundance of sand and mollusks in these beds suggests that the sampling site was relatively close to the paleo-shoreline at this time. Low water levels despite significant overland-water flow could indicate that evaporation was the main control on lake level at this time.

In the uppermost centimeters of Bed 8, TOC concentrations rise abruptly, and the sand modes disappear from grain-size-frequency curves of the sediments. These changes were probably caused by a sudden increase in water level at the onset of the Older Dryas Stadial. Bed 9 represents a short-lived phase of higher lake level corresponding to peak stadial conditions and Meltwater Pulse 1A at c. 14.4–14.0 cal ka BP (Menviel et al. 2011). This is consistent with sea-surface-temperature reconstructions from the eastern Mediterranean (Fig. 6f). However, preliminary ostracod analysis of the JRD sediments suggests an increase in water temperatures at this time (starting at c. 14.1 cal ka BP). Here, the warm-water species Ilyocypris salebrosa is present only above this stratigraphic level (Valdimarsson 2017). Ilyocypris salebrosa persisted in the lake until the end of the studied sedimentary sequence, including the Older and Younger Dryas stadials, although their abundance decreases during these periods. Thus, the regional climate may have cooled significantly during this time but the lake water remained at least seasonally warm enough to support these organisms due to generally shallow conditions (Valdimarsson 2017).

In Bed 10, the return to low lake level together with the abundance of warm water ostracod species suggest that environmental conditions in the Hula Valley during the Allerød Oscillation were likely warm and similar to those during the emplacement of Bed 8. Warm conditions during the Bølling and Allerød are in agreement with sea-surface-temperature reconstructions from the Eastern Mediterranean, where oxygen isotope ratios measured on planktonic foraminiferal tests indicate this was the warmest phase of the last deglaciation (Essallami et al. 2007; Fig. 6f). The region also experienced high spring solar insolation during this time (Berger and Loutre 1991; Fig. 6f).

Archeologically, this stage marks the appearance of the Natufian culture in the Levant. The Natufian are the first settlers and their sites, including Eynan (located only 9 km from JRD in the Hula Valley), are characterized by an explosion of technological innovations, art, and symbolism (Bar-Yosef and Valla 2013; Yaroshevich et al. 2016). Among these innovations, the spread of line-and-hook fishing technology is particularly well-preserved at JRD in the form of line sinkers and bone fish hooks (Sharon et al. 2020; Pedergnana et al. 2021). The timing of the deposition of the Natufian layers at JRD—between 14.8 ± 0.2 and 12.5 ± 0.1 cal ka BP—is consistent with the accepted chronology of the Late Epipaleolithic of the Southern Levant (c. 15.00 to 11.65 cal ka BP; Belfer-Cohen and Goring Morris 2020).

Bed 11: increased wetness at the end of the Allerød

Bed 11 represents the last few centuries of the Allerød interstadial, during which a small increase in effective moisture caused water level to rise too high for Melanopsis to thrive but not high enough to inhibit the recycling of organic carbon on the lakebed, suggesting that the environment here was biologically productive despite the low amount of preserved organic carbon. The presence of gyrogonites, teeth and bones—typically more common in even than odd numbered beds (Fig. 4)—supports the assessment of the site as a biologically productive, shallow-water environment at this time.

Bed 12: Early Younger Dryas lowstand (12.8 ± 0.1 to 12.5 ± 0.2 cal ka BP)

The deposition of the sandy, shell-rich Bed 12 (late Natufian Cultural Horizon 3-A) indicates low water level of Paleolake Hula during the onset of the Younger Dryas Stadial. Bed 12 likely represents the shallowest conditions observable at JRD, with the lake shoreline closest to the sampling site. The interpreted reduction in lake level here at the onset of the Younger Dryas may have occurred contemporaneously with a significant drop in water level reconstructed for the Sea of Galilee at c. 13 cal ka BP by Hazan et al. (2005) and the minimum stand of Lake Lisan between 13 and 12 cal ka BP (Bartov et al. 2002).

Artifacts found in Bed 12 include bladelets and lunates as well as extensive evidence for the use of line and hook fishing technology in the form of line sinkers and bone-fish hooks (Sharon et al. 2020). The earliest Pre-Pottery Neolithic artifacts at JRD are found at the top of this layer, in sediments dated to 12.5 ± 0.1 cal ka BP. This is considerably before the emergence of Neolithic cultures elsewhere in the Southern Levant (11.65 cal ka BP; Belfer-Cohen and Goring-Morris 2020). However, as stone tools are denser than the surrounding sediments these are expected to have dropped into the lake mud from above. Although no signs of soft sediment deformation were observed under these tools this is expected due to the structureless and uniform nature of the lake mud. The occurrence of Pre-Pottery Neolithic A artifacts in the uppermost centimeters of sediment in Bed 12 may also be related to the later archeological manipulation (reworking) of the Bed 13 sediments visible at other parts of the archeological site and proposed to be related to fishing activities in Sharon et al. (2020). Here, however, the sediments are not visibly disturbed and while the contact between Bed 12 and Bed 13 is sharp and well-defined it displays no signs of post-depositional modification or erosion.

Bed 13: Later Younger Dryas and Early Holocene modest lake-level rise (12.5 ± 0.1 to 11.3 ± 0.2 cal ka BP)

By 12.5 ± 0.1 cal ka BP, water level at JRD had risen modestly, resulting in the deposition of Bed 13 (early Neolithic Cultural Horizon 3-0) in slightly deeper water under peak stadial conditions. Bed 13 is the only odd-numbered bed at JRD to include archeological artifacts, suggesting that the water level was low enough during the time of its deposition for this area to be accessible and utilized by humans. Due to the abundance of clam shells and the presence of stone and bone tools in the sediments, the lake level during the deposition of Bed 13 is assumed to be higher than that of the other cultural horizons at JRD but lower than that of the other odd-numbered beds. The interpretation of a shore-proximal environment here is further supported by the presence of large amounts of fine sand in these sediments, comparable to the particle-size distributions of Bed 10.

Elsewhere, lake levels increased at both the Dead Sea and the Sea of Galilee during the Younger Dryas (Torfstein et al. 2013b; Lev et al. 2019), which we suggest was cool and wet at JRD. Consistently cool conditions and an increase in annual precipitation of approximately 40 mm, resulting primarily from an increase in summer and autumn precipitation, were also reconstructed in the palynological study of Langgut et al. (2021) during this period (Pollen Zone 5). Increased windiness during this time is also reported for the Eastern Mediterranean (Roskin et al. 2011). Here, it is possible that this increased windiness worked to counteract the increase in lake level that might have otherwise been expected with the increased annual precipitation.

The upper boundary of Bed 13 (contact between Unit I and Unit II) is sharp, uneven and interpreted as an erosional surface representing a significant gap in the sedimentary record. Here, sedimentation likely continued after 11.3 ± 0.2 cal ka BP, although it is impossible for us to estimate at which time sediment accumulation ceased or at which time the truncation of Unit II began.

Conclusions

The following main conclusions are inferred from the investigation of the JRD sedimentary sequence:

  • The refined age model presented here places the age of the JRD site at 21.1 to 11.3 cal ka BP. During this time, the archeological site was located within the margin of a shallow lake and was permanently submerged.

  • Artifacts left behind by four cultural entities are preserved in the sediments at JRD, as described in Sharon et al. (2020) and Pedergnana et al. (2021). The chronostratigraphy presented here can be used to estimate the timing of the site’s utilization by each of the three Epipaleolithic cultural entities and to refine estimates of their durations in the Southern Levant:

    1. o

      between 19.0 ± 0.1 to 18.4 ± 0.1 cal ka BP the site preserves evidence that JRD was utilized by an early Epipaleolithic (Masraqan?) cultural entity,

    2. o

      between 17.2 ± 0.1 to 16.0 ± 0.1 cal ka BP the site was utilized by a Geometric Kebaran population, and

    3. o

      between 14.8 ± 0.2 and 12.5 ± 0.1 cal ka BP the site preserves evidence of its usage by Natufian fisher-hunter gatherers.

  • The preservation of archeological artifacts at JRD is associated with periods of low water level at JRD, where the site was located close to the paleoshoreline: it is assumed that during periods of human utilization the site was submerged by water shallow enough that people were able to wade into the lake, leaving behind fishing equipment and other tools.

  • Despite changing paleoenvironmental conditions at the site the sedimentary sequence at JRD records the accumulation of lacustrine silty sediments at a high temporal resolution. All major late Pleistocene climate events can be identified at JRD and linked to changes in the water level/position of the shoreline.

  • A significant change in the size and nature of the lake, from a large and stable lake to a smaller, shallower, and more labile lake, occurred at approximately 17.2 cal ka BP.

  • The Younger Dryas Stadial at JRD includes facies deposited at varying water depths but sediments deposited under peak stadial conditions indicate higher Lake Hula water levels than at the onset of the stadial. Lake level at JRD, however, never rose high enough to inhibit the utilization of the site as a fishing locality, leading to the deposition of artifact-bearing sediments throughout the Younger Dryas Stadial.

  • Lake-level changes at Paleolake Hula broadly parallel those of both the Dead Sea and Sea of Galilee as well as their predecessor, Lake Lisan, following the LGM. Although very different in size and depth, these lakes were relatively larger and deeper during the LGM and generally smaller, shallower, and more labile during the late glacial.

The detailed chronostratigraphy presented here provides context for the analysis of the site’s archeological artifacts and allows for the comparison of the JRD sediments to other contemporary paleoenvironmental records. Furthermore, it enables hypothesis testing regarding the mechanisms by which late Pleistocene environmental change influenced human lifeways and behavior. Following completion of the excavation phase of the JRD project, we are now able to focus our sights on analyzing the cultures of the humans who utilized this exciting new site, as well as the environment and demography of the late Pleistocene Hula Valley.