Evidence for sophisticated raw material procurement strategies during the Lower Paleolithic—Hula Valley case study

The Hula Valley has two key Acheulian sites: Gesher Benot Ya'aqov (GBY), a large flake Acheulian site with hundreds of basalt bifaces and a significant number of flint handaxes, and Ma'ayan Barukh (MB), where more than 3500 flint handaxes were collected. Over the last one million years, the valley was filled by alluvium and basalt flows, devoid of flint sources suitable for handaxe production. We conducted archaeological and geological surveys combined with an inductively coupled plasma mass spectrometry geochemical study to determine the source(s) of flint, comparing elemental compositions of handaxes from GBY and MB with those of different flint sources using a novel statistical method. The results demonstrate that Hula Valley Acheulian flint handaxes were derived from Eocene flint. For GBY, the nearest matching source for its small number of excavated handaxes is a secondary deposit of the Dishon streambed found ~8 km northwest of the site. A more likely source for both GBY and the thousands of MB handaxes is the Dishon flint extraction and reduction complex located 20 km to the west, a possibility also supported by the near absence of production waste flakes at the sites themselves. These findings support direct procurement strategy as early as the Lower Paleolithic.


| INTRODUCTION
Acheulian sites worldwide typically include assemblages comprising numerous large cutting tools, primarily handaxes, and cleavers. The production of these tools required many tons of different raw material lithologies, resulting in large assemblages of waste products in sites where the tools were made (e.g., Bárez del Cueto et al., 2016;Paddayya et al., 2006;Petraglia et al., 1999). Because the degree of sophistication required for the advanced and effective procurement of raw materials is related both to human and social evolution, many studies have focused on the raw material acquisition strategies of Acheulian tool makers (c. 1.8-0.25 Ma BP). In some fortunate cases, sites are identified directly on the source of the quarried raw material (Bárez del Cueto et al., 2016;Petraglia et al., 1999). In others, mining of raw materials from underground sources was suggested (Barkai et al., 2006;Paddayya et al., 2006;Petraglia et al., 1999;Sampson, 2006;Verri et al., 2004).
Yet other assemblages comprising abundant tools were clearly not produced in situ but imported to the site as finished tools. For example, Layer II-6/L4 at Gesher Benot Ya'aqov (GBY) was uncovered as a dense "pavement" of bifacial tools with an average of 14 handaxes and cleavers per square meter and containing only a few flakes and debitage (Goren-Inbar et al., 2018). However, experimental work has demonstrated that the production of a single handaxe may result in tens to hundreds of waste flakes (Madsen & Goren-Inbar, 2004;Newcomer, 1971). Hence, the low number of flakes and debitage found in Layer II-6/L4 indicates that the bifacial tools were produced elsewhere and not in situ (Isaac, 1977;Potts, 1988;Potts et al., 1999).
Evidence for raw material acquisition by early humans is rarely preserved in the archaeological record. A primary approach to discussing raw material acquisition strategies during the Lower Paleolithic (LP) is evaluating the distance from the site to known raw material sources (e.g., Caruana et al., 2019;Wilkins, 2017). However, this approach is challenged by the difficulties of identifying the actual source outcrop due to the lithological and geochemical similarities of different exposures. For example, different basalt outcrops along the Northern Dead Sea Rift Valley share similar geochemical features (Mor, 1993(Mor, , 1994. In addition, the landscape and geomorphology in this tectonically active rift valley are not static; hence, some sources that were available during the LP are no longer accessible. The sources of obsidian used as raw materials for the production of stone tools are easier to identify chemically than basalt or flint sources. Provenance studies of Lower Paleolithic obsidian tools demonstrated the exploitation of both primary and secondary sources. Obsidian was obtained through quarrying as in the workshop at Melka Kunture, Ethiopia (Mussi et al., 2023) and also through embedded procurement by collecting pebbles from valley streams as in the site of NG1, Armenia (Frahm et al., 2020).
In the face of these sourcing challenges, this research employed archaeological, geological, and geochemical methods to study the available sources of raw materials and their acquisition strategies for the Acheulian sites of the Hula Valley. The findings are significant for our understanding of the behavioral patterns of early humans.
Two primary Acheulian sites are known from the Hula Valley-GBY located at the southern end of the valley (Goren-Inbar et al., 2000 and Ma'ayan Barukh (MB) at the northern edge (Sharon et al., 2022;Stekelis & Gilead, 1966). Acheulian handaxes were also collected in the mountainous sites of Yiron and Baram to the west of the valley (Ohel, 1986(Ohel, , 1991 and in smaller numbers, primarily as handaxes, in find spots in the Golan to the east (Goren, 1979;Goren-Inbar, 1985).
F I G U R E 1 Geography, geology, Lower Paleolithic sites, surveyed areas, and extraction and reduction complexes and locations in the research area (geological map after Sneh et al., 1998). Capital letters correspond to different exposures (paragraph 1.3) and surveyed areas (paragraph 3.1).
The Acheulian site of GBY is located on the banks at the outlet of the Upper Jordan River, south of the Hula Valley ( Figure 1). The terrain is basaltic and its numerous flows were likely the sources of raw materials, even if the exact source was not identified. The waterlogged archaeological horizons contain numerous stone tools and animal bones, embedded within the Benot Ya'akov Formation (BYF). This Early Middle Pleistocene formation is exposed along 2 km of the river bank, extending both to the north and to the south of the Benot Ya'aqov Bridge. Using the uranium-disequilibrium method, its youngest age in the Hula Valley was assigned to 222 Ka. (Moshkovitz & Magaritz, 1987). Primary excavation at the site was conducted during the 1990s by Goren-Inbar (Goren-Inbar et al., 2018). Additional localities have been tested, primarily north of the Benot Ya'aqov Bridge (Gilead, 1968;Sharon et al., 2010;Stekelis, 1960). The flint handaxes sampled in the current study were collected at the northern exposure of the BYF at the locality known as North of Bridge Acheulian (NBA; Grosman et al., 2011;Sharon et al., 2002Sharon et al., , 2010. The lithic assemblage of GBY is the type site for the Large Flake Acheulian facies (Sharon, 2007). It is characterized by a wealth of bifacial tools made of basalt, which is the primary and nearly exclusive raw material used for bifacial tools excavated at GBY. All cleavers at the site are made of basalt, while limestone and flint make up only 1%-2% of the handaxes. Of the 221 handaxes excavated in the exceptionally rich layer II-6 Level 4 (subdivided into Level 4 and Level 4b), only four are made of flint (<2%) and two of limestone (<1%). Of the 62 handaxes unearthed from Layer II-6 Level 1, only four are made of flint (<7%) and four of limestone (<7%) (Goren-Inbar et al., 2018). Other layers at the GBY site may have had higher concentrations of flint handaxes, but they were either removed from the layers by the site's inhabitants (Goren-Inbar & Sharon, 2006) or the tools were collected as surface finds hence hampering a robust quantification (Sharon et al., 2010). Notwithstanding, the number of flint handaxes collected from the different Acheulian localities along the BYF during 100 years of research is in the hundreds. While appearing in small percentages in excavated horizons, flint was a significant raw material in the production of handaxes at GBY. Moreover, the flint handaxes at GBY are of the highest technological quality and are very aesthetic. The people who made them mastered all aspects of bifacial knapping, demonstrating a high level of dexterity, knowledge, and proficiency. Given that the GBY inhabitants had access to flint sources, although not in close vicinity to the site (see below), it is important to note that the preference for basalt over flint by the GBY knappers was probably not due to a lack of technological knowledge but based upon lithic tradition preferences (Sharon, 2008).
The MB Acheulian site is located on the northern edge of the Hula Valley (Figure 1). Handaxe collection began in the 1930s; since then, thousands have been collected from the site and vicinity's red soil covering the northern basaltic slopes of the valley. The local microgeology of the site is not fully understood as the site was never excavated. Recent attempts to locate excavation-worthy localities failed, probably because the archaeological layer was close to the surface and completely disturbed by agricultural activity in this heavily farmed land (Ronen et al., 1980;Sharon et al., 2022).
Although MB is located on top of the Hazbani basalt that erupted at ca. 0.9 mya (Heimann & Sass, 1989), only 20 basalt handaxes were found among the ca. 3500 handaxes stored at the Upper Galilee Museum of Prehistory at Kibbutz MB (<1%); the remainder is made of flint (Lister et al., 2013;Rosenberg et al., 2015;Sharon et al., 2022).
As at GBY, only a handful of debitage flakes were collected, indicating that the handaxes were not produced on-site. An Eocene flint outcrop located some 6 km north of MB (Stekelis & Gilead, 1966) was postulated to be the source of raw materials for the MB handaxes. However, to the best of our knowledge, no such outcrop has since been clearly identified and described, and the current geopolitical climate does not permit further investigation of this source in southern Lebanon. Also similar to GBY, the thousands of exceptionally well-made flint handaxes testify to the technological expertise of the MB knappers.

| Potential flint sources in the Hula Valley during the Acheulian period
Although geologists tend to use the term "chert" for a variety of siliceous lithologies, for this interdisciplinary study the term "flint" as commonly used in archaeology is more appropriate. Here, we discuss the possible sources of flint used for the production of Acheulian handaxes in the Hula Valley. As noted, the GBY cleavers are all made of basalt and only a small percentage of the handaxes are made of flint (Goren-Inbar et al., 2018). The focus of this study is the flint used for the production of handaxes; hence, the non-biface flint artifacts in the GBY assemblages are not discussed. In contrast to GBY, at MB, only a small number of items are made of raw materials other than flint, and the number of nonflint handaxes is insignificant. Flint is also the only raw material used for handaxe production at the sites of the Upper Galilee and Golan except for a single basalt handaxe found at Berekhat Ram (Goren-Inbar, 1985). It should be noted that while the geology of the Hula Valley and its surroundings is well known, the geological setting and exposed lithologies north of the valley at the marginal regions in Syria and in Lebanon, in particular, have not been studied in detail.
More importantly, before any discussion of flint sources and flint procurement strategies is the fundamental question of suitable flint for handaxe production. Bifacial tools are a primary component of the Acheulian tool kit, and because suitable flint possessed advantages over basalt, it was desired for their production. One key quality of suitable flint is the nodule size, which must be sufficient to produce a large bifacial tool. Basalt bifacial tools were typically produced on large flakes (>10 cm) detached from giant cores, dictating a minimal size of suitable raw materials (Sharon, 2009). Flint handaxes, on the other hand, were more frequently shaped from cobbles, although large flakes were also used, requiring smaller rock fragments.
However, since the majority of handaxes in the assemblages are larger than 10 cm, the initial size of the rock fragment had to be FINKEL ET AL. | 651 significantly larger (Jones, 1979(Jones, , 1994Newcomer, 1971). In addition, when only low-quality, nonhomogenous flint was available, the Acheulian knappers possessed the craftsmanship to produce highquality tools from it, for example, at Revadim (Marder et al., 2006) and Jaljulia (Zupancich et al., 2021). Lastly, the quality of flint pebbles as raw materials for tool production deteriorates due to fluvial mobilization, such as rolling in streams, which causes battering and cracking. There is some evidence that prehistoric knappers used fluvial-transported flint cobbles from streams (Frahm et al., 2020); however, it is not possible to determine to what degree such fluvial flint cobbles were preferred for tool production by the Acheulian knappers.
Many researchers (see below) assume that cobbles collected from streams draining into the valley, mainly from the west, were the flint source for Acheulian and later knappers. These include A detailed geological description of flint exposures within these sources is provided in Appendix A (supplements). Possible secondary sources, flint pebbles found within streambeds, are specified in paragraph 2.1 (and seen in Figure 1). It is important to emphasize that the Late Pleistocene to recent streambeds and alluvium are younger than the BYF and the Acheulian sites. Therefore, flint pebbles and cobbles transported in these streams could not have been the source of raw materials for Acheulian flint tools. However, recent streambeds can serve as indicators for confirming or rejecting the presence of flint sources upstream from their drainage basin.

| METHODS
The research team used two complementary methods to evaluate the and Mount Meron's vicinity (West of B). We also surveyed exposures of conglomerates for possible flint sources, primarily the Hazor conglomerate exposed in the southwest part of the valley (Figure 1).
In addition to identifying and describing the flint present at each locality, we marked the flint dimensions and quality to assess its potential as raw materials for tool production.  (Finkel et al., 2019(Finkel et al., , 2022Finkel, Gopher, Sharon, et al., 2020).

| Elemental analysis
The samples were ground in two steps before their dissolution and the determination of their geochemical composition using inductively coupled plasma mass spectrometry (ICP-MS). First, the whole object, either collected at the outcrops or flaked from the GBY handaxes (30-100 g), was ground using a tungsten carbide heavy-duty grinder (Retsch BB 100 Jaw Crusher) and homogenized. After homogenization, an aliquot of~10-20 g was pulverized using a tungsten carbidebased mill (Retsch RS 200) set to 700 rpm for 1 min.
Approximately, 0.1 g (out of~30-100 g) of the pulverized homogenized samples was digested in PTFE (Savillex) beakers using distilled concentrated HF and HNO 3 (e.g., Ben Dor et al., 2018;Palchan et al., 2013). First, 1 mL of distilled HNO 3 (70%) was added to 0.1 g of the crushed sample. After 1 h, 2 mL of distilled HNO 3 (70%) and 2 mL of distilled HF (50%) were added, and the sample was heated in the closed beaker at 80°C for 24 h. Following this initial digestion, the sample was evaporated nearly to dryness and 2 mL of pure distilled HNO 3 was added and evaporated nearly to dryness once more, after which 5 m: HNO 3 (1%) was gradually added to the sample. The analyzed elements were: Li, Na, Mg, Al, K, Ca, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, As, Se, Rb, Sr, Y, Mo, Ag, Cd, Ba, Tl, Pb, Th, and U and the rare earth elements (REEs) La, Ce, Pr, Nd, Sm, Eu, Gd, Tb, Dy, Ho, Er, Tm, Yb, and Lu (Supporting Information: Table 1).

| Statistical analyses
The statistical analyses were designed to compare the handaxes substantially different, to be meaningfully clustered using the full set of measured elements. The pairwise distance matrix was calculated for the entire data set of provenance after performing a centered log-ratio transform (Aitchison, 1982;Buccianti & Grunsky, 2014 and references therein) and z-score normalization to zero mean and unity standard deviation using the scales determined for the source specimens (e.g., Acevedo, 2012;Ben Dor et al., 2023). In this approach, the entire data set of measured concentrations is used, rather than a small subset of preselected specific elements. After calculating the pairwise Euclidean distance between the specimens using the normalized CLR matrix (Bandyopadhyay et al., 2007;Kassambara, 2017), the distance matrix was used to hierarchically cluster the samples by linking the observations according to their similarity. This is depicted by the dendrogram that presents the relative similarity between the different observations (i.e., a "clustering tree"), using the average linkage method. After building the dendrogram, its cophenetic correlation was calculated to evaluate how well the dissimilarity (distance) of the samples is reflected by the two-dimensional structure of the dendrogram (e.g., Sokal & Rohlf, 1962).
The dendrogram was split into clusters based on the number of groups, but without consideration of the samples' position on the dendrogram, so that the position of the labeled archaeological artifacts reflects their likely grouping (see Finkel et al., 2022, for additional details).
Following the unsupervised clustering of the data, several classifier functions were trained and applied. Due to the limited amount of data, and a large number of measured features, the most meaningful elements for provenance determination were selected based on the p value of the Kruskal-Wallis test (Kruskal & Wallis, 1952) after accounting for multiple hypothesis testing (Šidák, 1967 were evaluated by calculating the class-wise percentage of correctly labeled items in the test data set (Supporting Information: Figure S1).
The GBY and MB artifacts were consequently classified using each of the trained models. The average classification for the GBY and MB artifacts was consequently determined by averaging the results of each of the models trained for each of the algorithms.
2.5 | Chondrite-normalized rare earth elements patterns REE normalization is another common method used in the study of prehistoric flint (Nathan et al., 1999  A. Cenomanian flint in the Naftali Mountains: Previously described by Glikson (1964Glikson ( , 1966b and Kafri (1991), these strata were C. Eocene flint in the Safed Mountains: The Eocene succession in the mountains of Canaan and Amuka previously described by Shiftan (1952) and Glikson (1964) (Mor, 1993(Mor, , 1994 was described previously by Michelson (1979). The survey of the upper part of the Shamir Eocene exposure revealed flint nodules, some up to 50 cm in size, occurring as beds within the Eocene limestone.
These nodules are of good quality and were, indeed, used for prehistoric tool making as evidenced by the nearby Middle Paleolithic (MP) Kela flint E&R locality (Finkel, Gopher, Sharon, et al., 2020). previously reported to be either devoid of flint or to contain only a few small angular fragments (Glikson, 1966a;Picard, 1952;Sneh & Weinberger, 2003;Sneh et al., 1998). This observation was confirmed in the surveys conducted as part of this study.
G. The Hazor conglomerate in the southwest Hula Valley: Picard (1952,1963), Glikson (1964), and Horowitz (1973) (2001) and Goren-Inbar et al. (2002) at the top of the GBY site sequence (the "Bar") is still exposed. It does not contain flint nodules nor flint fragments large enough for producing handaxes. It does contain small angular fragments of brown flint, only a few cm in size (Supporting Information: Figure S9). It should be noted that this conglomerate was deposited following the Acheulian layers of GBY, and hence some of the flint exposed may be of anthropogenic origin. Eocene origin is less certain, but not refuted for the 10th handaxe.  (Finkel et al., 2019;reanalyzed here). This is an important result when considering the variety of colors and different patination observed on the tools (see Finkel et al., 2019). It confirms that criteria such as color, pattern, patination, and visual similarity do not, by themselves, determine the source of flint for a given artifact.
Eocene flint of suitable size and quality for the production of handaxes is found in E&R piles in the Upper Galilee, located~18 km from GBY and~20 km from MB in horizontal distance. The difference in elevation between the floor of the Hula Valley at c. 70 m asl and the Upper Galilee outcrops at c. 800 m asl makes the speculated trip to flint sources a substantial journey. Eastern Upper Galilee flint nodules were used for the production of bifaces of the Yiron (Ohel, 1986) and Baram (Ohel, 1991) Acheulian sites. Handaxes were found on the surface (Finkel et al., 2016) and in Barkai and Gopher (2011) flint E&R tailing piles.
Producing an Acheulian flint handaxe typically generates numerous waste flakes, some of which are well-defined as biface manufacturing flakes (Newcomer, 1971). Hence, a considerable amount of raw materials was necessary to produce the c. 3500 handaxes found at MB (Sharon et al., 2022). The total weight in flint of finished tools given the average weight of~340 g per handaxe (Sharon, 2007) is c. 1200 kg. Using the accepted ratio of prehistoric bifacial item weight/initial raw material weight of 20% , we can estimate that at least 6 tons of raw nodules were necessary to produce the MB handaxes. This amount of waste has not been reported to date from any site in northern Israel or southern Lebanon, apart from the Dishon E&R complex. The total number of flakes per handaxe larger than 1 mm ranges from 4500 (Newcomer, 1971) to 20,000 (Madsen & Goren-Inbar, 2004 figure 20). Such numbers are exponentially higher than the number of waste flakes collected from MB (Sharon et al., 2022) or from any other Acheulian site in the Levant. Clearly, flint handaxes were not produced on-site and the lack of flakes at MB, and the absence of any report of flint reduction sites in its vicinity, including in southern Lebanon, suggests the Dishon E&R complex as the source of raw materials for the F I G U R E 6 Summary of classification results carried out using four different algorithms using 4-10 elements. All items (except for GBY2) were consistently attributed to Cenozoic sources, whereas item GBY2 was assigned to a Mesozoic source at a different level of confidence by the different algorithms. CTree, classification tree; LDA, linear discriminant; KNN, k-nearest neighbors; SVM, support vector machine.
production of the MB flint handaxes. However, the Dishon E&R complex lacks the expected number of biface reduction flakes that would indicate the E&R complex as the production site of the finished MB tools. Raw material extraction and the early stages of handaxe manufacturing, the roughout, likely took place at the Dishon E&R. The final stages of the reduction sequence, namely, the shaping and finishing stages (Newcomer, 1971), probably occurred elsewhere, creating large piles of typical biface flakes still waiting to be discovered. We cannot rule out the possibility that the raw material source for the flint handaxes at GBY could have been the Dishon fan even if we doubt that it was the source for the numerous handaxes collected from the surface. The c. 3500 handaxes of MB tell a different story than that of GBY, as our survey clearly demonstrates that the Dishon Stream and all other surveyed rock exposures and streams could not have been the source for such mass production. It should be noted that even if the GBY handaxe makers obtained their raw materials from the Dishon bed or from its fan, the journey to flint was still a few kilometers long and involved a careful search for suitable nodules.

| Implication for human behavior
The notion of "embedded procurement" (Binford, 1979(Binford, , 1980 during the Lower Paleolithic still dominates paleoanthropological research. Researchers claim that the acquisition of raw stone materials was not a separate, intentional activity, but rather integrated with planned subsistence activities such as game hunting. Based on the dominance of secondary source flint in Tabun Cave (Israel), Shimelmitz et al. (2020) suggest that flint was primarily acquired through embedded procurement strategies during most of the Lower Paleolithic. Embedded procurement is also claimed to be a primary method of flint acquisition for tools found in the Middle Paleolithic Amud Cave (Ekshtain et al., 2016).
The data presented here suggest otherwise. As already demonstrated for the basalt use strategies at GBY (Goren-Inbar et al., 2018), the Acheulian inhabitants of the Hula Valley applied highly sophisticated acquisition strategies to obtain raw flint materials to make handaxes at GBY and MB. The limited number of flint tools at GBY (Goren-Inbar et al., 2018) may indicate the exploitation of a sporadic, yet relatively close source of flint nodules such as the Dishon fan, situated 8 km from GBY.
A clearer case of direct procurement of flint as early as the Lower Paleolithic is strongly evident at the site of MB. More than 3500 handaxes were collected at that site; yet waste flakes, expected in the thousands due to the production of this large amount of handaxes (Newcomer, 1971), are almost absent (Sharon et al., 2022).
Moreover, no local source of flint suitable for the production of large items could be found in the vicinity of MB. Our results indicate that the raw material utilized for handaxe production at MB was of Eocene origin. We found abundant, large-sized high-quality Eocene flint raw materials and evidence for Lower Paleolithic knapping activity at the Dishon flint extraction and reduction complex (Finkel et al., 2016(Finkel et al., , 2019. This evidence supports the employment of a procurement strategy whereby the Acheulian handaxe makers walked a distance of c. 20 km from the site to the flint source.
Furthermore, the exceptionally large number of handaxes at MB, representing more than a ton in weight of stones to carry, suggests that such treks were likely conducted many times during the site occupation and formation. Such behavior suggests the involvement of multigeneration information transfer, both regarding navigation to the flint source as well as of methods for its exploitation.
Exceptionally large-scale Lower and Middle Paleolithic stone extraction and reduction activities were reported recently from Libya (Foley & Lahr, 2015) and the Arabian Peninsula (Groucutt et al., 2017;Jennings et al., 2015;Shipton et al., 2018). Such evidence may also imply direct procurement strategies applied by early knappers in those regions.
We suggest that the evidence from MB Acheulian handaxes and the Dishon E&R complex stand as an example of the continuous exploitation of stone raw material sources by humans, at outcrops far from the sites where the tools were used, as early as the Lower Paleolithic.

| CONCLUSIONS
We present here new data regarding the potential flint sources for two key Acheulian sites located in the Hula Valley, northern Israel: