Strontium isotope analysis (LA-MC-ICP-MS) of cattle teeth

Cattle teeth were provided by the Archaeological Service of the Canton Thurgau in Frauenfeld, Switzerland. Sampled skeletal elements were investigated for age at death at the Institute of Prehistory and Archaeological Science (IPAS), University of Basel, and published in 2004 [40]. An archaeozoological MNI-based approach to sampling was undertaken to ensure that the teeth and bones originated from different cattle. For Sr isotope analysis sample preparation in Basel, the surface of each tooth was carefully cleaned using a dental burr and hand drill. An enamel transect (~2 mm thick), covering the complete growth axis of the tooth, was cut out using a diamond-impregnated dental disk. The enamel was fixed in a round PTFE mount and embedded in epoxy resin (Biodur^® E12 + Biodur^® E1 hardener 100:28). After 12 hours under vacuum and subsequent hardening in an oven at 35°C for 48 hours, the surface of the resin of the sample mount was ground to expose the surface of the enamel, which was then polished with SiC paper (P1200). The prepared sample mount was then transferred to the National Oceanography Centre in Southampton (NOCS) for Sr isotopic analysis, which was performed on a Finnegan Neptune multi-collector ICP-MS with a New Wave 193 nm ArF homogenized excimer laser. Plasma conditions were optimized for low oxide formation by ensuring ²⁵⁴(UO)⁺/(²³⁸U⁺) < 0.1% [24, 25] to reduce the isobaric interference at m/z 87 primarily due to ⁴⁰Ca³¹P¹⁶O⁺ [41–43], which is a major constituent of the enamel matrix. Additional isobaric interferences can derive from doubly-charged rare-earth elements (REE), calcium-calcium and calcium-argide dimers, ⁸⁷Rb or ⁸⁶Kr. Contribution from ⁸⁶Kr was corrected for by running gas blanks, and an ⁸⁷Rb correction was applied assuming an ⁸⁷Rb/⁸⁵Rb ratio of 0.385617. Dimer interferences were monitored via the ⁸⁴Sr/⁸⁶Sr and ⁸⁹Y was measured as a proxy for REEs. Sr isotope time-series data were obtained by continuous laser ablation along the tooth’s growth axis (at 20 or 25 μms^-1, depending on the size of the tooth). The laser pulse repetition was 15 Hz, the spot size was 150 μm, and laser fluence was 8.6 Jcm^-2. Prior to the actual analysis, the laser track was laser-cleaned using the identical repetition rate and spot size, but at a higher traverse rate (100 μms^-1). Analysis of an in-house ashed bovine pellet standard (BP), with three repeated measurements after every third sample, revealed a mean offset (Laser ablation-TIMS) in ⁸⁷Sr/⁸⁶Sr of +44±33 ppm (1σ), determined from 279 analyses over 18 months. This is similar to the offset reported by other laboratories using a similar methodology [44]. While this shows that interferences could not be completely eliminated, they have been reduced to the point that the reproducibility was much smaller than the observed inter-tooth variability of 952 ppm, and is therefore considered insignificant to our interpretation of the isotope data. The accuracy of laser-ablation-derived ⁸⁷Sr/⁸⁶Sr ratios was further verified by comparison with discrete analyses of micro-drilled samples using conventional TIMS (S2 Fig).

Discrete strontium isotope analysis of baseline samples

For baseline ⁸⁷Sr/⁸⁶Sr determination, dentine of cattle and red deer (Cervus elaphus) was analysed using a LA-MC-ICP-MS in the laboratory facilities at the National Oceanography Centre Southampton, UK. The laser spot was initially cleaned for 5 seconds with a laser repetition rate of 15 Hz and spot size of 150 μm. Data were collected at the same spot in 60 one-second cycles with a laser repetition rate of 10 Hz and a spot size of 150 μm.

For discrete Sr isotope analysis of vegetation samples (collected in spring 2013 and 2014), leaves were rinsed with demineralized water soon after collection, and dried overnight at 40–50°C. Sample treatment followed established methods [45]. Briefly, dried leaves were ground manually and ashed in acid-washed crucibles at 550°C for 12 h. An aliquot of approximately 20 mg of ashed leaves was transferred into clean PFA (Savillex^™) vials and dissolved in 15 M HNO₃. 1 ml of 30% H₂O₂ was added to further dissolve the sample at 140–160°C. Samples were evaporated to dryness and dissolved in 2 ml 3M HNO₃, ultrasonicated and centrifuged, before loading onto ion-exchange columns. Water aliquots (collected in spring 2013 and 2014) were transferred to clean PFA (Savillex^™) vials. Sample preparation followed established methods [45]. Aliquots were evaporated to dryness, dissolved in 2 ml concentrated HCl and dried down. After adding 2 ml of concentrated HNO₃, samples were again dried down. Aliquots were then taken up in 1 or 2 ml 3 M HNO₃, ultrasonicated and centrifuged, prior to loading onto ion-exchange columns. The teeth of pigs (Sus domesticus) were sampled by cutting pieces of enamel representing the complete growth axis using a flexible diamond-impregnated dental disk and hand drill. The procedure is described in [46, 47]. Briefly, the inner and outer surfaces of the enamel were treated with a dental burr to remove any adhering contamination. Enamel samples were additionally cleaned with pure water in an ultrasonic bath, dried and then dissolved in 3 ml 7 M HNO₃. Samples were centrifuged to remove any adhering powder. The supernatant fluid was dried down and dissolved in 3 M HNO₃. An aliquot of this solution, representing 3 mg of solid enamel, was made up to 0.5 ml 3 M HNO₃, before loading onto ion-exchange columns. Dissolved Sr (from either plant, water or enamel samples) was separated using standard ion-exchange chromatography using 70 μl of Eichrom Sr spec resin (50–100 μm). Strontium was separated from the sample matrix by washing the column with 2.5 ml 3 M HNO₃ and eluted using 1.5 ml ultrapure water. After separation, the Sr-containing eluate was dried down. After dissolving in 1 μl 10% HNO₃, samples were loaded onto rhenium filaments, preconditioned with 1 μl TaCl₅ solution and 1 μl 10% H₃PO₄. Isotope ratios were measured on a Thermo Finnigan TRITON Thermal Ionization Mass Spectrometer (TIMS). Data were calibrated using the standard reference material NIST SRM 987 with a reported ⁸⁷Sr/⁸⁶Sr ratio of 0.710248 [48, 49]. All TIMS Sr-isotope measurements were performed in the laboratory facilities of the Department of Earth Sciences, University of Bristol, UK. The typical precision for ⁸⁷Sr/⁸⁶Sr measurements was ± 0.00001 (in archaeological tooth enamel).

Stable carbon isotope analysis of cattle bones

For carbon isotope analysis, collagen extraction followed established methods [50], with modifications [51] and omission of the ultrafiltration step [52, 53]. Chunks of bone were cut and subsamples were taken from a clean surface using a dental drill. 200–1000 mg of cleaned sample material was demineralized in 10 ml of 0.5 M HCl at 4°C for two weeks. Samples were then rinsed with ultrapure water before treatment with 10 ml of 0.1 M NaOH at 4°C for about 24 h. After rinsing, 4 ml of ultrapure water and 200 μl of 0.5 M HCl were added, and the samples gelatinized at 70°C for 48 h. The solution containing the dissolved collagen was filtered using Ezee Filter separators (Elkay, UK) with a pore size of 60–90 μm, frozen at -20°C, and lyophilized for 48 h. For carbon isotope analyses, between 0.5 and 1.0 mg of freeze-dried sample were weighed in duplicate into tin capsules, and introduced into an elemental analyser coupled to an isotope ratio mass spectrometer (EA-IRMS) (INTEGRA2, Sercon Ltd., Crewe, UK) in the laboratory facilities of the Department of Environmental Sciences, University of Basel, Switzerland. Carbon isotope data were calibrated using international reference materials (IAEA-CH-6, USGS40) and an in-house standard (EDTA), and reported in the conventional δ-notation as δ¹³C in ‰ relative to Vienna Pee Dee Belemnite (V-PDB). Based on replicate analyses of the isotopic reference materials, our in-house standard and a selected bone collagen sample which were analysed recurrently in each analytical sequence, the analytical reproducibility for δ¹³C was generally ≤ 0.2‰. Collagen preservation and purity was checked following established methods [54–56].

The landscape and its capacity for cattle husbandry

The interpretation of the cattle movement patterns requires a detailed understanding of the Neolithic landscape and its resources around Lake Constance. The surroundings of Arbon Bleiche 3 were dominated by mixed forests of beech (Fagus sylvatica), oak (Quercus sp.), silver fir (Abies alba), hazel (Corylus avellana), ash (Fraxinus excelsior), and birch (Betula sp.) [16]. As a consequence of long-lasting human activity [14], the landscape resembled a mosaic of densely forested environments and more open settings, both natural (windfall, flood plains, lake shores) and man-made (agricultural fields, fallow ground, orchards) [16]. Human activities, with severe impact on the landscape, created new pasture grounds [57, 58], which aimed at increasing the productivity and resource capacity of the environment [59]. The multifaceted exploitation of the land around settlement sites (e.g., for timber, coppicing, pollarding, hunting, gathering, agriculture, orchards) required sustainable land management practices [58], which implies a certain limitation of land available for animal husbandry in the adjacent forests. Neighbouring settlements with additional resource demands may have created additional pressure on the landscape in the nearby surroundings [60, 61]. Plant remains found in cattle dung demonstrate the importance of wood pasture at Arbon Bleiche 3, and suggest that grazing within and at the edges of forests, and at ruderal places (including lakeshores) was common [62]. Various herding regimes for cattle may have helped to avoid or mitigate overexploitation of the available resources, and diverse (and more distant) grazing grounds were likely targeted. Herd size is an important parameter in this context. A livestock of 30–60 cattle was previously estimated based on excavated remains at Arbon Bleiche 3 [63], which implies an average of one to two co-existing animals per house. Extrapolating the calculated herd size from the excavated area to the whole settlement, we can assume a total number of 60–120 animals. Given individual requirements of approximately 17 ha of land per animal, pasture grounds of 10–20 km² per year were needed to sustain the settlement’s combined herd (SI, section I). Ethnographic studies show that medium-sized cattle herds with more than one animal per house attest to communities that are specialised in animal husbandry or dairying, and which require elaborated herding strategies [64]. Indeed, lipid biomarker evidence on pot sherds from Arbon Bleiche 3 suggests dairying at the settlement [65, 66]. Interestingly, however, not all lactating mother cows seemed to have been kept in the immediate surrounding of Arbon Bleiche 3. Three first molars (ARB 16, 33, 110), representing the time in utero and shortly after birth, exhibited ⁸⁷Sr/⁸⁶Sr above the local range (S1 Fig), which point to some dairying in more outlying regions. We argue that the three mobility patterns in the Arbon Bleiche 3 cattle are reflective of different ecological niches utilized to sustain cattle numbers beyond the carrying capacity of the immediate environment of a settlement. This notion is supported by the observed stable carbon isotope ratios, which suggest that the exploited grazing grounds differed in their respective ecological attributes, e.g., with regards to humidity, forest cover or altitude [28, 29]. Although tooth-substance loss primarily relates to the animals’ age, differences in the soil or fodder composition, relating to cattle movement, can also influence distinct tooth-wear patterns (S4 Fig and Table 1) [67, 68]. The observed Sr-isotope pattern variability between the three groups underscores that cattle management in Arbon Bleiche 3 was highly differentiated, emphasizing the complexity of herding strategies.

Local herding and seasonal movement

The Sr isotope ratios of the individuals representing MP1, accounting for approximately one third of the cattle analysed, are indicative of a static local herd. This group of animals must have been kept in the vicinity of the settlement, not more than a few km away from the site during their enamel mineralization, i.e., the first 24 months. Their C isotope ratios were lowest and most similar to those of red deer (Fig 4), and can best be explained by either leaf and twig foddering in the settlement or feeding in nearby densely forested environments [29]. The observed C isotopic signature would also be consistent with grazing grounds in humid areas [69], e.g., at the shores of Lake Constance or in flood plains. Both ecological niches were readily available in the proximity of the settlement. The absence of any summer-plant remains in cattle dung from the site confirms that there was no year-round husbandry within the actual settlement [62]. Cattle foddering nearby the settlement site implies that local supply of summer browse and winter fodder was sufficient, and that the inevitable disadvantage of losing local fertile land for cattle grazing was compensated for by the obvious benefits of maintaining a local herd (e.g., for milking, slaughtering, and traction purposes). Dairy remains in potsherds [65, 66] as well as the evidence of old animals with traction-related pathologies [40] and a wooden yoke [70] lend clear evidence to the exploitation of these benefits at Arbon Bleiche 3.

In contrast to MP1, the Sr isotope ratios of MP2 individuals suggest periods of both local and non-local grazing. The observed patterns are consistent with a herding management that involves seasonal cattle movement away from the settlement to more Sr-radiogenic catchments, and back. The associated ¹³C/¹²C ratios were higher than for MP1 individuals, consistent with a temporary cattle translocation from the local forested and rather humid environment to less humid landscapes, scattered forests or even open pastures further afield [29]. The co-existence of local cattle (MP1) and cattle displaying isotope patterns of seasonal mobility (MP2) suggests that year-round cattle fodder resources near the site were likely to be fully exploited. That local resources were insufficient to sustain all herded animals is also supported by cattle that carry MP3 (year-round herding away from the site), which represent approximately half of the individuals sampled. MP2 does not necessarily represent long-distance seasonal transhumance, as bio-available Sr with relatively high ⁸⁷Sr/⁸⁶Sr in the same range as observed for cattle teeth occurs within 15 km from the site (S3 Fig). Hence, ⁸⁷Sr/⁸⁶Sr ratios that alternate between local values and values up to 0.7105, e.g., ARB 10, 14, likely reflect periodic mobility in the hilly hinterland of Arbon Bleiche 3. Grazing spots may have varied from one year to the other, but repetitive ⁸⁷Sr/⁸⁶Sr shifts in the composite records suggest that certain pastures were re-visited several times (S1 Fig).

Non-local herding and alpine transhumance

Alpine mountains, i.e., the Alpstein, are located in a distance of only ~30 km from the site, rendering Arbon Bleiche 3 an ideal environment to conduct alpine transhumance (i.e., the summer exploitation of high-altitude pastures). Botanical remains of stone pine (Pinus cembra), alpine speedwell (Veronica alpina) and bone finds of the alpine ibex (Capra ibex) at the study site indicate that the settlers of Arbon Bleiche 3 visited high-altitude alpine regions between 1500 and 2500 m a.s.l. [16]. Moreover, the remains of Swiss clubmoss (Selaginella helvetica), another plant most typical for high-altitude areas (but sporadically found also at lower altitudes and in the region around Lake Constance), were identified in cattle dung [62]. In combination, these findings provide putative evidence for alpine transhumance, exploiting pastures above the timber line (estimated at 1800 m a.s.l. for the mid-4^th millennium BC in the Pre-Alps [71]). The Alpstein area itself represents the closest suitable region for transhumant pasturing, but the topography is rather steep and hard to access. Moreover, isotopic values suggest it is unlikely that the Alpstein served as grazing grounds for cattle from Arbon Bleiche 3. The calcareous bedrock is characterized by ⁸⁷Sr/⁸⁶Sr ratios of 0.70812 ± 0.00032 (1σ) (S1 Table), inconsistent with the Sr isotope patterns observed for the MP3 cattle teeth, which are typically more radiogenic, with ⁸⁷Sr/⁸⁶Sr ratios of >0.7120 (ARB 16, 33, 117). Potential grazing spots with ⁸⁷Sr/⁸⁶Sr values of ~0.7120 are located 10–15 km southeast of the site, but they do not exceed an altitude of 1300 m a.s.l. The closest region at higher altitudes and equally high ⁸⁷Sr/⁸⁶Sr ratios (0.71242 ± 0.00086 (1σ)) is located south of Walensee in a distance of ~50 km from Arbon Bleiche 3 (SI, section II). Alternative pasture grounds with similarly old geological units are located east of the Alpine Rhine, slightly further away from the settlement (>60 km). Thus, the Sr isotopic patterns observed for individuals ARB 33 and especially ARB 117 may be best explained by seasonal translocation between different pastures in more Sr-radiogenic catchments, and therefore reflect a mobility pattern that is typical for alpine transhumance [72]. However, we do not see evidence for large-scale alpine cattle migration. The majority of teeth from the MP3 group show Sr isotope ratios that are consistent with grazing grounds away from, but within ~15 km distance of, the settlement site (i.e., ⁸⁷Sr/⁸⁶Sr values of 0.7083–0.7118), with little change in altitude. Moreover, herding grounds and food resources for animals that display MP3 appeared to be distinct from those of the MP2 and MP1 individuals, as indicated by the higher δ¹³C of bone collagen (Fig 4). The higher ¹³C/¹²C ratios may be related to prolonged grazing in open or less humid environments. For comparison, the low δ¹³C values determined for contemporaneous red deer are indicative of feeding grounds with a closed canopy [29]. The observed Sr isotopic fluctuations of MP3 are more complex than for MP2. Some individuals (e.g., ARB 29, 34, 114) suggest periods of stasis followed by periods of enhanced mobility on a timescale that is consistent with differential summer versus winter grazing (SI, section II). In earlier studies, ethnographic records of seasonal mobility provided evidence for complex mobility regimes, likely driven by grazing pressure and climatic constraints. Active cattle management involved multiple (3–8) herd relocations between spring and fall across different altitudes [72, 73]. Additional factors that may have contributed to the unexpectedly high number of MP3 are cattle exchange and the import of animals that had been born before the foundation of Arbon Bleiche 3, e.g., ARB 117 (house 1), ARB 16, 33 (house 3), and ARB 109 (house 14) (S5 Fig). The acquisition of animals from surrounding settlements (cattle exchange), e.g., bulls for breeding, draft animals for labour, or cattle to maintain and increase the herd size, is not unlikely, and social bonds within and beyond settlements are important for minimizing labour costs [64] and forging alliances (e.g., to gain access to land). Ritual activities, feasting, and the symbolic importance of cattle have been documented for the Neolithic and Bronze Age [22, 23, 74], and their relevance may also be indicated by bone middens and bucrania found at Arbon Bleiche 3 [40].

Specialised herding and differential access to resources

Evidence for specific mobility patterns (MP) appears to be systematically clustered within the settlement (Fig 2). MP1 was primarily observed for individuals found in the northern and central part, evidence for MP2 was focused on the central (and southern) portions of the settlement, and records of MP3 were found all over the settlement area, with a slight predominance in the north. We argue that the MP-clustering attests to differential access to resources by different individuals and groups in the settlement, and to specialised herding strategies to overcome local resource limitations. Local herding (i.e., MP1 and MP2) has likely been organized collectively, with patterns of isotopic variation being sufficiently similar to suggest the use of pastures on similar geological units. Shared herding, certainly the most efficient way of managing cattle, likely required complex interactions, cooperation and networks between the settlers or even between settlements [75–80]. We assume that social alliances were also needed to gain access to specific grazing grounds (e.g., through territorial rights) [81, 82], which apparently differed among the households in Arbon Bleiche 3, possibly depending on the period of occupation or social status. Although patterns of mobility were highly individual among MP3 animals (compared to MP1 and MP2), the range of ⁸⁷Sr/⁸⁶Sr ratios was similar for all cattle during periods of stasis (0.7095–0.7100) (S1 Fig). This suggests perhaps a collective overwintering strategy on the one hand, but the dispersal of animals in summer as small herds with differential access to grazing grounds on the other.