Sedimentary DNA and molecular evidence for early human occupation of the Faroe Islands

The Faroe Islands were a key stepping stone in Viking expansion across the North Atlantic, and the prevailing view has been that Norse settlers first occupied the islands between 800–900 CE. Yet archaeological and historical hints—including early-dated charred barley grains at A Sondum, medieval accounts of hermit occupation, Celtic-derived place names and grave markings, and asymmetries in modern Faroese genetic ancestry—suggest the possibility of pre-Norse human presence. Direct, continuous, and well-dated evidence has been lacking, in part because erosional reworking of catchment soils complicates radiocarbon chronologies in sedimentary archives. This study asks when humans first arrived in the Eiðisvatn watershed on Eysturoy and whether livestock introduction and associated landscape change predated the Norse. By combining unambiguous molecular indicators of mammals (fecal stanols) and livestock presence (sheep sedaDNA) with robust age control, the work aims to resolve the timing and ecological impacts of the earliest human activity on the Faroes.

Prior evidence for early occupation includes: (1) archaeological finds of charred barley grains at A Sondum dated to 351–543 CE, implying cultivation and human activity before Norse settlement; (2) a 9th-century text by the monk Dicuil suggesting northern islands were occupied by hermits for at least a century; (3) Celtic toponyms and grave markings on the Faroes; and (4) modern population genetics showing predominantly British Isles maternal ancestry and Scandinavian paternal ancestry. Paleoecological records from lakes and bogs document Holocene vegetation with woody taxa (Betula, Juniperus, Salix) and later transitions toward herbaceous dominance. Indicators such as Plantago lanceolata and cereal pollen have been used to infer anthropogenic influence, but age models have often been too uncertain due to erosion-driven incorporation of older organic material. Consequently, previous lines of evidence are suggestive but inconclusive regarding the precise timing of first human arrival.

Study site and coring: Sediment cores were recovered from the deepest part of Eiðisvatn (Eysturoy, Faroe Islands) in August 2015. Two gravity cores (EI-D-01-15, 83.7 cm; EI-D-02-15, 100.7 cm) preserved the sediment-water interface; a percussion-piston core (EI-P-01-15, 281.5 cm) was collected intact. One surface core and the piston core were split, imaged, and scanned with an ITRAX XRF core scanner (UMass Amherst). Cores were aligned on a composite depth scale using XRF and loss-on-ignition (LOI) profiles. LOI was measured at 1 cm3 resolution throughout. Chronology: An improved age model builds on prior work by adding radiocarbon and cryptotephra constraints. Twenty plant macrofossils were dated by AMS (NOSAMS and UC Irvine Keck). Five tephra layers were geochemically fingerprinted by electron microprobe. Age modeling used Bacon (R) with the IntCal13 calibration curve. Eleven age reversals were excluded from the final model. A modern 14C age at 18 cm was assigned 0 BP, and a linear model was applied from 18 cm to the core top (−65 BP) to accommodate a sedimentation-rate change associated with dam construction in 1980. The Landnám tephra (877 CE ± 1 yr) was tentatively identified at 33 cm; sensitivity tests excluding this marker indicated minimal impact on estimated ages for first human presence. Biomarker analyses: Sixty-eight 1-cm-thick subsamples spanning 29–53 cm (high-LOI interval) were processed. Freeze-dried, homogenized sediments underwent lipid extraction (Dionex ASE; 9:1 DCM:methanol, 100 °C). Total lipid extracts were saponified (KOH), and fractions were separated via liquid–liquid extraction (toluene), acidification, and recovery. Neutral fractions were further separated into apolar, ketone, and polar fractions on alumina. Fecal stanols were quantified by GC–MS. Coprostanol (5β-cholestan-5β-ol) and epicoprostanol (5β-cholestan-5α-ol) indicate human/omnivore presence; stigmastanol (5β-stigmastan-5β-ol) and epistigmastanol (5β-stigmastan-5α-ol) indicate herbivore inputs. Concentrations were normalized to organic matter content (wt% LOI) to account for changing sediment composition. Sedimentary ancient DNA (sedaDNA): DNA was extracted from 14 sediment samples at 11 depths spanning the disturbance interval. Processing occurred in a dedicated ancient DNA facility (UC Santa Cruz Paleogenomics Lab) in two batches. Metabarcoding targeted plants and mammals, with three replicate PCRs per metabarcode per extract. A minimum read threshold of five reads per taxon per PCR was applied to score presence, reducing false positives from sequencing error or contamination. Plant barcodes were resolved to genus; mammal reads were assigned to taxa where possible. Read counts and replicate consistency informed confidence in taxon detection. Batch one yielded only low human DNA in mammal assays (likely greater degradation sensitivity due to longer mammal amplicon). Batch two produced robust mammal detections including sheep. Supporting sedimentology and geochemistry: LOI and independent biomarker data (leaf wax concentrations and isotopes; soil-derived GDGTs) were used to infer erosion dynamics and organic input sources, particularly during a mid–late Holocene disturbance interval characterized by high LOI and radiocarbon age reversals.

- First evidence for livestock: Sheep DNA appears in lake sediments during the main disturbance interval at 50.88 cm depth, dated to 1458 BP (95% confidence interval 1580–1343 BP). Three samples in this interval contained sheep DNA in all three PCR replicates, with 7,521–355,512 reads assigned to sheep. - Fecal biomarkers increase: Coprostanol + epicoprostanol (human/omnivore indicator) and stigmastanol + epistigmastanol (herbivore indicator) concentrations, normalized to LOI, rise sharply at ~50 cm depth, dated to 1433 BP (95% CI 1317–1565 BP), exceeding pre-disturbance maxima and peaking again in modern samples. - Convergent age estimate for human arrival: The sedimentary horizon with first sheep DNA and fecal stanol increase is constrained to 1433–1458 BP, equivalent to 492–517 CE, with a latest possible age of 1317 BP (633 CE) at the 95% CI. The abstract’s integrated estimate places arrival at ~500 CE (95% CI 370–610 CE). - Vegetation change consistent with grazing: Plant sedaDNA shows increases in graminoids and forbs after livestock introduction (e.g., Juncus, Eriophorum, Carex, Agrostis, Ranunculus, Epilobium) and decreases in woody taxa (Betula, Juniperus) and certain forbs (Caprifoliaceae, Apiaceae). Salix persists through the disturbance interval. - Erosion and sediment dynamics: A high-LOI interval (≥20% LOI between ~53–29 cm; onset ~1514 BP, 95% CI 1415–1670 BP) coincides with multiple radiocarbon age reversals, interpreted as enhanced erosion of peaty soils delivering old organic matter to the lake—supported by elevated terrestrial leaf wax concentrations and δ13C, and increased soil-derived GDGTs. - Predates Norse settlement: The molecular evidence predates the first documented Norse activity at the nearby Argisbrekka shieling (mid-9th century CE) by ~300 years. The Landnám tephra (877 CE) lies stratigraphically above the molecular evidence, corroborating pre-Norse timing.

The co-occurrence of sheep sedaDNA and increased fecal stanol concentrations within a narrow stratigraphic interval provides definitive, independent lines of evidence for livestock presence and thus human activity in the Eiðisvatn watershed by ca. 500 CE. This directly addresses the long-standing question of pre-Norse occupation, demonstrating that humans and their animals were present several centuries before Viking-age settlements. The stronger rise of stigmastanol relative to coprostanol aligns with expected fecal stanol profiles for sheep (approximately 6:1 stigmastanol:coprostanol), indicating that herbivore inputs dominated and suggesting sheep were numerous relative to humans. The observed turnover in plant communities—decline of woody taxa and rise of graminoids and certain forbs—closely matches patterns expected from sustained grazing pressure and is consistent with pollen and macrofossil evidence from nearby sites. The timing of first livestock evidence overlaps with independent hints of early activity (e.g., early cereal pollen and macrocharcoal at regional sites, charred barley from Sandoy), integrating archaeological and paleoecological lines. Stratigraphic placement below the 877 CE Landnám tephra further supports a pre-Viking horizon. Together, these findings imply that human-driven landscape transformation on the Faroes began well before Norse settlement and that late Holocene vegetation change was primarily anthropogenic rather than climate-driven.

Combining fecal biomarker profiles with sedaDNA metabarcoding from lake sediments, this study provides conclusive evidence that humans introduced livestock to the Faroe Islands around 500 CE—approximately 300–350 years before Viking-age Norse settlements. The latest plausible arrival, given age-model uncertainties, is ~630 CE, still ~200 years before the earliest documented Norse activity. Livestock introduction triggered a rapid ecological shift from shrub/woody vegetation to graminoid-dominated communities, indicating anthropogenic shaping of Faroese landscapes in the late Holocene. The identity of the earliest settlers remains unresolved, but technological timelines for Norse sailing suggest they were unlikely to have been Norse. Future work could target additional watersheds, refine tephrochronological tie-points, expand sedaDNA taxonomic coverage, and integrate high-resolution archaeological surveys to identify the nature and extent of pre-Norse settlements.

- Chronological uncertainties: Despite extensive dating and tephrochronology, age reversals due to reworked old organic material required removal of dates and introduce uncertainty. The Landnám tephra identification is tentative (slight TiO2 offset), though sensitivity tests indicate minimal impact on key ages. - DNA preservation and batch effects: The first batch of sedaDNA samples yielded low mammal DNA, likely due to storage-related degradation and longer mammal amplicons, potentially underdetecting early low-abundance signals. Plant versus mammal detection differs in sensitivity and amplicon length. - Contamination/misassignment risk: Low read counts for human, cervid, and Bos taurus in isolated PCR replicates may reflect contamination or taxonomic misalignment; thresholds and replication reduce but do not eliminate such risks. - Inference limits: Molecular evidence confirms presence of humans and sheep but does not identify cultural affiliation or settlement permanence. Biomarker signals are watershed-integrated and cannot resolve site-specific activities. - Erosion overprint: Enhanced erosion complicates radiocarbon-based chronologies and may affect sediment composition, necessitating normalization and careful interpretation.