An eDNA-collecting AUV reveals the ichthyofauna of deep canyons off Banyuls-sur-Mer

Banyuls-sur-Mer (Côte Vermeille, north-western Mediterranean), within the Parc naturel marin du golfe du Lion, has long been recognised as a major site of marine biodiversity. This status reflects both the mosaic of coastal habitats and an exceptional scientific tradition anchored by the Laboratoire Arago, further strengthened by the establishment of the Réserve naturelle nationale marine de Cerbère-Banyuls in 1974. At the regional scale, the Mediterranean also emerges as a system in which certain island territories and protected areas may serve as refugia for threatened species, particularly elasmobranchs, reinforcing the value of non-invasive and temporally comparable inventories [1]. Yet along this steep coastline, where the shelf drops sharply into open water, a significant portion of biodiversity remains surprisingly poorly known: that of the deep canyons.

This knowledge gap is, above all, a problem of accessibility. Surveys still rely largely on traditional methods (diving, imaging, sampling), but these rapidly become impracticable as depth increases. Deep habitats lie beyond the reach of scuba diving, and the rugged topography of the canyons makes fishing gear difficult, hazardous, or outright incompatible with conservation objectives; regulatory constraints within and around marine protected areas further restrict extractive approaches. Yet these deep ecosystems, whose functional importance is disproportionate to their perceived remoteness, consistently rank among the most under-sampled worldwide [2–4]. In other words, a fraction of local biodiversity occupies a space that is rarely sampled — and rarely sampled well.

Here, an autonomous underwater vehicle (Fig. 1A) equipped with an environmental DNA (eDNA) collector was deployed to survey the ichthyofauna along a bathymetric gradient extending to 300 m off Banyuls (Figure 1B). eDNA refers to DNA extracted directly from environmental samples (e.g. water, soil or sediment) without prior isolation of organisms, enabling the identification of species present, including those that are rare or otherwise difficult to detect [5–7]. While eDNA has established itself as a key tool for biodiversity monitoring and conservation [6–8], it has also demonstrated a unique capacity to reveal the diversity of groups historically resistant to conventional inventorying, such as sharks [9]. Nevertheless, eDNA remains rarely applied in a truly integrative manner in structurally complex, deep environments — precisely where traditional approaches fail. By mounting an eDNA sampling head on a vehicle capable of surveying complex, deep, and at times regulated habitats, we test an operational strategy for exploring these environments without captures, and without damage to potentially vulnerable ecosystems.

Field Campaign & Methods

From 4 to 17 September 2025, we conducted 23 AUV transects of approximately 2 km each off Banyuls, covering a bathymetric gradient (≈35–300 m). Each transect was run by two AUVs a few minutes apart, providing two independent replicates. Seawater was filtered in situ via a filtration module¹ embedded in the AUV head, integrating a high-capacity filter (minimum filtration volume: 30 L). In addition, three surface transects were conducted as controls (peristaltic pump, two replicates per transect) to compare the surface eDNA signal with that obtained by AUV collection. Filters were preserved in a buffer solution and subsequently processed in the laboratory.

DNA was extracted, amplified using a universal fish primer pair (teleo [10]) targeting a region of the mitochondrial 12S gene, and sequenced using Illumina® high-throughput technology. The resulting sequences were processed with a bioinformatic pipeline optimised for eDNA analysis and compared against a local reference database comprising more than 70% of Mediterranean fish species. ‘Species groups’ corresponding to ambiguous assignments were treated as a single taxon, and read counts were interpreted as a proxy for eDNA signal intensity [11,12].

106 species were detected along the surface transects, while 125 species were recorded across the depth AUV transects, with only 61 species shared between the two approaches. AUV-based transects thus provide complementary information to surface transects, consistent with access to deep habitats that are typically absent from conventional inventories. The robustness of the approach is reinforced by comparison of AUV replicates: the two passes share a substantial proportion of detections yet remain strongly complementary (mean Jaccard index 0.43 ± 0.16), indicating that a single pass would miss a non-negligible fraction of diversity — particularly rare, highly mobile, or weakly detectable species. This complementarity justifies replication as a simple and effective strategy for improving completeness, a crucial consideration in heterogeneous and fragmented environments such as canyon slopes.

In samples collected by AUV between 35 and 300 m, detected species show marked bathymetric structuring: 109 species at depths shallower than 100 m, 53 species between 100 and 200 m, and 41 species below 200 m. Turnover is high, with 53 species confined to shallow depths, 12 strictly deep species (>200 m), and only 21 species present across the full bathymetric gradient (Figure 1C).

AUV transects thus reveal a distinct deep-water compartment, rather than a simple attenuated extension of coastal biodiversity. This contribution is particularly notable for the deep offshore component, which includes mesopelagic fauna rarely documented at the local scale. At depths >200 m, we consistently detect emblematic mesopelagic species, such as lanternfishes (Benthosema glaciale, Diaphus holti, Ceratoscopelus maderensis, Lampanyctus crocodilus, Myctophum punctatum), as well as hatchetfishes (Argyropelecus hemigymnus) and dragonfish (Chauliodus sloani, Stomias boa). The detection of these taxa at these depths is consistent with capturing a signal near the upper boundary of their diel vertical migrations, a key process in mesopelagic ecology [13,14]. In other words, the eDNA-collecting AUV renders legible a compartment that has historically been under-sampled and largely invisible.

Beyond species composition, semi-quantitative proxies (sequence read counts combined with the number of positive replicates) confirm this structuring. Despite the known limitations of read counts (variability in shedding rates, transport, and amplification), dominant profiles are remarkably consistent across replicates and transects. Shallow depths show strong and repeatable signatures from abundant neritic species and/or species with high eDNA production (e.g. Pagellus acarne, Sardina pilchardus, Sardinella aurita, Engraulis encrasicolus), whereas beyond 200 m the dominant signal shifts towards mesopelagic taxa. The 100–200 m zone thus emerges as an interface where persistent coastal signals coexist with the increasing contribution of outer shelf taxa. This concordance between taxonomic structuring and dominant signal structuring reinforces the interpretation of a functionally distinct deep-water compartment.

Chondrichthyans: Detecting the Invisible

We also detect a notable diversity of cartilaginous fishes, a group emblematic of the limitations of conventional approaches. In total, 13 chondrichthyan species (sharks and rays) were recorded out of 125 total species, representing 10.4% of the inventory — a result consistent with the idea that eDNA illuminates a significant share of diversity frequently missed in shark surveys [9]. Remarkably, eight of these conservation-priority species were not detected in the surface transects, highlighting the informational gain afforded by robotic access to depth. These species include benthic taxa and/or those associated with deeper substrates, for which eDNA signal is strongly depth-structured in the Mediterranean [15,17]. Among them are Dasyatis marmorata (NT), Mustelus asterias (NT) and Raja asterias (NT). Also detected are highly mobile species such as Alopias vulpinus (VU), Mobula mobular (EN) and Prionace glauca (NT), as well as a particularly elusive taxon, the bluntnose sixgill shark Hexanchus griseus (NT).

The capacity to detect and monitor threatened chondrichthyans is especially critical given that this group concentrates a major share of the risk of erosion of marine evolutionary history [18].

Conclusion

These observations highlight a simple but central point: deep ecosystems play a major functional role (matter fluxes, trophic connectivity, coast-ocean coupling), yet paradoxically remain among the least observed and least understood — primarily because they are difficult to access. In complex environments such as canyons and outer slopes, where diving, trawling, or certain instrumented approaches are impracticable, unsuited, or constrained by regulation, an eDNA-collecting AUV offers an alternative: to explore without degrading, to survey without capturing, and to render visible communities that have long been out of reach. Off Banyuls, this approach transforms a historical sampling frontier into a space of knowledge. Above all, it suggests a broader idea: the “blank zones” of marine biodiversity are not necessarily poor — they are often simply inaccessible [2–4].

By making them observable, this autonomous underwater sampling system opens a new field for understanding and monitoring deep ecosystems whose importance is recognised, but whose biodiversity remains largely yet to be discovered. This approach proves particularly relevant in the context of biodiversity monitoring within marine protected areas.

References

1. Pichot, F., Mouillot, D., Juhel, J.B., et al. (2025). Mediterranean islands as refugia for elasmobranch and threatened fishes. Diversity and Distributions 31(2), e13937.
2. Danovaro, R., Company, J.B., Corinaldesi, C., et al. (2010). Deep-sea biodiversity in the Mediterranean Sea: the known, the unknown, and the unknowable. PLoS ONE 5, e11832.
3. Puig, P., Canals, M., Company, J.B., et al. (2014). Submarine canyons and their role in deep-sea biodiversity and transport processes. Progress in Oceanography 128, 1–12.
4. Ramirez-Llodra, E., Brandt, A., Danovaro, R., et al. (2010). Deep, diverse and definitely different: unique attributes of the world’s largest ecosystem. Biogeosciences 7, 2851–2899.
5. Taberlet, P., Coissac, E., Hajibabaei, M., & Rieseberg, L.H. (2012). Environmental DNA. Molecular Ecology 21, 1789–1793.
6. Valentini, A., Taberlet, P., Miaud, C., et al. (2016). Next-generation monitoring of aquatic biodiversity using eDNA metabarcoding. Molecular Ecology 25, 929–942.
7. Thomsen, P.F., & Willerslev, E. (2015). Environmental DNA — an emerging tool in conservation for monitoring past and present biodiversity. Biological Conservation 183, 4–18.
8. Deiner, K., Bik, H.M., Mächler, E., et al. (2017). Environmental DNA metabarcoding: Transforming how we survey animal and plant communities. Molecular Ecology 26, 5872–5895.
9. Boussarie, G., Bakker, J., Wangensteen, O.S., et al. (2018). Environmental DNA illuminates the dark diversity of sharks. Science Advances 4(5), eaap9661.
10. Valentini, A., Taberlet, P., Miaud, C., et al. (2016). Next-generation monitoring of aquatic biodiversity using eDNA metabarcoding. Mol. Ecol. 25, 929–942.
11. Barnes, M.A., & Turner, C.R. (2016). The ecology of environmental DNA and implications for conservation genetics. Conservation Genetics 17, 1–17.
12. Rozanski, R., Velez, L., Hocdé, R., et al. (2024). Seasonal dynamics of Mediterranean fish communities revealed by eDNA. Ecological Indicators 166, 112290.
13. Robinson, C., Steinberg, D.K., Anderson, T.R., et al. (2010). Mesopelagic zone ecology and biogeochemistry — a synthesis. Biogeosciences 7, 3419–3436.
14. Brierley, A.S. (2014). Diel vertical migration. Current Biology 24, R1074–R1076.
15. Rozanski, R., et al. (2024). Seasonal dynamics of Mediterranean fish communities revealed by eDNA. Ecological Indicators 166, 112290.
16. Mission BioDivMed (2024). Guide d’utilisation de l’ADNe en milieu marin. Technical document.
17. Ministère de la Transition écologique (2024). Mission BioDivMed 2023.
18. Stein, R.W., Mull, C.G., Kuhn, T.S., et al. (2018). Global priorities for conserving the evolutionary history of sharks, rays and chimaeras. Nature Ecology & Evolution 2, 288–298.
19. Dalongeville, A., Boulanger, E., Marques, V., et al. (2022). Benchmarking eleven biodiversity indicators based on eDNA surveys. Journal of Applied Ecology 59(11), 2803–2813.
20. Mathon, L., Marques, V., Mouillot, D., et al. (2022). Cross-ocean patterns in fish biodiversity on coral reefs through eDNA metabarcoding. Proceedings of the Royal Society B 289(1973).