Environmental DNA: sea water reveals a beaked whale the SCANS surveys had never seen

Every animal leaves cells, skin fragments, mucus and faeces in the water. This environmental DNA stays detectable for a few hours to a few days, and filtering sea water is enough to find out who has passed by. What the method is worth against conventional counting remains to be established. A team from Vigo measured it by joining the largest cetacean census survey in Europe, producing along the way the largest environmental DNA dataset ever devoted to cetaceans. Its conclusion is more qualified than the one usually ascribed to it: the two methods agree, up to a certain scale.

The protocol

SCANS-IV is the fourth edition of the cetacean census of the European Atlantic waters. From 3 to 31 July 2022, aboard the research vessel Ramón Margalef, the team worked in Spanish oceanic waters to the north and north-west of the Iberian Peninsula, over 270,684 square kilometres split into three blocks: off Galicia, in the north-west Iberian area, and in the Bay of Biscay.

While two observers scanned the sea from a raised platform, with the naked eye and in sea states of 5 Beaufort or below, a pump drew water from 4 to 5 metres deep. Two 4-litre samples per station, pre-filtered at 60 micrometres then filtered on a 0.45 micrometre membrane. Twenty-four ultrapure water controls were processed each day to monitor contamination.

Sea water samples	258, at 129 stations
Kilometres of visual transect	4,667
Visual sightings	349, with 10 species identified
Species detected by DNA	9, at 114 stations (88%)
Combined total	13 species
Detected by both methods	6

The technique used is called metabarcoding. A short region of the genome is amplified, here a portion of the 16S ribosomal RNA gene of marine vertebrates, everything the sample contains is sequenced, then each sequence is compared with a reference database. The primers had been modified so as not to amplify the DNA of birds and sea turtles.

Of every hundred mammal DNA reads, ninety-one are human

Before the results, one figure that says everything about the difficulty of the exercise.

Of the sixteen and a half million raw reads, seven million passed the quality filters. Among them, 97.4% belong to vertebrates, roughly half to mammals and half to fish. But within the mammal half, cetaceans account for no more than 7.66% of the reads. All the rest comes from land mammals: 90.78% Homo sapiens, then pig, cattle, dog, cat and sheep.

In other words, the sequencer spends most of its time reading the crew, the ship’s galley and the discharges from the coast. Cetacean DNA accounts for no more than 3.74% of the usable total. The authors recommend using blocking primers for human DNA, and cetacean-specific primers. This technical detail explains a good part of the failures published in this field.

Even so, the yield is good: cetacean DNA was found at 114 of the 129 stations, with at least two species in 80 samples, and up to five species in two offshore samples.

Thirteen species, and an instructive split

Nine species by DNA, ten by eye, thirteen in all. The split tells more than the totals do.

Seen by both methods, six species: the striped dolphin, the common dolphin, the fin whale, the common bottlenose dolphin, Sowerby’s beaked whale and Risso’s dolphin.

Detected by DNA alone, three species: the long-finned pilot whale, the killer whale and True’s beaked whale. All three are otherwise known from the area, the pilot whale from earlier surveys, the killer whale from summer sightings off Galicia, but none was seen during these four weeks.

Seen by eye only, four species: Cuvier’s beaked whale, the sperm whale, the sei whale and the common minke whale.

This last group is readily explained, and the explanation is a useful one. Cuvier’s beaked whale and the sperm whale are deep divers that spend most of their time far below the surface. The water, though, was drawn at four or five metres, and DNA stratifies vertically in the water column. As for the sei whale and the common minke whale, each was seen only once: their rarity in the area is enough to account for the absence of a genetic trace.

True’s beaked whale, absent from all four SCANS surveys

One detection deserves to be set apart. The DNA of True’s beaked whale was found in five samples to the north-west of the Iberian Peninsula. This species has never been identified in the whole historical series of the SCANS surveys, which began in 1994.

This is not a discovery in the strict sense: its presence in the Bay of Biscay and off Galicia is attested by opportunistic sightings and by the stranding networks. But it is exactly the case in which the conventional census fails. Telling apart two beaked whales of the genus Mesoplodon at sea is very difficult, and the survey did in fact record ten sightings of “unidentified beaked whales” spread across the three blocks. DNA, for its part, settles the matter unambiguously.

For a family several of whose species have never been observed alive with any certainty, this is a considerable contribution: the method supplies an identification where the eye can produce only a category.

At regional scale the two methods agree

This is the point that commentary almost always skips over, and it is the one that decides what use to make of the method.

The four most frequently sighted species are also the most frequently detected in the DNA, at comparable frequencies. The fin whale totals 179 sightings and 39 genetic detections, the common dolphin 64 and 69, the common bottlenose dolphin 26 and 35. Species seldom seen are seldom detected: a single sighting of Risso’s dolphin and a single one of Sowerby’s beaked whale, very little DNA for either, and in the same block.

The density maps point the same way. For the fin whale, the DNA hotspots overlap the sightings across almost the entire study area, off Galicia as in the central Cantabrian Sea. The statistical analyses of composition by block give concordant results between the two methods.

At local scale they no longer overlap at all

The authors then narrowed the focus. Around each station they defined a local area of about 93 square kilometres, whose radius matches the greatest sighting distance recorded during the survey, 5,450 metres. They then compared, station by station, what the DNA had found and what the observers had seen within that circle.

The coincidence rate falls to 12%. Across all the cases, only 33 have a genetic detection matching a sighting. Against that, 212 cases where the DNA detects a species nobody saw, and 25 the other way round. Even for the best-monitored species, the fin whale, agreement peaks at 24%, and it falls to 6% for the common bottlenose dolphin.

This is not a contradiction, it follows from what each of them measures. The observer sees an animal present now in their field of view. The sample captures water that was alongside an animal a few hours earlier and has drifted since, in a sector the authors describe as crossed by very dynamic currents and transient eddies, a subject set out in detail on our page about ocean currents. The presence of an animal is a discrete fact; its genetic trace is a continuous phenomenon, diluted and transported.

The authors’ conclusion is unambiguous: it would be wrong to look for a direct correspondence between the presence of an animal and that of its DNA in the vicinity. But they add the essential point: at larger scale, with sufficient sampling coverage, this local instability dilutes and ceases to be a problem.

The practical lesson is therefore precise, and counter-intuitive. Environmental DNA is reliable for saying what lives in a region; it is not reliable for saying what is at a given point.

The case of the striped dolphin

One result for a single species deserves to be set apart, because it reveals a bias in conventional counting rather than a quality of DNA.

The striped dolphin is the species most frequently detected by DNA, in 81 samples spread across the three blocks. Yet it was identified only 21 times by the observers, far behind the fin whale and the common dolphin.

The explanation lies partly in its sociability, a large group releasing more DNA than a lone animal. But it lies above all in a problem of identification. At medium and long range, the striped dolphin is readily confused with the common dolphin. Observers then record it as “unidentified dolphin” or “common or striped dolphin”: the survey counts 28 sightings classed in this way, 8% of the total. Those sightings drop out mechanically from the analyses by species and from the abundance estimates.

In other words, a common species may be under-represented in the official data not because it is rare or inconspicuous, but because it resembles its neighbour. This is a bias of method, and it can be corrected, provided it has been measured. That is precisely what the DNA allows here.

Two base pairs between two species

DNA has blind spots of its own, and the study documents them without indulgence.

On the gene fragment amplified, some delphinids differ by no more than two base pairs. That is the case for the common bottlenose dolphin against the common dolphin, and for the common bottlenose dolphin against the striped dolphin. Nine sequences could therefore not be assigned automatically to a species, and two remained undetermined after manual examination.

Worse, two sequences were assigned by the algorithm to the pantropical spotted dolphin, a tropical species absent from the area. On checking, they matched the common bottlenose dolphin just as well, but the ranking order of the results had tipped the machine the wrong way. The authors draw a rule from this: every assignment, automatic as much as manual, has to be checked case by case.

They also flag the risk carried by the correction itself. Ruling out non-resident species on principle is convenient, but species move and ranges change under the effect of climate change. A method that can find only what it expects to find will never detect an arrival.

The same obstacle holds elsewhere. Metabarcoding does not reliably separate the three large rorquals of the North Atlantic, blue whale, fin whale and sei whale, which made it necessary to design dedicated assays for them, species by species, published in August 2026. The method sees what it has been asked to look for.

Four strengths and three structural limits

It detects the elusive. That is its main contribution, and True’s beaked whale is the demonstration of it. The pilot whale and the killer whale are the same case: present in the area, absent from the sightings of the survey, recovered from the water.

It depends on neither light nor weather. Visual observation stops at nightfall and beyond 5 Beaufort. A water sample does not. Over a survey, that represents a considerable share of sea time otherwise lost.

It calls for no specialist on board. A crew trained in a few hours can take samples, which opens the door to ferries and merchant vessels, and so to time series that would otherwise be impossible.

It gives the diet into the bargain. Nearly half the reads from this survey belonged to fish. A Portuguese study published in PLOS One in October 2024 exploited this property systematically: from 64 samples taken by bucket along a transect, it identified five cetacean species against four by eye, and above all nineteen fish species in the same samples, among them sardine, sea bass and anchovy. Neither observation nor acoustics supplies that, and it bears directly on our pages about diets.

Set against that, three structural limits.

1. It does not count. That is the decisive limit. The quantity of DNA recovered depends on the number of animals, but also on the time elapsed, the currents, the temperature, behaviour and group size. Going from a concentration to a population figure remains out of reach, whereas that is precisely what the visual surveys produce and what managers ask for.

2. It does not localise finely. The 12% local agreement says as much. Our distribution maps rest on a spatial precision that DNA does not supply.

3. It does not see at depth. Sampling at four metres means missing the animals that live at a thousand. The authors propose sampling at several depths, down to the bottom, to check whether this improves the detection of divers.

The volume filtered changes everything

A technical detail that is nothing of the sort, and which explains a good part of the failures published.

The Spanish team filtered eight litres per station. A study carried out in British Columbia and published in 2023 in Environmental DNA did the opposite: fifteen stations in areas known for their use by cetaceans, with samples of low volume, about 250 millilitres, a large glass of water. Out of 120 paired samples, a single one gave a positive detection, for a humpback whale.

The detection rates there work out as follows: 18% for visual observation, 4% for acoustics, less than 3% for DNA. The authors conclude that larger volumes and species-specific approaches are needed. The gap with the SCANS-IV survey, which finds nine species at 88% of its stations, is therefore not a matter of luck: it is a matter of protocol.

Three methods, three blind spots

Cetology now has three families of tools at its disposal, and this comparison argues for using them together rather than setting them against each other.

Visual observation gives population figures and behaviour, but only by day, in fair weather, and it stumbles on species that look alike. Passive acoustics covers the night and the divers, but counts detections rather than animals, as our page devoted to acoustic monitoring explains. Environmental DNA detects the elusive, separates look-alike species and informs on diet, but counts nothing, does not localise finely and stays blind below the surface layer.

None of them is sufficient on its own, and the very title of the Vigo study says so: DNA is a complementary tool. Its authors note moreover that in each of the three blocks the number of species recorded rises as soon as the two approaches are added together, going for instance from nine to eleven in the north-west Iberian block. Our page on tracking technologies compares these approaches, and the one on demographics shows why counting remains the central requirement, the one that DNA still does not meet.

Sources: Álvarez-González M., Saavedra C., Rotllant J., Pierce G. J., Gutiérrez-Muñoz P., Vázquez Bonales J. A. and Suarez-Bregua P., “Environmental DNA (eDNA) metabarcoding as a complementary monitoring tool in cetacean visual surveys: from broad to fine-scale spatial analysis in open waters”, Marine Environmental Research, volume 217, article 107985, 12 March 2026, DOI 10.1016/j.marenvres.2026.107985, Creative Commons Attribution licence. Instituto Español de Oceanografía and Instituto de Investigaciones Marinas, Vigo. Robinson C., Migneault A., Dracott K. and Glover R., “Seas the DNA? Limited detection of cetaceans by low-volume environmental DNA transect surveys”, Environmental DNA, volume 5, 2023, DOI 10.1002/edn3.485. Afonso L., Costa J., Correia A. M., Valente R., Lopes E., Tomasino M. P. et al., “Environmental DNA as a complementary tool for biodiversity monitoring: A multi-technique and multi-trophic approach to investigate cetacean distribution and feeding ecology”, PLOS One, 16 October 2024, DOI 10.1371/journal.pone.0300992.