Krill were sampled as part of a line-transect survey carried out from RV Investigator using a calibrated scientific echosounder operating at 120 kHz. The echosounder transducer was drop keel-mounted and faced vertically downward. Acoustic echoes arising from krill were isolated using the swarms-based method described in Krafft et al. (2021) to a maximum depth of 250 m. Krill swarm metrics describing morphology (e.g. length), energy (e.g. mean volume backscattering strength (Sv, dB re 1 m-1; see MacLennan et al. (2002)) and position (e.g. mean depth of swarm) were extracted for each krill swarm using Echoview version 13.1 (Echoview, Hobart, Australia) using validated methods (e.g. Tarling et al., 2009; Cox et al., 2011). Krill echoes were then integrated along transect on a one nautical mile by 250 m grid. The integrated values, called Nautical Area Scattering Coefficient (NASC; MacLennan et al., 2002) with units (m²/NM²), are a linear measure of average acoustic energy scattered by krill and were converted to a real krill biomass density (g wetmass/m²) following standard methods (Krafft et al., 2021; Cox et al., 2022).

A Canon HFG10 camera in a waterproof housing was attached to the CTD rosette to enable filming at each CTD station. Once the CTD was approximately 30m from the sea floor and still descending, the camera and lights were switched on and filming commenced. Filming continued as the CTD reached approximately five m from the seafloor at which point it stopped descending and maintained this depth for a further five minutes. After five minutes at the seafloor the camera and lights were switched off and the CTD made its ascent to the surface. Once the CTD reached approximately five m from the surface, the process was repeated, so that 5 minutes of footage at both the bottom and the surface was obtained at each CTD station.

In total, surface and seafloor videos from 50 CTDs were used for analysis (255 minutes footage at both the bottom and surface). Abundance data was obtained first as a binary yes/no occurrence, whereby videos were watched in their entirety and then categorized as either containing at least a single krill (‘yes’) or no krill (‘no’). ‘Yes’ videos were then re-watched twice to gain abundance data. To account for movement of the camera and currents, we used two measures of abundance to form the basis of a categorical abundance classification. First, we measured MaxN - the maximum number of individuals in a single frame, which is usually used for stationary baited underwater videos with a single maximum obtained over a set filming period (e.g. Gardner and Struthers, 2013; Campbell et al., 2015; Schobernd et al., 2014). Since MaxN measures are typically taken by cameras fixed to the substrate, this method alone underestimated the abundance when the camera was substantially moving. Therefore, we also used absolute abundance by counting all krill occurrences throughout the 5-minute videos, which is usually used when calculating abundance from moving transects (e.g. French et al., 2021; Pelletier et al., 2011). A combination of these methods meant that our data fell broadly into three abundance categories: “low”, < 10 krill; “medium”, 10–100 krill; and “high”, > 100 krill.

Trawls were conducted with a Rectangular Midwater Trawl (RMT 8 + 1). This incorporated two nets: a primary net which was optimized for adult and juvenile krill with an 8 m² mouth opening equipped with a flow meter, and a 4 mm mesh aperture (RMT 8) and a smaller secondary net optimized for zooplankton and larval krill with a 1 m² mouth opening and a 300 µm mesh aperture (RMT 1). At some locations, only one of the two nets were successfully deployed. Trawls were either conducted as target trawls (n_RMT8 = 15; n_RMT1 = 13) which were acoustically targeted at krill swarms to capture a sample of the target (10 to 108 m deep), or routine trawls (n_RMT8 = 38; n_RMT1 = 35) at pre-determined stations to examine the distribution of zooplankton, which were performed as oblique trawls from 200 meters depth to 10 meters depth. RMT 8 samples were sorted into broad taxonomic groups, photographed, individuals counted and preserved. RMT 8 target trawls capturing E. superba were subsampled with a 1 liter jug and 200 individuals were randomly selected for measuring body length, sex and maturity stage. The number of individuals of other krill species caught in the net during routine trawls were counted. When the number of individual krill was likely to be more than ~1,000 individuals, approximate number of individuals were estimated by measuring total volume of the species in milliliters using a volumetric cylinder, and multiplied by conversion factors of 16.1, 29.7, 16.1 for E. crystallorophias, Thysanoessa macrura, and E. triacantha respectively (assuming wet weights of 62.2, 33.7, and 62.2 mg, with density of 1), to derive individual numbers. Numerical densities of these species were expressed in individuals per 1000 m³. RMT 1 samples were preserved in 5% buffered formaldehyde and seawater solution for later analysis in the laboratory at the Institute for Marine and Antarctic Studies, Hobart, Australia (Weldrick et al., 2024). After rinsing in filtered seawater and splitting into subsamples (Motoda, 1985), individual euphausiids were enumerated and identified to species level for E. superba and T. macrura, or grouped into “other Euphausiids”, using a Leica M165 C stereoscopic microscope. The counts were adjusted based on the splitting ratio and the calibration value from flowmeter readings. Numerical densities were calculated for each euphausiid group per sampling site and expressed as the number of individuals per 1000 m³.

Antarctic krill eDNA detection and abundance was explored using the short marker, as this marker was more sensitive than the long marker and detected Antarctic krill in more samples than the long marker. eDNA age was calculated using both species-specific qPCR markers (short and long), and both eDNA abundance and age were visualized separately for CTD surface and seafloor samples using R packages SOmap (Maschette et al., 2019) and “ggplot2” (Wickham, 2016). To assess whether sampling depth (surface vs seafloor) affected the likelihood of detecting Antarctic krill eDNA in CTD samples, we used a binomial generalized linear model with logit link and with krill eDNA detection (eDNA present/absent) as the dependent and sampling depth (surface/seafloor) as the explanatory variable. To assess whether eDNA amounts decreased with increasing distance from two separate krill swarms, a linear model with eDNA amounts as the dependent variable and distance from swarm and separate swarms as explanatory variables was fitted. Using this model, we extrapolated the distance from swarms at which the swarm’s eDNA signal may still be detected, using the lowest value of a recent eDNA detection in CTD samples as the threshold of detectability (26 copies).

Using the euphausiid-specific metabarcoding marker, presence of different krill species, relative abundance within samples and total estimated euphausiid eDNA abundance was calculated and visualized separately for CTD surface and seafloor samples as well as for the two krill swarms where samples were taken at increasing distances.

To compare the two eDNA analysis methods (qPCR and metabarcoding), the co-occurrence of Antarctic krill eDNA detections was investigated and visualized, taking eDNA abundance and age into account. Specifically, the overlap of Antarctic krill eDNA detections was visualized with an Euler diagram (R package “eulerr”, Larsson, 2020). The relationship of eDNA age, abundance and detection overlap between methods was further visualized using stacked barplots in R package “ggplot2”, separately for overlapping and non-overlapping eDNA detections. The relationship of metabarcoding eDNA abundance with qPCR eDNA abundance was explored using a linear regression model, with sampling depth (surface or seafloor) included as an explanatory variable.

Surface eDNA samples collected from CTDs were compared to acoustic areal krill biomass estimates in the immediately surrounding waters. The acoustic data were converted to presence/absence by thresholding at the 25th percentile of the non-zero acoustic biomass values. For each surface eDNA sample, acoustic presence was collated up to 15km either side of the CTD station location. The relationship between acoustic presence and eDNA age class was examined by fitting a binomial generalized additive model with logit link using the mgcv R package (Wood, 2017), with a cyclic smooth term for normalized time of day and distance from station, and a factor term for eDNA age class. Models with combinations of those terms were compared using Akaike’s information criterion (AIC) and comparisons between age classes were conducted using Tukey’s multiple comparisons method and the emmeans R package (Lenth, 2024). These analyses were repeated with distances of 5 and 10 km, and thresholds of 0 and the 10^th percentile of the non-zero acoustic biomass values, with equivalent results (not presented here).

Krill are known to exhibit diurnal vertical migration as well as variations in swarming behavior (Tarling et al., 2018; Bahlburg et al., 2023). The normalized time of day term in the above model allows for acoustic detection probability to vary according to the solar cycle. The times of midnight, sunrise, noon, and sunset around each acoustic biomass estimate were calculated using the suncalc R package (Thieurmel and Elmarhraoui, 2022). The normalized time of day was calculated by linear interpolation of the acoustic sampling time to a value in the range of 0 to 2, where 0 corresponds to midnight, 0.5 to sunrise, 1.0 to noon, 1.5 to sunset, and 2 to midnight.

Visual detections and abundance estimates of Antarctic krill were plotted separately for surface and seafloor samples using R packages “SOmap” and “ggplot2”. To assess whether sampling depth (surface vs seafloor) affected the likelihood of detecting Antarctic krill visually, we used a binomial generalized linear model with logit link, and with krill visual detection (krill present/absent) as the dependent and sampling depth (surface/seafloor) as the explanatory variable. The relationship of visual abundance, eDNA abundance and eDNA age was explored using samples of known age (recent or old, undetermined age was excluded). Recent or old eDNA samples were only associated with low or medium visual abundances at the surface, therefore a binomial generalized linear model with logit link and with visual abundance as the dependent and eDNA abundance and age (recent or old) as explanatory variables was used. This data was further visualized using stacked barplots (R package ggplot2). There was not enough overlap between visual and eDNA detections at the seafloor to conduct meaningful statistical comparisons of eDNA abundance and age with visual abundance.

Trawl detections and abundance estimates of euphausiid species were visualized separately for adult or juvenile krill (data from RMT 8 catches) and larval krill (data from RMT 1 catches) using R packages “SOmap” and “ggplot2”, separately for target and regular trawls. Differences of krill densities between target and regular trawls were explored separately for RMT 8 and RMT trawls by summarizing detections (number of detections; minimum, maximum and average densities detected) by trawl type and species detected. Trawl locations were matched to CTD locations if they were within 2,794 m of each other (see results – estimated distance of eDNA detectability from krill swarm), and trawl results were compared to eDNA results for trawls where both RMT 8 and RMT 1 data was successfully collected. Overlap of krill detections between trawls and eDNA were visualized using an Euler diagram (R package “eulerr”), and the relationship of krill densities determined by trawls with eDNA data was explored using a linear regression model with trawl krill density as the dependent variable and krill eDNA abundance, age and trawl type as the explanatory variables.

Limit of detection, limit of quantification, primer efficiency and standard curve intercept for the short and long marker are listed in Supplementary Table S1 and illustrated in Supplementary Figure S1 .

Using the combination of the short and long marker, Antarctic krill was detected in most CTD surface samples (44 out of 50 samples), with short marker copy numbers varying from 26 to 2734 (median: 276) copies per sample ( Figure 2A ; Supplementary Figure S2 ). Based on eDNA fragmentation, 15 samples were classified as “recent”, 21 as “old” and eight as “undetermined”. Recent samples were largely found in transect 2 and further south, whereas many old samples were detected in transect 4 and further north ( Figure 2A ).

In CTD samples collected near the seafloor, substantially fewer samples contained Antarctic krill eDNA (29 out of 48, Figure 2B ) than surface samples (44 out of 50, z = 2.985, p = 0.003, Figure 2A ). eDNA abundance in seafloor samples was generally lower than in surface samples: apart from of one sample (3609 copies, recent), seafloor samples had less than 1000 copies (ranging from 20 to 955, median 158, Supplementary Figure S2 ). When compared to the surface, fewer samples were classified as recent, but a larger proportion of samples were classified as old in seafloor samples (seafloor: six recent, 15 old, eight undetermined, Figure 2B ), however, due to small sample numbers, eDNA abundance and age patterns could not be formally assessed. Recent eDNA was generally detected further south near the continental slope, while older DNA was detected throughout the survey area. Antarctic krill eDNA was detected at great depth: nine detections were between 3000 m and 4000 m depth (one recent, six old, two undetermined age), four detections below 4000 m (one recent, one old, two undetermined age), with the deepest detection at 4327 m depth (undetermined age) and the deepest recent eDNA detection at 4300 m depth.

At two target trawl krill swarm locations (T15 and T18, Figure 1 ), eDNA was collected while the ship was positioned above the detected krill swarms (two samples each) and at several short intervals as the ship was moving away from/around the swarms. Preliminary analyses indicated that T15 likely falls into the ‘superswarm’ category, while T18 was of smaller, ‘standard swarm’ size (Tarling et al., 2009). While the ship was above these krill swarms, very high short marker copy numbers were detected (T15: 1,213,447 and 117,420 copies; T18: 49,360 and 4,428 copies) – higher than the highest copy numbers for any CTD samples at the surface (2,734 copies). The amount of eDNA significantly decreased with increasing distance from the swarms, and overall at T18 significantly lower eDNA concentrations were detected than at T15 ( Figure 2E ; Supplementary Table S2 ). At both swarm locations, the short marker copy numbers at the furthest distance measured from the swarm (T15: 837 m; T18: 739 m) was still greater than any surface detections from CTD samples (T15: 10,976 copies; T18: 3,730 copies). When the linear model was extrapolated to the lowest short marker copy number of any recent Antarctic krill eDNA detection (26 copies), the estimated distance at which the recent eDNA signal may still be detectable from a krill swarm was 2,794 m (T15) and 2,107 m (T18).

None of the negative PCR controls (36 NTC qPCRs) or the negative DNA extraction controls (54 qPCRs) amplified with either species-specific marker. Of the 11 field controls (66 qPCR amplifications), one qPCR amplification was positive. As only one of three technical replicas amplified, this control sample was still considered to be krill negative.

Across all samples, five euphausiid species were detected: Euphausia superba, E. frigida Hansen, E. crystallorophias Holt & Tattersall, Thysanoessa macrura G.O. Sars and T. gregaria G.O. Sars. In the 50 CTD surface samples, predominantly two species were detected: E. superba (in 24 samples) and T. macrura (in 13 samples, Figure 2C ). Traces of eDNA of E. frigida were found in one sample and of T. gregaria in two samples. The two southernmost samples in transect five (near 68°S) contained eDNA of E. crystallorophias.

In seafloor samples, predominantly E. superba was detected (25 samples), with traces of three other species present (T. macrura: three samples; T. gregaria: two samples; E. crystallorophias: one sample; Figure 2D ). When estimates of semi-quantitative copy numbers were considered, generally less eDNA was detected in seafloor samples when compared to surface samples.

At the krill swarm locations where eDNA was sampled as the ship was moving away from the swarm, E. superba was the dominant krill species detected at swarm T15, however, at increasing distances from the swarm, substantial amounts of T. macrura eDNA were detected ( Figure 2F, top ). At swarm T18, only small proportions of reads were assigned to E. superba, while T. macrura was the dominant species detected ( Figure 2F , bottom). At swarm T15, eDNA traces of E. crystallorophias (1 sample) and E. frigida (3 samples) were detected, while at swarm T18 eDNA traces of E. crystallorophias (1 sample) and E. frigida (4 samples) were detected.

There was a large overlap of samples where the species-specific assay (qPCR) and the metabarcoding assay both detected Antarctic krill eDNA (57 samples). However, in 30 additional samples, only the species-specific assay detected Antarctic krill eDNA ( Supplementary Figure S3 ). These detections contained relatively low eDNA concentrations, and eDNA was often fragmented (26–861 copies; 20 samples old, six undetermined, four recent, see also Supplementary Figure S4A ). In six samples, Antarctic krill eDNA was only detected with the metabarcoding marker ( Supplementary Figure S3 ). The semi-quantitative estimate of metabarcoding copy numbers indicated that the eDNA concentrations in these samples was low for Antarctic krill (9–67 copies, median across all samples: 343 copies, see also Supplementary Figure S4B ).

Antarctic krill eDNA copy numbers measured with the short species-specific marker positively correlated with copy numbers estimated from the metabarcoding assay, however, there was also a strong effect of sampling depth ( Supplementary Table S3 ; Supplementary Figure S5 ), indicating that metabarcoding tended to overestimate the amount of Antarctic krill eDNA in surface samples, but not in seafloor samples, when compared to the species-specific marker.

Acoustic detections of Antarctic krill across the survey area were previously reported in Cox et al. (2022) and are shown in Figure 3 . Detection of acoustic biomass was best explained by the binomial generalized linear model that included terms for time of day and eDNA age class (decrease in AIC of 74.6 relative to the intercept-only model, Supplementary Table S4 ). Krill swarms were most likely to be detected around noon ( Supplementary Figure S6 ). There was little evidence to suggest that adding distance (within the 15km limit) to this model further improved the fit (decrease in AIC of 1.4). Pairwise contrasts indicated that the probability of detecting acoustic biomass within 15 km of a “recent” eDNA sample (probability 0.35, averaged over the diel cycle) was significantly higher than for an “old” eDNA sample (probability 0.18; difference t(723) = 4.285, p < 0.001). The probability of detecting acoustic biomass within 15 km of an eDNA sample in which krill was not detected was 0.23, and this was not found to be significantly different from either “recent” or “old” eDNA samples (recent: t(723) = 2.303, p > 0.05; old: t(723) = 1.166, p > 0.05), but we note the low number (n = 6) of “no krill” eDNA samples.

At surface CTD stations, Antarctic krill were visually detected at 18 out of 50 surveyed locations, largely south of the Polar Front, with some higher concentrations detected at the southern ends of transect two and three ( Figure 4A ). Antarctic krill eDNA was detected at every surface CTD station where Antarctic krill was visually detected (18 stations). In addition, Antarctic krill eDNA was also detected at 26 additional CTD surface locations where Antarctic krill was visually absent. At surface stations, there was no evidence that Antarctic krill eDNA abundance was associated with low or medium visual Antarctic krill abundance (likelihood ratio test 0.505, df = 1, p = 0.477, see also Supplementary Table S5 ). However, eDNA age was associated with visual detections: the probability of visually detecting medium krill abundances at surface stations was higher when the eDNA age was “recent” (likelihood ratio test 4.253, df = 1, p = 0.039, Supplementary Figure 7A ).

Similar to the pattern observed with eDNA, visual krill detections at the seafloor were less frequent (seven of 48 CTD station) than visual detections at the surface (18 out of 50 CTD stations, z = 2.365, p = 0.018, Figures 4A, B ). The distribution of visual detections at depth was also similar to “recent” eDNA detections, i.e. concentrated at the southern end of the survey area (south of the Southern Boundary Front), near the continental slope ( Figure 4B ). Visual abundances at the seafloor were generally lower than at the surface, except for a krill swarm detected at the southern end of transect five at 381 m depth, where hundreds of Antarctic krill were caught on camera ( Figure 4B ; Supplementary Figure 7B ) – the same location where exceptionally high concentrations of recent eDNA were detected (see above). Two visual detections were below 3000 m depth, with the deepest detection at 3080 m. At five of the seven visual detection locations, Antarctic krill eDNA was present. At one location (960 m depth, transect one), Antarctic krill eDNA was only detected with the metabarcoding assay, but not with the qPCR markers. At the deepest visual detection location (3080 m, transect three), eDNA was not detected with either method (qPCR or metabarcoding). At both those locations, visual krill abundance was low (1–10 krill, Supplementary Figure 7B ). Antarctic krill eDNA was furthermore detected at 24 seafloor locations where visual detections were absent. There was not enough overlap between visual and eDNA detections at the seafloor to conduct meaningful statistical comparisons of eDNA abundance and age with visual abundance but see Supplementary Figure 7B for a visualization of the data.

Krill densities (individuals per 1000 m³) were estimated for juvenile and adult krill from RMT 8 trawls, and for larval krill from RMT 1 trawls, at routine locations (pre-determined survey design; n_RMT8 = 38; n_RMT1 = 35) and at target locations (after acoustic detection of krill swarms; n_RMT8 = 15; n_RMT1 = 13). Similar to eDNA metabarcoding results, the most commonly detected juvenile/adult krill species across all RMT 8 trawls was E. superba (35 of 53 trawls, mean density of 281.91 krill per 1000 m³), followed by T. macrura (27 of 53 trawls, mean density 25.80 krill per 1000 m³, Supplementary Table S6 ). Results from RMT 8 trawls differed substantially between target and routine locations: At target trawls, E. superba was the dominant krill species: detected densities were higher than at routine trawl locations (Target trawls: mean density of 1023.75 krill per 1000 m³, Regular trawls: mean density of 28.99 per 1000 m³), and krill were detected more frequently (Target trawls: 14/15 locations, regular trawls: 21/38 locations, see Figures 5A, B ; Supplementary Table S6 ). At regular trawls, other krill species were more commonly detected than at target trawls: T. macrura was detected at 25 locations throughout the survey area, E. crystallorophias at four southern locations and E. triacantha at eight northern/eastern locations ( Figure 5A ; Supplementary Table S6 ).

Species detection patterns of RMT 1 trawl data (larval krill) differed both from RMT 8 (juvenile/adult) as well as eDNA metabarcoding data, however, there were no significant differences between routine and target trawl larval detections ( Supplementary Table S7 ).Thysanoessa macrura was the most common euphausiid larval species, detected in 29 of 48 RMT 1 trawls throughout the survey area, both at routine and target locations, with an average density of 234.1 larvae per 1000 m³ ( Figures 5C, D ; Supplementary Table S7 ). Euphausia superba larvae were detected in nine of 48 RMT 1 trawls (both routine and target), and all detections were at the southern end of the survey area near the continental shelf ( Figure 5B ; Supplementary Table S7 ). Trawl results were further compared to eDNA results for trawls where both RMT 8 and RMT 1 data was successfully collected, and where a surface eDNA sample was collected from a CTD within 2,794 m distance. At most locations where juvenile/adult krill were present, and at all locations where larval krill were detected, Antarctic krill eDNA was also detected (19 of 21 and 5 of5, respectively, Figure 5E ). In addition, Antarctic krill eDNA was also detected at six further locations where Antarctic krill was absent from trawls. Antarctic krill density from trawls did not correspond to eDNA abundance or eDNA age, neither for juvenile/adult krill ( Figure 5F ) nor for larval krill ( Figure 5G ).

Antarctic krill eDNA was detected using species-specific qPCR assays (short and long marker) in most surface samples throughout the survey area. These results reflect that although Antarctic krill distribution is patchy in the Southern Ocean, krill swarms move around actively and with currents, and continuously release high concentrations of eDNA, for example through their molts and faeces. This eDNA can persist in the environment long after krill swarms have passed, and can spread further with currents, resulting in a nearly ubiquitous Antarctic krill eDNA presence in Southern Ocean surface waters, particularly of short eDNA fragments.

However, when eDNA abundance and eDNA age (estimated from eDNA fragmentation) were considered, a more diversified picture emerged: when eDNA samples were collected within or immediately above krill swarms, Antarctic krill eDNA concentrations were much higher than at other locations, reflecting the high eDNA shedding rates of dense krill swarms. With increasing distance to the swarms, eDNA concentrations declined at similar rates for both swarms. We estimated the distance at which recent eDNA shed from a krill swarm could still be detected to be up to approximately 2,794 m, however, as we did not collect samples beyond 800 m from the krill swarms, this estimate will need further examination with greater sampling density at greater distances from krill swarms, ideally also taking swarm size, krill density as well as the direction and strength of currents into account (Tarling et al., 2009; Andruszkiewicz et al., 2019). At T15, likely a very large ‘superswarm’, Antarctic krill eDNA concentrations were much higher than at T18, likely a smaller ‘standard swarm’, indicating that swarm size may relate to eDNA concentrations within the swarm. It is not yet clear whether Antarctic krill eDNA concentrations can provide quantitative information that would contribute to Antarctic krill biomass estimates across a survey area. However, as a first step, more frequent eDNA sampling along survey transects and further measurements of Antarctic krill eDNA concentrations within more krill swarms of different sizes would assist in characterizing the relationship between local biomass and eDNA concentrations.

Several aspects of our results suggested that recent eDNA detections indicated the presence of krill swarms nearby: in the vicinity (within 15 km) of recent eDNA detections, the probability of detecting krill swarms with acoustic surveys was greater than in the vicinity of old eDNA detections. In addition, higher visual krill abundance on video footage was related to more recent eDNA detections. In the acoustic data, the likelihood of detecting krill swarms was significantly higher around noon than other times of the day. This pattern is consistent with previous results from the same voyage in which krill densities during daytime were found to be higher than at night (Cox et al., 2022), and with results from other krill surveys of East Antarctica: during the BROKE-West voyage in 2006, lower krill densities were detected at night and krill surface swarms were visually detected during the day (Jarvis et al., 2010), and during the ENRICH voyage in 2019, more krill swarms were encountered during daytime around solar noon than during the night (Miller et al., 2019).

Based on the comparisons to the trawl data, Antarctic krill eDNA detections at the surface largely originated from juvenile or adult krill, with only a small proportion of the detections originating from larval krill. Larval Antarctic krill detections in trawls were restricted to the southern end of the survey, corresponding to the continental slope area, matching findings from Kawaguchi et al. (2010), who found larval krill distributions restricted to the shelf break.

For Antarctic krill surveys, metabarcoding of all krill species provided less differentiated data on Antarctic krill than the species-specific approach: the detection sensitivity was lower, eDNA abundance estimates were less reliable and eDNA fragmentation could not be measured (see also section 4.3). However, euphausiid metabarcoding was very useful to gain an understanding of the distribution, relative abundance and interaction of other krill species. In both eDNA metabarcoding and trawl data, E. superba was the most common species, followed by T. macrura, both following distribution and abundance expectations (Cuzin-Roudy et al., 2014). In RMT 8 trawls, E. superba densities and detection frequencies were significantly higher at target trawl locations than at regular trawl locations, reflecting the nature of the trawls: target trawls were conducted at locations where krill swarms (juvenile/adult krill) were acoustically detected, whereas regular trawls were conducted at pre-determined locations independent of acoustic presence of krill. The trawl type had no influence on the presence or abundance of other krill species, highlighting the specificity of acoustic krill swarm detections. While E. superba eDNA detections likely largely originated from juvenile/adult krill, a large proportion of T. macrura detections may have originated from larval krill: T. macrura was the most abundant larval krill species in RMT 1 trawls, matching findings of previous surveys (Hosie, 1991). Larval T. macrura were more abundant than adult T. macrura, and often co-occurred with adult E. superba. Thysanoessa macrura eDNA was also common around the two E. superba target swarms, and while we did not have RMT 1 trawl data for those locations due to net-opening failure, no adult T. macrura were detected there, thus the T. macrura eDNA likely originated from larvae.

In addition to these two dominant krill species, E. crystallorophias was detected in southernmost samples near the ice edge, the expected habitat of this species (Pakhomov and Perissinotto, 1996), both in eDNA samples (n = 2) and in trawls (n = 4). Traces of E. frigida eDNA were detected in one CTD sample and at both target swarms within the expected habitat (Hosie, 1991), and T. gregaria traces were found in two samples, outside of its expected more northerly range (Kirkwood, 1982). Neither E. frigida nor T. gregaria were detected in trawls, however, E. triacantha was detected in low densities in eight regular trawls, but not detected in eDNA samples. Mismatches between trawl and eDNA species detections could be related to the low densities/abundances of these species, or the mismatch of sampling depth (eDNA: surface; trawls: down to 200 m) or geographic location (CTD and trawl locations were up to 2.8 km apart). In addition, larval krill detections in trawls were only resolved to E. superba T. macrura and “other euphausiid”, thus other species detected with eDNA but not in trawls could have been present in un-identifiable larval form (compare to e.g. Hosie, 1991). Thysanoessa gregaria was the only krill species found outside its expected distribution range. Whether some T. gregaria individuals were lost in the South, whether their eDNA was carried south, e.g. through consumption by southwards moving predators (Reidy et al., 2022; Kawamura, 1980), or whether T. gregaria mitochondrial DNA introgressed into the T. macrura genome via the overlapping ranges of these closely related species (Qi et al., 2014; Seixas et al., 2018) requires further investigation.

Similar to surface samples, determining eDNA age in seafloor samples was essential to understand the presence and distribution of Antarctic krill in the seafloor environment: while low concentrations of Antarctic krill eDNA were detected throughout the survey area, recent eDNA detections were concentrated in the south at the continental slope area, only extending further north in transect two. Visual detections of Antarctic krill at the seafloor were also confined to the continental slope area, suggesting that recent eDNA detections in seafloor samples indicated the presence of Antarctic krill at depth. Recent eDNA detections towards the north of transect two may be related to different water properties discovered along this transect: Foppert et al. (2024) discovered higher dissolved oxygen and lower salinity at the seafloor along transect two, potentially related to an export pathway of Antarctic bottom water from the Cape Darnley Polynya through the Daly Canyon (Foppert et al., 2024). The exact relationship of krill at depth with Antarctic bottom water chemistry will require further in-depth investigation. Both recent eDNA and visual detections were found down to great depths, at 3080 m for the deepest visual detections, and at 4300 m for the deepest recent eDNA detections –considerably deeper than the deepest recorded presence of Antarctic krill to date (3500 m, Clarke and Tyler, 2008). The distribution of Antarctic krill at depth in the continental shelf area may be related to the seabed in this area providing an attractive and attainable alternative feeding ground for Antarctic krill (Schmidt et al., 2011). For example, phytodetritus can accumulate and persist on the continental shelf, providing a long-lasting food source for benthic feeders (Mincks et al., 2005; Smith et al., 2008). In times of relatively high predator abundance and low or patchy concentrations of phytoplankton in surface waters the seabed may present a preferable feeding ground for krill (Clarke and Tyler, 2008; Schmidt et al., 2011). Both individual and cooperative feeding behaviors have been described from benthic camera observations, from including the individual krill “skimming” the surface of the sediment to collect organic material to “balls” of krill resuspending sediment (Kane et al., 2021), and krill have been observed to excrete at depth, contributing to carbon export at depth as well as releasing eDNA into the environment (Smith et al., 2025). At one location at the southern end of transect five, an exceptionally large number of Antarctic krill were captured on camera (> 1000 krill at 381 m depth). At this location, the Antarctic krill eDNA concentration was the highest of any CTD sample (3609 copies), including surface samples, suggesting that high eDNA concentrations at depth corresponded to the presence of high krill numbers. In all other seafloor samples, Antarctic eDNA concentrations and visual krill densities were generally lower than in surface samples, suggesting that fewer Antarctic krill were present in the seafloor habitat, and/or krill were more spread out (generally not forming dense swarms at depth), resulting in lower eDNA concentrations at depth. The formation of dense krill swarms in the epipelagic zone is thought to be a mechanism to avoid visual predators (Olson et al., 2013) – in the darkness of the bathypelagic or even abyssopelagic zones swarming may be less useful to Antarctic krill, and individuals may spread out more at these depths.

Antarctic krill eDNA detections classified as “old” were spread throughout the survey area, including further north than recent detections. While these detections indicate the presence of Antarctic krill eDNA at depth, they may not necessarily indicate the presence of Antarctic krill, particularly beyond the continental slope: fecal pellets and molts released in great quantities by krill swarms in the epipelagic zone are known to sink (Cavan et al., 2019). While during the sinking process, particles may break up or be absorbed by microorganisms (Cavan et al., 2021), considerable amounts of krill debris may still reach the seafloor (Belcher et al., 2017; Smith et al., 2025), contributing to carbon export, and importantly carrying Antarctic krill eDNA to these depths. The sinking rate of krill fecal pellets is estimated to be 27–1218 m d⁻¹ (median 304 m d⁻¹) (Atkinson et al., 2012), and krill molts are estimated to sink at 52–1020 m d⁻¹ (mean 674 m d⁻¹) (Nicol and Stolp, 1989). To reach 3000 m depth, it could take 2.5–111 days (median 9.9 days) for fecal pellets, and 2.9–58 days (mean 5.5 days) for molts. During this sinking period the eDNA contained in the krill debris would likely fragment and reach the seafloor as “old” eDNA (Suter et al., 2023). Detection and quantification of old Antarctic krill eDNA could therefore indicate the amount and location of krill debris reaching the seafloor, which could contribute to estimates of krill-derived high-particulate organic carbon fluxes to the deep ocean (Manno et al., 2020).

The dominant euphausiid species detected in seafloor samples with the euphausiid-specific metabarcoding marker was E. superba. This likely reflects that of the detected krill species, E. superba is the only one known to descend to and actively inhabit the deep-sea environment. In addition, E. superba discarded matter may be more likely to reach the seafloor than the debris of other krill species: unlike other common Southern Ocean krill species (e.g. T. macrura), E. superba can form swarms of immense proportions. The generated localized large amounts of discarded krill matter may over-saturate scavenging communities, making it more likely for the discarded matter to reach the seafloor (Manno et al., 2020; Belcher et al., 2019). In addition, larger molts (such as adult E. superba molts) may sink faster than smaller molts (e.g. juvenile molts or molts of smaller euphausiid species such as T. macrura) (Nicol and Stolp, 1989), and molts can disintegrate within a few days, further reducing the likelihood of small molts reaching the seafloor (Nicol and Stolp, 1989).