Taxa and morphotype occurrences of 68 select taxa from the 2022-2024 near-seafloor optical imagery transects in proposed conservation network sites in Southwest New Brunswick, Bay of Fundy, Canada.

Study Extent

Southwest New Brunswick, specifically benthic habitats of the Fundy Isles region, containing numerous proposed conservation network sites and Ecologically and Biologically Significant Areas. This includes Passamaquoddy Bay, Deer Island, Campobello Island, The Wolves Islands, and Grand Manan Island. Imagery surveys were conducted across 15 survey days from September 2022 to October 2024.

Sampling

At each target location surveyed, a drift transect protocol was employed using the FOBIS survey package (see Teed et al. 2025a – their Figure 5) maintaining the system ~1 m above the seafloor, with a maximum allowable depth deployment of 200 m (a formal technical report detailing FOBIS set-up, protocols, and analyses is currently underway). A full frame digital camera (D850, Nikon, Mississauga, ON, Canada), oriented downwards, captured individual digital still images (generally ~25 MB, but up to 45 MP; 8256 x 5504 pixels) at periodic intervals. Target image acquisition rate was typically every 30 s; though dependent on speed over ground (SOG), in order not to generate either multiple closely-spaced images (at very low SOG), or only a few widely-spaced images (very high SOG). Primary image settings (ISO speed, aperture, focusing mode – stored in each image file’s metadata; Table 1) were generally set at the start of each annual survey period. A downward-facing high-definition video camera (1Cam Mk6, SubC Imaging, Clarenville, NL, Canada) provided continuous video imagery of the seafloor. In 2023 and 2024 (unavailable in 2022), a forward-facing high-definition video camera (HD Multi SeaCam, Deep Sea Power & Light, San Diego, CA, USA) provided additional video imagery. The forward-facing video imagery allowed the FOBIS operator to make real-time adjustment of the camera height above the seafloor, however, the forward-facing camera only transmitted composite (standard-definition) video to the surface. Table 1 Overview of still camera settings for the Nikon D850 camera deployed as part of the FOBIS survey package in Southwest New Brunswick optical imagery transects from 2022 to 2024. Survey year Transects Focal length Focus ISO speed Shutter speed Aperture 2022 CON002-008 24 mm Auto 800 1/100 s f16 2023 CON029-052 24 mm Auto 500 to 800 1/200 to 1/250 s f13 to f16 2024 CON080-121 24 mm Auto 500 to 800 1/200 s f14 Both video and still image files were recorded and automatically named according to transect number (CON###), date and UTC time (e.g., CON002-20220921T124202). A VA500 altimeter (Teledyne Valeport Ltd, Totnes, UK) provided records of the height of the survey package off-bottom (once every second), and a pair of lasers (10-cm spacing) provided a means to evaluate the field of view (FOV) in both the downward video and still imagery (estimated using ImageJ2 – Rueden et al. 2017). The FOV was estimated using a subset of images and video frames for each year to account for minor system adjustments. For downward-facing video, 10-cm lasers were in view for altitude off-bottom ranging from 0.5 to 2.7 m, while for digital stills, FOV calculations were only possible when altitudes were 0.7 to 1.7 m. To generate an estimate of area covered (m2) for each transect, the average FOV from the downward-facing video across each transect was multiplied by the estimated distance travelled (m). These estimates were then summed across transects to generate overall seafloor area coverage. For the first two years (2022-2023) of the survey, an available underwater ultra-short baseline acoustic positioning system was used (Tracklink 1500HA transceiver with Tracklink 1505B transponder; Linkquest Inc., San Diego, USA). This system proved to be unreliable, and so year one and two survey vessel geographic positions were recorded using a SeaNav 300 Global Navigation Satellite System receiver at the surface (Kongsberg Gruppen ASA, Kirkegårdsveien, Norway). The track line of each transect can be expected to correspond to the actual distance travelled by the camera package over the seafloor, although the offset of the camera system from the vessel was not defined. In year three (2024) a new underwater acoustic positioning system was implemented (uPAP 201-H transceiver/transducer, along with cNODE Micro transponders/beacons, Kongsberg Gruppen ASA, Kirkegårdsveien, Norway). The upgraded positioning system allowed for transects to be mapped according to the FOBIS position (<5-m horizontal accuracy). Year three also allowed for direct comparison of the camera system position in relation to the vessel’s position, where it was found that spatial offsets from the survey vessel ranged up to 50 m astern and 1-10 m port or starboard depending on depth and tidal influences (akin to observations from Teed et al. 2025a). Reported depths (in m Below Chart Datum, BCD) corresponding to the location of the camera track were extracted from the 1-10-m multibeam echosounder and LiDAR bathymetry data (OMG 2007; Figure 3). Depths of video transect observations were categorized into the following depth classes for depth distribution visualization purposes: 10-35 m, 35-100 m, and 100-200 m, selected to reflect survey classification approaches outlined in Lawton and Teed (2026). Drift camera tracks for all transects were plotted in ArcGIS Pro (v. 2.8) corresponding to the date and time when the camera package was in a near-seafloor survey mode. The actual distance traveled (m) by the camera package over the seafloor was calculated using the z-field geometry tool in ArcGIS Pro (ESRI 2021), calculated at the transect level. The distance between still images along a transect was calculated using the “Split Line at Point” tool in ArcMap (v.10.8, Advanced licensing).

Quality Control

A sub-set of digital still images was selected from each FOBIS drift transect, at approximately 20-m spacing, using buffer tools in ArcGIS Pro v.2.8 (ESRI 2021). This aimed to reduce the workload for annotating the presence of organisms of interest (selected a priori), along with anthropogenic debris such as fishing gear and garbage. Still images were deemed unusable and removed from the dataset if there were insufficient light levels, too much turbidity, or if the altitude off-bottom exceeded 1.9 m. This yielded 1699 still images for analysis (of the 5081 collected).

Method steps

1. Organism occurrences in selected still images were analyzed using the image annotation software BIIGLE 2.0 (Langenkamper et al. 2017; Zuroweitz and Nattkemper 2021). Annotations corresponded to a designated label of a taxa (species or higher level: genus, family, order, etc.) or a morphotype. Identifications were limited by the bottom conditions (light penetration, water, turbidity) and the image resolution, focus, and blurriness. As such, only taxa and morphotypes of interest (i.e., organisms that are easily identifiable and distinguishable in imaging surveys; often ≥2 cm in size) were annotated, mostly to the species or genus level. We report ‘taxa count’ as the number of taxa present, of the taxa selected for analysis, within analysed images. For morphotypes, labels were based on visual assessment of morphological traits, for example ‘encrusting sponges’. Taxa and morphotypes of interest were limited to sessile and mobile benthic taxa, with a few burrowing organisms (i.e., cerianthids, sandfan worms). Taxonomic names were obtained from the World Register of Marine Species (WoRMS 2025). Annotations were organized as a label tree in BIIGLE (named SWNB2024), comprising a hierarchical set of groupings of taxa and morphotypes, using the 41 taxa of interest from the 2016-2017 SWNB imagery survey as a starting basis (Lawton 2022; Lawton et al. 2026). Some species and general taxa labels (where species could not be confirmed) were added to the label tree as they were encountered, including labels for fishes and commonly observed sponges in which the annotator was confident in identifying. In the prior survey, fishes were only classified as sculpins and allies (Cottoidei) or unidentified fish, while the current survey aimed to identify all fishes to the lowest possible taxonomic classification (e.g., Pholis gunnellus). As of this report, unconfirmed taxa may later be confirmed by taxonomic experts, including feedback on image examples posted on the community site, iNaturalist.org (iNaturalist 2025). Selected images of organisms were uploaded to iNaturalist (https://www.inaturalist.org/projects/southwest-new-brunswick-imaging-surveys-dfo-science). Feedback from community members resulted in some suggested changes to identifications (namely fishes), which were then applied in BIIGLE. Overall, images were analyzed for occurrences of 68 taxa of interest assigning a ‘0’ for absent or ‘1’ for present. Taxa counts per image were binned into the following categories for visualization purposes similar to Lawton et al. (2026); 0, 1, 2-5, 6-10, and >10 taxa. For the current survey, these bins correspond to 0, 1.5, 3-7, 8-15, and >15% of all taxa analyzed (n=68), respectively. Occurrence was then reported as the number and proportion (%) of still images where a taxa or morphotype was present across a transect (Table 2). Abundance was estimated for select taxa of specific conservation priority (Boltenia ovifera, Modiolus modiolus, Gersemia rubiformis, and Pachycerianthus borealis) and commercial interest (Homarus americanus, Cucumaria frondosa, Strongylocentrotus droebachiensis, Placopecten magellanicus). Abundance (number of individuals visible in each image) was also converted to density by dividing the count by the still image FOV (number of individuals – ind. m-2). We compare our density estimates with other data sources for some species, however, in general we report abundance, binned into categories for visualization; 1: 1 ind. image-1 , 2: 2-5 ind. image-1, 3: 6-10 ind. image-1, 4: 11-20 ind. image-1, 5: 21-50 ind. image-1, 6: 51-100 ind. image-1, 7: 101-200 ind. image-1, and 8: >200 ind. image-1.

68 select benthic taxa

1. Chromista
   rank: kingdom
2. Plantae
   rank: kingdom
3. Animalia
   rank: kingdom

Southwest New Brunswick, in benthic habitats within and surrounding Passamaquoddy Bay, Deer Island, Campobello Island, The Wolves Islands, and Grand Manan Island.