Diversity and time-series analyses of Caribbean deep-sea coral and sponge assemblages on the tropical island slope of Isla de Roatán, Honduras

Shore-based submersible operations, from 2006 to 2020, have allowed us to examine megabenthic assemblages along the island margin of Isla de Roatán from depths of about 150 to 750 m, including repeated observations of the same organisms. These dives were used to photo-document a diverse benthic assemblage and observe the health and condition of the sessile fauna in a well-explored but relatively undocumented area of the Mesoamerican Reef. Samples were collected by dip net, and some dives profiled the water column chemistry in the year 2011. The deep-sea coral assemblage observed off Roatan exhibits high abundance and diversity. The sessile habitat-forming taxa consist primarily of at least 20 different octocorals (e.g., Plexauridae, Primnoidae, Coralliidae, Isididae, and Ellisellidae) and 20 different sponges each (Demospongiae and Hexactinellida), with several known and unknown taxa of Zoantharia, Antipatharia (Bathypathes spp), and Scleractinia (e.g., Desmophyllum pertusum, Dendrophyllia alternata, Madracis myriaster, and solitary taxa). Crinoidea were also abundant and diverse, represented by at least nine species. Epifaunal assemblages associated with corals include at least 24 macroinvertebrate species dominated by Asteroschema laeve (Ophiuroidea) and Chirostylus spp. (Decapoda: Anomura). Repeated observations of a few large octocoral colonies over many years illustrate patterns of predation, recolonization, and epibiont host fidelity, including a 14-year record of decline in a plexaurid octocoral (putatively Paramuricea sp.) and loss of its resident ophiuroids. The shore-based submersible provides a practical and relatively inexpensive platform from which to study coral and sponge assemblages on a deep tropical island slope. The deep-sea coral gardens are likely to harbor new species and new discoveries if more samples can be acquired and made available for taxonomic research.


Introduction
Our knowledge and understanding of the ecology, physiology, biodiversity, and conservation of habitat-forming, deepand cold-water coral and sponge assemblages has broadened substantially over the last two decades (e.g., Freiwald et al. 2004;Davies et al. 2008;Davies and Guinotte 2011;Roberts et al. 2006;Dahl et al. 2012;Larsson et al. 2013a, b;Cathalot et al. 2015;Hourigan et al. 2017;Quattrini et al. 2017;Hebbeln et al. 2019). However, the majority of research has been carried out at higher latitudes in temperate seas, leaving assemblages at subtropical and tropical latitudes less well-investigated (e.g., Messing et al. 1990;Grasmueck et al. 2006;Reed et al. 2006Reed et al. , 2013Demopoulos et al. 2014;Ross et al. 2015;Ramiro-Sánchez et al. 2019). Fewer studies have incorporated time-series observations of the same sites and organisms over extended periods (e.g., Brooke and Young 2009;Kahn et al. 2012;Kuhnz et al. 2014;Purser 2015;Osterloff et al. 2019). The increasingly widespread use of privately funded research submersibles (e.g., OceanX, Nekton Foundation, Curasub) offers the opportunity to conduct and expand in situ deep-sea time-series investigations. Here we report the preliminary results of ecological observations made using a privately owned submersible over a 14-year period.
Isla de Roatán, off the north coast of Honduras, Central America, lies in the western Caribbean Sea at the southern end of the Mesoamerican Reef (Fig. 1), on the southern margin of a left-lateral strike-slip fault, the Swan Islands Transform Fault, near the west end of the Cayman Trough (Rosencrantz and Mann 1991). The transform fault is part of the boundary between the North American and Caribbean plates. The deep island margin supports a high diversity of geologic and topographic features, beginning with the deep reef wall, and, below it, numerous slump blocks and a wide variety of outcrops and escarpments-chiefly limestone above ~ 400-450 m and basalt below-both exposed and sediment veneered and accompanied by extensive areas of unconsolidated sediment. This is the first detailed report of living deep-sea coral and sponge assemblages from Isla de Roatán and from the Mesoamerican Barrier Reef in general. The data discussed herein grew out of the felicitous confluence of independent projects that used the same research platform, the submersible Idabel, owned and piloted by coauthor Karl Stanley and based at the Roatan Institute for Deepsea Exploration, West End, Roatán. The submersible Idabel has been used previously by marine researchers to study the growth of deep-sea Crinoidea (e.g., Syverson et al. 2014), genetics of slit-shell gastropods (J. Harasewych, Smithsonian Institution, National Museum of Natural History (NMNH), pers. comm.), and to describe new fish species (Lipogramma idabeli) named in honor of the sub (Tornabene et al. 2018).

Materials and methods
Submersible Idabel is 4-m long, 2.0-m wide, 2.1-m tall, with space for the pilot and two observers (maximum capacity 220 kg), a 1.05-m diameter observation compartment with a 76-cm diameter viewport, and a depth rating of 915 m (Fig. 2). The dock where Idabel is based is 250 m from the trench wall, which eliminates the need for a surface support vessel for most dives. The lithium batteries in the sub allow for a 16 km range. The sub was towed as far as 5 km, from its base at Half Moon Bay to Anthony's Key Reef (see Fig. 1).
Three field expeditions using Idabel were conducted in 2010 and 2011. These were called Deep Coral and Associated Species Taxonomy (DeepCAST) expeditions (Etnoyer et al. 2011). Over the course of 2 years, thirteen submersible dives were made at eight different locations (Fig. 1). Of these, eight dives proved useful for image-based analysis of species distribution (Table 1). In addition, a brief multibeam survey was conducted by RV Falkor on July 9-10, 2013, to map the steep terrain from 150 to 1500 m depth (NOAA 2013). Subsequent dives were conducted to study the ecology of Crinoidea (Veitch and Baumiller 2021). Nineteen dives focused on crinoids were made from June 2012 to June 2016 in the general vicinity of previous dives. Some dives repeated photo observations of large sea fan corals from previous years. All the above data resources are assembled Dives were made at these and unnamed sites in the same region between 2008 and 2020. Map created in ArcGIS 10.4, using bathymetry data from RV Falkor FK-008T (see NOAA 2013) 1 3 here to illustrate the diversity of sessile benthic epifauna, in a relatively pristine deep-sea environment with remarkable proximity to shore (within 2 km).
DeepCAST I was a 3-day reconnaissance in August 2010. DeepCAST II conducted quantitative photo and video transects of deep coral and sponge habitats at six independent sites in May 2011 with observations spanning a depth range of ~ 60-670 m (Etnoyer et al. 2011). DeepCAST III conducted quantitative dives at three additional sites in December 2011 to a maximum depth of 700 m; a fourth chiefly shallower dive is not discussed here (Lavelle 2012).
Beginning with DeepCAST II, the sub was outfitted with laser scale, cameras, an oceanographic sensor, and a dip net for sample collections. The dip net consisted of a 14-ft (4.3-m) pole mounted to Idabel's port side rail with a steel ring ~ 20 cm across at its front end with a long mesh bag weighted with stones to make it hang. A Sony HDR-HC9 video camera mounted on the forward rail was operated continuously throughout each 4 to 6-h dive. Interior cameras included a Sony HDR-HC9 video camera on a tripod and a Canon SLR still camera, with exterior strobes synchronized to the camera. The still camera operator took ~ 1 photo min −1 with particular attention to coral and sponge colonies and included both wide and medium angle shots of the seafloor. The video operator recorded continuous high-definition video on 1-h-long MiniDV tapes. A housing containing a pair of scaling lasers 10-cm apart was also mounted on the forward rail. Both camera operators attempted to frame images to include the pair of green laser dots projected from the submersible onto the seafloor.
A Sea-Bird SEACAT 19 conductivity-temperature-depthoxygen (CTD-O) sensor with a supplementary pH sensor was mounted on Idabel's starboard side rail to generate continuous, along-track water chemistry profiles for each dive, especially around Desmophyllum pertusum reefs (formerly Lophelia pertusa; Addamo et al. 2016). The unit was activated before each dive, and data were downloaded on recovery. The unit did not record for the entire transit on most dives, but one complete track was recorded. All watches and cameras were synchronized. Divers produced brief summaries of each dive with dive chronologies providing information on dive start time, start coordinates, maximum and minimum depths, habitat, debris, and animals observed (see Etnoyer et al. 2011 for details).
A few biological samples were also collected in deep water with a baited trap-a green plastic basket baited with lionfish heads, wrapped in mesh fabric, and weighted with rebar. A yellow rope threaded through a block of buoyant syntactic foam and attached to the basket created a retrieval loop. The trap was successfully retrieved using a long shepherd's crook device mounted on the port side rail in place  Nineteen dives from June 2012 to June 2016 used interior mounted and handheld Canon SLR and GoPro cameras, with exterior-mounted scaling lasers on some dives. The maximum depth was 640 m with most observations made between 150 and 400 m. For collecting small sessile crinoids, a small steel blade was welded to the forward end of the dip net ring to permit scraping along with hard substrates, and the net was lined with finer material.
The analysis of species distribution for corals reviewed video and images from select dives that met criteria where depth and location were known, still images were semi-periodic, and notes were available. These 8 dives are hereafter referred to as "quantitative dives" (see Table 1). The sponge and crinoid observations are based on a larger set of observations from "nonquantitative dives" in 2012-2016. For the quantitative dives in 2010 and 2011, the on-bottom video was reviewed for counts of large (> 10 cm) habitat-forming corals. Some rare, unidentifiable, and hyperabundant taxa (e.g., Stichopathes) were excluded. Depth zones in Table 2 are based upon discontinuities in faunal distributions identified using a one-way incomplete blocking design ANOVA (not discussed here) (Lavelle 2012).
Still images were reviewed independently for a more general census of presence and absence of all taxa, and attribution of depth ranges. Taxonomic identifications of corals and sponges were made nearly exclusively using these still images except in the case of a few fragile coral species that could be collected (i.e., Hemicorallium niobe (Bayer, 1964), Madracis myriaster (Milne Edwards & Haime, 1850), Desmophyllum pertusum (Linnaeus, 1758) (formerly Lophelia pertusa). As such, many coral and sponge taxa were identified at higher levels (family or genus) based upon gross morphology, using online image references, and working in consultation with experts (see Acknowledgements). The crinoids were identified based upon tissue samples collected in 2012-2016, so their scientific names are more specific.

Results
Dives recorded at least 32 morphological species (i.e., phenotypes) of deep-water corals (6 colonial Scleractinia, 17 Octocorallia, 9 Antipatharia) at depths of ~ 150 m or greater. Of the deep taxa, 16 were recorded during quantitative dive transects (4 Scleractinia, 8 Octocorallia, 4 Antipatharia) ( Table 2). Taxa recorded exclusively shallower than 150 m included Swiftia exserta and Nicella spp. (both Octocorallia). Of the two relatively shallow (< 165 m) colonial scleractinians observed during nonquantitative dives, Madracis myriaster was common, whereas only a few colonies of Madrepora carolina (Pourtalès, 1871) were observed (Fig. 3). Of the deeperwater taxa, Dendrophyllia alternata Pourtalès, 1880, occurred in substantial groves and in two color forms. Desmophyllum pertusum occurred as more or less individual colonies, sometimes as thickets over 3-5 m in diameter, other times as living colonies on dead branches attached to vertical or overhanging substrates, or as scattered small living branches on a substantial accumulation of dead branches. A single colony of Enallopsammia rostrata (Pourtalès, 1878) was recorded at 588 m. Solitary cup corals, including Javania sp., were observed but not examined in detail.
Among octocorals, two large Plexauridae sea fans-one yellow, one purple-were most widespread and abundant (Table 2, Fig. 4). Colony sizes ranged from 40 to 160 cm in height and width. These have been attributed to Paramuricea due to their resemblance to similar colonies in the Gulf of Mexico (Doughty et al. 2014). Some steep-walled areas were characterized by large numbers of Primnoidae (subfamily Calytrophorinae: presumably Narella and Paracalytrophora) with dozens of colonies over 1-m tall, some up to 1.6 m in height. The precious octocoral Hemicorallium niobe also occurred in small aggregations on mixed substrates near 600 m depth, with colony sizes 40-80 cm in width. In shallower waters, on steep walls near 150 m depth, we observed large Nicella sp. (Etnoyer et al. 2011), the largest ~ 2.5 m across. Among antipatharians, both fanshaped taxa and orange whip corals (Stichopathes sp.) were common (Fig. 5), whereas large pinnate Bathypathes corals were rare. Other cnidarians included a few zoantharian taxa (e.g., Parazoanthus spp.), all overgrowing the axes of other anthozoans; actinians (e.g., Alicia mirabilis Johnson, 1861, Liponema sp., and Sagartiidae), and stylasterid hydrocorals. Although the latter are usually treated broadly as corals, they have not been included here in any counts, as no morphospecies were distinguished and no colonies quantified.
Few taxa were observed often enough to extract detailed estimates of bathymetric distributions from quantitative dives. However, of the 305 records of the 16 coral morphospecies made during quantitative dives, 201 records of ten species were noted at depths of 335-449 m, followed by 69 records of 13 species in 550-700 m (Table 3, from Lavelle 2012). Of these, seven taxa occurred in both depth zones (i.e., the two Paramuricea morphs, three antipatharians, and Desmophyllum pertusum). Observations in the two other depth ranges included far fewer records: 15 observations of three coral taxa in 150-334 m, and 20 observations of seven taxa in 450-549 m. Three taxa observed at both shallower and greater depths were not recorded in the 450 to 549-m range. In addition, Fig. 8 illustrates how benthic sessile and semi-sessile megafaunal assemblages can exhibit highly restricted distributions and can differ widely over relatively short distances within the same depth range (here only 341-396 m).
The two Paramuricea morphospecies were recorded over strongly overlapping depth ranges, with the purple found in 305-671 m (n = 78) and the yellow in 250-664 m (n = 81). By contrast, members of Primnoidae were chiefly restricted to depths greater than ~ 400 m, whereas Ellisellidae occurred only shallower than this. Large flabellate stylasterid hydrocoral colonies were chiefly observed at depths less than 160 m, with smaller colonies in deeper water. Many taxa were observed only in restricted local habitats within their depth ranges, perhaps associated with specific substrates, topography, and near-bottom hydrodynamic conditions. Desmophyllum pertusum, corallistid sponges, and barnaclelike cyrtocrinid crinoids H. mikihe and C. pourtalesii were observed chiefly on vertical or near-vertical surfaces, versus Paramuricea spp. typically on projecting rocky substrates or where high-relief topography may be focusing near-bottom flow.
Repeated observations of a few octocoral colonies revealed variations in predation, recolonization, and epibiont host fidelity. Figure 10 shows a Paramuricea sp. that suffered no obvious predation over at least a 9-year period (2010-2019), whereas Fig. 11 shows a different colony that illustrates a multiyear decline, including branch loss, denuded skeleton, regeneration of polyps, and subsequent regrazing of polyps.
The pencil urchin Cidaris abyssicola was observed on the colony in 2008 and again later, adjacent to its base in 2013 (Fig. 11). Evidence of predation was present already, in the form of denuded branches, before the urchin was first observed on the colony. Cidarid pencil urchins appear to be among the most likely corallivores in this ecosystem. The urchins were observed upon colonies of Dendrophyllia alternata and Bathypathes sp., but these colonies showed no evidence of injury. Figure 11 also shows the movement and eventual loss of two Asteroschema laeve brittle stars on this Paramuricea colony over a 14-year period. In 2006, one clung on the left-and one on the right-hand branch. They exhibited evidence of predation on the tips of their arms where the brittle stars appeared to be regenerating new, smaller arm segments. In 2008, two were on the left-hand branch, also with regenerating arm tips. In 2010, only one remained but with polyp regrowth on the coral colony by 2012; a second had returned on the right-hand branch and remained through 2013. With substantial polyp loss by 2014, both ophiuroids had vacated. However, in an image taken later in 2014 (not shown), the small clusters of polyps on the far right-hand branch (upper, middle, and lower) had all  Table 1

Discussion
The observations presented here contribute to a growing body of literature using human-occupied vehicles (HOVs or submersibles) to explore the Caribbean deep sea and to study the benthic ecology of the Mesoamerican Reef (MAR), in particular, which is the second-longest fringing barrier reef in the world. Scientific expeditions of the MAR since the 1900's deployed trawls and dredges from the US Fisheries Commission steamers Blake and Albatross and from the University of Miami's R/V Pillsbury in the 1960s. These expeditions provided tissue samples from the deep-sea floor that are now well illustrated in books and reside preserved in national archives.
What is different about the submersible-based research is that we are seeing these animals alive for the first time, observing their behavior and tracking their survival over time using images. This is being done in a selective and targeted manner,  with negligible impact on the benthic community. The submersible can access a steep rocky habitat that is common in the 150-800 m depth range throughout the western Caribbean (James and Ginsburg 1979). The steep grade is relatively inaccessible to other types of sampling gear like trawls and dredges and is actually very difficult to map with precision.
The collaborative nature of the 10-year study was also somewhat unique. Over time, we attracted assets like the Research Vessel Falkor to map Roatan's seafloor using multibeam echo sounders (NOAA 2013), and experts in taxonomy from NMNH to identify invertebrate predators. We were able to collect seafloor images, water chemistry data, bathymetry data, and tissue samples of corals, gastropods, and crinoids. We did not attempt to collect sediments or tissue samples of fishes or soft corals. This would require a multifunction manipulator, as well as the funds and time to do so. There is considerably more sampling work to be done.
Images of species in their natural habitat do not provide DNA, but they can provide details of morphology often (usually) destroyed by trawling or dredging. Photographs and video observations are collected along a swath that documents spatial relationships among species and assemblages, with relatively high precision. The images also provide evidence of the orientation of individuals in situ with regard to their habitat, currents, and other species. Trawls, dredges, and grabs do not permit time-series analyses of the same individual specimens to record age, predation, recovery and regrowth, or the fate of their epibionts. These latter methods do allow sampling and systematic assessments over very large areas (e.g., Santodomingo et al. 2013).
The Mesoamerican Reef has been explored using submersibles since the 1970s but rarely as deep as the observations shown here. The Nekton submersible explored the geology of the MAR to 300 m using film cameras (James and Ginsburg 1979). Since 1990, no less than 100 submersible dives have taken place between Cozumel and Curacao (NOAA 2019). Johnson Sea Link I submersible (JSL I) made 32 dives to survey along the Yucatan Peninsula from Cozumel to Belize, working to a maximum depth near 365 m in 1990. The JSL II submersible made another 30 dives to a depth of 600 m in 1997. Together, these dives represent a 50-year body of submersible exploration spanning > 1000 km.
However, data on diversity and abundance of sessile benthic megafauna deeper than 300 m are relatively new to scientific discourse in the western Caribbean. A 10-year series of epibenthic trawl surveys sponsored by Marine and Coastal Research Institute (INVEMAR) off the Columbian Caribbean coast explored depths from 20 down to 500 m. A total of 142 species of azooxanthellate corals was collected and archived through the research, and those deeper than 200 m were found to be relatively widespread (Santodomingo et al. 2013).
The present study begins with DeepCAST in the year 2010 and dives to depths between 150-750 m to find relatively few taxa in high abundance and document them alive in their habitat over time and in detail. This depth range is deeper than most submersible surveys in the Caribbean Sea but nearly equivalent to JSL II dives from Curacao in May of 2000. A similarly inspired effort called Deep Reef Observing Program (DROP) began coincidentally in the Dutch Antilles in 2010, at nearly the same time as the present study, working to a depth of 300 m. The DROP documented many new species of fishes, including blennies, sea bass, and gobies Robertson 2013, 2014;Tornabene et al. 2016), as well as cusk eels, octopods (Pratt et al. 2019), and hermit crabs (Lemaitre et al. 2017). These descriptions and observations are documented in no less than two dozen papers.
We used the Idabel sub to illustrate the diversity of the benthic assemblages off Isla de Roatán at a coarse taxonomic level, in an effort to highlight the importance of structureforming corals for mobile invertebrates in the deep sea and offer insights into the stability and variability of benthic megafaunal assemblages on a Caribbean deep island slope Image data allowed us to document the diversity, distribution, and abundance of key species. Repeat observations allowed us to monitor growth and behavior, and characterize species distributions.
The topography of the deep island slope off West End, Roatán, is quite variable and in many places highly complex, rugged, and incised by a series of sedimentary channels. Although the location of these sites along the western island slope protects them from strong flow associated with the northwestward flowing Caribbean Current (Richardson 2005), there appears to be enough productivity and near-bottom flow to maintain diverse and often dense assemblages of suspension-feeding megafauna. Orientation of fan-shaped stony coral, octocoral, and black coral colonies reflect longterm responses to near-bottom flow conditions, both acrossslope (e.g., Figs. 3b, 4a-g, and 5a-d) and upslope (Fig. 6h). The latter and other images (not shown) of horizontally oriented, concave-downslope coral colonies suggest the existence of localized upwelling. Also, the highly localized and in many cases restricted distributions of many taxa likely reflect important small-scale variations in near-bottom flow and food availability.
Observations of several Paramuricea sp. colonies over time reveal variations in predation ranging from no obvious predation (Fig. 10) to virtually complete colony death (Fig. 11). These observations are important because they provide evidence of a long, slow decline in deep-sea coral colonies in the absence of known anthropogenic impacts such as fishing gear entanglement or oil spills. The partial regeneration of one of the octocorals suggests that recovery is possible, but it can be negatively affected by grazing. Injured sea fans in the Gulf of Mexico also showed limited ability to regenerate after Deepwater Horizon (Fisher et al. 2014;Etnoyer et al. 2016;Girard and Fisher 2018). However, Girard et al. (2016) described a mutualistic symbiosis in which brittle stars Asteroschema clavigerum Verrill, 1894, on Paramuricea biscaya Grasshoff, 1977, benefitted the octocoral colonies following the oil spill, likely via the ophiuroids removing material deposited on the polyps and possibly by inhibiting settlement of hydroids.
A similar mutualist relationship was described between the snake star Astrobrachion constrictum (Farquhar) and the black coral Antipathes fiordensis in Doubtful Sound, New Zealand (Stewart 1998). One method of feeding by the euryalid brittle star was for its arms to wipe across the branches of the host coral colony, gathering plankters and  detritus along with mucus from the coral polyps. The coral polyps were not harmed. The snake stars apparently assisted in keeping the host coral colony free from epizoic organisms and from detritus which otherwise may have smothered the colony or impeded feeding. In contrast, our study suggests that epibionts such as Asteroschema laeve, Novodinia antillensis, feather stars (e.g., Neocomatella puchella and Crinometra brevipinna), and chyrostyloid squat lobsters ultimately confer no protection to Paramuricea sp. colonies over the long term, at least to megafaunal predation (Grange 1991;Stewart 1998). Cidarid pencil urchins are considered here to be among the most likely predators on these colonies off Roatán. However, in more temperate waters of the North Pacific and North Atlantic, asteroids are more commonly observed feeding on deep-sea corals (Mah et al. 2010). Asteroids similar to Hippasteria were observed on several of our dives but cannot be identified from our photographs and video images (C. Mah, Smithsonian NMNH, pers. comm.).
The diets of deep-sea regular echinoids are poorly studied, but they are presumably broad generalists and carnivores on sluggish or sessile organisms (Jangoux and Lawrence 1982), including predation by cidarids on stalked crinoids (Baumiller et al. 2008). The shallow-water echinoid Arbacia lixula (Linnaeus, 1758), a dominant sea urchin in the Mediterranean, was long considered to be an herbivore but later found via stable isotope analysis to be omnivores tending to carnivory and feeding on a wide variety of invertebrate taxa, including hydrozoans (Wangensteen et al. 2011). Although we have only video observations of the echinoid Cidaris abyssicola ostensibly feeding on a variety of invertebrate taxa, the frequency of these observations over time, suggests that it may be a predator of some deep-sea corals, particularly yellow Paramuricea sp. We hypothesize that Cidaris abyssicola might use a homing strategy perhaps similar to that of the sea urchin Diadema antillarum Philippi, 1845, in shallow water, in which the urchins avoid repeatedly feeding on exactly the same areas of the seafloor (Carpenter 1984).
Future scientific studies should conduct sampling to verify and improve the taxonomic resolution reported here and investigate the predator-prey field for deep-sea corals in disturbed ecosystems, such as the Gulf of Mexico and more natural ecosystems, like the Bay Islands of Honduras. Our study illustrates the precision and utility of submersibles for repeat observations over long time frames. Given the scientific and economic value of these platforms, we advocate for their broader use, particularly where the deep sea is readily accessible from shore.
Sampling and field studies All necessary permits for sampling and observational field studies have been obtained by the authors from the competent authorities and are mentioned in the Materials and methods section, if applicable. The study is compliant with CBD and Nagoya protocols.

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author upon reasonable request.
Author contribution PE and TS conceived and designed the primary research project. CM and TB contributed data from separate research projects. PE, TS, TB, and CM acquired funding. All authors assisted in data collection and logistics. KS piloted the sub, assisted with site selection, and provided image data. KL, TS, TB, and CM analyzed data. CM, PJE, and TS wrote the manuscript. All authors read and approved the manuscript.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will 1 3 need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.