Visual faunistic exploration of geomorphological human-impacted deep-sea areas of the north-western Mediterranean Sea

This study reports the composition and distribution of demersal megafauna from various north-western Mediterranean submarine areas such as canyons, seamounts and landslides between 60–800 m depth, based on remotely operated vehicle (ROV) observations. From a total of 30 h of video, 4534 faunistic observations were made and analysed in relationship to environmental factors (i.e. topography, substrate type and depth). In addition, anthropogenic impact was quantified by grouping observations in four categories: fishing nets, longlines, trawl marks and other litter. The different targeted environments showed similarities in faunal composition according to substrate, depth and topography. Our results also indicated the presence of anthropogenic impact in all the sampled areas in which litter and trawl marks were the most observed artefacts.

The identification of deep-sea "essential habitats" is currently a major focus of 39 European Community research programs with the aim of furthering the conservation and 40 management of benthic biodiversity (Salomon 2009). In this context, faunistic surveys in cold 41 seeps, mud volcanoes, seamounts, and canyons as "hot spots" for local biodiversity are of  was equipped with two wide-angle color CCD cameras with a resolution of 3.2 Mpixel, 1Gb, 98 offering a frontal and a lateral view, plus a third with a macro-zoom. Lighting was provided 99 by 2 x 100 W HID lights and 4 x 150 W quartz lights. The ROV speed and height above the 100 seabed during filming operations were approximately 1.2 knots and 1.5-2.0 m, respectively.

101
The resolution was constant along transects. The limit of detection depended on the ROV 102 distance to the bottom. In some cases, the presence of mud clouds could result in a diminution 103 of the limit detection.

104
Seven transects (here after termed as "dives") were conducted for a total of 14.5 km 105 surveyed (equivalent to a total of 30 hours of video; Table 1). Three different NW  All video footages considered for animal taxonomic identification and counting were 118 obtained with the frontal camera and inspected in a time-lapse mode (i.e. at 50% of 119 acquisition rate). Video analysis was conducted using the software application Intervideo 120 WinDVD 9.0 (Windows). All observed organisms larger than 5 cm were identified as 121 faunistic entries (i.e. smaller animals were not visible), being classified to the lowest 122 taxonomical level as possible. For a more precise taxonomic determination, digital frames 123 were extracted after video partitioning. Classification was accomplished by the use of current 124 taxonomic guides for the Mediterranean (Zariquiey 1968, Riedl 1983, Mercader et al. 2001).

125
Data on faunal composition were annotated according to their timing of occurrence in 126 the video footage (hence allowing correlation with ROV navigation data for a precise 127 geographic positioning) along with concomitant annotations on the substrate type, classified 128 as mud, rock, sand, and coral rubble, as well as on anthropogenic artifacts.

129
Data analyses were carried out considering faunal entries grouped within classes, to 130 avoid those classification mistakes that may occur in ROV studies when attempting a more 131 precise classification when no concomitant sampled specimens are available for comparison.

132
Faunistic comparisons among different substrate types and depth ranges were carried out by     162 We observed a total of 4534 individuals, considered different faunistic entries (Table   163 2) in the various geomorphological areas surveyed (i.e. canyon, seamount, and landslide) (see 164 Figure 1). A comprehensive list of these entries, classified to the species level (when 165 possible), is provided in Appendix 1. The fauna belonging to the classes Actinopterygii, 166 Malacostraca, and Anthozoa were the most abundant, representing 24%, 20%, and 14% of all 167 observations, respectively (see Figure 2). The class Demospongiae was less abundant (12%), 168 with an occurrence similar to those of Rhynchonellata (11%) and Scyphozoa (9%). The 169 abundance of all other remaining invertebrate classes was less than 3% each.

170
NMDS results showed the presence of a significant effect of depth on species 171 ordination, taking all the inspected areas both together and within each area (see Figure 3).

172
Area and sediment type were significantly related to the class ordination only when areas 173 were considered together (Table 3; see Case 1). No significant effect of anthropogenic factors 174 was found. When we considered all the classes in the three areas taken together (see Figure   175 3A), we observed that Asteroidea, Echinoidea, and Holothuroidea were associated with 176 shallower sandy areas, whereas Ophiuroidea, Crinoidea, and Cephalopoda occurred primarily 177 in deeper zones on muddy flat slopes.  Table 2 and Figure 1). Both dives were similar in setting, with the exception 182 of the southern dive 2, which crossed an area with a steeper slope in its deepest section. Two   Table 3, Case 2) but not of sediment typology and 203 anthropogenic impact. At the faunistic level, there is a highly similar group composed by 204 Actinopterygii, Malacostraca, Anthozoa and Elasmobranchii. This group of taxa is dissimilar 205 to Demospongiae, Hydrozoa, Gastropoda and Thaliacea (see Figure 3B).  Table 1). The first seamount presented a conical 210 morphology surrounded by a muddy plain. A total of 10 hours of images were recorded at this    Table 2). The most commonly observed groups were the benthic 246 classes Demospongiae (24%) and Rhynchonellata (19%) (see Figure 4B). These groups were 247 followed by Actinopterygii (16%) and Anthozoa (15%). The classes Malacostraca, Scyphozoa 248 and Holothuroidea were less abundant in the seamount dives, representing 16%, 8% and 2%, 249 respectively of the total observations.

250
The NMDS analysis conducted in the seamount showed that the factors depth and 251 sediment were significantly affecting the distribution of the classes (see Table 3, Case 3).

252
Classes Holothuroidea, Polychaeta, and Thaliacea were associated with shallower sandy 253 areas, while Rhynchonellata and Demospongia were associated with medium depths and 254 rocky areas. In contrast, Crinoidea, Gastropoda, and Ophiuroidea showed a preference for 255 deeper muddy areas (see Figure 3C).     cans, and bottles (see Figure 5A). Trawl marks were also consistently observed (see Figure   292 5B). Finally, lost or discarded fishing gears were also detected, including longlines (see 293 Figure 5C) and the remains of hauling fishing nets (see Figure 5D).

296
In the canyon head, plastic bags and bottles represented 79% of the total observations, 297 whereas longlines represented only 14%. A minority of the observations (7%) were related to 298 trawl marks. No fishing nets were detected (see Figure 6B). On the seamounts and their 299 surrounding areas, 58% of the anthropogenic impact referred to the presence of longlines, 300 with a significant amount of other litter (22%), trawl marks (16%), and only 4% of discarded 301 fishing nets (see Figure 6C). On the landslides, approximately half (45%) of the total 302 anthropogenic observations were represented by trawl marks and other litter (44%), with 303 longlines (9%) and fishing nets (3%) less representatives (see Figure 6D).         approach generates a different bias-dependent effect on observations. In our case, ROV does 435 not seem perceived as a potential threatening stimulus by some species. 436 We observed 4 individuals of Paromola cuvieri as carrying human artifacts, as already           90x46mm (300 x 300 DPI)