Notes on a remotely operated vehicle survey to describe reef ichthyofauna and habitats – Agulhas Bank , South Africa

Surveys of the marine ichthyofauna associated with temperate reefs are challenging because, unlike soft sediment, these habitats cannot be surveyed with trawl gear (Sink et al. 2012). Reefs often harbour complex species assemblages, stratified according to physical and biological gradients (Ross & Quattrini 2007; Söffker, Sloman & Hall-Spencer 2011). Survey methods for such habitats have mostly been confined to diver-based underwater visual census (UVC) for shallow depths (<30 m) (Assis et al. 2013) and traps, hooks and line fishing for deeper depths (>30 m) (Söffker et al. 2011). Recently developed remote video-based methods have produced promising results (Cappo, Speare & De’ath 2004; Stoner et al. 2008). Unlike stationary video sampling methods such as baited remote underwater video (BRUV) surveys (Cappo, Speare & De’ath 2007), remotely operated vehicle (ROV) surveys, like diver-based UVC, allow for a comprehensive inspection of the sampling area and can also be used within a standardised sampling design in the form of transects or fixed, baited stationary deployments. Here we report on the first ROV dives to investigate the deep reef ichthyofauna and its associated habitats on the central Agulhas Bank. In addition to general habitat observations, we were interested in (1) fish behaviour towards a ROV, (2) deployment times sufficient to detect representative species assemblages and (3) the number of fish species detected relative to other surveys.


Introduction
Surveys of the marine ichthyofauna associated with temperate reefs are challenging because, unlike soft sediment, these habitats cannot be surveyed with trawl gear (Sink et al. 2012). Reefs often harbour complex species assemblages, stratified according to physical and biological gradients (Ross & Quattrini 2007;Söffker, Sloman & Hall-Spencer 2011). Survey methods for such habitats have mostly been confined to diver-based underwater visual census (UVC) for shallow depths (<30 m) (Assis et al. 2013) and traps, hooks and line fishing for deeper depths (>30 m) (Söffker et al. 2011). Recently developed remote video-based methods have produced promising results (Cappo, Speare & De'ath 2004;Stoner et al. 2008). Unlike stationary video sampling methods such as baited remote underwater video (BRUV) surveys (Cappo, Speare & De'ath 2007), remotely operated vehicle (ROV) surveys, like diver-based UVC, allow for a comprehensive inspection of the sampling area and can also be used within a standardised sampling design in the form of transects or fixed, baited stationary deployments. Here we report on the first ROV dives to investigate the deep reef ichthyofauna and its associated habitats on the central Agulhas Bank. In addition to general habitat observations, we were interested in (1) fish behaviour towards a ROV, (2) deployment times sufficient to detect representative species assemblages and (3) the number of fish species detected relative to other surveys.
Six sites were surveyed on the central Agulhas Bank during summer 2009 to 2010, namely Vlak Bank, Martha's Reef, 12-Mile Bank, Alphard Bank, 45-Mile Bank and 72-Mile Bank. The former four were surveyed during dedicated research voyages with the RV Ellen Khuzwayo; the latter two during a 2009 survey with the Seacor Achiever dedicated to assessing the impact of petroleum infrastructure on the area (Sink et al. 2010) (Figure 1).

Survey dives
A Seaeye Falcon ROV (SAAB, system 12177) was used to carry out the surveys. The ROV dives from the RV Ellen Khuzwayo were carried out on anchor, and the Seacor Achiever was able to maintain position above the dive stations. At each of the six sites, four ROV dives of between 45 and 60 minutes each were conducted consecutively at adjacent locations 0.5 km apart. Cumulative dive times per site amounted to between 180 and 240 minutes, and dives usually covered an area of approximately 800 m 2 . Although the pilot attempted to fly the device close (~2 m) to the seafloor to allow for recognition of the sessile marine life, the difficulty of maintaining a constant distance to the seafloor over diverse, high-profile reef in considerable currents precluded standardised transects. The general habitat characteristics for each site were assessed following the methodology of Götz et al. (2009), in terms of substrate composition, reef profile (low, medium and high), rugosity (low and high) and benthic cover.
At the site with the highest ichthyofauna diversity (Martha's Reef), species behaviour towards the ROV was recorded as follows: positive, if the fishes were attracted and swam towards the vehicle; negative, if the fishes were visibly disturbed and altered their behaviour to avoid the vehicle; neutral, if no behavioural reaction was evident. In addition, ROV motion at the time of observation was noted as 'stationary' or 'moving'. At four of the six sites, species saturation was determined by noting individual video times when new species were detected. At two sites, after an initial reconnaissance period, a weighted bait bag with pilchard (Sardinops sagax) was lowered down at the end of a 10-mm rope to the sea floor in the vicinity of the ROV to attract more fishes to the survey area. Overall ichthyofaunal diversity was compared to fishing survey reports and commercial fishing data from the same sites (

Results and discussion
Although ROVs have been used to study abundance and behaviour of fish communities beyond SCUBA diving depth since the early 2000s (Harvey et al. 2007;Widder et al. 2005), their use in South Africa has been limited (Roberts et al. 2006). Although our findings should be regarded as preliminary because of the small sample size and the exploratory nature of the observations, they represent the first ichthyofaunal (and habitat) ROV survey on the Agulhas Bank mesophotic reefs. Previously, these habitats were assessed only with indirect methods like fishing (Japp, Sims & Smale 1994) and dredging Samaai, Janson & Kelly 2012).

General description of study sites
The ROV surveys in this study revealed complex, diverse reef habitats (Figure 2), which differed considerably in profile, rugosity, sand inundation and benthic invertebrate cover (Table 1). The survey at the Vlak Bank between 20-and 30-m water depths revealed patchy, low-profile reef habitat interspersed with sandy areas. The reef itself was of low rugosity and heavily inundated with sand. Massive (up to 50-cm height) wall sponges (Spirastrella spp), gorgonians (Eunicella spp) and red algae dominated the benthic biota.

Fish behaviour towards ROV
Few studies have evaluated fish behaviour towards ROVs (Ross & Quattrini 2007). Stoner et al. (2008) reviewed behavioural responses to ROVs, which they broadly classified as attraction or avoidance. Responses may not be only species and size specific, but also depend on size, illumination, sound emission and speed of the vehicle. Our findings suggest that most species (68%) behaved neutrally towards the vehicle (Table 2). Positive reaction mostly occurred when sediment and benthic growth were stirred up by downthrusting, attracting omnivorous and herbivorous small shoaling species such as fransmadam (Boopsoidea inornata) and hottentot seabream (Pachymetopon blochii). Only yellowbelly rockcod (Epinephelus marginatus) were repelled by the ROV and hid in holes in the reef. These effects need to be taken into account in future surveys, but alternative survey methods have their own inherent bias (Bernard et al. 2014;Watson et al. 2005).

Species detection
Species richness and saturation varied considerably among dives (Figure 3). Minimum and maximum time of species saturation was reached at 34 and 133 minutes, respectively, depending on location (Figure 3). Additional deployment of bait tested during two dives shortened the species saturation time (Figure 3b and c). A total of 35 species were detected, with the number per individual site ranging between 9 and 19 species. Overall detection numbers are similar to those of research fishing surveys ) and to those reported by the commercial line fishery (30 species reported over 10 years) ( Table 3, DAFF unpublished data). Five species were detected solely by ROV, but two could only be identified to genus and family level. Most of the common, resident species were detected (Table 3), and omissions of shoaling species such as horse mackerel (Trachurus capensis) and large, solitary chondrichthyans might be ascribed to the short survey time. Although the deployment of bait accelerated species detection (Sink et al. 2010), ROV surveys have the added advantage of direct exploration of the seascape and the capability of detecting cryptic species and species that are not attracted to bait.

Conclusions
This survey has shown that ROV-based assessments of the ichthyofauna associated with temperate reefs in South Africa may present an alternative to existing methods and the only one that can directly relate ichthyofaunal observations to the surrounding seafloor environment. Unlike stationary video sampling methods such as BRUV surveys ), ROV surveys allow for a comprehensive inspection of the sampling area. Considering that ROV surveys yielded similar species numbers compared to previous survey methods based on fishing and species accumulation curves compare well with those of BRUV surveys ), this approach is potentially useful where rapid assessments of species diversity within a complex habitat are required.     Source: comparative data -DAFF, unpublished data