First fish fauna assessment in the Fernando de Noronha Archipelago with BRUVS: Species catalog with underwater imagery

fish fauna assessment in the Archipelago with Species catalog with underwater imagery Abstract: Fish fauna assessments with traditional catch methods are undesired in marine reserves. Underwater visual census on the other hand is biased due to fish-diver interactions, such as shyness or avoidance behavior of large-bodied target species. This study presents the first marine ichthyofauna inventory of the Fernando de Noronha Archipelago, off northeastern Brazil, sampled with non-destructive, independent Baited Remote Underwater Video Systems (BRUVS). High fish species richness, totaling 66 species from 28 families and 10 orders, including five elasmobranch species and other large predatory fish, such as barracudas, groupers and snappers, were recorded from 81 underwater video samples. Additionally, all of the 4,398 individuals sighted were associated to up to four different benthic habitat types. A catalog of underwater fish images and a detailed species list with additional information, such as conservation status, are provided.


Introduction
The Fernando de Noronha Archipelago (FN) located in the Equatorial Western Atlantic Ocean, off Brazil's northeastern coast, contains a Marine National Park (PARNAMAR) to protect its important marine and terrestrial biodiversity (Soto, 2001;Floeter et al., 2001;IBAMA, 1990;ICMBio, 2017). The reef fish fauna of FN presents a considerable level of endemism (6.3%) (Floeter & Gasparini, 2000;Floeter et al., 2001). Recently, Pinheiro et al. (2018) found that Fernando de Noronha Archipelago may share endemic species with insular regions as far away as Trindade, located 1.860 km south of FN, highlighting the importance of the island in harboring a unique and rich marine biodiversity.
One of the most complete ichthyofauna inventories of the Fernando de Noronha Archipelago and its surroundings was developed based on existing literature and three scientific expeditions (ARFENOR I, II and III), where a combination of visual census with scuba, snorkeling, and fisheries inventory was employed (Soto, 2001). More recent inventories were carried out using visual census with scuba gear to assess the reef fishes of FN (Krajewski & Floeter, 2011;Medeiros et al., 2011), or snorkeling, focusing on the fish fauna of shallow reef areas (Ilarri et al., 2017). Additionally, the intertidal fishes of FN were studied by Andrades et al. (2018) in an assemblage structure approach.
Stressful sampling techniques, inducing the capture and sacrifice of high numbers of specimens is undesirable in marine protected areas, specifically in aquatic ecosystems harboring threatened and endemic species, such as FN (Floeter & Gasparini, 2000;Floeter et al., 2001;Andrades et al., 2018). In regard to traditional underwater visual census, a widespread technique to study reef fishes, fish-scuba diver interactions are a common methodological bias (Kulbicki, 1998;Harvey et al., 2002). This is specifically problematic when studying large bodied species, such as groupers, snappers and parrotfishes (Gotanda et al., 2009;Januchowski-Hartley et al., 2011), commonly targeted along the tropical coast of Brazil (Nunes et al., 2012(Nunes et al., , 2016. Sampling the highest possible number of species inhabiting a certain study area is paramount in a fish fauna inventory. As exposed above with studies in FN, focusing on distinct components of the local fish fauna, this may require a combination of different sampling methods, in order to achieve this. In this sense, all possible method-related limitations should be avoided, by also employing alternative methods (Willis & Babcock, 2000;Cappo et al., 2003). Baited remote baited underwater video systems (BRUVS) can offer such an alternative, non-destructive, non-lethal and efficient fish fauna sampling technique, that additionally bypasses the limitations and risks inherent to scuba diving (Willis & Babcock, 2000;Cappo et al., 2003;Lindfield et al., 2014). In this context, the present study provides an inventory of the ichthyofauna of the Fernando de Noronha Archipelago, using data from BRUVS sampling in different benthic habitats, and supplementary species images.

Study Area
This study was conducted in the Fernando de Noronha Archipelago (FN), an isolated group of volcanic islands, located in the western tropical Atlantic 345 km off the north-eastern coast of Brazil (03°52'S; 32°25'W) ( Figure 1). About 70% of the main island and the waters spanning its coastline to the 50 m isobath constitute as a notake zone (Figure 1), declared a National Marine Park (corresponding to IUCN Category II) in 1988 (IBAMA, 1990, Maida, Ferreira, 1997, Dudley, 2008. The remaining portion of FN is an Environmental Protection Area (EPA) designated for sustainable use (corresponding to IUCN Category V) established in 1986 (IBAMA, 1990, Maida, Ferreira, 1997, Dudley, 2008. The archipelago is under the influence of the South Equatorial Current, with annual averages of water temperature and salinity of 26°C and 36‰, respectively. The archipelago has two main underwater landscapes: the windward side, characterized by extensive algal-vermetid reef barriers along rocky shorelines, and the leeward side, mainly composed of descending slopes along a rocky shoreline with large scattered boulders and sparse and small reefs (Maida, Ferreira, 1997).

BRUVS sampling
In August 2017, 81 single BRUVS (one camera) deployments were realized in order to sample the FN ichthyofauna. Deployments were made randomly over a variety of benthic habitats, in depths ranging from approximately 5 to 35 m. BRUVS sampling was conducted at least 1 hour after sunrise and 1 hour before sunset due to visibility conditions and to remove the effect of crepuscular behavior. Soak time for each deployment was approximately 90 min. BRUVS were equipped with high-definition (HD) GoPro Hero 3+ action cameras and baited with 1,000 g of a mix of crushed sardine and fish offal. Only the windward (northwestern) side of FN was sampled due to its calmer sea conditions at this time of the year.

Video and data analysis
All video samples were analyzed using the free VLC media player software (www.videolan.org). All fish were identified to species level, with few exceptions identified to genus level (e.g. Halichoeres spp.).
The maximum number of individuals within a paused still frame at any time of the video (MaxN) was adopted as relative abundance measure, in order to avoid recounting individual entering several times in the camera's field of view (Cappo et al., 2004). Total relative abundances (TMaxN), summed from all BRUVS samples, per species are presented in a matrix (Table 1) with information on the international and Brazilian conservation status (IUCN, 2019; ICMBio, 2018. Mean relative abundance and species richness per BRUVS deployment was also calculated. Four habitat types were visually identified: rocky reefs (RO), rocky bottom with macroalgae (MA) and rhodolith beds (RH), as well as the unconsolidated sandy bottom habitats (SA) (Figure 2). At last, species images were extracted from the video samples, in order to produce a catalog of underwater images of the species (Figures 3-28).

Results and Discussion
From the 81 BRUVS deployments, 4,398 individuals of 66 fish species, belonging to 28 families and 10 orders were identified. The overall mean relative abundance (mean MaxN ±SD) and species richness (mean S ±SD) sampled per BRUVS deployment were 54.3 ±46.2 and 13.9 ±8.8, respectively. In the rocky reefs, the mean relative abundance and species richness were 96.4 ±38.2 and 22.5 ±5.1, in rocky bottom with macroalgae 65.8 ±39.8 and 17.6 ±9.7, in rhodolith beds 32.5 ±40.3 and 7.5 ±6.4 and in sandy bottom habitats 25.3 ±25.9 and 8.0 ±5.3, respectively. BRUVS were deployed randomly, without prior identification of the habitat types at the sampling locations, which contributed to a highly unbalanced number of samples per habitat (SA: 44/ RO: 30 / MA: 5 / RH: 2), unsuitable for further statistical analysis and testing, yet still serving as an additional information, that may support future species-specific or habitatrelated studies in the Archipelago.
Compared to former fish fauna inventories in this area (Soto, 2011;Krajewski & Floeter, 2011;Medeiros et al., 2011;Ilarri et al., 2017;Andrades et al., 2018), a considerably high number of species (66) was detected within the 81 BRUVS samples. Krajewski & Floeter (2011) detected 60 species in 91 belt transects with scuba, sampling both sides of FN and the same depth range as this study. Ilari et al. (2017) recorded 50 species from 27 families in a study employing an effort of 140 transects in waters up to 6 m depth. Soto (2001) recorded 167 species, including pelagic species in an inventory using visual census and fisheries surveys supplemented with literature records. Furthermore, seven species of six different families have been recorded with BRUVS in this study, which were not registered in the former, above mentioned, studies. Being them Acanthurus bahianus (Acanthuridae), Bothus maculiferus (Bothidae), Caranx ruber and Trachinotus falcatus (both Carangidae), Hirundichthys speculiger (Exocoetidae), Clepticus brasiliensis and Halichoeres dimidiatus (Labridae).
Based on these results, BRUVS have shown to be a valuable alternative sampling method, that may eventually complement or even partially substitute traditionally used methods, in FN and other tropical marine protected areas, depending on which portion of the fish assemblage is studied. In this study, notably large predatory fish such as sharks, groupers and snappers were registered in high abundances, e.g. Carcharinus perezii and Negaprion brevirostris, each with 45 individuals, highlighting the efficiency of BRUVS to sample large-bodied, olfactory driven predatory fish. Studies developed with nonlethal methods are especially recommended for research on elasmobranchs, as this group includes some of the most vulnerable marine species (Musick et al., 2000).
Twelve families were represented by a single species (Table 1). The most abundant species were Melichthys niger (TMaxN: 628), Abudefduf saxatilis (TMaxN: 613), Chromis multilineata (TMaxN: 263) and Cephalopholis fulva (TMaxN: 254). For 10 species only a single individual was recorded in all video samples (Table 1). Furthermore, five different species of elasmobranchs of three orders and four families were also recorded, being them: Carcharinus perezii, Negaprion brevirostris, Ginglymostoma cirratum, Aetobatus narinari and Hypanus berthalutzae. All elasmobranchs showed high total relative abundances (TMaxN) ranging from 39 to 45 individuals (Table 1), except for one single individual of Aetobatus narinari (Myliobatiformes). From the 66 fish species recorded, a total of 49 species images were suitable to produce an underwater image species catalog (Figures 3-27). The specific image number for each species is indicated in the main matrix (Table 1).
Several threatened fish species were recorded, including the Goliath grouper Epinephelus itajara, classified as "Vulnerable" by the IUCN and already "Critically Endangered" by the Brazilian environmental agency (ICMBio). Three shark species classified as "Vulnerable" by the ICMBio, the Caribbean reef shark Carcharinus perezii, Lemon shark Negaprion brevirostris and the Nurse shark Ginglymostoma cirratum, were registered in large numbers, with 45, 45 and 39 individuals, respectively. Overall, more than 17% of all species registered in the present study are threatened (CR, VU) or near threatened (NT) according to IUCN and/or ICMBio (Table 1). The relatively high numbers of large predatory fish, threatened along most of the Brazilian coast, are certainly a positive response to three decades of protection from fishing by the PARNAMAR.
Finally, we recommend future research, ideally with stereo-BRUVS that additionally enables fish length measurements, applied both in benthic and pelagic habitats of FN and surrounding areas. The information of species richness and relative abundance within FN provides a baseline for future monitoring efforts in order to evaluate the impact of long-term environmental changes. Yet, in order to achieve a much complete fish fauna inventory of FN, we still recommend employing a combination of BRUVS with other conventional sampling methods, to compensate for the BRUV's disadvantage in detecting, for instance, small cryptic living fish species.