A photographic method to identify benthic assemblages based on demersal trawler discards

G Model

ARTICLE IN PRESS

FISH-4224; No. of Pages 10

Fisheries Research xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Fisheries Research journal homepage: www.elsevier.com/locate/fishres

A photographic method to identify benthic assemblages based on demersal trawler discards Camilla Piras a , Monica Mion a , Tomaso Fortibuoni a , Gianluca Franceschini a , Elisa Punzo b , Pierluigi Strafella b , Marija Despalatovic´ c , Ivan Cvitkovic´ c , Saˇsa Raicevich a,b,∗ a

Institute for Environmental Protection and Research (ISPRA), Loc. Brondolo, 30015 Chioggia, Italy National Research Council (CNR), Institute of Marine Sciences (ISMAR), Largo Fiera della Pesca, 2, 60125 Ancona, Italy c Institute of Oceanography and Fisheries, Sˇ et. I. Meˇstrovi´ca 63, 21000 Split, Croatia b

a r t i c l e

i n f o

Article history: Received 26 February 2015 Received in revised form 6 August 2015 Accepted 19 August 2015 Available online xxx Keywords: Photographic method Discard Benthic habitat Fishery-dependent data Marine Strategy Framework Directive

a b s t r a c t Knowledge of the distribution of benthic assemblages is important for monitoring the environmental status of the seafloor and understanding the spatial pattern of demersal fish species and their essential habitats. This knowledge may allow for the enforcement of spatially explicit management approaches, such as those requested by the Marine Strategy Framework Directive. However, traditional methods for benthic fauna sampling are time consuming and expensive, especially when surveying wide areas and relying on expensive sampling platforms, such as research vessels. We developed and tested a photographic method based on mega-epifauna identification and quantification through the analysis of pictures of demersal trawler discards. The method was tested with samples collected in the Northern Adriatic Sea (Mediterranean Sea) at different spatial scales. In this framework, species compositions and abundance were determined through analysis of discard pictures. These samples were compared to those derived from discard samples simultaneously collected during field activities whose specific composition was analysed in the laboratory. The direct comparison between the photographic and laboratory data highlighted a significantly strong correlation in abundance estimates, although the photographic method was less effective for the detection of small-sized or hidden species. The multivariate comparisons of speciessite matrices obtained with the two methods also showed a strong, significant correlation, and the spatial patterns of assemblages were significantly consistent. Our results indicate that epifauna discarded by commercial demersal trawlers can be efficiently characterized and quantified using the photographic method, thereby halving the time needed for sample processing and easing practical barriers for sample collection and storage. These data may be used to identify different benthic assemblages and their distributions. This approach could take advantage of ongoing monitoring of commercial fishing activities and/or direct involvement of the fishing industry to allow the collection of benthic species/assemblage data over a wide spatial scale and with a high spatial/temporal resolution, thus making use of fishing vessels as an efficient sampling platform for benthic habitat investigations. © 2015 Elsevier B.V. All rights reserved.

1. Introduction

∗ Corresponding author at: Institute for Environmental Protection and Research (ISPRA, Istituto Superiore per la Protezione e la Ricerca Ambientale), Loc. Brondolo, 30015 Chioggia, Italy. Fax: +39 0415547897; National Research Council (CNR), Institute of Marine Sciences (ISMAR), Largo Fiera della Pesca, 2, 60125 Ancona, Italy. E-mail addresses: [email protected] (C. Piras), [email protected] (M. Mion), [email protected] (T. Fortibuoni), [email protected] (G. Franceschini), [email protected] (E. Punzo), [email protected] ´ [email protected] (I. Cvitkovic), ´ (P. Strafella), [email protected] (M. Despalatovic), [email protected] (S. Raicevich).

Over the last few decades, concern about the ecological significance of fishing effects on ecosystems has led to the development of the so-called Ecosystem Approach to Fisheries Management (EAFM) (Garcia et al., 2003; Garcia and Cochrane, 2005). In this context, increasing importance is given to studies focused on fishing effects on habitats, as well as the study and characterization of the habitats themselves (Garcia et al., 2003; Jennings and Kaiser, 1998). In particular, the study of Essential Fish Habitats (EFH), defined as “(. . .) those waters and substrate necessary to fish for spawning,

http://dx.doi.org/10.1016/j.fishres.2015.08.019 0165-7836/© 2015 Elsevier B.V. All rights reserved.

Please cite this article in press as: Piras, C., et al., A photographic method to identify benthic assemblages based on demersal trawler discards. Fish. Res. (2015), http://dx.doi.org/10.1016/j.fishres.2015.08.019

G Model FISH-4224; No. of Pages 10 2

ARTICLE IN PRESS C. Piras et al. / Fisheries Research xxx (2015) xxx–xxx

breeding, feeding, or growth to maturity” (Gleason et al., 2013), has received growing attention in the scientific and management areas (Auster and Langton, 1998). The recognition of the value of EFH on the policy agenda was demonstrated by the Magnuson-Stevensen Fishery Conservation and Management Act (MFCMA). The MFCMA was enforced in the USA in 1996 and requires the identification of EFH designated for each managed species and the development of actions to conserve and enhance their protection. However, it is necessary to understand the association between fish and their habitats in order to define and characterize EFH (Kaiser et al., 1999); this link is increasingly considered to be important for sustaining fish production (Collie et al., 2000). In the European context, the importance of protecting estuarine and marine habitats was confirmed by the enforcement of different policies (e.g., the Habitats Directive (92/43/EEC) and the recently adopted Marine Strategy Framework Directive (MSFD; 2008/56/EC)). The latter has the goal of achieving Good Environmental Status (GES) across Europe’s marine waters by 2020 through the application of ecosystem-based management. In this Directive, benthic habitats and species are considered within Descriptor 1 (Biodiversity) and Descriptor 6 (Seafloor integrity), while the other descriptors mainly address the quantification of the pressures exerted on them (e.g., Descriptor 3–commercial fish and shellfish in relation to fishing pressure). The implementation of the MSFD highlighted the need for better data on habitat and benthic species distributions and revealed the presence of a relevant information gap in many European seas that needs to be filled with the next monitoring activities (Zampoukas et al., 2014). Moreover, the accurate spatial delineation of habitats is considered to be a fundamental step in establishing reference points for indicators and targets for the benthic domain because such attribute are known to inherently vary across habitats (Van Hoey et al., 2013). Studies on benthic fauna are commonly conducted on infauna species rather than on megafauna because the sampling is easier and more accurate (Kaiser et al., 1998; Rodrigues et al., 2006 Zenetos et al., 2000). However, data acquisition is usually time consuming and costly in terms of sampling gear (e.g., grabs, semi-quantitative dredges), sampling platforms (i.e., research vessels) and sample processing. Accordingly, multitaxa field studies adopting such sampling approaches are rarely undertaken over broad spatial/temporal scales (ICES, 2011a). In contrast, bottom-trawl surveys could be suitable for collecting data on a high range of benthic megaepifauna species over wide areas and serve as a valuable source of data for implementing the EAFM (Cotter et al., 2009) with a good taxonomic sufficiency of megaepifauna (Brind’Amour et al., 2014). However, logistic and economic constraints limit the data collection of investigations on a wide spatial scale to a low spatial and temporal resolution. Moreover, the sampling scheme is most often conceived for the study of demersal fish abundance and distribution (e.g., random stratified sampling approach) and not for habitats. Demersal gears (such as otter and beam trawls) allow for the collection of megaepifauna samples (Ellis et al., 2000; Froglia and Orel, 1971; Kaiser et al., 2004; Marano et al., 1989) and investigations into habitat-related demersal species’ spatial distributions (Kaiser et al., 1999). Most often, megaepifauna samples are collected from the total catch following standardized procedures (standard sampling gears and hauls), and sample composition (i.e., species’ abundance and biomass) is assessed directly on board (or in the laboratory in the case of time and logistic constraints). This sampling approach has also been applied to commercial fishing gears to characterize the discarded catch of demersal fishing activities; moreover, it was used to infer information on benthic assemblage compositions through observational or manipulative

studies of fishing effects on the benthic domain (Hall-Spencer et al., 1999; Kaiser et al., 1999; Pranovi et al., 2001). Thus, fishing vessels could be used to collect samples of benthic megaepifauna to characterize benthic habitats provided that the interaction of the fishing gear with the seabed is known and georeferenced samples are classified using standard procedures. Until now, the use of fishing vessels as a sampling platform for benthic discards (as a proxy to describe benthic assemblages) has received little attention, possibly due to barriers including, among others: (i) practical barriers for sample collection and storage; (ii) data with mixed taxonomic resolution (e.g., species and morphospecies, see Brind’Amour et al., 2014); (iii) the inherent fishery-dependent allocation of sampling sites; and (iv) lack of detailed information on commercial fishing gear/net features and the effects of the fishing gear on the selectivity for megaepibentic discarded species. In this study, we report the development and testing of a photographic method capable of overcoming and controlling some of the above mentioned barriers for the use of fishing vessels. This approach is proposed as an alternative/complementary method (compared to the laboratory method) for the evaluation of benthic discard composition. It has been conceived to reduce the time needed for sample collection, species identification and abundance estimations, thus potentially increasing the sampling effort. Sampling activities were designed to assess the efficacy of the photographic method in revealing benthic species and their relative abundance in fishery discards, as well as discriminating benthic assemblage distributions compared to the “traditional” method based on the field collection of benthic samples from the fishery discarded catch and the later identification and quantification of abundance/biomass in the laboratory (hereafter called the “laboratory method”). The feasibility and merits/drawbacks of the photographic method compared to the traditional discard sampling and laboratory analysis were also investigated. The study was developed in the Northern Adriatic Sea (Mediterranean Sea) and benefited from the collaboration with the fishing industry established in the framework of the GAP2 project and trawl survey activities performed in the area with a standard beamtrawl (i.e., rapido trawl) to assess flatfish (i.e., SoleMON trawl survey; Scarcella et al., 2014). The aims of the study were as follows: • To develop a photographic method to describe demersal trawler’s discard compositions and relative abundance; • To test and compare the performance of this method to the traditional laboratory procedure; • To assess the efficacy of the photographic method in discriminating benthic assemblages and their spatial distributions compared to the traditional laboratory method. 2. Materials and methods 2.1. Study area The study area is characterized by the predominance of soft bottoms, varying from sand to mud to detritic sediments with increasing distance from the coast (Brambati et al., 1983). The sediments’ spatial pattern is mirrored by the complexity of the benthic assemblages (Gamulin-Brida, 1967), with a higher degree of heterogeneity as the spatial scale increases. To evaluate the performance of the photographic method with increasing levels of assemblage heterogeneity, the comparison between the two sampling techniques (photographic vs. laboratory methods) was performed at two different spatial scales (i.e., local and regional). For this purpose, the “local spatial scale” was defined as the sub-portion of the Northern Adriatic Sea represented

Please cite this article in press as: Piras, C., et al., A photographic method to identify benthic assemblages based on demersal trawler discards. Fish. Res. (2015), http://dx.doi.org/10.1016/j.fishres.2015.08.019

G Model FISH-4224; No. of Pages 10

ARTICLE IN PRESS C. Piras et al. / Fisheries Research xxx (2015) xxx–xxx

3

Fig. 1. Study area: sampling site distributions at the local spatial scale (Veneto region administrative waters) (a) and regional spatial scale (Northern Adriatic Sea) (b); 䊉 = otter trawl;  = rapido trawl.

by the Veneto Region administrative waters 3 nautical miles (nM) eastward from the Italian seashore (from the Po river mouth to the Tagliamento river mouth) up to 12 nM, corresponding to a surface of 3,485 km2 (depth range: 15–30 m; Fig. 1a). In contrast, the “regional spatial scale” was defined as the whole Northern Adriatic Sea, whose area extends northward from Ancona up to Trieste and eastward from 3 nM off the Italian seashore up to the Croatian national water limits (i.e., 12 nM westward from the eastern Adriatic Sea coast) over an estimated area of 19,200 Km2 (depth range: 10–80 m; Fig. 1b). 2.2. Sampling platforms and effort Discard samples and digital pictures were collected using both fishing and research vessels towing demersal fishing gears (i.e., otter trawl and rapido trawl (a sort of beam trawl, see Hall-Spencer et al., 1999)). Samples were collected in the context of commercial fishing activities (fishery-dependent data) and trawl surveys (fishery-independent data). In the latter case, the sampling gear design was fully comparable to commercial fishing gear, although a smaller mesh size in the codend was adopted. At the local spatial scale the photographic and laboratory samples were collected from 73 sampling stations from fisherydepended and fishery independent survey carried out with otter-trawl and rapido trawl (Table 1; Fig. 1a) while at the regional spatial scale samples were collected from 34 sampling stations by the rapido trawl in the framework of a fishery-independent survey (SoleMON trawl-survey; Table 1; Fig. 1b). 2.3. Sampling procedure At the end of each haul (i.e., sampling station), the total catch was weighted using an electronic dynamometer (Dynafor LLX2, precision: ±3.2 kg). After the commercial catch was sorted, counted and weighed, five digital pictures were taken on different randomly selected portions of the discard. Pictures were collected using a digital 8 Mpixel camera (Sony DSC-W130) by positioning a squareshaped frame (0.5 × 0.5 m) above the discard distributed on the deck (Fig. 2a). Biological samples of megaepifauna (10–20 kg) were also randomly collected from the total discard at the end of each haul and immediately analysed (i.e., SoleMON trawl-survey) or stored in a freezer (−20 ◦ C) until the analyses were performed in the laboratory. The analyses were performed to assess the abundance and biomass of each benthic species/taxon at the highest possible taxonomic resolution, resulting in the compilation of mixed species and morphotype matrix data (Brind’Amour et al., 2014). Abundance estimates for some samples were not recorded due to the species’ deterioration during transportation and freezing.

2.4. Photographic method implementation Pictures were arranged for the subsequent image analysis after correcting for possible perspective distortions at the moment the picture was taken. The square-shaped frame was divided into sectors using a grid (500 × 500 pixels) with GIMP 2.6.11 (open source software) (Fig. 2b). The species’ abundance estimation for each picture was assessed using the Discard Analyzer 1.0 software (developed for this study), which allows marking of each specimen identified in the picture by a trained observer, thus preventing multiple counts. The software also records the total abundance of each species per single picture in a database that was later used for the statistical analysis of the samples’ species composition. The adequate number of pictures to be analysed per sample (i.e., haul) was estimated by computing cumulative plots (Clarke and Warwick, 2001) with a threshold level of 85% total richness. To this purpose, 10 pictures were randomly taken from the benthic discard from three randomly allocated sampling sites, and cumulative plots were estimated (PRIMER-E 6.1, Clarke and Gorley, 2006). To control for potential differences determined by the fishing gear effect, this approach was replicated for both the otter trawl and rapido trawl discards. 2.5. Comparison of the effectiveness of the photographic and laboratory methods The effectiveness of the photographic method in detecting benthic species in discards was assessed by comparing species’ composition and abundance estimates from biological samples and the corresponding pictures, both of which were collected on the same portions of the discard catch (i.e., the same portion of discard was first photographed and then collected). This sampling approach was replicated in 4 randomly allocated sampling sites per fishing gear (rapido vs. otter-trawl). The relationship between ranked species’ abundance estimates from the photographic and laboratory methods per sampling site/gear was investigated using the Spearman rank correlation (STATISTICA 7). Additionally, a logistic regression analysis was performed by assigning each sampled species to a dichotomous variable: “1” (species identified with both the methods) and “0” (species that were not identified in the pictures but were present in the samples according to the laboratory method). The following predicting variables were tested as the main effects: species’ abundance and average weight (continuous) and fishing gear (categorical). Moreover, the “ percentage” index was estimated, as difference between percentage abundance estimated from the photographic and the corresponding laboratory values for each taxonomic Class (i.e. Class percentage abundance assessed in the

Please cite this article in press as: Piras, C., et al., A photographic method to identify benthic assemblages based on demersal trawler discards. Fish. Res. (2015), http://dx.doi.org/10.1016/j.fishres.2015.08.019

G Model

ARTICLE IN PRESS

FISH-4224; No. of Pages 10

C. Piras et al. / Fisheries Research xxx (2015) xxx–xxx

4

Table 1 Number of sampling stations, fishing gear and survey features where photographic and laboratory samples were collected to the purposes of this study at the local and regional scale. OTB: otter trawl; TBB: rapido trawl. Spatial scale

N. of sampling stations

Gear

Codend dimond mesh size (mm)

Survey

Sampling strategy

Sampling years

Local

19

TBB

56

TBB

40

fishery-dependent (opportunistic) fishery-independent (opportunistic)

2012–2013

11

33

OTB

50

10

OTB

28

GAP2 onboard sampling SoleMON trawl survey (Grati et al., 2013) GAP2 onboard sampling GAP2 trawl survey

16

TBB

40

18

TBB

40

Regional

SoleMON trawl survey (Grati et al., 2013) SoleMON trawl survey (Grati et al., 2013)

Fall 2012; Fall 2013

fishery-dependent (random-stratified) fishery-independent (systematic) fishery-independent (random-stratified)

2012–2013 Summer 2012; Summer 2013 Fall 2012; Fall 2013

fishery-independent (random-stratified)

Fall 2012; Fall 2013

Fig. 2. Square-shaped frame (0.5 × 0.5 m) used to collect discard pictures (a) and a screenshot from Discard Analyzer 1.0 (b).

photographic method — Class percentage abundance assessed in the laboratory method). The index was estimated for each sample using only percentage data based on species identified by both methods. Higher values of  percentage represent a larger difference in abundance estimation between the methods; the index thus allows to identify the Classes more subjected to bias in abundance estimates when applying the photographic method. The Wilcoxon matched pairs test was applied to assess the presence of significant differences between the two methods in the time needed for sample processing. The threshold level of 0.05 was selected as probability level of significance for all the tests performed to assess the comparison of the effectiveness of the photographic and laboratory methods.

2.6. Multivariate analysis and species’ assemblage spatial distributions Matrices for the species’ abundance per site based on the laboratory and photographic methods were constructed for both the local and regional scale datasets. Data were subjected to different transformations (i.e., presence/absence, percentage, square root, double square root and log (x + 1)) to evaluate which method was most suitable for data analysis; then, the Bray–Curtis similarity matrices were calculated. To test whether the multivariate patterns reconstructed by the photographic and laboratory methods were correlated, the RELATE test (PRIMER-E 6.1; Clarke and Gorley, 2006) was applied by computing the Spearman correlation index (␳) between the Bray–Curtis similarity matrices and its p-value (Clarke and Warwick, 2001).

To examine whether the data collection methods revealed similar spatial patterns of benthic assemblages’ distributions, the samples were classified into assemblages based on a cluster analysis (Clarke and Warwick, 2001) of the two abundance matrices (percentage data; average group method) (Ellis et al., 2000; Kaiser et al., 2004; Parravicini et al., 2010). The correspondence and significance between the classifications obtained with the two methods at definite similarity cut-off levels was evaluated using a binomial test. To this end, a vector was defined by labelling the cases in which each station was ascribed to the same cluster (according to both the photographic and laboratory methods) with a value of 1, while the other cases were given a value of 0. By applying the ␹2 test to the binomial distribution, we tested the null hypothesis (H0) that this vector was not significantly different from a vector obtained by randomly assigning each station to a cluster, maintaining the proportion of stations per cluster as identified in our laboratory cluster analysis. At the local and regional spatial scales, the PERMANOVA non parametric randomization test (Anderson, 2001; Anderson et al., 2008) was applied to ascertain the presence of significant differences between the two methods in detecting changes in multivariate benthic assemblage compositions associated with different sedimentary features of the sampling sites using sediment classification derived by Brambati et al., 1983. The taxa with the largest contribution to dissimilarity between neighbouring assemblages (identified by cluster analysis for the two methods) were identified by applying the SIMPER (Similarity Percentages) analysis and selecting taxa presenting a dissimilarity/standard deviation ratio greater than 1.4 (Clarke and Warwick, 2001) or a percentage cumulative contribution greater than 10%.

Please cite this article in press as: Piras, C., et al., A photographic method to identify benthic assemblages based on demersal trawler discards. Fish. Res. (2015), http://dx.doi.org/10.1016/j.fishres.2015.08.019

G Model FISH-4224; No. of Pages 10

ARTICLE IN PRESS C. Piras et al. / Fisheries Research xxx (2015) xxx–xxx

Fig. 3. Photographic method: cumulative richness based on the increase in the number of picture replicates. Richness is expressed as the mean (n = 3) percentage of asymptotic richness and standard deviation; 䊉 = otter trawl;  = rapido trawl.

5

Fig. 4. Relationship between the average biomass (wet weight; g) and abundance:  = species detected by both the photographic and laboratory methods; 䊉 = species not detected by the photographic method.

The threshold level of 0.001 was selected as the probability level of significance for all the tests undertaken for the multivariate analysis and species’ assemblage spatial distributions comparisons. 3. Results 3.1. Photographic method implementation For both the otter and rapido trawls, the cumulative plots showed that the analysis of one picture allowed identification of at least 60% of the species detected with 10 pictures (replicates), while 100% of the species were identified with a minimum of 8 replicates. On average, three pictures allowed detection of 90% and 85% of the total richness for the otter and rapido trawls respectively (Fig. 3). Accordingly, three replicates (pictures) per station were used in the following analyses to characterize discard compositions; these results were later compared with data derived from the laboratory analysis. The comparison of the species’ compositions assessed using the photographic method and the laboratory analysis from four different sampling sites per fishing gear showed that the photographic method allowed on average the detection of 78 ± 21% of the species identified in the otter trawl discard samples and 68 ± 17% of the species identified in the rapido-trawl discard samples (Table 2). The discrepancy was probably caused by the challenge in obtaining a higher taxonomic resolution in some cases (i.e., species determination) due to the difficulty in evaluating some species’ distinctive morphological features through the picture. It is worth noting that the species identified with the photographic method represented up to 97 ± 1% and 96 ± 3% of the total abundance recorded in the otter-trawl and rapido trawl laboratory discard samples, respectively, thus only the very rare species were not detected by the photographic method. Moreover, the photographic method allowed the detection of species that were not detected by laboratory analysis in a limited number of cases (i.e., an average of 1 and 0.75 species per sampling station compared to the average number of species detected by laboratory analysis of 21.5 and 23.5 in the otter-trawl and rapido trawl samples, respectively) (Table 2). 3.2. Comparison between the photographic and laboratory methods Overall, there was a good and significant correlation between the species’ abundance estimates for the two methods for each gear-type, with correlation values (␳) ranging between 0.63 to 0.83 and 0.60 to 0.77 for the otter-trawl and rapido trawl samples,

Fig. 5. Median and upper-lower quartile of the  percentage (% abundance lab–% abundance pictures) for each taxonomic class identified by the photographic method.

respectively (Spearman rank correlation; p < 0.05 in all sampling stations for both types of gear). Species not detected by the photographic method (Fig. 4; black dots) were commonly characterized by low abundance and low average individual weight. This result was confirmed by the outcomes of the logistic regression analysis: among three predictors, only the continuous variables influenced the effectiveness of the photographic method in detecting discard species. Both the laboratory abundance and average weight showed positive slope values, with higher Walden statistics and lower significant p-values recorded for the abundance predictor compared to the average weight (Walden statistic 13.571 and 4.571; p < 0.001 and p < 0.05, respectively). Conversely, no significant gear effect was detected (p > 0.05). The photographic method showed a tendency to underestimate Arthropoda (Malacostraca) possibly due to the prevalence of Paguridae in the samples, which are difficult to detect due to their cryptic behaviour (e.g. specimens living inside sponges, shells); in contrast, a tendency toward the overestimation of Porifera was observed (Fig. 5). On average, the processing of the picture samples required 38 ± 6 min compared to 75 ± 9 min needed for the laboratory analysis (Wilcoxon test: Z = 8.46; p < 0.001). 3.3. Multivariate comparisons and spatial distribution of species’ assemblages The application of the RELATE routine showed the presence of strong and significant correlations between multivariate patterns based on the photographic and laboratory methods in both the

Please cite this article in press as: Piras, C., et al., A photographic method to identify benthic assemblages based on demersal trawler discards. Fish. Res. (2015), http://dx.doi.org/10.1016/j.fishres.2015.08.019

G Model

ARTICLE IN PRESS

FISH-4224; No. of Pages 10

C. Piras et al. / Fisheries Research xxx (2015) xxx–xxx

6

Table 2 Comparison of abundance and richness estimates from the laboratory and photographic methods. Significant correlations (p < 0.05) are highlighted in bold. OTB: otter trawl; TBB: rapido trawl Gear/Station

Laboratory method

Photographic method

OTB1 OTB2 OTB3 OTB4 Mean OTB TBB1 TBB2 TBB3 TBB4 Mean TBB

Richness 11 20 24 31 21.5 ± 8.3 24 28 29 13 23.5 ± 7.3

Richness 11 10 18 27 16.5 ± 7.9 16 17 15 12 15 ± 2.2

N. of ind. 52 227 586 345 302.5 ± 224 219 320 297 390 306.5 ± 70.5

N. of ind. 45 51 623 211 232.5 ± 271.4 142 166 212 1039 389.75 ± 433.8

Table 3 Multivariate correlation (␳) between the similarity/dissimilarity matrices based on the photographic and laboratory methods according to different data transformations of abundance data. Significant results (p < 0.001) are highlighted in bold Data transformation

Spatial scale

Percentage log (x + 1) Square root Double square root Presence/absence

Local (␳) 0.863 0.870 0.874 0.837 0.711

Regional (␳) 0.903 0.872 0.871 0.851 0.801

local and regional datasets (p < 0.001 for all of the comparisons, Table 3). The degree of correlation varied according to the applied data transformations; the correlation was the lowest for the presence/absence of data at both spatial scales ( = 0.71 and 0.81 at local and regional scale, respectively) and the highest for square root and percentage transformation at the local ( = 0.87) and regional scales ( = 0.90), respectively. 3.4. Local spatial scale Four main groups were identified through cluster analysis for both sampling techniques (cut-off threshold: 30% similarity). The percentage of correspondence between the classification obtained with the two methods was high (79%) and differed significantly from a random assignment of stations into clusters, as confirmed by the binomial test (p < 0.001). Thus, the overall spatial pattern of assemblage distributions was consistent between the methods despite, the lack of full correspondence in the classification (Fig. 6). Both methods detected consistent assemblage patterns in the southern part of the investigated area (the Po river, characterized by muddy sediments, clusters D and A) and the central part of the Veneto Region Administrative waters (approximately 6 nM from the coast to the offshore area, associated with the presence of detritic bottoms, cluster B). Only the group of stations closer to the seashore, and in particular those assigned to cluster C according to the laboratory method, did not show a stable classification and were assigned to a different group based on data collected with the photographic method. The PERMANOVA test also confirmed that the sampling method did not alter the multivariate repartition of assemblage compositions as determined by sediment features. The interaction between the two factors was not significant (pseudo-F: 1.309; p = 0.092), while both factors (method and sediment type) were significant (pseudo-F: 9.504; p < 0.001 and pseudo-F: 15.65; p < 0.001, respectively).

% of species identified by photographic method

% of individuals identified by photographic method

n. of species identified only by photographic method

Correlation between abundance estimated (lab. vs photograph.)

100.0 50.0 75.0 87.1 78.0 ± 21.3 66.7 60.7 51.7 92.3 67.8 ± 17.4

96.2 95.2 98.3 98.0 96.9 ± 1.5 96.8 95.3 91.6 99.5 95.8 ± 3.3

2 0 0 2 1.0 ± 1.2

0.60 0.82 0.63 0.67 0.60 0.75 0.75 0.77

0.7 ± 0.5

The SIMPER analysis performed for each method showed good agreement in the identification of taxa that significantly contributed to dissimilarity between neighbouring assemblages (i.e., clusters) according to the two methods (Appendix A1; A2 of Supplementary information). Specifically, the dissimilarity between assemblages B and D was due to Aporrhais pespelecani (more abundant in assemblage D) and Suberites domuncula (more abundant in assemblage B) for both sampling methods. Moreover, the laboratory dataset revealed a greater abundance of Paguristes eremita in cluster B (concordant with S. domuncula) and Astropecten irregularis (more abundant in assemblage D). Assemblage B was also characterized by a higher abundance of Porifera spp. compared to assemblage C (photographic method). Significant differences between assemblages C and D for the photographic dataset were due to the greater abundance of Turritella communis and Aequipecten opercularis in assemblage C and the greater abundance of A. pespelecani in assemblage D. Differences in the laboratory dataset were due to A. irregularis (predominant in assemblage D) and Liocarcinus vernalis (predominant in assemblage C) (Appendix A.2 of Supplementary information). Assemblage A was characterized by the predominance of Liocarcinus depurator and Ocnus planci compared to all neighbouring assemblages (B, C, D) using the laboratory method. Even the photographic method characterized assemblage A by L. depurator compared to all of the neighbouring assemblages instead of O. planci, which was found to be significant only in the comparison with assemblage D.

3.5. Regional spatial scale Five different assemblages for each sampling method (cut-off level of similarity: 18% in both analyses) were identified through cluster analyses at the regional spatial scale. Among the 34 sampling stations, 32 were ascribed to the same groups with both methods (94% correspondence), resulting in a significant matching of the classification (binomial test, p < 0.001). Accordingly, an almost complete correspondence in the assemblages’ spatial distributions was obtained with both methods (Fig. 7). The spatial assemblages identified by both methods showed a clear spatial pattern with the presence of two different spatial gradients: the increase in depth from the Italian coastline to the open sea and a latitude gradient from North to South. However, it is worth noting that the lack of correspondence between classifications obtained with the two methods was strictly related to coastal stations, similar to the results obtained at the local scale. The PERMANOVA analysis showed that the sampling method did not affect the multivariate composition of the assemblages

Please cite this article in press as: Piras, C., et al., A photographic method to identify benthic assemblages based on demersal trawler discards. Fish. Res. (2015), http://dx.doi.org/10.1016/j.fishres.2015.08.019

G Model FISH-4224; No. of Pages 10

ARTICLE IN PRESS C. Piras et al. / Fisheries Research xxx (2015) xxx–xxx

7

Fig. 6. Spatial distribution of assemblages classified at the local spatial scale according to cluster analysis (group average) performed on abundance data (% transformed) estimated by the laboratory (a) and photographic (b) methods.

Fig. 7. Spatial distribution of assemblages classified at the regional scale by cluster analysis (group average) performed on the abundance data (% transformed) obtained using the laboratory (a) and photographic (b) methods.

(pseudo-F: 0.44; p = 0.945), while the sedimentary features of the stations were a significant factor (pseudo-F: 3.309; p < 0.001). Moreover, PERMANOVA showed the lack of a significant interaction between the two factors (pseudo-F: 0.16; p = 1). For the laboratory dataset, the SIMPER analysis (Appendix B.1, B.2 of Supplementary information) highlighted the presence of a greater abundance of Holothuria forskali, Alcyonium palmatum, Porifera spp. and Pilumnus spinifer in assemblage C compared to all neighbouring assemblages (B, D, E,). The photographic dataset confirmed the greater abundance of H. forskali but also highlighted the role of the Bryozoa Amathia semiconvoluta in characterizing assemblage C. Assemblage D resulted in a greater abundance of T. communis and L. depurator compared to assemblage E; this result was in agreement between the two methods. Additionally, the photographic method demonstrated a greater abundance of O. planci. The two methods were in agreement in assessing the greater abundance of A. pespelecani and A. irregularis for assemblage E. 4. Discussion 4.1. Performance of the photographic method in assessing discarded species compositions and relative abundance Image analysis methods (e.g., underwater camera, Remotely Operated underwater Vehicle — ROV) have been applied in marine ecology research in a variety of studies ranging from research focused on the study of sea-bottom epibenthic fauna to the investigation of demersal fish fauna compositions and behaviour (Beuchel et al., 2010; Jorgensen and Gulliksen, 2001; Heagney et al., 2007; Kollmann and Stachowitsch, 2001; Parravicini et al., 2010;

Ponti et al., 2011). These techniques have also been adopted in field studies to evaluate the impact of bottom fishing on benthic communities (Collie et al., 2000) and have been used to assess the environmental impact of several anthropogenic activities on ecosystems (Carbines and Cole, 2009; Fraschetti et al., 2001). More recently, the use of Closed Circuit Television (CCTV) has been put forward for the assessment of commercial catch species and size compositions, mainly for control and monitoring purposes in commercial fishing activities (Kindt-Larsen et al., 2011). However, to the best of our knowledge image analysis techniques have not been applied to the analysis of species compositions and the relative abundance of the megaepifaunal fraction of discards. This is in contrast to the most often utilized “traditional” laboratory approach (Brind’Amour et al., 2014; Ellis et al., 2000, 2011; Froglia and Orel, 1971; Hall-Spencer et al., 1999; Kaiser et al., 2004; Marano et al., 1989; Pranovi et al., 2000, 2001; Relini et al., 1986; Sanchez et al., 2008) which takes advantage from the use of modified dredges and trawl fishing gears as sampling devices (Eleftheriou and Moore, 2005). In this context research and fishing vessels are used as sampling platforms, and the sample compositions are analysed in the laboratory (fishing vessels) or directly on board during trawl surveys (research vessels). The “traditional” process of megaepifaunal sample collection and analysis relies on gathering information on benthic habitats over wide areas, although with a limited spatial and temporal resolution; however, this process is time consuming and expensive. As a consequence of logistic and economic restrictions, multi-taxa field studies over broad spatial and temporal scales have rarely been undertaken (ICES, 2011a). A recent study from Brind’Amour and colleagues (2014) showed that data collected during trawl-surveys

Please cite this article in press as: Piras, C., et al., A photographic method to identify benthic assemblages based on demersal trawler discards. Fish. Res. (2015), http://dx.doi.org/10.1016/j.fishres.2015.08.019

G Model FISH-4224; No. of Pages 10 8

ARTICLE IN PRESS C. Piras et al. / Fisheries Research xxx (2015) xxx–xxx

could result in low taxonomic resolution in species identification (i.e., morphotypes) and that this limitation did not affect the quality of data in support of habitat identification. Therefore, the wide adoption of trawl surveys to fill information gaps has been proposed (i.e., in the framework of MSFD implementation) (Brind’Amour et al., 2014; Zampoukas et al., 2014). The development of the photographic method aims to overcome some of the practical restrictions related to the collection and laboratory analysis of biological discard samples, potentially allowing more intensive data collection using fishing vessels as sampling platforms. The photographic method showed that the analysis of 3 pictures (describing on average 85% of the cumulative species richness accessible by pictures, Fig. 4) allowed the detection of approximately 70% of the species present in the discard samples and a significant proportion of the estimated abundance compared with traditional laboratory methods. This correlation was obtained despite some limitations of the photographic method, such as the lower capability to detect species and morphospecies characterized by small size and low biomass. These species mainly belong to specific taxa such as the Paguridea (Arthropoda), which live inside shells often covered by the sponge S. domuncula, or Mollusca Gasteropoda (e.g., T. communis), which may be hidden by larger organisms. In contrast, the photographic method allowed the identification of some species whose assessment could not be efficiently determined in the laboratory due to sample deterioration (e.g., the frosting-defrosting process, as in the case of some Porifera) or overdispersion in the catch. 4.2. Multivariate pattern in benthic assemblages according to the photographic and laboratory methods The overall good and significant multivariate correlation observed between the abundance matrices for the two methods at both spatial scales showed that they convey consistent information. This is a relevant outcome indicating the efficacy of the photographic method in discriminating benthic assemblages. The RELATE test results showed different degrees of correlation between datasets based on different data transformations. Indeed, allowing a progressively greater contribution from the rarer species (not well detected by the photographic method) led to a decrease in the correlation between sampling techniques (Clarke, 1993). This result partially contrasts with the outcomes of the comparison of different visual techniques for epibenthic community assessment (Parravicini et al., 2010). Importantly, a similar classification of stations into assemblages and a concordant spatial distribution of benthic assemblages were obtained regardless of the method applied, although a better spatial overlap was achieved at the regional scale. This is probably due to the inherent differences in the two datasets being employed at different scales. Indeed, at the regional spatial scale discard samples were collected under fully controlled operating conditions such as adopting a standard sampling protocol (e.g., standard rapido gear and tows, performed in the same sampling season) (Scarcella et al., 2014), thereby minimizing potential sources of variation. This feature was also combined with a wide sampling area, a higher average distance among sampling stations and a broader gradient of environmental parameters (i.e., depth and sediment type) (Brambati et al., 1983; Gamulin-Brida, 1967) that in turn allowed the identification of more differentiated assemblages and resulted in the lack of significant differences between methods in the multivariate composition of data revealed by PERMANOVA analysis. At the local spatial scale most data were collected during fisherydependent surveys with limited control over the main sources of variability: different fishing gears were adopted, samples were

collected in different seasons and the sampling protocol was based on commercial practices. Additionally, the sampling effort by gear was spatially uneven (being linked to commercial practices) and the average distance between sampling sites was lower compared to the regional scale data. Moreover, the commercial fishing haul duration was usually longer compared to scientific surveys, thus implying that the swept area may encompass habitats with different features. In this context, the application of the photographic method on a local scale represents a direct test of this method under real conditions where the prevailing sampling platform is largely represented by fishing vessels. The results confirmed fairly good agreement between the two methods; the lack of controls for the above mentioned sources of variation strongly support the potential use of the photographic method using fishing vessels as the sampling platform. Moreover, at both spatial scales the observed lack of full correspondence was strictly related to the coastal zone, while the classification obtained for the off-shore areas was more consistent. This is due to the greater stability in the off-shore sediments and environmental parameters compared to the coastal environment, which is characterized by a succession of sedimentary features on a fine spatial scale (Brambati et al., 1983). Additionally, otter trawls have a tendency to achieve better sampling of epifauna species compared with rapido trawls, which are more efficient in the capture of infaunal species (Pranovi et al., 2001; Botter et al., 2006). Our analysis demonstrates that the photographic method has a similar performance in describing discard compositions regardless of the fishing gear adopted. Our results are in agreement with the existing literature on Adriatic Sea benthic biocenosis (Gamulin-Brida, 1967), although the gradients identified in the present work in some areas seem to be less strong compared to the report by Gamulin-Brida (1967). These differences could be due to a simplification of benthic communities that occurred in the following years due to the chronic fishing disturbances that characterise the Northern Adriatic Sea (Kaiser et al., 2000; Pranovi et al., 2001) or the anoxic events that occurred between 1970 and 1980 in the area (Ott, 1992). These types of differences may also depend on the use of different sampling methods because Gamulin-Brida (1967) used a grab to collect samples. 4.3. The potential use of fishing vessels as scientific platforms for benthic habitat description It is important to consider in the evaluation of the efficacy of the photographic method that its implementation allows for halving the time spent in sample processing compared to the laboratory analysis. However, the time needed to process samples is highly dependent on the training of observers for both methodologies. Moreover, it would be useful to collect some samples to be processed in the laboratory to validate the consistency of the taxonomic identification of species using the photographic method. The photographic method allows the user to easily overcome the most frequent logistical restrictions related to the collection of biological discard samples (i.e., the collection, transport and storage of voluminous samples). The simplicity of the approach and its time effectiveness suggest that it could be successfully used to collect information on benthic compositions over wide areas at a high spatial resolution, taking advantage of the use of fishing vessels as convenient sampling platforms. Picture collections could be implemented either by involving observers participating in fishery-dependent monitoring programs, such as the biological data collection performed under the Data Collection Framework (DCF) as suggested by ICES (2013). However, a substantial increase in sample collection would be achieved by involving the fishing industry in self-sampling activ-

Please cite this article in press as: Piras, C., et al., A photographic method to identify benthic assemblages based on demersal trawler discards. Fish. Res. (2015), http://dx.doi.org/10.1016/j.fishres.2015.08.019

G Model FISH-4224; No. of Pages 10

ARTICLE IN PRESS C. Piras et al. / Fisheries Research xxx (2015) xxx–xxx

ities after providing the necessary training (Kraan et al., 2013). Current knowledge on the spatial distribution of fishing efforts through VMS data shows that the spatial coverage of such activities potentially allows the collection of benthic data from wide areas and at high spatial resolutions from both in-shore and off-shore areas. Further implementation could be achieved by the use of CCTV, thereby extending its adoption beyond the target of assessing commercial and undersized discarded species (Kindt-Larsen et al., 2011). This approach could also provide information on those areas that are not easily accessible by trawl-surveys and have a high ecological value, including vulnerable marine ecosystems or relevant habitat-forming species (Brind’Amour et al., 2014). The potential of this approach can only be achieved by further methodological assessments to improve the method. In particular, its efficacy should be tested on different ecosystems and in commercial fishing contexts; moreover, the role of the sources of variability that were not fully controlled in this study (e.g., fishing gear effect, swept area, and seasonality) should be considered by comparing photographic data to quantitative benthic sampling. A relevant drawback of percentage data is that they cannot be used for the evaluation of the most familiar measures of abundance/biomass of taxa despite being useful for multivariate assessments (Parravicini et al., 2010). This drawback limits the capability to derive robust indicators for monitoring purposes; therefore, issues related to standardization and to the derivation of quantitative (i.e., additive) values should be further investigated. Overall the results of this study show that photographic method could be used to collect an appropriate amount of samples distributed over time and space. It could be used to classify habitats, delineate their boundaries and guide quantitative sampling. This last point is particularly important to provide a confident assessment of the status in an ecosystem approach context, such as that enforced in the MSFD context (Van Hoey et al., 2013). Enhanced knowledge of habitats and the spatial distribution of benthic species would increase our understanding of their relevance in support of the species’ life cycle (Gleason et al., 2013) and provide additional explanatory variables to be used for the interpretation of CPUE (catch per unit effort) over time and fishing effort spatio-temporal distributions (ICES, 2011b). 5. Conclusions and recommendations Our results indicate that epifauna discarded by commercial demersal trawlers can be efficiently characterized and quantified using the photographic method, showing an example of new and possible innovative approaches to using commercial fishing vessels as scientific platforms for research. Such method could complement other traditional approaches used to identify different benthic assemblages and their distributions. In particular: • Data collected by means of the photographic method applied onboard fishing vessels allowed identification of the megaepifauna discarded by demersal trawlers. • The direct comparison of this technique with the traditional laboratory method highlighted that its efficacy is validated despite some limits related to the underestimation of small-sized/low abundance and hidden species. The assessment of these sources of bias allows for a better understanding of the results and their potential applications. • The evaluation of the discard composition by means of the photographic method showed the identification of benthic assemblage distributions to be in agreement with the laboratory method, as confirmed by the highly significant agreement between the multivariate patterns and the reference literature on benthic assemblage spatial distributions in the study area.

9

• The application of the photographic method halves the processing time required by the laboratory method and could allow an enhancement of data collection following the removal of some practical limitations of sample collection by involving the fishing industry in picture collection during their fishing trips or by using CCTV onboard of fishing vessels. • Although the results of this study point to the high potential of this method to efficiently contribute to the definition of habitat and their distributions at high spatial resolution over a large spatial scale, further research is recommended to implement the method and verify the role of different potential sources of bias. • Statistical analyses should be undertaken to this method to contribute to the setting of appropriate indicators, management targets and measures to achieve good environmental status in marine ecosystems. Acknowledgements This work was developed in the framework of the international GAP2 Project (’Bridging the gap between science, stakeholders and policy makers. Phase 2: Integration of evidence-based knowledge and its application to science and management of fisheries and the marine environment’; www.gap2.eu) supported by the European Union under the FP7 Science in Society (SiS) Framework, SiS-20101.0-1, Mobilization and Mutual Learning Actions, Coordinating and support actions, Grant Agreement n. 266544. The Project aims to improve the actual management and develop a more adaptive system by actively involving fishermen, researchers, stakeholders and policy-makers in this process. The authors thank all fishermen and skippers involved in the GAP2 Project and the crew and scientific investigators of the Research Vessel G. Dallaporta for their helpful contribution to data collection during the SoleMON trawl surveys in 2012–2013 supported by FAO ADRIAMED. We are thankful to Stefano Lazzarini for the development of the Discard Analyzer 1.0 software. We are also thankful to the anonymous reviewer for the constructive revision that contributed to improve the quality of the manuscript. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.fishres.2015.08. 019. References Anderson, M.J., 2001. A new method for non-parametric multivariate analysis of variance. Aust. J. Ecol. 26, 32–46. Anderson, M.J., Gorley, R.N., Clarke, K.R., 2008. Permanova + for Primer: Guide to Software and Statistical Methods. PRIMER-E, Plymouth. Auster, P.J., Langton, R.W., 1998. The effects of fishing on fish habitat. In: Benaka, L. (Ed.), Fish Habitat: Essential Fish Habitat and Rehabilitation. American Fisheries Society, Maryland, pp. 150–187. Beuchel, F., Primicerio, R., Lonne, O.J., Gulliksen, B., Birkley, S., 2010. Counting and Measuring Epibenthic Organism from Digital Photographs: a Semiautomated Approach, 8. Limnol Oceanogr-meth, pp. 229–240. Botter, L., Nerlovic, V., Franceschini, G., Da Ponte, F., Pranovi, F., Giovanardi, O., Raicevich, S., 2006. Valutazione comparativa dello scarto degli attrezzi da pesca in Adriatico settentrionale. Biol. Mar. Medit 13 (1), 814–816. Brambati A., Ciabatti M., Fanzutti G.P., Marabini F., Marocco R., 1983. New sedimentological textural map of the Northern and Central Adriatic Sea. Boll. Oceanol. Teor. Applic. 1, 4, 267–271. Brind’Amour, A., Laffargue, P., Morin, J., Vaz, S., Foveau, A., Le Bris, H., 2014. Morphospecies and taxonomic sufficiency of benthic megafauna in scientific bottom trawl surveys. Cont. Shelf. Res. 72, 1–9. Carbines, G., Cole, R.G., 2009. Using a remote drift underwater video (DUV) to examine dredge impacts on demersal fishes and benthic habitat complexity in Foveaux Strait, Southern New Zealand. Fish. Res. 96, 230–237. Clarke, K.R., 1993. Non-parametric multivariate analyses of changes in community structure. Aust. J. Ecol. 18, 117–143. Clarke, K.R., Warwick, R.M., 2001. Change in Marine Communities: An Approach to Statistical Analysis and Interpretation, 2nd edition. PRIMER-E, Plymouth.

Please cite this article in press as: Piras, C., et al., A photographic method to identify benthic assemblages based on demersal trawler discards. Fish. Res. (2015), http://dx.doi.org/10.1016/j.fishres.2015.08.019

G Model FISH-4224; No. of Pages 10 10

ARTICLE IN PRESS C. Piras et al. / Fisheries Research xxx (2015) xxx–xxx

Clarke, K.R., Gorley, R.N., 2006. PRIMER v6: User Manual/Tutorial. PRIMER-E, Plymouth. Collie, J.S., Escanero, G.A., Valentine, P.C., 2000. Photographic evaluation of the impacts of bottom fishing on benthic epifauna. ICES J. Mar. Sci. 57, 987–1001. Cotter, J., Petitgas, P., Abella, A., Apostolaki, P., Mesnil, B., Politou, C.Y., Rivoirard, J., Rochet, M.J., Spedicato, M.T., Trenkel, V.M., Woillex, M., 2009. Towards an ecosystem approach to fisheries management (EAFM) when trawl surveys provide the main source of information. Aquat. Living Resour. 22, 243–254. Eleftheriou, A., Moore, D.C., 2005. Macrofauna techniques. In: Eleftheriou, A. (Ed.), Methods for the Study of Marine Benthos. John Wiley & Sons Ltd, pp. 160–228. Ellis, J.R., Rogers, S.I., Freeman, S.M., 2000. Demersal assemblages in the Irish Sea, St Georges Channel and Bristol Channel. Estuar. Coast. Shelf S. 51, 299–315. Ellis, J.R., Maxwell, T., Schratzberger, M., Rogers, S.I., 2011. The benthos and fish of offshore sandbank habitats in the southern North Sea. J. Mar. Biol. Assoc. U. K. 91 (6), 1319–1335. Fraschetti, S., Nike Bianchi, C., Terlizzi, A., Fanelli, G., Morri, C., Boero, F., 2001. Spatial variability and human disturbance in shallow subtidal hard substrate assemblages: a regional approach. Mar. Ecol. Prog. Ser. 212, 1–12. Froglia, C., Orel, G., 1971. Possibilità di delimitazione di biocenosi bentoniche in base agli organismi catturati con rete a strascico. Quad. Lab. Tecnol. Pesca. 1 (2), 37–41. Gamulin-Brida, H., 1967. The benthic fauna of the Adriatic Sea. Oceanogr. Mar. Biol. 55, 535–568. Garcia, S.M., Zerbi, A., Aliaume, C., Do Chi, T., Lasserre, G., 2003. Fisheries Impact on the Ecosystems, In: The Ecosystem Approach to Fisheries. Issues, Terminology, Principles, Institutional Foundations, Implementation and Outlook Fao Fish. Tech. Pap, 443. FAO, Rome, pp. 9–11. Garcia, S.M., Cochrane, K.L., 2005. Ecosystem approach to fisheries: a review of implementation guidelines. ICES J. Mar. Sci. 62, 311–318. Gleason, M., Feller, E.M., Merrifield, M., Copps, S., Fujita, R., Bell, M., Rienecke, S., Cook, C., 2013. A transactional and collaborative approach to reducing effects of bottom trawling. Conserv. Biol. 27, 470–479. Grati, F., Scarcella, G., Polidori, P., Domenichetti, F., Bolognini, L., Gramolini, R., Vasapollo, C., Giovanardi, O., Raicevich, S., Celic, I., Vrgoc, N., Isajlobic, I., Jenic, A., Marceta, B., Fabi, G., 2013. Multi-annual investigation of the spatial distributions of juvenile and adult sole (Solea solea L.) in the Adriatic Sea (northern Mediterranean). J. Sea Res. 84, 122–132. Hall-Spencer, J.M., Froglia, C., Atkinson, R.J., Moore, P.G., 1999. The impact of Rapido trawling for scallops, Pecten jacobaeus (L.), on the benthos of the Gulf of Venice. ICES J. Mar. Sci. 56, 111–124. Heagney, E.C., Lynch, T.P., Babcock, R.C., Suthers, I.M., 2007. Pelagic fish assemblages assessed using mid-water baited video: standardising fish counts using bait plume size. Mar. Ecol-Prog. Ser. 350, 255–266. ICES, 2011a. Studies on the wider biodiversity of marine habitats. In: Report of the Working Group on Biodiversity Science (WGBIODIV), 21-25 February 2011, Copenhagen, Denmark. ICES CM 2011/SSGEF:02, 94 pp. ICES, 2011b. Report of the Workshop on the utility of commercial CPUE and VMS data in assessment (WKCPUEFFORT). In: 5–7 April 2011. Copenhagen, Denmark, ICES CM 2011/ACOM:49, 71 pp. ICES, 2013. Report of the Workshop on Data Collection — Assessments of non-fishery impacts(WKDCF-NF). In: 8–10 October 2013. Copenhagen, ICES CM 2013/ACOM:74, 64 pp. Jennings, S., Kaiser, M.J., 1998. The effects of Fishing on Marine Ecosystems. Adv. Mar. Biol. 34, 201–351. Jorgensen, L.L., Gulliksen, B., 2001. Rocky bottom fauna in arctic Kongsfjord (Svalbard) studied by means of suction sampling and photography. Polar Biol. 24, 113–121. Kaiser, M.J., Edwars, D.B., Armstrong, P.J., Radford, K., Lough, N.E.L., Flatt, R.P., Jones, H.D., 1998. Changes in megafaunal benthic communities in different habitats after trawling disturbance. ICES J. Mar. Sci. 55, 353–361. Kaiser, M.J., Roger, S.I., Ellis, J.R., 1999. Importance of benthic habitat complexity for demersal fish assemblages. Am. Fish. S. S. 22, 212–223. Kaiser, M.J., Ramsay, K., Richardson, C.A., Spence, F.E., Brand, A.R., 2000. Chronic fishing disturbance has changed shelf sea benthic community structure. J. Anim. Ecol. 69, 494–503.

Kaiser, M.J., Bergmann, M., Hinz, H., Galanidi, M., Shucerksmith, R., Rees, E.I.S., Darbyshire, T., Ramsay, K., 2004. Demersal fish and epifauna associated with sandbank habitats. Estuar. Coast. Shelf S. 60, 445–456. Kindt-Larsen, L., Kirkegaard, E., Dalskov, J., 2011. Fully documented fishery: a tool to support a catch quota management system. ICES J. Mar. Sci. 68 (8), 1606–1610. Kollmann, H., Stachowitsch, M., 2001. Long-term changes in the benthos of the Northern Adriatic Sea: a phototransect approach. Mar. Ecol. 22 (1–2), 135–154. Kraan, M., Uhlmann, S., Steenbergen, J., Van Helmond, A.T.M., Van Hoof, L., 2013. The optimal process of self-sampling in fisheries: lessons learned in the Netherlands. J. Fish Biol. 83, 963–973. Marano, G., Ungaro, N., Vaccarella, R., 1989. Nota preliminare sulle comunità di macroinvertebrati dei fondi strascicabili dell’Adriatico pugliese. Thalassia Salentina 19, 3–19. Ott, J.A., 1992. The Adriatic benthos: problems and perspectives. In: Colombo, G., Ferrari, I., Ceccherelli, V., Rossi, R. (Eds.), Marine Eutrophication and Population Dynamics. Olsen & Olsen, Fredensborg, pp. 367–378. Parravicini, V., Micheli, F., Montefalcone, M., Villa, E., Morri, C., Bianchi, C.N., 2010. Rapid assessment of epibenthic communities: a comparison between two visual sampling techniques. J. Exp. Mar. Biol. Ecol. 395, 21–29. Ponti, M., Fava, F., Abbiati, M., 2011. Spatial-temporal variability of epibenthic assemblages on subtidal biogenic reefs in the northern Adriatic Sea. Mar. Biol. 158, 1447–1459. Pranovi, F., Raicevich, S., Franceschini, G., Farrace, M.G., Giovanardi, O., 2000. Rapido trawling in the northern Adriatic Sea: effects on benthic communities in an experimental area. ICES J. Mar. Sci. 57, 517–524. Pranovi, F., Raicevich, S., Franceschini, G., Torricelli, P., Giovanardi, O., 2001. Discards analysis and damage to non-target species in the rapido trawl fishery. Mar. Biol. 139, 863–875. Relini, G., Peirano, A., Tunesi, L., 1986. Osservazioni sulle comunità dei fondi strascicabili del Mar Ligure Centro-Orientale. Bollettino dei musei istituti biologici –Università degli Studi di Genova, 52 suppl., 139–161. Rodrigues, A.M., Meireles, S., Pereira, T., Gama, A., Quintino, V., 2006. Spatial patterns of benthic macroinvertebrates in intertidal areas of a Southern European estuary: the Taguas, Portugal. Hydrobiologia 555, 99–113. Sanchez, F., Serrano, A., Parra, S., Ballesteros, M., Cartes, J.E., 2008. Habitat characteristics as determinant of the structure and spatial distribution of epibenthic and demersal communities of Le Danois Bank (Cantabrian Sea, N. Spain). J. Mar. Syst. 72, 64–86. Scarcella, G., Grati, F., Raicevich, S., Russo, T., Gramolini, R., Scott, R.D., Polidori, P., Domenichetti, F., Bolognini, L., Giovanardi, O., Celic, I., Sabatini, L., Vrgoc, N., Isajlovic, I., Marceta, B., Fabi, G., 2014. Common sole in the Northern Adriatic Sea: possible spatial management scenarios to rebuild the stock. J. Sea Res. 89, 12–22. Van Hoey, G., Permuy, D.C., Vandendriessche, S., Vincx, M., Hostens, K., 2013. An ecological quality status assessment procedure for soft-sediment benthic habitats: weighing alternative approaches. Ecol. Indic. 25, 266–278. Zampoukas N., Palialexis A., Duffek A., Graveland J., Giorgi G., Hagebro C., Hanke G., Korpinen S., Tasker M., Tornero V., Abaza V., Battaglia P., Caparis M., Dekeling R., Frias Vega M., Haarich M., Katsanevakis S., Klein H., Krzyminski W., Laamanen M., Le Gac J.C., Leppanen J.M., Lips U., Maes T., Magaletti E., Malcolm S., Marques J.M., Mihail O., Moxon R., O’Brien C., Panagiotidis P., Penna M., Piroddi C., Probst W.N., Raicevich S., Trabucco B., Tunesi L., Van der Graaf S., Weiss A., Wernersson A.S. and Zevenboom W., 2014. Technical guidance on monitoring for the Marine Strategy Framework Directive. European Commission, Joint Research Centre, Institute for Environment and Sustainability, Report EUR 26,499 EN, pp. 175. doi:10.2788/70344 Zenetos, A., Revkov, N.K., Konsulova, T., Sergeeva, N., Simboura, N., Todorova, V.R., Zaika, V.E., 2000. Coastal benthic diversity in the black and Aegean Seas. Mediterr. Mar. Sci. 1/2, 105–117.

Please cite this article in press as: Piras, C., et al., A photographic method to identify benthic assemblages based on demersal trawler discards. Fish. Res. (2015), http://dx.doi.org/10.1016/j.fishres.2015.08.019