TY - JOUR
AB - Abstract In-trawl camera systems promise to improve the resolution of trawl sampling used to ground-truth the interpretation of acoustic survey data. In this study, the residence time of fish in front of the Deep Vision camera system, used to identify, measure and count fish inside the trawl, was analysed to determine the reliability of spatial distribution recorded by the system. Although Atlantic herring (Clupea harengus), haddock (Melanogrammus aeglefinus), and most Atlantic cod (Gadus morhua) moved quickly back through the aft part of the pelagic trawl, saithe (Pollachius virens) spent up to 4 min in front of the system. The residence time increased for saithe and cod when other individuals were present, and cod swimming in the low water flow close to the trawl netting spent longer there than cod at the centre of the trawl. Surprisingly, residence time was not related to the size of the fish, which may be explained by the collective behaviour of shoaling fish. Our findings suggest that while in-trawl images can be used to identify, measure and count most species, when sampling fast-swimming species such as saithe the position inferred from when they were imaged may not reflect the actual spatial distribution prior to capture. Introduction Fisheries surveys monitor both ocean communities and individual species in order to understand inter-species relationships, the status of commercial stocks, and areas that may be particularly impacted by commercial fishing activity. The trawl is a tool that is frequently used to collect biological information, estimate population abundance and verify acoustic data (Gunderson, 1993; Mehl et al., 2015). To verify acoustic data, the trawl is placed at the depth of interest and the organisms collected in the codend are correlated with the backscatter on the echosounder (Simmonds and Maclennan, 2005). Single-codend trawls collect all the organisms together while multiple-codend trawls (Multisampler; Engås et al., 1997) allow for several separate codends to sample discrete time periods and depths. Increasing the number of sample intervals may improve the correlation between the acoustic targets and the actual catch. For this reason, new tools that allow for continuous sampling are being tested as alternatives to the single- and multi-codends. The Deep Vision system (Rosen et al., 2013) is one example of such tools, consisting of a stereo camera with strobe lighting placed between the trawl’s extension and codend to collect a continuous record of organisms passing. However, more precise sampling brings a need to better understand how organisms move through the trawl to the codend. It is essential that the time and depth at which the camera records an organism along the trawl path correspond to the position at which the organism was initially recorded on the echosounder. Previous studies in pelagic trawls have shown that adult Atlantic cod (Gadus morhua) orient themselves in the direction of tow and swim forward. Cod had a residence time in the trawl that was longer than predicted simply by the trawl’s speed through water and the passage rate decreased as individuals move through the narrowing of the trawl to the codend (Rosen et al., 2012). The Deep Vision system is narrower than a typical extension or codend, and may therefore further delay the passage of fish through to the codend. The residence time may be expected to differ between species due to variations in their swimming capacity and behaviour (see reviews by Videler, 1993; Videler and He, 2010). Artificial lighting may also affect the rate at which fish move through the trawl. Previous studies in demersal trawls have demonstrated that additional factors such as fish length and density influence fish behaviour, which may affect residence time (see reviews by Wardle, 1993; Engås, 1994; Godø, 1994; Winger et al., 2010). These factors are likely relevant in pelagic trawls, however they have not been investigated in detail. To be able to estimate the time and location where fish were initially captured, it is essential to evaluate the amount of time that different species spend in front of Deep Vision prior to using the system as a tool to verify acoustic targets. Our goal was to quantify how long it took four common and commercially important Northeast Atlantic species [Atlantic herring (Clupea harengus), haddock (Melanogrammus aeglefinus), Atlantic cod, and saithe (Pollachius virens)] to pass through the aftmost portion of the pelagic trawl ahead of and leading into the Deep Vision section. We concentrated our observations at the area in front of Deep Vision for two reasons: First, it is a likely area for individuals to congregate due to the reduced trawl cross-section and the strobe lighting used with Deep Vision. Second, the limited range and field of view of cameras placed in trawls mean that placement farther forward in the trawl do not afford a full overview by the cameras. Material and methods Experiments were conducted along the coast of northern Norway in February 2016 on board RV “G.O. Sars” (Figure 1). A standard pelagic sampling trawl used by the Institute of Marine Research (Åkra trawl; Valdemarsen and Misund, 1994) was towed in a straight line for 1 h at an average speed through the water of 1.5 ms−1 (SD ±0.1). When the sensor used to measure speed through water (Trawl Speed Sensor, SCANMAR AS, Åsgårstrand, Norway) occasionally lost contact with the vessel, warp tension and speed over ground (acquired from GPS) were used to maintain towing resistance consistent with a speed through the water of 1.5 ms−1. An 8-m long extension constructed of 40-mm mesh was added to the end of the trawl in order to create a smooth taper into Deep Vision (Figure 2) and eliminate the “pocket” observed in a previous cruise when Deep Vision was joined directly to the trawl. The extension’s performance was verified by a camera mounted on a towed underwater vehicle (FOCUS 400, MacArtney A/S, Esbjerg, Denmark). Figure 1. View largeDownload slide Position of pelagic hauls used for behavioural observations in front of Deep Vision along the northern coast of Norway in February 2016. Figure 1. View largeDownload slide Position of pelagic hauls used for behavioural observations in front of Deep Vision along the northern coast of Norway in February 2016. Figure 2. View largeDownload slide Schematic representation of the GoPro cameras in relation to the Deep Vision section, extension and Åkra trawl. The dashed arrows indicate orientation of the cameras with example frames. Camera position A covered fish from the extension to Deep Vision’s photo chamber, while camera position B covered interactions close to Deep Vision in greater detail. A steady white light was placed at camera position B during night hauls and when fish were aggregated at depths with insufficient natural illumination for the cameras. An ADCP (black rectangle at camera position B) was also placed 3 m in front of the Deep Vision section to record water flow at camera position B. Residence time of individuals was recorded in Section 1 (time they entered field of view of camera position A until time they entered field of view of camera position B or swam forward again, out of the field of view of camera position A) and Section 2 (time they entered field of view of camera position B until time they entered Deep Vision or swam forward again, out of the field of view of the camera position B). The lower left panel is a frame from the footage from camera position A. The arrows in the lower left panel show the seams connecting the extension to the trawl and Deep Vision to the extension, which are visible from the footage in camera position A. The lower right panel shows Deep Vision when connected to the trawl, observed from a remotely operated towed vehicle. Figure 2. View largeDownload slide Schematic representation of the GoPro cameras in relation to the Deep Vision section, extension and Åkra trawl. The dashed arrows indicate orientation of the cameras with example frames. Camera position A covered fish from the extension to Deep Vision’s photo chamber, while camera position B covered interactions close to Deep Vision in greater detail. A steady white light was placed at camera position B during night hauls and when fish were aggregated at depths with insufficient natural illumination for the cameras. An ADCP (black rectangle at camera position B) was also placed 3 m in front of the Deep Vision section to record water flow at camera position B. Residence time of individuals was recorded in Section 1 (time they entered field of view of camera position A until time they entered field of view of camera position B or swam forward again, out of the field of view of camera position A) and Section 2 (time they entered field of view of camera position B until time they entered Deep Vision or swam forward again, out of the field of view of the camera position B). The lower left panel is a frame from the footage from camera position A. The arrows in the lower left panel show the seams connecting the extension to the trawl and Deep Vision to the extension, which are visible from the footage in camera position A. The lower right panel shows Deep Vision when connected to the trawl, observed from a remotely operated towed vehicle. The Deep Vision camera system has been described in detail by Rosen and Holst (2013), Rosen et al. (2013), and Underwood et al. (2014). It consists of a stereo camera and strobe lighting system (5 Hz) mounted inside a large frame to provide well-illuminated colour images of all objects for species identification and length measurement. A tapered panel guides individual fish from a 1 m2 cross-section of the trawl to within the field of view of the cameras (trapezoidal cross-section, 0.24 m2). The vertical opening of the trawl net, trawl speed through water, door spread and height were measured with acoustic trawl instrumentation (SCANMAR AS, Åsgårdstand, Norway; SIMRAD Kongsberg Maritime AS, Horten, Norway). The trawl was towed at 50–130 m depths, matching where fish were observed on the echosounder. Temperature at the fishing depth was measured between hauls using a CTD (SAIV AS, Bergen, Norway) and ranged from 5.7 to 7.5 °C. A total of 18 hauls were carried out, 15 of which recorded fish passages that could be used for behavioural observations (Table 1). GoPro HERO4 cameras (GoPro Inc, San Mateo, USA) in 300 m rated housings (iQ Sub, Ostrava, Czech Republic) were mounted in protective frames on the lower panel of the trawl at two positions (Figure 2). One camera was mounted 1.5 m in front of the seam between the Åkra trawl and the 8-m tapered extension. The camera was oriented to look aft and observe fish in most of the extension (Section 1). The other camera was placed 7.3 m farther aft and observed fish in front of Deep Vision (Section 2). A steady white light (Luxus Compact LED; MacArtney A/S, Esbjerg, Denmark) was turned on in Section 2 during night hauls and when fish were aggregated at depths with insufficient natural illumination for the cameras (Table 1). Table 1. Station data and number of fish of each species observed in front of the Deep Vision. Haul Time (UTC) Steady light placed at Position B Section 1 Observations Start Depth (m) Trawl speed range during observations (ms−1) Average trawl speed during observations (± SD) Atlantic herring Haddock Atlantic cod Saithe 74 8:34 161 21 75 11:45 163 1.39–1.55 1.46 (±0.08) 3 76 15:38 X 244 1.25–1.67 1.45 (±0.09) 31 9 77 17:59 X 121 0.96–1.56 1.10 (±0.15) 43 34 3 78 7:34 X X 200 1.24–1.68 1.52 (±0.09) 43 13 19 82 7:36 X 162 1.41–1.98 1.73 (±0.19) 6 2 83 9:33 X 173 1.42–1.68 1.54 (±0.07) 15 82 5 84 12:13 177 1.45–1.45 1.45 1 86 23:28 X X 177 1.50–1.97 1.67 (±0.12) 5 21 10 88 9:42 179 0.94–1.22 1.13 (±0.08) 6 28 1 89 11:52 179 1.48–1.86 1.77 (±0.13) 2 2 3 90 21:32 X X 177 59 3 95 23:47 X X 162 1.28–1.88 1.55 (±0.09) 60 3 98 7:45 X 145 1.68–1.68 1.68 1 107 7:52 X 104 1.61–1.61 1.61 (±0.00) 2 Total 205 28 224 79 Haul Time (UTC) Steady light placed at Position B Section 1 Observations Start Depth (m) Trawl speed range during observations (ms−1) Average trawl speed during observations (± SD) Atlantic herring Haddock Atlantic cod Saithe 74 8:34 161 21 75 11:45 163 1.39–1.55 1.46 (±0.08) 3 76 15:38 X 244 1.25–1.67 1.45 (±0.09) 31 9 77 17:59 X 121 0.96–1.56 1.10 (±0.15) 43 34 3 78 7:34 X X 200 1.24–1.68 1.52 (±0.09) 43 13 19 82 7:36 X 162 1.41–1.98 1.73 (±0.19) 6 2 83 9:33 X 173 1.42–1.68 1.54 (±0.07) 15 82 5 84 12:13 177 1.45–1.45 1.45 1 86 23:28 X X 177 1.50–1.97 1.67 (±0.12) 5 21 10 88 9:42 179 0.94–1.22 1.13 (±0.08) 6 28 1 89 11:52 179 1.48–1.86 1.77 (±0.13) 2 2 3 90 21:32 X X 177 59 3 95 23:47 X X 162 1.28–1.88 1.55 (±0.09) 60 3 98 7:45 X 145 1.68–1.68 1.68 1 107 7:52 X 104 1.61–1.61 1.61 (±0.00) 2 Total 205 28 224 79 Table 1. Station data and number of fish of each species observed in front of the Deep Vision. Haul Time (UTC) Steady light placed at Position B Section 1 Observations Start Depth (m) Trawl speed range during observations (ms−1) Average trawl speed during observations (± SD) Atlantic herring Haddock Atlantic cod Saithe 74 8:34 161 21 75 11:45 163 1.39–1.55 1.46 (±0.08) 3 76 15:38 X 244 1.25–1.67 1.45 (±0.09) 31 9 77 17:59 X 121 0.96–1.56 1.10 (±0.15) 43 34 3 78 7:34 X X 200 1.24–1.68 1.52 (±0.09) 43 13 19 82 7:36 X 162 1.41–1.98 1.73 (±0.19) 6 2 83 9:33 X 173 1.42–1.68 1.54 (±0.07) 15 82 5 84 12:13 177 1.45–1.45 1.45 1 86 23:28 X X 177 1.50–1.97 1.67 (±0.12) 5 21 10 88 9:42 179 0.94–1.22 1.13 (±0.08) 6 28 1 89 11:52 179 1.48–1.86 1.77 (±0.13) 2 2 3 90 21:32 X X 177 59 3 95 23:47 X X 162 1.28–1.88 1.55 (±0.09) 60 3 98 7:45 X 145 1.68–1.68 1.68 1 107 7:52 X 104 1.61–1.61 1.61 (±0.00) 2 Total 205 28 224 79 Haul Time (UTC) Steady light placed at Position B Section 1 Observations Start Depth (m) Trawl speed range during observations (ms−1) Average trawl speed during observations (± SD) Atlantic herring Haddock Atlantic cod Saithe 74 8:34 161 21 75 11:45 163 1.39–1.55 1.46 (±0.08) 3 76 15:38 X 244 1.25–1.67 1.45 (±0.09) 31 9 77 17:59 X 121 0.96–1.56 1.10 (±0.15) 43 34 3 78 7:34 X X 200 1.24–1.68 1.52 (±0.09) 43 13 19 82 7:36 X 162 1.41–1.98 1.73 (±0.19) 6 2 83 9:33 X 173 1.42–1.68 1.54 (±0.07) 15 82 5 84 12:13 177 1.45–1.45 1.45 1 86 23:28 X X 177 1.50–1.97 1.67 (±0.12) 5 21 10 88 9:42 179 0.94–1.22 1.13 (±0.08) 6 28 1 89 11:52 179 1.48–1.86 1.77 (±0.13) 2 2 3 90 21:32 X X 177 59 3 95 23:47 X X 162 1.28–1.88 1.55 (±0.09) 60 3 98 7:45 X 145 1.68–1.68 1.68 1 107 7:52 X 104 1.61–1.61 1.61 (±0.00) 2 Total 205 28 224 79 A trawl-mounted acoustic Doppler current profiler (ADCP; 2 MHz Aquadopp Profiler; Nortek AS, Rud, Norway) was used to measure the water flow in 10 cm bins across the trawl’s cross-section for one haul (haul 107). The ADCP was mounted inside a protective frame and attached to the outside of the lower panel of the trawl, 3 m in front of the seam between the extension and Deep Vision section (diameter measured at 1.2 m during trawling; Figure 2). This placed the transducer 10–15 cm from the netting panel, which meant that the water flow for the bin closest to the net panel (5–15 cm) was not recorded reliably, as it included the netting. The catch was measured using standard biological sampling procedures at the Institute of Marine Research (Mjanger et al., 2014). In addition, the fork length of each individual gadoid that passed into the codend (fish length) was measured post-cruise from the images collected from Deep Vision (Rosen et al., 2013). All fish could be identified to species and 84% of the gadoids passing through the camera system could be measured. Video analysis Video footage from the GoPro cameras was analysed using Observer XT software, vers. 12.1 (Noldus Information Technology, Wageningen, The Netherlands). The behaviours of all individual gadoids were registered, while the more numerous herring were subsampled by recording the behaviour of the first 20 non-overlapping individuals from the start of each 26-min video file. As three video files were recorded per haul, this resulted in subsampling herring at multiple instances throughout the haul. The residence time of an individual was recorded as the number of whole seconds in which an individual was present in a section. In the first section (Figure 2; Section 1; 12.5–5.2 m in front of Deep Vision), the residence time was recorded from the time the individual entered the field of view of camera position A until it entered the field of view of camera position B or swam forward again, out of the field of view of camera position A. In the second section (Figure 2; Section 2; up to 5.2 m in front of Deep Vision), the residence time was recoded from the time the individual entered the field of view of camera position B until it entered Deep Vision or swam forward again, out of the field of view of camera position B. Footage from camera position A (Section 1) was only available for eight hauls, while footage from camera position B (Section 2) was available for all hauls (Table 1). The behaviour of individual herring, haddock, cod, and saithe was recorded in Section 2 only. In Section 1, the long distance from the camera (up to 9.5 m) and the low contrast made accurate classification of fish farthest away from the camera difficult. Each observation in Section 2 was classified according to four categories: orientation, swimming direction, position and fish density. Orientation of the fish body was described on the basis of the direction of the head, with individuals classified as facing the “towing direction” (head oriented towards the vessel) or “Deep Vision” (head oriented towards the codend). If the orientation changed in the course of the observation, it was recorded as “varied”. Swimming direction was classified as “backward”—moving steadily towards Deep Vision, “forward”—moving forward out of the field of view towards the trawl opening, “keeping pace”—maintaining position in the trawl, or “irregular”—erratic up/down or side-to-side burst swimming towards the netting walls. Position relative to the trawl netting was recorded as “centre of the trawl” if the individual maintained its distance from the trawl netting, or “close to netting” if in contact with or within roughly 20 cm of the netting. Orientation, swimming direction and position were combined to create an overall behaviour. For example, “towing direction-backward-close to netting (TD-B-N)” would classify fish with their head oriented towards the vessel moving steadily towards Deep Vision and in contact with or within roughly 20 cm of the netting (Table 2). Overall behaviours with <8% of the total number of each species were grouped together into an “other” category (Table 3). Fish density reflects the total number of individuals present in Section 2 (regardless of species) and was classified as “single” if no additional individuals were present during the observation or “multiple” when other fish were present at any time during the observation. Table 2. Counts of overall behaviours of each species. Overall behaviour (orientation-swimming direction-position) Herring Haddock Cod Saithe Orientated in the towing direction, swimming back towards the Deep Vision in the centre of the trawl (TD-B-C) 18 67 19 Orientated in the towing direction, swimming back towards the Deep Vision close to netting (TD-B-N) 2 6 119 6 Orientated in the towing direction, keeping pace with the trawl in the centre of the trawl (TD-K-C) 2 4 40 Orientated in the towing direction, keeping pace with the trawl close to netting 5 12 1 Orientated in the towing direction-irregular-close to netting 3 1 1 2 Orientated towards Deep Vision, swimming back towards the Deep Vision in the centre of the trawl 5 6 4 Orientated towards Deep Vision, swimming back towards the Deep Vision close to netting 2 12 4 Varied orientation, swimming back towards the Deep Vision in centre of the trawl 1 Varied orientation, swimming back towards the Deep Vision close to netting 2 Varied orientation, keeping pace with the trawl close to netting 2 Varied orientation-irregular-close to netting 186 1 1 2 Overall behaviour (orientation-swimming direction-position) Herring Haddock Cod Saithe Orientated in the towing direction, swimming back towards the Deep Vision in the centre of the trawl (TD-B-C) 18 67 19 Orientated in the towing direction, swimming back towards the Deep Vision close to netting (TD-B-N) 2 6 119 6 Orientated in the towing direction, keeping pace with the trawl in the centre of the trawl (TD-K-C) 2 4 40 Orientated in the towing direction, keeping pace with the trawl close to netting 5 12 1 Orientated in the towing direction-irregular-close to netting 3 1 1 2 Orientated towards Deep Vision, swimming back towards the Deep Vision in the centre of the trawl 5 6 4 Orientated towards Deep Vision, swimming back towards the Deep Vision close to netting 2 12 4 Varied orientation, swimming back towards the Deep Vision in centre of the trawl 1 Varied orientation, swimming back towards the Deep Vision close to netting 2 Varied orientation, keeping pace with the trawl close to netting 2 Varied orientation-irregular-close to netting 186 1 1 2 The fish drawings illustrate the overall behaviour. Table 2. Counts of overall behaviours of each species. Overall behaviour (orientation-swimming direction-position) Herring Haddock Cod Saithe Orientated in the towing direction, swimming back towards the Deep Vision in the centre of the trawl (TD-B-C) 18 67 19 Orientated in the towing direction, swimming back towards the Deep Vision close to netting (TD-B-N) 2 6 119 6 Orientated in the towing direction, keeping pace with the trawl in the centre of the trawl (TD-K-C) 2 4 40 Orientated in the towing direction, keeping pace with the trawl close to netting 5 12 1 Orientated in the towing direction-irregular-close to netting 3 1 1 2 Orientated towards Deep Vision, swimming back towards the Deep Vision in the centre of the trawl 5 6 4 Orientated towards Deep Vision, swimming back towards the Deep Vision close to netting 2 12 4 Varied orientation, swimming back towards the Deep Vision in centre of the trawl 1 Varied orientation, swimming back towards the Deep Vision close to netting 2 Varied orientation, keeping pace with the trawl close to netting 2 Varied orientation-irregular-close to netting 186 1 1 2 Overall behaviour (orientation-swimming direction-position) Herring Haddock Cod Saithe Orientated in the towing direction, swimming back towards the Deep Vision in the centre of the trawl (TD-B-C) 18 67 19 Orientated in the towing direction, swimming back towards the Deep Vision close to netting (TD-B-N) 2 6 119 6 Orientated in the towing direction, keeping pace with the trawl in the centre of the trawl (TD-K-C) 2 4 40 Orientated in the towing direction, keeping pace with the trawl close to netting 5 12 1 Orientated in the towing direction-irregular-close to netting 3 1 1 2 Orientated towards Deep Vision, swimming back towards the Deep Vision in the centre of the trawl 5 6 4 Orientated towards Deep Vision, swimming back towards the Deep Vision close to netting 2 12 4 Varied orientation, swimming back towards the Deep Vision in centre of the trawl 1 Varied orientation, swimming back towards the Deep Vision close to netting 2 Varied orientation, keeping pace with the trawl close to netting 2 Varied orientation-irregular-close to netting 186 1 1 2 The fish drawings illustrate the overall behaviour. Table 3. Summary of average residence times (s) and behaviours of each species. Herring Haddock Cod Saithe Residence time (s) Counts Residence time (s) Counts Residence time (s) Counts Residence time (s) Counts Section 1 6.0 (±3.1) 43 6.4 (±4.1) 17 8.4 (±4.8) 75 30.4 (±51.8) 38 Section 2 7.1 (±6.8) 205 12.8 (±11.2) 28 12.5 (±19.0) 224 41.7 (±60.2) 79 Sections 1 and 2 17.2 (±11.5) 43 20.5 (±12.5) 17 24.7 (±29.3) 75 57.9 (±63.0) 38 Section 2 categories: Density Multiple fish 6.8 (±6.8) 178 13.4 (±8.8) 11 15.1 (±24.5) 141 53.6 (±70.5) 52 Single fish 9.2 (±6.8) 27 12.4 (±12.8) 17 8.9 (±7.2) 83 18.9 (±18.0) 27 Steady light Present 7.1 (±6.8) 205 20.6 (±16) 5 14.3 (±25.9) 105 25.8 (±44.8) 41 Absent 0 11.1 (±9.5) 23 11.4 (±13.0) 119 58.9 (±70.0) 38 Orientation Deep Vision 6.0 (±4.2) 7 0 5.0 (±3.3) 18 5.0 (±1.4) 8 Towing direction 23.8 (±15.1) 10 12.7 (±11.4) 27 13.6 (±20.9) 203 47.7 (±63.0) 68 Varied 6.3 (±4.8) 188 16.0 1 5.0 (±3.5) 3 5.0 (±1.7) 3 Position Centre of the trawl 6.8 (±4.8) 5 11.3 (±10.1) 20 8.9 (±9.8) 77 48.8 (±64.9) 64 Close to netting 7.1 (±6.8) 200 16.6 (±13.7) 8 14.8 (±23.5) 147 11.5 (±7.6) 15 Swimming direction Backward 7.1 (±4.2) 9 9.7 (±7.4) 24 10.3 (±16.1) 206 9.1 (±6.8) 34 Keeping pace 28.9 (±14.6) 7 31.0 (±0.0) 2 44.8 (±35.9) 16 72.2 (±71.1) 41 Irregular 6.3 (±4.9) 189 32.0 (±22.6) 2 14.0 (±7.1) 2 7.3 (±2.9) 4 Measured length NA 23 188 66 Overall behavior (orientation- swimming direction-position) TD-B-N 11.5 (±5.8) 6 12.6 (±20.7) 119 TD-B-C 9.1 (±7.9) 18 7.7 (±3.0) 67 9.2 (±7.4) 19 TD-K-C 73.3 (±71.7) 40 V-I-N 6.2 (±4.8) 186 Other 15.9 (±14.0) 19 31.5 (±13.1) 4 22.1 (±30.3) 38 9.6 (±7.4) 20 Herring Haddock Cod Saithe Residence time (s) Counts Residence time (s) Counts Residence time (s) Counts Residence time (s) Counts Section 1 6.0 (±3.1) 43 6.4 (±4.1) 17 8.4 (±4.8) 75 30.4 (±51.8) 38 Section 2 7.1 (±6.8) 205 12.8 (±11.2) 28 12.5 (±19.0) 224 41.7 (±60.2) 79 Sections 1 and 2 17.2 (±11.5) 43 20.5 (±12.5) 17 24.7 (±29.3) 75 57.9 (±63.0) 38 Section 2 categories: Density Multiple fish 6.8 (±6.8) 178 13.4 (±8.8) 11 15.1 (±24.5) 141 53.6 (±70.5) 52 Single fish 9.2 (±6.8) 27 12.4 (±12.8) 17 8.9 (±7.2) 83 18.9 (±18.0) 27 Steady light Present 7.1 (±6.8) 205 20.6 (±16) 5 14.3 (±25.9) 105 25.8 (±44.8) 41 Absent 0 11.1 (±9.5) 23 11.4 (±13.0) 119 58.9 (±70.0) 38 Orientation Deep Vision 6.0 (±4.2) 7 0 5.0 (±3.3) 18 5.0 (±1.4) 8 Towing direction 23.8 (±15.1) 10 12.7 (±11.4) 27 13.6 (±20.9) 203 47.7 (±63.0) 68 Varied 6.3 (±4.8) 188 16.0 1 5.0 (±3.5) 3 5.0 (±1.7) 3 Position Centre of the trawl 6.8 (±4.8) 5 11.3 (±10.1) 20 8.9 (±9.8) 77 48.8 (±64.9) 64 Close to netting 7.1 (±6.8) 200 16.6 (±13.7) 8 14.8 (±23.5) 147 11.5 (±7.6) 15 Swimming direction Backward 7.1 (±4.2) 9 9.7 (±7.4) 24 10.3 (±16.1) 206 9.1 (±6.8) 34 Keeping pace 28.9 (±14.6) 7 31.0 (±0.0) 2 44.8 (±35.9) 16 72.2 (±71.1) 41 Irregular 6.3 (±4.9) 189 32.0 (±22.6) 2 14.0 (±7.1) 2 7.3 (±2.9) 4 Measured length NA 23 188 66 Overall behavior (orientation- swimming direction-position) TD-B-N 11.5 (±5.8) 6 12.6 (±20.7) 119 TD-B-C 9.1 (±7.9) 18 7.7 (±3.0) 67 9.2 (±7.4) 19 TD-K-C 73.3 (±71.7) 40 V-I-N 6.2 (±4.8) 186 Other 15.9 (±14.0) 19 31.5 (±13.1) 4 22.1 (±30.3) 38 9.6 (±7.4) 20 SDs are in parentheses. TD is orientated in the “towing direction”, and V is a “varied” orientation. B is moving “backward”, K is “keeping pace” with the trawl, and I is swimming in an “irregular” direction (erratic up/down or side-to-side). C is positioned in the “centre of the trawl”, and N is positioned “close to netting”.Note: The residence times between each species significantly differ in Section 2 except for cod vs. haddock (Wilcoxon rank-sum tests, p < 0.05). Table 3. Summary of average residence times (s) and behaviours of each species. Herring Haddock Cod Saithe Residence time (s) Counts Residence time (s) Counts Residence time (s) Counts Residence time (s) Counts Section 1 6.0 (±3.1) 43 6.4 (±4.1) 17 8.4 (±4.8) 75 30.4 (±51.8) 38 Section 2 7.1 (±6.8) 205 12.8 (±11.2) 28 12.5 (±19.0) 224 41.7 (±60.2) 79 Sections 1 and 2 17.2 (±11.5) 43 20.5 (±12.5) 17 24.7 (±29.3) 75 57.9 (±63.0) 38 Section 2 categories: Density Multiple fish 6.8 (±6.8) 178 13.4 (±8.8) 11 15.1 (±24.5) 141 53.6 (±70.5) 52 Single fish 9.2 (±6.8) 27 12.4 (±12.8) 17 8.9 (±7.2) 83 18.9 (±18.0) 27 Steady light Present 7.1 (±6.8) 205 20.6 (±16) 5 14.3 (±25.9) 105 25.8 (±44.8) 41 Absent 0 11.1 (±9.5) 23 11.4 (±13.0) 119 58.9 (±70.0) 38 Orientation Deep Vision 6.0 (±4.2) 7 0 5.0 (±3.3) 18 5.0 (±1.4) 8 Towing direction 23.8 (±15.1) 10 12.7 (±11.4) 27 13.6 (±20.9) 203 47.7 (±63.0) 68 Varied 6.3 (±4.8) 188 16.0 1 5.0 (±3.5) 3 5.0 (±1.7) 3 Position Centre of the trawl 6.8 (±4.8) 5 11.3 (±10.1) 20 8.9 (±9.8) 77 48.8 (±64.9) 64 Close to netting 7.1 (±6.8) 200 16.6 (±13.7) 8 14.8 (±23.5) 147 11.5 (±7.6) 15 Swimming direction Backward 7.1 (±4.2) 9 9.7 (±7.4) 24 10.3 (±16.1) 206 9.1 (±6.8) 34 Keeping pace 28.9 (±14.6) 7 31.0 (±0.0) 2 44.8 (±35.9) 16 72.2 (±71.1) 41 Irregular 6.3 (±4.9) 189 32.0 (±22.6) 2 14.0 (±7.1) 2 7.3 (±2.9) 4 Measured length NA 23 188 66 Overall behavior (orientation- swimming direction-position) TD-B-N 11.5 (±5.8) 6 12.6 (±20.7) 119 TD-B-C 9.1 (±7.9) 18 7.7 (±3.0) 67 9.2 (±7.4) 19 TD-K-C 73.3 (±71.7) 40 V-I-N 6.2 (±4.8) 186 Other 15.9 (±14.0) 19 31.5 (±13.1) 4 22.1 (±30.3) 38 9.6 (±7.4) 20 Herring Haddock Cod Saithe Residence time (s) Counts Residence time (s) Counts Residence time (s) Counts Residence time (s) Counts Section 1 6.0 (±3.1) 43 6.4 (±4.1) 17 8.4 (±4.8) 75 30.4 (±51.8) 38 Section 2 7.1 (±6.8) 205 12.8 (±11.2) 28 12.5 (±19.0) 224 41.7 (±60.2) 79 Sections 1 and 2 17.2 (±11.5) 43 20.5 (±12.5) 17 24.7 (±29.3) 75 57.9 (±63.0) 38 Section 2 categories: Density Multiple fish 6.8 (±6.8) 178 13.4 (±8.8) 11 15.1 (±24.5) 141 53.6 (±70.5) 52 Single fish 9.2 (±6.8) 27 12.4 (±12.8) 17 8.9 (±7.2) 83 18.9 (±18.0) 27 Steady light Present 7.1 (±6.8) 205 20.6 (±16) 5 14.3 (±25.9) 105 25.8 (±44.8) 41 Absent 0 11.1 (±9.5) 23 11.4 (±13.0) 119 58.9 (±70.0) 38 Orientation Deep Vision 6.0 (±4.2) 7 0 5.0 (±3.3) 18 5.0 (±1.4) 8 Towing direction 23.8 (±15.1) 10 12.7 (±11.4) 27 13.6 (±20.9) 203 47.7 (±63.0) 68 Varied 6.3 (±4.8) 188 16.0 1 5.0 (±3.5) 3 5.0 (±1.7) 3 Position Centre of the trawl 6.8 (±4.8) 5 11.3 (±10.1) 20 8.9 (±9.8) 77 48.8 (±64.9) 64 Close to netting 7.1 (±6.8) 200 16.6 (±13.7) 8 14.8 (±23.5) 147 11.5 (±7.6) 15 Swimming direction Backward 7.1 (±4.2) 9 9.7 (±7.4) 24 10.3 (±16.1) 206 9.1 (±6.8) 34 Keeping pace 28.9 (±14.6) 7 31.0 (±0.0) 2 44.8 (±35.9) 16 72.2 (±71.1) 41 Irregular 6.3 (±4.9) 189 32.0 (±22.6) 2 14.0 (±7.1) 2 7.3 (±2.9) 4 Measured length NA 23 188 66 Overall behavior (orientation- swimming direction-position) TD-B-N 11.5 (±5.8) 6 12.6 (±20.7) 119 TD-B-C 9.1 (±7.9) 18 7.7 (±3.0) 67 9.2 (±7.4) 19 TD-K-C 73.3 (±71.7) 40 V-I-N 6.2 (±4.8) 186 Other 15.9 (±14.0) 19 31.5 (±13.1) 4 22.1 (±30.3) 38 9.6 (±7.4) 20 SDs are in parentheses. TD is orientated in the “towing direction”, and V is a “varied” orientation. B is moving “backward”, K is “keeping pace” with the trawl, and I is swimming in an “irregular” direction (erratic up/down or side-to-side). C is positioned in the “centre of the trawl”, and N is positioned “close to netting”.Note: The residence times between each species significantly differ in Section 2 except for cod vs. haddock (Wilcoxon rank-sum tests, p < 0.05). Statistical analysis To test if the residence time was the same for the two sections, paired Wilcoxon signed-rank tests were used for herring (n = 43), cod (n = 75), and saithe (n = 38). We initially used a Kruskal-Wallis test to assess whether there were significant differences in Section 2 residence times between species. Wilcoxon rank-sum tests were used for comparisons between species. We applied the standard Bonferroni technique to adjust for multiple comparisons (p = 0.008; Bonferroni, 1936; Miller, 1981). Section 2 residence time was then modelled as a function of the fixed effects; overall behaviour, fish density, fish length and the presence of steady light, with haul as a random factor: Section 2 residence time∼overall behaviour+fish density+fish length+steady light+haul We used generalized linear mixed models (GLMM) with negative binomial distribution (log link function). Haul was included as a random factor to take the variance between hauls and pseudo replication into account (Millar and Anderson, 2004). The data on haddock were not modelled due the small sample size (n = 28 fish). The length of some individuals could not be measured by Deep Vision, which reduced the number of observations for the models to 66 for saithe and 188 for cod. Fish length was not measured for herring and herring were only observed in hauls with steady lights. Neither fish length nor steady light was included in the models for herring. We initially tested a GLMM with Poisson error; however, our models were over dispersed, which led us to apply a negative binomial distribution instead (Zuur et al., 2009). GLMMs were performed using the lme4 package (Bates et al., 2015) in R, vers. 3.3.1 (R Core Team, 2016). Fixed effects in the models were removed using backward step-wise elimination until a reduced model with only those variables that explained a significant amount of the variation remained (likelihood ratio test, p < 0.05; Crawley, 2007). As the trawl speed sensor at times lost contact with the vessel, simultaneous speed measurements were available for only 60% of the fish passages. In order to retain the maximum number of observations for each of the species modelled, the effect of towing speed (the trawl’s speed through water) on residence time was tested separately. Any significant fixed effects from the reduced models were included as fixed effects along with trawl speed. Results Water flow The ADCP recorded a marked decrease in the water flow near the netting compared with the centre of the trawl’s cross-section (Figure 3). Flow through the centre of the trawl was similar to the speed measured at the headline (vertical grey area and dotted line, Figure 3), while the flow within 25 cm of the netting was reduced by >0.5 ms−1 to about 65% of the trawl’s speed through water. Figure 3. View largeDownload slide Profile of water speed as measured by an ADCP mounted on the lower panel of the trawl. The boxes represent the 25th–75th percentiles, the solid black vertical line is the median and the whiskers illustrate 1.5× the interquartile range. The dotted vertical line is the median value of the trawl speed measured at the trawl opening and the grey area represents the 25th–75th percentiles. Figure 3. View largeDownload slide Profile of water speed as measured by an ADCP mounted on the lower panel of the trawl. The boxes represent the 25th–75th percentiles, the solid black vertical line is the median and the whiskers illustrate 1.5× the interquartile range. The dotted vertical line is the median value of the trawl speed measured at the trawl opening and the grey area represents the 25th–75th percentiles. Atlantic herring Herring were present in four hauls; all conducted with steady light illuminating the entrance of Deep Vision. It was the only species observed to enter the observational area in high numbers (∼120 fish per min). Individuals ranged from 22 to 39 cm in length, measured from the catch data. Herring quickly passed Section 1 (mean = 6.0 s, SD = 3.1, Table 3). A fifth of the individuals (n = 43) passing through Section 1 could be identified again in Section 2 closest to Deep Vision. The residence time in Section 2 averaged 7.1 s (SD = 6.8, Figure 4; not significantly different from Section 1: Wilcoxon signed-rank test, p = 0.90). Differences in residence times between species were highly significant (Kruskal-Wallis, p < 0.001) and herring had the shortest residence time of all the species observed (Wilcoxon rank-sum tests, p < 0.001). Residence time in Section 2 was not influenced by fish density or speed of the trawl through the water but depended on the overall behaviour (Table 4). Over 90% of the 205 herring observed displayed similar overall behaviour, moving through the trawl at different orientations (“varied” orientation) and displaying an irregular swimming pattern around the circumference of the trawl, (i.e. staying close to the netting and circling in a corkscrew path, either clockwise or anti-clockwise or changing direction during the observation (V-I-N), Figure 5). Residence time of individuals not displaying V-I-N behaviour (i.e. “other” in Table 3) spent significantly longer in front of Deep vision (V-I-N mean = 6.2 s and other mean = 15.9 s, Table 4, Figure 6a). Table 4. GLMM of residence time in Section 2 (in front of Deep Vision) for Atlantic herring (C. harengus) in the aft part of a pelagic trawl. Number Estimate SE z-value Pr (>z) Retained fixed effects  (Intercept) 205 1.939 0.140 13.829 <0.001  Overall behaviour: V-I-N vs. Other 0.623 0.135 4.602 <0.001 Removed fixed effects Order of removal Δ deviance Pr (>χ2)  Fish density 1 1.168 0.280 Retained variables  (Intercept) 81 2.019 0.114 17.780 <0.001  Overall behaviour: V-I-N vs. Other 0.647 0.204 3.168 0.002 Removed variables Order of removal Δ deviance Pr (>χ2)  Trawl speed 1 0.098 0.754 Number Estimate SE z-value Pr (>z) Retained fixed effects  (Intercept) 205 1.939 0.140 13.829 <0.001  Overall behaviour: V-I-N vs. Other 0.623 0.135 4.602 <0.001 Removed fixed effects Order of removal Δ deviance Pr (>χ2)  Fish density 1 1.168 0.280 Retained variables  (Intercept) 81 2.019 0.114 17.780 <0.001  Overall behaviour: V-I-N vs. Other 0.647 0.204 3.168 0.002 Removed variables Order of removal Δ deviance Pr (>χ2)  Trawl speed 1 0.098 0.754 All models used haul as a random factor and a negative binomial distribution with log link function. Retained fixed effects show the results of the reduced models. z-value is the Wald-Z test. Removed fixed effects show the order of the variables that were removed and the change in deviance from the previous model [likelihood ratio test (χ2), p < 0.05]. Table 4. GLMM of residence time in Section 2 (in front of Deep Vision) for Atlantic herring (C. harengus) in the aft part of a pelagic trawl. Number Estimate SE z-value Pr (>z) Retained fixed effects  (Intercept) 205 1.939 0.140 13.829 <0.001  Overall behaviour: V-I-N vs. Other 0.623 0.135 4.602 <0.001 Removed fixed effects Order of removal Δ deviance Pr (>χ2)  Fish density 1 1.168 0.280 Retained variables  (Intercept) 81 2.019 0.114 17.780 <0.001  Overall behaviour: V-I-N vs. Other 0.647 0.204 3.168 0.002 Removed variables Order of removal Δ deviance Pr (>χ2)  Trawl speed 1 0.098 0.754 Number Estimate SE z-value Pr (>z) Retained fixed effects  (Intercept) 205 1.939 0.140 13.829 <0.001  Overall behaviour: V-I-N vs. Other 0.623 0.135 4.602 <0.001 Removed fixed effects Order of removal Δ deviance Pr (>χ2)  Fish density 1 1.168 0.280 Retained variables  (Intercept) 81 2.019 0.114 17.780 <0.001  Overall behaviour: V-I-N vs. Other 0.647 0.204 3.168 0.002 Removed variables Order of removal Δ deviance Pr (>χ2)  Trawl speed 1 0.098 0.754 All models used haul as a random factor and a negative binomial distribution with log link function. Retained fixed effects show the results of the reduced models. z-value is the Wald-Z test. Removed fixed effects show the order of the variables that were removed and the change in deviance from the previous model [likelihood ratio test (χ2), p < 0.05]. Figure 4. View largeDownload slide Residence time of Atlantic herring (C. harengus), haddock (M. aeglefinus), Atlantic cod (G. morhua), and saithe (P. virens) in Section 2 (in front of Deep Vision). Width of shaded “violin” plots indicates number of individuals. The boxes represent the 25th–75th percentiles, the solid black vertical line is the median and the whiskers illustrate 1.5× the interquartile range. Figure 4. View largeDownload slide Residence time of Atlantic herring (C. harengus), haddock (M. aeglefinus), Atlantic cod (G. morhua), and saithe (P. virens) in Section 2 (in front of Deep Vision). Width of shaded “violin” plots indicates number of individuals. The boxes represent the 25th–75th percentiles, the solid black vertical line is the median and the whiskers illustrate 1.5× the interquartile range. Figure 5. View largeDownload slide Percentage of overall behaviours of Atlantic herring (C. harengus), haddock (M. aeglefinus), Atlantic cod (G. morhua) and saithe (P. virens) in Section 2 (in front of Deep Vision). TD is orientated in the “towing direction”, DV towards “Deep Vision”, and V is a ‘varied’ orientation. B is moving “backward”, K is “keeping pace” with the trawl, and I is swimming in an “irregular” direction (erratic up/down or side-to-side). C is positioned in the “centre of the trawl”, and N is positioned “close to netting”. The fish drawings illustrate the overall behaviour. Figure 5. View largeDownload slide Percentage of overall behaviours of Atlantic herring (C. harengus), haddock (M. aeglefinus), Atlantic cod (G. morhua) and saithe (P. virens) in Section 2 (in front of Deep Vision). TD is orientated in the “towing direction”, DV towards “Deep Vision”, and V is a ‘varied’ orientation. B is moving “backward”, K is “keeping pace” with the trawl, and I is swimming in an “irregular” direction (erratic up/down or side-to-side). C is positioned in the “centre of the trawl”, and N is positioned “close to netting”. The fish drawings illustrate the overall behaviour. Figure 6. View largeDownload slide Residence time of (a) Atlantic herring (C. harengus) and (b) Atlantic cod (G. morhua) in Section 2 (in front of Deep Vision) in relation to overall behaviour. TD is orientated in the “towing direction”, V is a varied orientation, B is moving “backward”, I is an irregular swimming style, C is positioned in the “centre of the trawl”, and N is positioned “close to netting”. Width of the shaded “violin” plots in indicates number of individuals. The boxes represent the 25th–75th percentiles, the solid black vertical line is the median and the whiskers illustrate 1.5× the interquartile range. The fish drawings illustrate the overall behaviour. Figure 6. View largeDownload slide Residence time of (a) Atlantic herring (C. harengus) and (b) Atlantic cod (G. morhua) in Section 2 (in front of Deep Vision) in relation to overall behaviour. TD is orientated in the “towing direction”, V is a varied orientation, B is moving “backward”, I is an irregular swimming style, C is positioned in the “centre of the trawl”, and N is positioned “close to netting”. Width of the shaded “violin” plots in indicates number of individuals. The boxes represent the 25th–75th percentiles, the solid black vertical line is the median and the whiskers illustrate 1.5× the interquartile range. The fish drawings illustrate the overall behaviour. Haddock Twenty-eight haddock were observed. Length measurements were recorded for 23 individuals (46–67 cm in length). Haddock quickly passed Section 1 (mean = 6.4 s, SD = 4.1, Table 3). More than half of the individuals (n = 17) passing through Section 1 could be followed into Section 2. The residence time in Section 2 (mean = 12.8 s, SD = 11.2; Figure 4) was longer than that of herring (Wilcoxon rank-sum test, p < 0.001) but was not significantly different from that of cod or saithe (Wilcoxon rank-sum tests, p = 0.9 and 0.014, respectively). Eighty-six percent of the haddock were orientated in the towing direction while moving backward towards Deep Vision, either close to the lower netting (25%) or in the centre of the trawl (75%; TD-B-N and TD-B-C, Figure 5). Atlantic cod Cod were the most abundant gadoid, with 224 individuals observed. Length measurements were recorded for 188 individuals (58–133 cm in length). Cod quickly passed through Section 1 (mean = 8.4 s, SD = 4.8, Table 3), and a third of the individuals (n = 75) passing through Section 1 could be identified again in Section 2. The residence time in Section 2 averaged 12.5 s (SD = 19.0; Figure 4), which was longer than the time in Section 1 (Wilcoxon signed-rank test, p = 0.004). Cod spent significantly less time in Section 2 than saithe (Wilcoxon rank-sum test, p < 0.001). This was longer than herring (Wilcoxon rank-sum test, p < 0.001) but was not significantly different from haddock (Wilcoxon rank-sum tests, p = 0.9). Residence time in Section 2 was not influenced by fish length, speed of the trawl through the water or by the steady light illuminating the entrance of Deep Vision but depended on the overall behaviour as well as by fish density (Table 5). The majority of the fish (84%) were orientated in the towing direction and moved backwards into Deep Vision, either close to the lower netting (64%) or in the centre of the trawl (36%; TD-B-N and TD-B-C, Figure 5). Cod that were orientated in the towing direction and moved backwards into Deep Vision spent more time in Section 2 when they were near the netting (within ∼20 cm) than in the centre of the trawl (12.6 and 7.7 s, respectively; TD-B-N vs. TD-B-C in Table 5; Figure 6b). Individuals also spent significantly more time in Section 2 when other fish were present (multiple fish mean = 15.1 s, single fish mean = 8.9 s; Table 5; Figure 7). Table 5. GLMM of residence time in Section 2 (in front of Deep Vision) for Atlantic cod (G. morhua) in the aft part of a pelagic trawl. Number Estimate SE z-value Pr (>z) Retained fixed effects  (Intercept) 188 2.274 0.182 12.494 <0.001  Overall behaviour: TD-B-C vs. TD-B-N 0.444 0.129 3.431 <0.001  Overall behaviour: TD-B-C vs. Other 0.828 0.160 5.173 <0.001  Fish density −0.473 0.110 −4.291 <0.001 Removed fixed effects Order of removal Δ deviance Pr (>χ2)  Fish length 1 0.037 0.848  Steady light 2 0.328 0.567 Retained variables  (Intercept) 164 2.194 0.204 10.750 <0.001  Overall behaviour: TD-B-C vs. TD-B-N 0.486 0.163 2.984 0.003  Overall behaviour: TD-B-C vs. Other 0.999 0.200 5.005 <0.001  Fish density −0.512 0.136 −3.770 <0.001 Removed variables Order of removal Δ deviance Pr (>χ2)  Trawl speed 1 0.308 0.579 Number Estimate SE z-value Pr (>z) Retained fixed effects  (Intercept) 188 2.274 0.182 12.494 <0.001  Overall behaviour: TD-B-C vs. TD-B-N 0.444 0.129 3.431 <0.001  Overall behaviour: TD-B-C vs. Other 0.828 0.160 5.173 <0.001  Fish density −0.473 0.110 −4.291 <0.001 Removed fixed effects Order of removal Δ deviance Pr (>χ2)  Fish length 1 0.037 0.848  Steady light 2 0.328 0.567 Retained variables  (Intercept) 164 2.194 0.204 10.750 <0.001  Overall behaviour: TD-B-C vs. TD-B-N 0.486 0.163 2.984 0.003  Overall behaviour: TD-B-C vs. Other 0.999 0.200 5.005 <0.001  Fish density −0.512 0.136 −3.770 <0.001 Removed variables Order of removal Δ deviance Pr (>χ2)  Trawl speed 1 0.308 0.579 All models used haul as a random factor and a negative binomial distribution with log link function. Retained fixed effects show the results of the reduced models. z-value is the Wald-Z test. Removed fixed effects show the order of the variables that were removed and the change in deviance from the previous model [likelihood ratio test (χ2), p < 0.05]. Table 5. GLMM of residence time in Section 2 (in front of Deep Vision) for Atlantic cod (G. morhua) in the aft part of a pelagic trawl. Number Estimate SE z-value Pr (>z) Retained fixed effects  (Intercept) 188 2.274 0.182 12.494 <0.001  Overall behaviour: TD-B-C vs. TD-B-N 0.444 0.129 3.431 <0.001  Overall behaviour: TD-B-C vs. Other 0.828 0.160 5.173 <0.001  Fish density −0.473 0.110 −4.291 <0.001 Removed fixed effects Order of removal Δ deviance Pr (>χ2)  Fish length 1 0.037 0.848  Steady light 2 0.328 0.567 Retained variables  (Intercept) 164 2.194 0.204 10.750 <0.001  Overall behaviour: TD-B-C vs. TD-B-N 0.486 0.163 2.984 0.003  Overall behaviour: TD-B-C vs. Other 0.999 0.200 5.005 <0.001  Fish density −0.512 0.136 −3.770 <0.001 Removed variables Order of removal Δ deviance Pr (>χ2)  Trawl speed 1 0.308 0.579 Number Estimate SE z-value Pr (>z) Retained fixed effects  (Intercept) 188 2.274 0.182 12.494 <0.001  Overall behaviour: TD-B-C vs. TD-B-N 0.444 0.129 3.431 <0.001  Overall behaviour: TD-B-C vs. Other 0.828 0.160 5.173 <0.001  Fish density −0.473 0.110 −4.291 <0.001 Removed fixed effects Order of removal Δ deviance Pr (>χ2)  Fish length 1 0.037 0.848  Steady light 2 0.328 0.567 Retained variables  (Intercept) 164 2.194 0.204 10.750 <0.001  Overall behaviour: TD-B-C vs. TD-B-N 0.486 0.163 2.984 0.003  Overall behaviour: TD-B-C vs. Other 0.999 0.200 5.005 <0.001  Fish density −0.512 0.136 −3.770 <0.001 Removed variables Order of removal Δ deviance Pr (>χ2)  Trawl speed 1 0.308 0.579 All models used haul as a random factor and a negative binomial distribution with log link function. Retained fixed effects show the results of the reduced models. z-value is the Wald-Z test. Removed fixed effects show the order of the variables that were removed and the change in deviance from the previous model [likelihood ratio test (χ2), p < 0.05]. Figure 7. View largeDownload slide Residence time in relation to fish length and density in Section 2 (in front of Deep Vision). The upper panel is for Atlantic cod (G. morhua) and the lower panel for saithe (P. virens). The crosses indicate individuals that were observed with other individuals (multiple fish) and the circles represent individuals that were observed alone (single fish) in front of Deep Vision. Figure 7. View largeDownload slide Residence time in relation to fish length and density in Section 2 (in front of Deep Vision). The upper panel is for Atlantic cod (G. morhua) and the lower panel for saithe (P. virens). The crosses indicate individuals that were observed with other individuals (multiple fish) and the circles represent individuals that were observed alone (single fish) in front of Deep Vision. Saithe Seventy-nine saithe were observed and length measurements were recorded for 66 individuals (34–120 cm). Half of the individuals (n = 38) passing through Section 1 could be followed into Section 2. Residence times were not significantly different between the two sections (Section 1: mean = 30.4 s, SD = 51.8; Section 2: mean = 41.7 s, SD = 60.2; Wilcoxon signed-rank test, p = 0.98; Table 3). Interestingly, it was not the same individuals that contributed to the long residence times in Sections 1 and 2: they either spent a long time in one section or the other but not in both (Figure 8a). Saithe had significantly longer (Wilcoxon rank-sum tests, p < 0.001) and more variable residences times than herring and cod but was not significantly different from haddock (Wilcoxon rank-sum tests, p = 0.014). Figure 8. View largeDownload slide Residence time of saithe (P. virens) in Section 2 (in front of Deep Vision) in relation to (a) residence time in Section 1 and (b) overall behaviour. TD is orientated in the “towing direction”, B is moving “backward”, K is “keeping pace” with the trawl, and C is positioned in the “centre of the trawl”. Width of shaded “violin” plots in B indicates number of individuals. The boxes represent the 25th–75th percentiles, the solid black vertical line is the median and the whiskers illustrate 1.5× the interquartile range. The fish drawings in (b) illustrate the overall behaviour. Figure 8. View largeDownload slide Residence time of saithe (P. virens) in Section 2 (in front of Deep Vision) in relation to (a) residence time in Section 1 and (b) overall behaviour. TD is orientated in the “towing direction”, B is moving “backward”, K is “keeping pace” with the trawl, and C is positioned in the “centre of the trawl”. Width of shaded “violin” plots in B indicates number of individuals. The boxes represent the 25th–75th percentiles, the solid black vertical line is the median and the whiskers illustrate 1.5× the interquartile range. The fish drawings in (b) illustrate the overall behaviour. The residence time in Section 2 was not significantly dependent on fish length, presence of a steady light or speed of the trawl through the water (Table 6). However, the residence time was influenced by the overall behaviour. The fish were primarily (75%) orientated in the towing direction, in the centre of the trawl. Sixty-five percent of these individuals kept pace with the trawl while 35% moved backward toward the codend (TD-K-C and TD-B-C, Figure 5). Twenty-five percent of saithe displayed different overall behaviours (“other” in Table 3) including staying close to the netting. However, residence time of the “other” group was not significantly different from the saithe orientated in the towing direction, moving backwards in the centre of the trawl (TD-B-C vs. Other in Table 6; Figure 8b). Residence time in Section 2 was also influenced by fish density, with saithe spending more time in Section 2 when in the presence of other fish (multiple fish mean = 53.6 s, single fish mean = 18.9 s; Table 6; Figure 7). Two small individuals (39 and 68 cm) remained in Section 2 for over 4 min. Another individual (49 cm) was observed to leave the codend and swim forward out of the field of view and was not seen again for the rest of the haul or during heaving, and it presumably escaped from the forward part of the trawl. Table 6. GLMM of residence time in Section 2 (in front of Deep Vision) for saithe (P. virens) in the aft part of a pelagic trawl. Number Estimate SE z-value Pr (>z) Retained fixed effects  (Intercept) 66 2.302 0.281 8.197 <0.001  Overall behaviour: TD-B-C vs. TD-K-C 1.802 0.302 5.955 <0.001  Overall behaviour: TD-B-C vs. Other 0.182 0.370 0.491 0.624  Fish density −0.278 0.113 −2.457 0.014 Removed fixed effects Order of removal Δ deviance Pr (>χ2)  Fish length 1 0.053 0.890  Steady light 2 0.072 0.788 Retained variables  (Intercept) 54 2.351 0.276 8.507 <0.001  Overall behaviour: TD-B-C vs. TD-K-C 1.306 0.277 4.712 <0.001  Overall behaviour: TD-B-C vs. Other −0.086 0.312 −0.275 0.783  Fish density −0.201 0.094 −2.150 0.032 Removed variables Order of removal Δ deviance Pr (>χ2)  Trawl speed 1 1.5 0.220 Number Estimate SE z-value Pr (>z) Retained fixed effects  (Intercept) 66 2.302 0.281 8.197 <0.001  Overall behaviour: TD-B-C vs. TD-K-C 1.802 0.302 5.955 <0.001  Overall behaviour: TD-B-C vs. Other 0.182 0.370 0.491 0.624  Fish density −0.278 0.113 −2.457 0.014 Removed fixed effects Order of removal Δ deviance Pr (>χ2)  Fish length 1 0.053 0.890  Steady light 2 0.072 0.788 Retained variables  (Intercept) 54 2.351 0.276 8.507 <0.001  Overall behaviour: TD-B-C vs. TD-K-C 1.306 0.277 4.712 <0.001  Overall behaviour: TD-B-C vs. Other −0.086 0.312 −0.275 0.783  Fish density −0.201 0.094 −2.150 0.032 Removed variables Order of removal Δ deviance Pr (>χ2)  Trawl speed 1 1.5 0.220 All models used haul as a random factor and a negative binomial distribution with log link function. Retained fixed effects show the results of the reduced models. z-value is the Wald-Z test. Removed fixed effects show the order of the variables that were removed and the change in deviance from the previous model [likelihood ratio test (χ2), p < 0.05]. Table 6. GLMM of residence time in Section 2 (in front of Deep Vision) for saithe (P. virens) in the aft part of a pelagic trawl. Number Estimate SE z-value Pr (>z) Retained fixed effects  (Intercept) 66 2.302 0.281 8.197 <0.001  Overall behaviour: TD-B-C vs. TD-K-C 1.802 0.302 5.955 <0.001  Overall behaviour: TD-B-C vs. Other 0.182 0.370 0.491 0.624  Fish density −0.278 0.113 −2.457 0.014 Removed fixed effects Order of removal Δ deviance Pr (>χ2)  Fish length 1 0.053 0.890  Steady light 2 0.072 0.788 Retained variables  (Intercept) 54 2.351 0.276 8.507 <0.001  Overall behaviour: TD-B-C vs. TD-K-C 1.306 0.277 4.712 <0.001  Overall behaviour: TD-B-C vs. Other −0.086 0.312 −0.275 0.783  Fish density −0.201 0.094 −2.150 0.032 Removed variables Order of removal Δ deviance Pr (>χ2)  Trawl speed 1 1.5 0.220 Number Estimate SE z-value Pr (>z) Retained fixed effects  (Intercept) 66 2.302 0.281 8.197 <0.001  Overall behaviour: TD-B-C vs. TD-K-C 1.802 0.302 5.955 <0.001  Overall behaviour: TD-B-C vs. Other 0.182 0.370 0.491 0.624  Fish density −0.278 0.113 −2.457 0.014 Removed fixed effects Order of removal Δ deviance Pr (>χ2)  Fish length 1 0.053 0.890  Steady light 2 0.072 0.788 Retained variables  (Intercept) 54 2.351 0.276 8.507 <0.001  Overall behaviour: TD-B-C vs. TD-K-C 1.306 0.277 4.712 <0.001  Overall behaviour: TD-B-C vs. Other −0.086 0.312 −0.275 0.783  Fish density −0.201 0.094 −2.150 0.032 Removed variables Order of removal Δ deviance Pr (>χ2)  Trawl speed 1 1.5 0.220 All models used haul as a random factor and a negative binomial distribution with log link function. Retained fixed effects show the results of the reduced models. z-value is the Wald-Z test. Removed fixed effects show the order of the variables that were removed and the change in deviance from the previous model [likelihood ratio test (χ2), p < 0.05]. Discussion This study showed that the residence time in front of the Deep Vision system (up to 12.5 m in front) differed between the studied species. Herring and haddock moved through the aft portion of the survey trawl without much delay (means = 7.1 and 12.8 s, respectively). Cod also mostly moved through without delay (mean = 12.5 s); though swimming close to the netting or the presence of other individuals increased the delay. In contrast, saithe spent up to 4 min in front of Deep Vision (mean = 41.7 s), and one individual even escaped the trawl by swimming forward. Saithe also spent longer in the aft of the trawl when in the presence of other individuals. With the current standard survey towing speed of 1.5 ms−1, the Deep Vision system could be used to infer the spatial pre-capture distribution of herring, haddock and cod, in order to improve the interpretation of acoustic data. The results for saithe indicate that using Deep Vision’s data to back-calculate their position prior to capture in the trawl may result in significant spatial and temporal errors. For example, a fish swimming at the trawling speed of 1.5 ms−1 for 4 min will be displaced by 360 m along the trawling track. No observations of the studied species were made farther forward in the trawl, and delays might also occur here. However, a previous study of cod in a pelagic trawl concluded that fish passed through the forward portion of the trawl at nearly the trawl’s speed through water and that the delay increases as fish move towards the aft of the trawl (Rosen et al., 2012). The steady artificial lights used on some hauls to illuminate Section 2 (the entrance to Deep Vision) did not significantly influence the residence time of cod or saithe inside the trawl. Herring were only captured in night hauls that used steady artificial lights, and only five haddock were observed during hauls with steady artificial lights. It is therefore not possible within this study to draw inferences as to whether the addition of steady lights affected the residence time of these species. The limited time available did not allow us to determine whether the strobe lights that are essential to the operation of the Deep Vision system affect the performance of the trawl. Future studies should take place on fish behaviour at the front of the Deep Vision system with the strobe lights turned off. The behaviour of cod and saithe Over 80% of cod were orientated in the towing direction moving backwards, differing only by the position in the trawl’s cross-section (64% close to the trawl netting, 36% in the centre). The position of cod influenced residence time in the survey trawl; however, the trawling speed had no clear effect at the range of speeds tested. At higher trawling speeds, cod would be expected to have had shorter residence times (pass backwards more quickly) as swimming endurance generally decreases at higher swimming speeds (Beamish, 1966; He, 1991; Winger et al., 2000). At 5–8 °C, the temperatures encountered in our study, Winger et al. (2000) found that cod swimming at 1.3 ms−1 reached exhaustion after an average of 1.3 min (predicted maximum ∼5.5 min). We trawled at greater speed (1.5 ms−1). At this velocity, the maximum endurance is expected to be 2.8 min (Winger et al., 2000) which is consistent with our findings of cod spending a maximum of 3.1 min before moving to the codend. Cod staying in the lower water flow near the netting could have masked the expected effect of trawling speed on the residence time. Cod of all sizes appeared to conserve energy by remaining close to the netting where the water flow was approximately 0.5 ms−1 slower than at the centre of the trawl. Here, cod were on average able to spend 1.6 times longer in front of the Deep Vision (12.6 s close to the netting vs. 7.7 s in the centre of the trawl). A higher towing speed than used in this study may increase the water flow near the netting, and move cod quickly back into the codend. Saithe are strong swimmers, able to swim nearly twice as fast as cod for the same duration (Beamish, 1966; He and Wardle, 1988), and were able to maintain pace with the trawl for up to 4 min in front of Deep Vision. Unlike cod, which had the longest residence times when staying close to the netting, saithe had the longest residence times (mean = 48.8 s) when keeping pace in the centre of the trawl where the water flow was highest. Swimming in the centre of the trawl away from the netting may allow saithe to swim more freely without hitting the netting. Similar to cod, the differences in trawl speed did not affect the residence time of saithe. Most saithe moved back within a minute, while 28% spent longer time in Sections 1 or 2. Potential reasons for saithe to have stayed in front of Deep Vision include repulsion from the strobe lights from Deep Vision and/or the narrow passage in front of the cameras. Saithe did not move more slowly back because they were more likely to be oriented towards the trawl opening (in fact cod and haddock were more likely to have this orientation), rather we believe their greater swimming capacity gives them the physical capacity to stay in front of Deep Vision when the trawl is towed at 1.5 ms−1, while the other species observed do not. Even the smallest saithe (34 cm) was able to maintain pace with the trawl. Information from the trawl monitoring sensors suggests in fact that far more fish entered the trawl than were retained in the codend or recorded on the observation cameras or Deep Vision’s cameras. Recordings of fish entrances believed to represent saithe by a Trawleye sensor mounted on the headrope estimated more than twice the number of fish than were caught in the (small-meshed) codend. This suggests that a significant proportion of saithe that entered the trawl escaped (the mesh size at the front of the trawl was up to 3.2 m), possibly when the speed was reduced during heaving. Single fish would move towards Deep Vision at a relatively quick rate and enter the codend, but when accompanied by other fish, both cod and saithe moved more slowly toward Deep Vision. Though little information is available on the influence of fish density with pelagic trawls, cod and haddock in the mouth of a demersal trawl displayed higher turnover rates when only one or two individuals were present than when in larger aggregations (Aglen et al., 1997; Godø et al., 1999). The presence of another fish may mutually influence both fish, with each individual trying to match the other’s speed. In addition, swimming in the wake behind another fish may provide hydrodynamic benefits that increase swimming endurance (Weihs, 1973; Polverino et al., 2013; Hemelrijk et al., 2015). There were wide variations in fish length, with the largest cod and saithe being two to three times longer than the smallest individuals. Still, residence time was not related to fish length. This is surprising, as the swimming speed of fish is usually highly correlated with length (Beamish, 1978; Blake, 1983; Videler, 1993). The collective behaviour of fish in groups may also explain this observation. The fish probably fall back in the trawl when they reach a certain threshold of fatigue before they are completely exhausted. This may be influenced by the presence of other fish, with smaller fish raising their fatigue threshold when in the presence of larger individuals in order to avoid losing contact with the group. At the same time, the threshold of larger fish may be lowered in order to match the speed of the smaller individuals. The end result could be that all fish keep together and thereby exhibit similar residence times despite differences in absolute swimming capacity. Survey implications Trawl-mounted cameras such as Deep Vision have the potential to improve the spatial resolution of trawl sampling, but a better understanding of how fish behaviour in the trawl affects interpretation of the results and whether the cameras, lights and physical structure around the cameras affect the behaviour and catchability is needed prior to implementation. For example, Deep Vision can be used to record the presence, abundance and size distribution of small individuals or species that move rapidly through the trawl (Rosen and Holst, 2013; Rosen et al., 2013; Underwood et al., 2014). Under the conditions in the present study, the presence of the Deep Vision system did not introduce significant delays in the passage of haddock or herring in the aft part of the trawl, nor did it affect the trawl’s ability to retain these species. Most cod were only slightly delayed though there were some exceptions (<5% of the cod were delayed by more than a minute). Saithe presented a more complicated case, with a longer and more variable delay in the aft portion of the trawl, suggesting that the standard trawling speed of 1.5 ms−1 is insufficient to make all individuals move quickly through the trawl or even to retain all the individuals that enter the trawl. Further studies using cameras or acoustic sensors farther forward in the trawl are needed to evaluate the potential effects of camera systems in the aft portion of the trawl and to determine whether higher towing speeds are required to effectively capture and retain saithe. Misinterpretation of the spatial component of Deep Vision’s data could lead to incorrect categorisation of acoustic backscatter used to calculate species abundance and size distribution. Herring and haddock had similar, short residence times in front of Deep Vision, and a correction for the short delay may therefore be computed for each species when verifying acoustic targets. As the variations in the residence time of saithe was high at the towing speed of 1.5 ms−1, and cod and saithe residence times were dependent on density, comparing targets in the echosounder image to Deep Vision’s data becomes difficult. The displacement of one species, but not another, can alter the estimated species composition of a shoal observed acoustically. Trawling towards large shoals of fish seen in the echosounder to verify the targets would create large densities in front of Deep Vision, and on the basis of our findings, would increase the residence time of individuals and increase the spatial and temporal displacement when they are imaged. In our study, saithe swimming in the presence of other individuals had a residence time of up to over 4 min. Still, such challenges for verifying acoustic targets through trawl sampling are not unique to Deep Vision. Delays in movement of species through the trawl could also affect the interpretation of the data from multiple codend systems that take discrete samples during a single trawl haul. If fish do not pass quickly through the trawl or swim for long periods of time in front of the trawl mouth, they may be retained in a codend that does not correspond to the depth or location where they were initially present. Although tools to collect data at finer spatial and temporal resolution have obvious potential benefits for verifying acoustic targets, it is still of great importance to consider and evaluate potential sampling biases they may introduce. Acknowledgements We thank the crew of the RV “G.O. Sars” for their assistance and hospitality at sea, as well as Egil Ona, Jan Tore Øvredal, Hildegunn Mjanger and Anne Sæverud (IMR), and Thor Bærhaugen (SIMRAD Kongsberg Maritime AS) for their technical assistance. Lastly, we thank the anonymous reviewers for their comments on earlier versions. The research was funded by the Centre for Research-based Innovation in Sustainable fish capture and Processing technology, the Research Council of Norway (Project number 203477). References Aglen A. , Engås A. , Godø O. R. , Mccallum B. R. , Stansbury D. , Walsh S. J. 1997 . Density dependent catchability in bottom trawl surveys. ICES CM 1997/W: 16; 1997. Bates D. , Maechler M. , Bolker B. , Walker S. 2015 . Fitting linear mixed-effects models using lme4 . Journal of Statistial Software , 67 : 1 – 48 . Beamish F. W. H. 1966 . Swimming endurance of some northwest Atlantic fishes . Journal of the Fisheries Research Board of Canada , 23 : 341 – 347 . Google Scholar CrossRef Search ADS Beamish F. W. H. 1978 . Swimming capacity. In Fish physiology: locomotion , pp. 101–187. Ed. by Hoar W. S. , Randall D. J. . Academic Press , Cambridge, MA, USA . 576 pp. Blake R. W. 1983 . Fish locomotion , Cambridge University Press , Cambridge . 208 pp. Bonferroni C. E. 1936 . Teoria statistica delle classi e calcolo delle probabilità. Pubblicazioni del R Istituto Superiore di Scienze Economiche e Commerciali di Firenze , 8 : 3 – 62 . Crawley M. J. 2007 . The R Book . John Wiley & Sons Ltd , Chichester, UK . 942 pp. Google Scholar CrossRef Search ADS Engås A. 1994 . The effects of trawl performance and fish behaviour on the catching efficiency of demersal sampling trawls. In Marine fish behaviour in capture and abundance estimation , pp. 45–68. Ed. by Fernö A. , Olsen S. . Fishing News Books , Oxford . 221 pp. Engås A. , Skeide R. , West C. W. 1997 . The ‘MultiSampler’: a system for remotely opening and closing multiple codends on a sampling trawl . Fisheries Research , 29 : 295 – 298 . Google Scholar CrossRef Search ADS Godø O. R. 1994 . Factors affecting the reliability of groundfish abundance estimates from bottom trawl surveys. In Marine fish behaviour in capture and abundance estimation , pp. 166–199. Ed. by Fernö A. , Olsen S. . Fishing News Books , Oxford . 221 pp. Godø O. R. , Walsh S. J. , Engås A. 1999 . Investigating density-dependent catchability in bottom-trawl surveys . ICES Journal of Marine Science , 56 : 292 – 298 . Google Scholar CrossRef Search ADS Gunderson D. R. 1993 . Surveys of fisheries resources , John Wiley & Sons . 257 pp. He P. 1991 . Swimming endurance of the Atlantic cod, Gadus morhua L., at low temperatures . Fisheries Research , 12 : 65 – 73 . Google Scholar CrossRef Search ADS He P. , Wardle C. 1988 . Endurance at intermediate swimming speeds of Atlantic mackerel, Scomber scombrus L., herring, Clupea harengus L., and saithe, Pollachius virens L . Journal of Fish Biology , 33 : 255 – 266 . Google Scholar CrossRef Search ADS Hemelrijk C. K. , Reid D. A. P. , Hildenbrandt H. , Padding J. T. 2015 . The increased efficiency of fish swimming in a school . Fish and Fisheries , 16 : 511 – 521 . Google Scholar CrossRef Search ADS Mehl S. , Aglen A. , Berg E. , Dingsør G. E. , Korsbrekke K. 2015 . Acoustic abundance of saithe, coastal cod and haddock Finnmark – Møre, Autumn 2015. Cruise report Nr. 4 - 2015, Institute of Marine Research, Bergen. 38 pp. (in Norwegian). Millar R. B. , Anderson M. J. 2004 . Remedies for pseudoreplication . Fisheries Research , 70 : 397 – 407 . Google Scholar CrossRef Search ADS Miller R. G. 1981 . Simultaneous Statistical Inference . Springer , New York . 299 pp. Google Scholar CrossRef Search ADS Mjanger H. , Hestenes K. , Svendsen B. , De Lange Wenneck T. 2014 . Handbook for sampling fish and crustaceans. Version 3.16. Bergen. 16 pp. (in Norwegian). Polverino G. , Phamduy P. , Porfiri M. 2013 . Fish and robots swimming together in a water tunnel: robot color and tail-beat frequency influence fish behavior . PloS ONE , 8 : e77589. Google Scholar CrossRef Search ADS PubMed R Core Team 2016 . R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. Rosen S. , Engås A. , Fernö A. , Jörgensen T. 2012 . The reactions of shoaling adult cod to a pelagic trawl: implications for commercial trawling . ICES Journal of Marine Science , 69 : 303 – 312 . Google Scholar CrossRef Search ADS Rosen S. , Holst J. C. 2013 . DeepVision in-trawl imaging: Sampling the water column in four dimensions . Fisheries Research , 148 : 64 – 73 . Google Scholar CrossRef Search ADS Rosen S. , Jörgensen T. , Hammersland-White D. , Holst J. C. , Grant J. 2013 . DeepVision: a stereo camera system provides highly accurate counts and lengths of fish passing inside a trawl . Canadian Journal of Fisheries and Aquatic Sciences , 70 : 1456 – 1467 . Google Scholar CrossRef Search ADS Simmonds J. , Maclennan D. N. 2005 . Fisheries Acoustics: Theory and Practice , 2nd edn . Blackwell Science Ltd , Oxford . 456 pp. Google Scholar CrossRef Search ADS Underwood M. J. , Rosen S. , Engås A. , Eriksen E. 2014 . Deep vision: An in-trawl stereo camera makes a step forward in monitoring the pelagic community . PLoS ONE , 9 : e112304. Google Scholar CrossRef Search ADS PubMed Valdemarsen J. W. , Misund O. A. 1994 . Trawl designs and techniques used by Norwegian research vessels to sample fish in the pelagic zone. Proceedings of the Sixth PRlW-PEW0 Symposium, Bergen, 14–17 June 1994. Videler J. J. 1993 . Fish Swimming , Chapman & Hall , London . 260 pp. Google Scholar CrossRef Search ADS Videler J. J. , He P. 2010 . Swimming in marine fish. In Behavior of Marine Fishes: Capture Processes and Conservation Challenges , pp. 3–24. Ed. by He P. . Wiley-Blackwell , Oxford . 392 pp. Wardle C. S. 1993 . Fish behaviour and fishing gear. In Behaviour of Teleost Fishes , 2nd edn ., pp. 609–643. Ed. by Pitcher J. T. . Chapman and Hall , London . 716 pp. Weihs D. 1973 . Hydromechanics of Fish Schooling . Nature 241 : 290 – 291 . Google Scholar CrossRef Search ADS Winger P. D. , Eayrs S. , Glass C. W. 2010 . Fish behaviour near bottom trawls. In Behavior of Marine Fishes: Capture Processes and Conservation Challenges , pp. 67–103. Ed. by He P. . Wiley-Blackwell , Oxford . 392 pp. Winger P. D. , He P. , Walsh S. J. 2000 . Factors affecting the swimming endurance and catchability of Atlantic cod (Gadus morhua) . Canadian Journal of Fisheries and Aquatic Sciences , 57 : 1200 – 1207 . Google Scholar CrossRef Search ADS Zuur A. , Ieno E. , Walker N. , Saveliev A. , Smith G. 2009 . Mixed Effects Models and Extensions in Ecology with R . Springer , New York, NY . 574 pp. Google Scholar CrossRef Search ADS © International Council for the Exploration of the Sea 2018. All rights reserved. For Permissions, please email: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)
TI - Species-specific residence times in the aft part of a pelagic survey trawl: implications for inference of pre-capture spatial distribution using the Deep Vision system
JF - ICES Journal of Marine Science
DO - 10.1093/icesjms/fsx233
DA - 2018-01-29
UR - https://www.deepdyve.com/lp/oxford-university-press/species-specific-residence-times-in-the-aft-part-of-a-pelagic-survey-KVdJs0p68f
SP - 1
EP - 1404
VL - Advance Article
IS - 4
DP - DeepDyve
ER -