Slow sink rate in floated-demersal longline and implications for seabird bycatch risk.

Bycatch of birds in longline fisheries is a global conservation issue, with between 160,000-320,000 seabirds killed each year, primarily through being caught and drowned as they attempt to snatch baits off hooks as they are set. This conservation issue has received significant recognition in southern hemisphere longline fisheries over the past several decades, largely due to the impact on highly charismatic and highly threatened birds, notably Albatrosses. As a result, the use of effective mitigation measures has been subject to fisheries regulations to reduce seabird bycatch from longliners in a number of national jurisdictions and in several Regional Fisheries Management Organisations (RMFOs). While mitigation measures have been mandated in a number of north Pacific longline fisheries, this is largely not the case in north Atlantic longline fisheries. This includes vessels using floated-demersal longlines in the North-East Atlantic longline fishery targeting European Hake Merluccius merluccius, in which high levels of seabird bycatch are estimated. In this paper, we analysed the sinking speed of a floated-demersal longline used to target European Hake in the offshore waters of Scotland, to determine potential bycatch risks to seabirds. We deployed Time Depth Recorder devices at different points of the gear. We assessed how this gear performed in comparison to the best practice minimum sink rate of 0.3 m/s recommended by the Agreement on the Conservation of Albatrosses and Petrels (ACAP) to limit bird access to baited hooks. We found that the average sinking speed of the floated-demersal longline was substantially slower than the ACAP recommendation, between two and nine times slower in non-weighted parts of the gear down to 10m water depth. Our work also found that the sink rate is particularly slow in the top 2m of the water column, increasing with depth and stabilizing at depths over 10m, presumably a consequence of propeller wash behind the vessel. We calculated that the distance astern of the vessel for hooks to sink beyond susceptible seabirds' reach largely exceeds optimum coverage of best practice design Bird Scaring Lines (100 m). Our results indicate that hooks from floated-demersal longlines are therefore readily open to seabird attacks, and as a result, present a clear bycatch risk. Research is needed to adapt existing mitigation measures to floated-longlines and to develop novel mitigation approaches to improve the sink rate of the gear without impacting target fish catch.

Introduction

Seabirds are one of the most threatened groups of birds [1] with bycatch in fisheries recognized as one of the major threats to their conservation [2]. Worldwide, longline fisheries are estimated to cause mortality of 160,000–320,000 seabirds annually, with birds primarily caught and drowned as they attempt to steal baited hooks as the line is set [3]. This global estimate includes 56,000 from the North-Eastern Atlantic Spanish demersal longline fishery alone, primarily Great Shearwater Ardenna gravis, though this historical figure is based on limited data and has since been reviewed down (Pep Arcos 2020, personal communication). Primarily focussed on an area of the Celtic Seas to the southwest of Ireland known as the ‘Gran Sol’, this fleet targets European Hake Merluccius merluccius and other demersal fish such as Common Ling Molva molva and Atlantic Cod Gadus morhua. Alongside Spanish-flagged vessels, this fishery also includes–for the most part–vessels flagged to France and the United Kingdom, together representing over 85% of all European Hake landings in 2018 for ICES area 27 [4]. Recent publication estimates that up to 10,000 seabirds, including 2,600–9,000 Northern Fulmars Fulmarus glacialis [5], could be bycaught each year by the British flagged segment of the fleet, which represents only about 13% of the fleet’s Hake landings [4]. The total magnitude of seabird bycatch across this multi-national fleet is unknown but is likely to result in the mortality of tens of thousands of seabirds each year [3].

Several simple and effective seabird bycatch mitigation measures for longline fisheries already exist [6] but most have been developed and tested in the southern hemisphere longline fisheries [7-10]. To date, limited work has been conducted to test, adapt or implement these measures for the unique set of challenges posed by ‘floated’ demersal longlines (also referred to as ‘Spanish longline’ [11], ‘piedra bola’ [12] or ‘semi-pelagic longline’ [13]), which are predominantly used in the North-East Atlantic to target Hake and other whitefish fisheries. This fishery operates mainly in the ecoregion known as the Celtic Seas (ranging from the north of the Shetland Islands to Brittany in the south [14]) and towards the Bay of Biscay, including the area referred to as the ‘Gran Sol’ [3,15]. ‘Floated’ demersal longlines are lofted off the seabed through the addition of subsurface floats on the main line (Fig 1). Due to the added buoyancy, this gear presents an elevated risk to seabirds because the floats counteract the weights attached to the line, resulting in hooks sinking slowly and making baits available to seabirds’ attacks for longer periods of time. One study reported birds attacking baited hooks on floated longlines ten times more compared to longlines without floats [16].

Fig 1

Schematic of floated-demersal longline.

This ‘typical’ configuration is used to target European Hake in the North-East Atlantic fishery, composed by a sequence of floats, weights and dropper lines (connecting the main line–with hooks–to the trace line). A “unit” (or box) corresponds to 200m of line and 80 hooks, and a longline is comprised of up to 130 units in this fishery. Noting variation in some of the specific elements depending on fishing conditions and skipper preference.

In terms of regulation, use of bird-scaring lines have been mandated for Alaskan longline fleets since 2004, which helped reduce bycatch of surface-foraging seabirds by 88–100% in the Sablefish and Pacific cod fisheries [17,18]. In Europe, the European Union (EU) published in 2012 an Action Plan for reducing bycatch of seabirds in fishing gears, notably requiring that member states implement at least two proven mitigation measures in longline fisheries–specifically mentioning “the Gran Sol […] and non-EU waters”–such as Night setting, Bird-scaring lines or Line weighting in accordance with minimum technical standards as set out in ACAP guidelines [19]. In spite of this action plan, Mitchell et al. [20] found that those aforementioned mitigation measures are still not required to be implemented in most EU longline fisheries and “no action has been taken to address seabird bycatch in the Gran Sol fishery”. The 2019 EU Technical Measures Regulation should have provided an additional impetus for efforts to address bycatch, yet actions are still severely lacking [20]. Since the withdrawal of the UK from the EU in 2020, no concrete actions have been mandated or encouraged to tackle seabird bycatch in UK waters, by national or foreign fleets. Although a UK Seabird Bycatch Plan of Action has been in development over the past couple of years, and collaborative research platforms recently put in place (Clean Catch UK) [21], it remains the case that there is no clear pathway or timeline for implementation of best practice mitigation measures in the longline fleets operating in UK waters (personal communication with Defra representatives, 2021). Some vessels in the UK fleet are voluntarily using night setting and bird-scaring lines to reduce seabird bycatch [5], though the efficacy of these lines is currently unknown when paired with floated demersal longlines.

In this study, we used time-depth recorders (TDRs) to better understand the sink rate of the typical gear used by the north east Atlantic floated demersal longline fleet to inform relevant actions to reduce seabird bycatch. The Agreement on the Conservation of Albatrosses and Petrels (ACAP) defines the best practice longline sink rate as at least 0.3 meters per second to avoid seabird captures [22,23]. This can be achieved through a range of weighting options, but 5 to 8 kg weights spaced every 40 m is a configuration that is mandated for fisheries operating in the Convention on the Conservation of Antarctic Marine Living Resources (CCAMLR) area [24]. Weights that were used in the fishery in the present study can vary but, generally speaking, are substantially lighter and spaced at wider intervals; 3kg weights spaced every 100m appears as a common configuration in this fishery (Fig 1) (Juan Pablo Forti Buratti, fishery observer, personal observations 2020). With the extra buoyancy created by the regular deployment of floats along the line, it is expected that the sink rate would be substantially slower compared to best practice recommendations.

Bird behaviour and diving proficiency determines a “safe depth”, below which seabird bycatch risk is greatly reduced [25]. Northern Fulmars, being surface feeders with limited diving capacity [26], are most at risk of bycatch in the upper 2 metres of the water column [27]. Other species such as Great Shearwaters can dive down to nearly 20m, although about half of dives occur within the first two meters of water, with nearly all remaining dives within 2–10 m [28]. In this fishery, most bycatch is therefore likely occurring within the top 10m of the water column, and is therefore the key area of interest when examining the sink rate of the gear.

Materials and methods

Ethics statement

This research did not involve animal or human subjects, and therefore did not require ethical approval. The research was conducted in accordance with the Centre for Environment, Fisheries and Aquaculture Science (Cefas) Seafood Innovation Fund (SIF) guidelines for the project FS031, as well as the Royal Society for the Protection of Birds (RSPB) Ethics Advisory Committee guidelines

Field work methodology

To measure the sink rate of a floated-demersal longline, we collaborated with a 40 m long commercial longliner targeting European Hake in the North-East Atlantic waters. Data collection took place between February 24th and March 5th 2020, at Latitudes 60.1442/60.2876 and Longitudes -4.4957/-4.0434 (Fig 2). All data were collected by a single observer and onboard the same vessel. We deployed a series of G5 Long-life 8MB CEFAS Time Depth Recorders (TDRs), at four different positions across the fishing gear: Dropper, Float, Middle and Weight (Fig 3). Configuration of the fishing gear, such as the spacing of weights and floats, is subject to change depending on fishing conditions and the skipper’s experience, and as such there is no ‘standard’ configuration of a floated-demersal longline in this fleet. However, these trials were conducted on the most common configuration of the gear according to the fishery observer, and echoes the configuration recorded in other semi-pelagic Hake fisheries across the Atlantic [29-31].

Fig 2

Location of data collection (red square), which took place on a longliner targeting European Hake in Scottish waters, between February and March 2020.

Each black cross indicates the start of a longline deployment with TDRs attached.

Fig 3

Time Depth Recorders’ deployment on the fishing gear.

The red cylinders indicate where the Time Depth Recorders were deployed on the floated demersal longline.

Fishers were asked to deploy gear as they would in a normal fishing operation. Eight to ten TDRs were deployed on each experimental set, depending on fishing conditions. After each set was hauled, TDR data were downloaded using the DST Host software from CEFAS (https://www.cefastechnology.co.uk/downloads), then recalibrated for the next fishing operation. Each TDR was calibrated with the Dive Logging option, using wet/dry sensors (activation by entry into the water) and 2Hz log rate (recording every 0.5 sec), as well as 12-bit Data Points defined resolution for the Pressure sensor. TDRs were calibrated with 50 bar sensors, providing a pressure resolution better than 15cm with the above setting.

We analysed the data according to the “bycatch risk depths” for Northern Fulmar and Great Shearwater, which have constituted over 90% of all bycaught birds in published studies of this fishery [3,5], though it should be noted that overall this is a very data-poor fishery with regard to seabird bycatch. Other seabird species are being bycaught, though these are also primarily surface-feeding and pursuit-plunging species yet capable of deep diving [27,28]. We therefore split the water column into different depth sections to better understand the sinking dynamics of the gear in relation to the most affected seabird’s feeding strategies [32]. Those sections are 0 to 2m (to represent surface-feeders) and 0 to 10m (for pursuit-plunging divers) [33]. We also decided to include a 0 to 5m section to provide a more precise understanding of the bycatch risk across the water column. Analysis at 2 to 5m and 5 to 10m, will also help understand the sinking dynamic of the line at different depth sections and inform potential re-design of the gear.

Statistical analysis

Sink rate and distance from vessel

Sink rates were calculated as the time taken in seconds for TDRs set at different positions on the longline to traverse a specific water depth range (0–2 metres, 0–5 metres, 0–10 metres, 2–5 metres and 5–10 metres); separate models were run for each water depth range. The time taken to traverse each specified water depth range (sink time) was log-transformed prior to analysis and modelled using a Bayesian random effects models in the R Environment (R Version 4.0.4 [34]) using the MCMCglmm package [35]. Sink time was log-transformed to prevent models predicting negative values which are unrealistic in this context. When reporting the results from Bayesian analysis we report parameter estimates (posterior means) and corresponding equal-tail 95% credibility interval (CRI) values.

As predictor variables in our initial sink rate models, we included the position of the TDR on the line (Dropper, Float, Middle, Weight) as well as the wind speed recorded during setting (wind measured on the Beaufort scale), setting speed, water depth and the log number of hooks on the line. We chose to model the number of hooks on the log scale as the raw number of hooks per net was typically very high with some discontinuity between values (range: 4000–10400) and we felt results would therefore be more interpretable on the log-scale. For ease of interpretation, setting speed and water depth were standardized (mean centred and divided by their standard deviation) prior to modelling and wind speed was centred to the modal wind speed recorded [36]. Because we were specifically interested in calculating sink rates at different positions of the gear, we did not consider the position term for removal when performing model selection. However, we decided to retain or remove the other predictors examined using leave-one-out cross-validation (LOO CV) and assessing whether their addition improved predictive performance versus a model in which only TDR position was included as a predictor. TDRs set on the same line on the same day are unlikely to be completely independent, therefore we included a random intercept for Day in our models. As one longline was set per day during the study (Table 1) this ensures measurements from all TDRs from the same deployment are grouped together to avoid potential pseudo-replication issues. Initial visual inspection of raw time to depth data suggested that the variation in sink time was greater at some positions of the gear than others. Therefore, we allowed the residual variance in our models to vary in relation to the position of the TDR to avoid potential problems with heteroscedasticity [37,38]. As before, LOO CV was used to examine whether a model in which the residual variance for each TDR position was modelled separately performed better than one in which a single, pooled estimate for the residual variation was calculated. Because time to depth was modelled as log-transformed it was back-transformed to the original scale using standard formulae [39,40]. Here we report marginal estimates from our models by performing back-transformation over the random effects distribution in our models. Marginal estimates of time to depth were also converted to estimates of sink rates (m / s). Because we are using Bayesian models sink rates can be calculated across model posterior distributions allowing easy calculation of 95% credible intervals (95% CRI).

Table 1

Summary of TDRs deployment with gear configuration and fishing conditions.

Day	Date	Start Fishing Operation	Shooting speed [knots]	Water depth [m]	Beaufort	# hooks set	# working TDRs recording
1	24/02/2020	03:00:00	7.6	277	2	9600	9*
2	25/02/2020	03:15:00	7.7	271	2	9600	8
3	26/02/2020	04:40:00	7.8	250	3	2000	7**
4	27/02/2020	04:00:00	9.1	255	4	2800	8
5	28/02/2020	04:40:00	8.2	262	3	5600	7**
6	29/02/2020	03:00:00	8.5	265	7	10400	8
7	01/03/2020	04:30:00	8.4	275	6	4000	7**
8	02/03/2020	04:20:00	8.4	270	4	4000	8
9	03/03/2020	04:10:00	8.5	270	2	4000	10
10	04/03/2020	03:20:00	6.5	260	2	10400	9*
11	05/03/2020	03:10:00	6.4	270	2	10000	10

*a TDR was lost during the deployment.

**a TDR failed at recording.

A similar approach was used to model the expected distance from the stern that different positions within the longline would attain after sinking to a specific water depth (2m, 5m, 10m). Distance from the stern in metres was estimated from raw TDR data by multiplying the number of seconds it took each TDR to reach a specified water depth by the shoot speed (m / s). Distance from stern was modelled as a log-transformed variable using the same fixed and random effects structures and modelling procedure as described above.

For all the fixed effect predictors in our models we used normal priors with a mean of 0 and a variance of 100. Priors for variance components were set using an inverse-Wishart distribution. For a single variance component, the inverse-Wishart distribution was dictated by two parameters, V and v, in MCMCglmm notation following Hadfield [35]. Hadfield refers to ν as the ‘degree of belief parameter’ and smaller values of v denote weaker belief in prior values for the unknown variance parameter V. We set v = 0.002 to create a diffuse prior for unknown variance components and ran 3 MCMC chains that began at dispersed values for V (0.05, 0.5, and 1.0 respectively, with v fixed at 0.002). Convergence of chains was examined through inspection of trace plots and calculation of the Gelman-Rubin statistic [41]. Chains were run for 20000 rounds, with a burn-in of 3000 rounds and a thinning interval of 10 to ensure each parameter had an effective sample between 1000–2000 per MCMC chain.

Change in depth over time

In addition to investigating the time it took to reach specific water depths, we modelled TDR depth over time in a continuous manner to highlight potential non-linearities in sink rate. A generalized additive mixed model (GAMM) was used with log-transformed TDR depth recorded at regular one second intervals as the response variable [42,43]. As before, log transformation was used to prevent models predicting negative values for depth and model predictions were subsequently back-transformed to the data scale. We used time elapsed since shot (seconds) as a predictor variable. The relationship between log-depth and time elapsed since shot was modelled using a spline smoother (cubic regression spine with shrinkage). We used estimated separate smoothers for each TDR position category and included a random intercept for Day. Temporal autocorrelation in depth over time was modelled by including an auto-correlation structure of order 1 (AR-1 [44]) for measures taken from each unique TDR deployment (a combination of TDR position and day of deployment). The extent of auto-correlation was then estimated separately across each TDR position category.

Results

TDRs were deployed 97 times on the gear across 11 fishing operations. Each fishing operation consisted of a full longline deployment, ranging from 2,800 to 10,400 hooks per operation. Between seven and ten TDR recordings were collected for each fishing operation (Table 1). Two TDRs were lost across the project duration (due to fishing gear breakage) and a further four recordings failed due to errors in setting the recording periods. In total, 91 TDR deployments provided recordings (~94% deployment success) of sink rate at the four different gear positions: Dropper n = 21, Float n = 21, Middle n = 6 and Weight n = 43.

Sink rate

With the exception of wind speed, none of the predictor variables we initially included alongside TDR position were retained in our final models after our model selection procedure (see model selection tables in the Supporting Information). Wind speed was found to be positively associated with the time taken to sink from 0–5 metres (S4 Table), from 0–10 metres (S6 Table) and from 2–5 metres (S8 Table). Model coefficients suggest a one-unit increase on the Beaufort scale is predicted to lead to a small decrease in sink rate of ~ 3–5% in these models. Wind speed was also found to be positively associated with the expected distance travelled from stern before TDRs reached 2 metres (S14 Table), 5 metres (S16 Table), and 10 metres of depth (S18 Table). Model coefficients suggest a one-unit increase on the Beaufort scale is predicted to lead to an increase in the distance travelled before reaching these depths as ~ 6–8% in these models.

TDRs at the weight reached 2 m, 5 m and 10 m much faster than those at other positions (Table 2). Average sinking speed was significantly slower than the ACAP minimum threshold of 0.3 m/s for most parts of the gear and at any depth (Table 3, Fig 4). Only TDRs closer to the weights achieved >0.3m/s sinking speed (with the likely exception at depth 0–2 m), while TDRs set at the dropper, float or middle positions, were much slower: achieving, respectively, only 16%, 11% and 24% (on average) of the recommended sinking speed within the first two meters of the water column, 50%, 60% and 73% respectively in the 2 to 5m water depth section, and 32%, 63% and 83% in the 5 to 10m water depth section (Table 3).

Table 2

Summary table of time taken in seconds for TDRs set at different longline locations to reach specific water depths.

	Dropper	Float	Middle	Weight
2m	40.1	58.8	28.07	7.55
	(30.96–51.71)	(46.33–72.71)	(20.67–42.81)	(5.87–9.49)
5m	59.12	74.43	41.56	11.86
	(49.58–70.33)	(61.84–89.55)	(30.48–61.14)	(9.81–14.29)
10m	107.76	101.65	61.14	22.17
	(80.53–142.22)	(76.93–136.06)	(45.23–86.09)	(16.52–29.27)

Values correspond to modelled estimates of average taken across all deployed TDR with the relevant longline location category. Bayesian 95% CRI are also displayed in brackets. For more detailed tables see S2, S4 and S6 Tables in Supporting Information.

Table 3

Estimated sinking speed across different water depth ranges expressed as the average sink rate (m/s) at each TDR location together with 95% CRI in brackets.

Depth	Dropper	% of 0.3m/s	Float	% of 0.3m/s	Middle	% of 0.3m/s	Weight	% of 0.3m/s
0-2m	0.049	16	0.034	11	0.071	24	0.26	87
	(0.039–0.064)		(0.027–0.043)		(0.046–0.099)		(0.21–0.34)
2-5m	0.15	50	0.18	60	0.22	73	0.67	223
	(0.12–0.19)		(0.14–0.22)		(0.16–0.28)		(0.54–0.86)
5-10m	0.097	32	0.19	63	0.25	83	0.49	163
	(0.067–0.13)		(0.15–0.25)		(0.18–0.32)		(0.41–0.61)

The associated percentage of ACAP sinking speed recommendation achieved by TDRs based upon modelled sink speeds also provided. For more detailed tables see S8 & S10 Tables in Supporting Information.

Fig 4

Box plots showing observed sink rates (raw, unmodelled data) obtained from TDR loggers at each TDR location across specified depth ranges.

Red dotted lines indicate ACAP sinking speed recommendation (0.3 m/s).

Based on the outputs of our GAMM modelling Fig 5 shows the typical sinking profile over time of TDRs deployed at each longline location (see S11 Table for detailed output from GAMM modelling). The average sink rate at the dropper and middle positions never exceeded 0.3 m / s. Typically, the average sink rate at the float position was under 0.3 m / s as well. Between 100 and 125 seconds after setting, the average sink rate at the float was predicted to be at 0.3 m / s or above, but only at depths already out of susceptible seabirds’ reach (around 10 m and over). Additionally, this should be interpreted with caution as the 95% CI were also wider at such depths, highlighting a greater level of uncertainty about model prediction at this point in the curve. The average sink rate at the weight location was below 0.3 m / s in the first 2 meters of the water column, then rapidly exceeded this tresholds from approximately 2 m in depth until around 13 m, then dropping again below 0. 3 m / s thereafter. However, the depth attained by a TDR at the weight location still exceeded the depth that would be attained by a longline sinking at a constant rate of 0.3 m / s during the first 50 seconds of deployment. The raw data points displayed on the plots also highlight variation between sink rates observed across different TDR deployments at the same location but performed on different days / locations.

Fig 5

The relationship between TDR depth and time elapsed since line was set at each TDR location along the longline.

The solid line represents the average relationship across all recorded TDRs deployed at a specific location estimated via a GAMM. The solid line is black when predicted sink rate is below 0.3 m/s and red when predicted sink rate is above 0.3 m / s. The blue polygon around the line encompasses the 95% confidence interval (95% CI) around model predictions. The raw data from each TDR deployment is also displayed as a grey line to highlight variation between different longline deployments on different days. The dashed diagonal line on the plot represents the trajectory that would be achieved by a hypothetical TDR sinking at a constant rate of 0.3 m / s. See S6 Table for more details.

Distance from vessel and hooks sinking dynamics

Assuming the portions of the longline that are situated between two different sinking points of the gear (e.g. a weight and a float) follow a linear sinking trend across its length, we calculated that for the 50 m portion of the fishing line between a weight and a buoy (or dropper) which hosts 20 hooks (Fig 1), the percentage of hooks with a sink rate faster or equal to 0.3 m / s would equal 0% at 2 m depth, 20% at 5 m depth and 25% at 10m depth (S12 Table). With a moving vessel setting the longline at roughly 8 knots (n = 11; 7.9 knots in average during our trials (range: 6.5–9.1)), we calculated that the distance astern from the vessel for the longline to reach 2m water depth ranged from 29.66 m (weight) to 231.17 m (float), 47.41 m (weight) to 297.47 m (float) to reach 5 m, and 88.69 m (weight) to 443.73 m (dropper) to reach 10 m (Table 4). See S14, S16 & S18 Tables for more details.

Table 4

Summary table of expected distance travelled from vessel in metres for TDRs set at different locations to reach specific water depths.

	Distance from vessel to reach 2m deep (in m)	Distance from vessel to reach 5m deep (in m)	Distance from vessel to reach 10m deep (in m)
Dropper	158.58	237.87	443.73
	(129.21–200.63)	(188.56–304.21)	(332.91–599.04)
Float	231.17	297.47	405.39
	(188.29–290.66)	(230.94–382.81)	(299.61–548.48)
Middle	109.66	161.55	246.07
	(78.47–171.17)	(115.65–243.45)	(176.89–365.09)
Weight	29.66	47.41	88.69
	(23.91–36.62)	(37.51–60.21)	(66.53–118.04)

Bayesian 95% CRI are also displayed in brackets. For more detailed tables see Supporting Information.

Discussion

Two characteristics of the sink profile are particularly important in assessing hook and bait accessibility for seabirds while the longline sinks within the top 10 m of water: the sink rate (or speed of descent) and the distance from the vessel’s stern. The first because this relates to the diving range of the species most susceptible to capture–particularly Northern Fulmar but also Skuas, Shearwaters and Gannets–and how “easy” it is to catch a moving bait for these birds. The second because birds won’t approach a vessel too closely, but mainly because current ‘best practice’ bird-scaring line designs only cover an area 100m astern of the vessel.

Our analysis indicates that only the hooks close to the weight-lines are likely to sink at the recommended minimum sink rate of 0.3 m / s to limit seabird bycatch. It also indicates that the sink rate slightly increases with depth, stabilizing at depths over 10m, presumably a consequence of propeller wash behind the vessel as noted in other similar studies [45]. In the surface waters (0–2 m depth), virtually all the hooks are below this sinking threshold.

We estimated that most of the longline (and therefore hooks) were, on average, between 100 m and 230 m astern of the vessel before reaching a depth of 2m, and between 150 m to 300 m, before reaching a depth of 5 m. In the most extreme cases–near the floats and Droppers–the longline may be up to 290 m astern of the vessel before reaching 2 m, 380 m before reaching 5 m, and up to 600 m before reaching a depth of 10 m. We acknowledge that such estimates are fairly coarse (multiplying setting speed by time to depth) and that additional factors might dictate how far a longline can travel horizontally. Although environmental factors were recorded during field work, we were not able to assess their influence on sinking speed. Wind speed has showed a positive relationship with hooks sinking rate in a drifting-pelagic longline fishery [46], but the drastically different weighting regime renders any assumptions hazardous in our case. Slower vessel setting speed is also known to reduce the amount of time hooks remain at the surface and at depths that seabirds can reach [47], although its effect on a floated-demersal longline and the practicality of reduced vessel speed during fishing operations needs to be investigated.

Albeit potentially incomplete, our results strongly suggest that with floated-demersal longlines, seabirds remain at high risk of bycatch for some distance astern of the vessel. Even with a 150 m long bird scaring line–the recommended minimal length for demersal longline fisheries [48]—deployed at the correct height from the stern, only 100 m behind the vessel would be covered, leaving a large proportion of hooks available to seabirds. In spite of the reported use of bird-scaring lines by some vessels in this fleet (the exact specifications and efficacy of which are unknown in the absence of paired sea trials), the slow sinking speed of this gear–even if best practice bird-scaring lines were deployed–is likely the driver of the high seabird bycatch estimates reported by Anderson et al. [3] and Northridge et al. [5].

Another recommended mitigation measure for longline fisheries is night setting, but its effectiveness can be reduced during bright moonlight and when using powerful deck lights [49]. More importantly, the floated-demersal longline fishery frequently operates in high latitudes of the northeast Atlantic Ocean, including in the Faroe-Shetland channel, where during the summer months the time between nautical dusk and dawn is virtually absent—or at best—very limited. It has also been found that in the Alaskan Longline fisheries, Northern Fulmars were caught at significantly higher rates at night [50], highlighting further that best practice mitigations can vary by species assemblage and fishery.

In the absence of an easy-to-implement solution, investment in gear modification research is urgently needed to adapt existing and novel mitigation measures to this fishery. Although changes in the floated-demersal longline configuration would not be without challenges for fishers and likely be unpopular with industry. For instance, a switch from current 3kg to 5kg weights for a set of ten thousand hooks would represent roughly an additional half a tonne to be carried onboard and handled by fishers. Cortés & González-Solís [51] also found that in artisanal demersal longlines, more weight is susceptible to increase entanglement risks between the branchlines and hooks during the setting operations. However, change in the longline weighting regime is the single measure most likely to deliver a significant reduction in seabird bycatch, and a series of measures that could help reduce operational issues should be investigated such as alternative spacing between branchlines and between weights [52], use of steel weights instead of concrete [29], etc. When paired with bird scaring lines, appropriate weighting regime can virtually eliminate seabird bycatch from demersal longline fisheries [18].

Developing and testing a floated-demersal longline with a significantly improved sinking speed more in line with ACAP’s recommendations, whilst maintaining or potentially improving economic returns, would be a win-win scenario that is more likely to foster interest and support from the fishery. Seabirds are often able to steal numerous baits before being hooked, and the resulting loss in fish catching potential reduces the efficiency of the fishing operation [53]. Reducing bycatch could therefore bring direct economic returns, both through increased fish catch and decreased hook loss. Further, by improving the sustainability of the fishing operation, the potential for these fisheries to pass third-party sustainability certification schemes is likely to increase.

Model selection tables for time taken to reach 2 metres in depth.

Table shows model tested and corresponding Root Mean Square Error (RMSE) calculated using Leave-One-Out Cross-Validation (LOO-CV). Best performing model highlighted in bold. ε [TDR Position] denotes a model in which separate estimates of the residual variance were made for each Position.

(PNG)

Click here for additional data file.

Time taken in seconds for TDRs set at different locations to reach 2 metres depth.

Table display coefficients from a model in which time was modelled using a log transformation. Back-transformed coefficients for sink speed also displayed in original units as well as expressed as the average sink rate (m / s) from 0–2 metre depth. σ Day is the random effect associated with day on which longlines were deployed. σε is the residual variation in the model–note that different residual variation parameters were estimated for each location in this model. n = 91 observations, 11 days. Model marginal R2 = 0.93 (95% CRI: 0.89–0.95); Model conditional R2 = 0.94 (0.91–0.96).

(PNG)

Click here for additional data file.

Model selection tables for time taken to reach 5 metres in depth.

Table shows model tested and corresponding Root Mean Square Error (RMSE) calculated using Leave-One-Out Cross-Validation (LOO-CV). Best performing model highlighted in bold. ε [TDR Position] denotes a model in which separate estimates of the residual variance were made for each Position.

(PNG)

Click here for additional data file.

Time taken in seconds for TDRs set at different locations to reach 5 metres depth.

Table display coefficients from a model in which time was modelled using a log transformation. Back-transformed coefficients for sink speed also displayed in original units as well as expressed as the average sink rate (m / s) from 0–5 metre depth. Back-transformed estimates assume Beaufort scale is set at its modal value. σ Day is the random effect associated with day on which longlines were deployed. σε is the residual variation in the model–note that different residual variation parameters were estimated for each location in this model. n = 91 observations, 11 days. Model marginal R2 = 0.93 (95% CRI: 0.89–0.95); Model conditional R2 = 0.94 (0.90–0.96).

(PNG)

Click here for additional data file.

Model selection tables for time taken to reach 10 metres in depth.

Table shows model tested and corresponding Root Mean Square Error (RMSE) calculated using Leave-One-Out Cross-Validation (LOO-CV). Best performing model highlighted in bold. ε [TDR Position] denotes a model in which separate estimates of the residual variance were made for each Position.

(PNG)

Click here for additional data file.

Time taken in seconds for TDRs set at different locations to reach 10 metres depth.

Table display coefficients from a model in which time was modelled using a log transformation. Back-transformed coefficients for sink speed also displayed in original units as well as expressed as the average sink rate (m / s) from 0–10 metre depth. Back-transformed estimates assume Beaufort scale is set at its modal value σ Day is the random effect associated with day on which longlines were deployed. σε is the residual variation in the model. n = 91 observations, 11 days. Model marginal R2 = 0.92 (95% CRI: 0.88–0.94); Model conditional R2 = 0.93 (0.91–0.95).

(PNG)

Click here for additional data file.

Model selection tables for time taken to travel between 2 and 5 metres of depth.

Table shows model tested and corresponding Root Mean Square Error (RMSE) calculated using Leave-One-Out Cross-Validation (LOO-CV). Best performing model highlighted in bold. ε [TDR Position] denotes a model in which separate estimates of the residual variance were made for each Position.

(PNG)

Click here for additional data file.

Time taken in seconds for TDRs set at different locations to travel from 2 to 5 metres depth.

Table display coefficients from a model in which time was modelled using a log transformation. Back-transformed coefficients for sink speed also displayed in original units as well as expressed as the average sink rate (m / s) from 2–5 metre depth. Back-transformed estimates assume Beaufort scale is set at its modal value. σ Day is the random effect associated with day on which nets were deployed. σε is the residual variation in the model–note that different residual variation parameters were estimated for each net location in this model. n = 91 observations, 11 days. Model marginal R2 = 0.84 (95% CRI: 0.72–0.91); Model conditional R2 = 0.86 (0.74–0.92).

(PNG)

Click here for additional data file.

Table shows model tested and corresponding Root Mean Square Error (RMSE) calculated using Leave-One-Out Cross-Validation (LOO-CV). Best performing model highlighted in bold. ε [TDR Position] denotes a model in which separate estimates of the residual variance were made for each Position.

(PNG)

Click here for additional data file.

Time taken in seconds for TDRs set at different locations to travel from 5–10 metres depth.

Table display coefficients from a model in which time was modelled using a log transformation. Back-transformed coefficients for sink speed also displayed in original units as well as expressed as the average sink rate (m / s) from 5–10 metre depth. σ Day is the random effect associated with day on which longlines were deployed. σε is the residual variation in the model–note that different residual variation parameters were estimated for each location in this model. n = 91 observations, 11 days. Model marginal R2 = 0.77 (95% CRI: 0.68–0.82); Model conditional R2 = 0.79 (0.69–0.83).

(PNG)

Click here for additional data file.

Results from a GAM modelling the relationship between time since deployment and depth reached at different parts of the gear.

σ Day represents the random effect associated with longlines deployed on different days. The temporal autocorrelation in depth over time is represented as ø and was estimated separately for each gear location. Different smoothers were fitted for each gear location and details on these smoothers (estimated degrees of freedom and k index are also displayed). Model R2 = 0.88.

(PNG)

Click here for additional data file.

Calculated average sinking speed (in m/s) per hook in a 20 hooks floated-demersal longline section.

Using a linear series from TDRs recordings at different positions of the gear (in grey shading). In bold are rates equal or over ACAP recommendation (0.3 m/s).

(PNG)

Click here for additional data file.

Model selection tables for the expected distance travelled from stern in metres for TDRs set at different positions to reach 2 metres depth.

Table shows model tested and corresponding Root Mean Square Error (RMSE) calculated using Leave-One-Out Cross-Validation (LOO-CV). Best performing model highlighted in bold. ε [TDR Position] denotes a model in which separate estimates of the residual variance were made for each Position.

(PNG)

Click here for additional data file.

Expected distance travelled from stern in metres for TDRs set at different locations to reach 2 metres depth.

Table display coefficients from a model in which distance from stern was modelled using a log transformation. Back-transformed coefficients for distance travelled also displayed in original units. Back-transformed estimates assume Beaufort scale is set at its modal value. σ Day is the random effect associated with day on which longlines were deployed. σε is the residual variation in the model–note that different residual variation parameters were estimated for each location in this model. n = 91 observations, 11 days. Model marginal R2 = 0.93 (95% CRI: 0.91–0.95); Model conditional R2 = 0.95 (0.94–0.96).

(PNG)

Click here for additional data file.

Model selection tables for the expected distance travelled from stern in metres for TDRs set at different positions to reach 5 metres depth.

Table shows model tested and corresponding Root Mean Square Error (RMSE) calculated using Leave-One-Out Cross-Validation (LOO-CV). Best performing model highlighted in bold. ε [TDR Position] denotes a model in which separate estimates of the residual variance were made for each Position.

(PNG)

Click here for additional data file.

Expected distance travelled from stern in metres for TDRs set at different locations to reach 5 metres depth.

Table display coefficients from a model in which distance from stern was modelled using a log transformation. Back-transformed coefficients for distance travelled also displayed in original units. Back-transformed estimates assume Beaufort scale is set at its model value. σ Day is the random effect associated with day on which longlines were deployed. σε is the residual variation in the model–note that different residual variation parameters were estimated for each location in this model. n = 91 observations, 11 days. Model marginal R2 = 0.92 (95% CRI: 0.89–0.95); Model conditional R2 = 0.95 (0.93–0.96).

(PNG)

Click here for additional data file.

Model selection tables for the expected distance travelled from stern in metres for TDRs set at different positions to reach 10 metres depth.

Table shows model tested and corresponding Root Mean Square Error (RMSE) calculated using Leave-One-Out Cross-Validation (LOO-CV). Best performing model highlighted in bold. ε [TDR Position] denotes a model in which separate estimates of the residual variance were made for each Position.

(PNG)

Click here for additional data file.

Expected distance travelled from stern in metres for TDRs set at different locations to reach 10 metres depth.

Table display coefficients from a model in which distance from stern was modelled using a log transformation. Back-transformed coefficients for distance travelled also displayed in original units. Back-transformed estimates assume Beaufort scale is set at its modal value. σ Day is the random effect associated with day on which longline were deployed. σε is the residual variation in the model. n = 91 observations, 11 days. Model marginal R2 = 0.91 (95% CRI: 0.81–0.94); Model conditional R2 = 0.94 (0.91–0.96).

(PNG)

Click here for additional data file.

6 Jan 2022

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Vitor Hugo Rodrigues Paiva, Ph.D.

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. Thank you for stating the following in the Acknowledgments Section of your manuscript:

"We are particularly grateful for the funding support from Centre for Environment, Fisheries and Aquaculture Science (CEFAS) UK Seafood Innovation Fund, which allowed us to carry out this work, as well as the University of St. Andrews Sea Mammal Research Unit, which helped in securing capacity for onboard observation and data collection. Sincere thanks to Hooktone Limited, the Longliner’s crew and skipper who granted us permission and technical support to operate on their vessel during real fishing conditions, without whom this project would have not been possible."

We note that you have provided funding information that is not currently declared in your Funding Statement. However, funding information should not appear in the Acknowledgments section or other areas of your manuscript. We will only publish funding information present in the Funding Statement section of the online submission form.

Please remove any funding-related text from the manuscript and let us know how you would like to update your Funding Statement. Currently, your Funding Statement reads as follows:

"YR - Funding was received from the Centre for Environment, Fisheries and Aquaculture Science (CEFAS) UK Seafood Innovation Fund, on the Feasibility study “Developing a floated demersal longline design that minimises seabird bycatch (FS031)” [https://www.seafoodinnovation.fund/]. Funding allowed the acquisition of the Time Depth Recorder devices, their at-sea deployment on an commercial longliner and the data collection over a series of fishing operations. None of the funders had any role in study design, collection, analysis, and interpretation of data, in the writing of the report or in the decision to submit this manuscript to publication."

Please include your amended statements within your cover letter; we will change the online submission form on your behalf.

3. In your Data Availability statement, you have not specified where the minimal data set underlying the results described in your manuscript can be found. PLOS defines a study's minimal data set as the underlying data used to reach the conclusions drawn in the manuscript and any additional data required to replicate the reported study findings in their entirety. All PLOS journals require that the minimal data set be made fully available. For more information about our data policy, please see http://journals.plos.org/plosone/s/data-availability.

Upon re-submitting your revised manuscript, please upload your study’s minimal underlying data set as either Supporting Information files or to a stable, public repository and include the relevant URLs, DOIs, or accession numbers within your revised cover letter. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories. Any potentially identifying patient information must be fully anonymized.

Important: If there are ethical or legal restrictions to sharing your data publicly, please explain these restrictions in detail. Please see our guidelines for more information on what we consider unacceptable restrictions to publicly sharing data: http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions. Note that it is not acceptable for the authors to be the sole named individuals responsible for ensuring data access.

We will update your Data Availability statement to reflect the information you provide in your cover letter.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

Reviewer #2: Yes

Reviewer #3: Yes

Reviewer #4: Partly

Reviewer #5: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: No

Reviewer #2: Yes

Reviewer #3: I Don't Know

Reviewer #4: Yes

Reviewer #5: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

Reviewer #2: Yes

Reviewer #3: No

Reviewer #4: Yes

Reviewer #5: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

Reviewer #4: Yes

Reviewer #5: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: In this study, the authors collected and analyzed sink rate data from a floated demersal longliner in the Eastern North Atlantic. In my opinion, this research topic is important for seabird bycatch mitigation. For example, see Dietrich et al, 2008. The study design is appropriate, although there are several places detailed below that needs to be further clarified. The statistical analysis as used in this manuscript is not satisfactory, and both the analysis and the interpretation of the results can be further improved.

In this study, the use of spline is good as it smoothes out the noise and it is also straight forward to calculate the sink rate, which is the 1st derivative of the spline. Both Fig5a and 5b look good. However, Fig5c and 5d look problematic. First, the interval estimates on the right-hand side are way too wide. Second, most data tracks stop midway. Either the depth readings didn't change afterwards or the device stopped recording at certain depth. I think the authors just chopped the second half of the data if depth readings remained stable. This is understandable, but the authors need to document it in the methods section. Assuming this is the case, the use of spline smoothing is not appropriate for the second half of the data in Fig5c and d due to this data cutting. For example, in Fig5c, to the right of ~110 on the x-axis, the spline flattens while none of the tracks flattens, and the flattening is simply due to inappropriate smoothing across multiple chopped tracks. The authors need to be careful here. Even though this portion of the data (>10m) has little practical significance to this study, both the methods and the interpretation of the result need to be correct.

In addition, what is the observed target catch and bycatch rates for these experimental trials? If the authors plan to save these statistics for another paper, at least provide a summary here.

The model selection results are missing. Add a table with the loo score for each candidate model.

Specific comments:

line 127: even though there is no standard configuration, the authors need to provide evidence that the configuration presented here is representative of the actual situation.

line 180: Where is your settings for the priors, which is an essential component of every Bayesian model?

line 181: which type of CI did you pick, symmetric or HDI?

line 203: Are these 97 trials conducted on the same fishing vessel with the help of a single observer? This is important because otherwise "observer id" and "vessel id" would be two important factors affecting the sink rate and they should be incorporated into the model.

line 262: Rephrase this sentence. I can guess what you are trying to say by looking at the table in the Appendix. Also, it is calculation not simulation.

Fig5a: Looking at this figure, I can see that either the device used in this study has a low depth resolution or the data have been truncated. State in the method section what is the design accuracy of this device and the resolution of the recorded data.

The surface layer is the most important in this study. What is the design accuracy of this device for the surface layer? Did you calibrate the device (pressure sensor) before using it or after the trials? Readings may drift under harsh environmental conditions. A seperate section on the validity of this device and its readings would further strenghen this study.

Editorial comments:

line 45: Do not capitalize the second part of a scientific name.

line 49: Use different dashes consistently

line 64: What do you mean here? "lofted off the seabird"?

line 179: not grammatically correct.

line 240: typo

Fig2: which base map did you use, where is the attribution, what projection did you use? Instead of a star, I think a rectangle shape is more appropriate. Also, it might be better to use a zoomed-in main map and keep this one as an inset.

Some references that might be useful:

Dietrich K., et al. 2008. Integrated weight longlines with paired streamer lines – Best practice to prevent seabird bycatch in demersal longline fisheries. Biological Conservation.

Reviewer #2: This is an interesting study that address an important aspect when it comes to pinpointing reasons, and consequently also effective mitigation actions, for high bycatch rates in floated-demersal longlines in the north Atlantic. I think the results are highly relevant in a broader perspective, as it gives some general estimates to the sink rate in these types of fisheries, as well as shoeing the importance of weighted lines in reducing the time period when the bait is available for seabirds. The statistics/methods seems to appropriate, and I only have minor comments. Addressing them will probably remove my small concern about the seemingly very small N for the rather complex models used. Good work.

Minor comments:

Title: Suggest to moderate the title, as it doesn’t seems like you have actually measured any changes in bycatch risk, although this is implied/assumed by the observed sink rate.

Line 51: Would analysis be a better word than “data”, in “Recent data estimate that”?

Line 62: In my printed version it seems that there is a different font or color within this line, please check

Line 120-142: It is not clear to me whether all observation is done from one vessel, or if there are multiple vessels? Suggest to include this information in the text.

Line 141-142: It is a bit difficult to understand the rationale behind the different sections chosen as basis for the different models? Especially since there seems to be no focus on this in the discussion. Why overlapping intervals, from the text it seems to me that 0-2 (surface feeders) meters and 2-10 (diving) meter would be logical choices given the two different feeding regimes described? Suggest to clarify.

Line 161: How was it standardized, and why does this standardization ease interpretation? My initial though is that it would be easier to interpret according to actual setting speed and depth rather than a (mean?)standardized version of the variables for example, as I don’t know the range of the data.

Line 166: Not sure if I understand why predictive performance is in focus in the model selection, when the results seems to be predicted over mean’s (or the range, not explained) of all other variables than the TDR-positions. Why the model selection if all variables are relevant to account for when measuring effects on sink speed? Suggest to clarify. Also, I might have overlooked it, but I cant seem to find the results from the model selection?

Line 168: would it be relevant to consider a vessel random effect as well?

Line 180: Change “used” to “using”

192: I don’t understand why a log transformation was used to prevent the model to predict negative values for depth. Or rather, I don’t understand why the model would predict negative values for depth if the proper distribution is assumed in the models. Did you assume a normal distribution, and is this assumption correct? Would a truncated distribution be a better choice? Suggest to include information about the type of model.

Line 151-200: It’s a bit unclear to me why you chose to both construct models with sink rate as a response, and models with depth over time? Why not just the latter, as it seems to provide all the information (sink rates could easily be derived from these), and why not include TDR location as an interaction effect with time elapsed rather than creating 4 different models?

Also, from the results it seems that you only have 91 lines in total (and much less per TDR location (?)), did you actually have enough statistical power to make sound estimate for all models/combinations. Your model-output seems to suggest that you have, but the model structure seems rather complex given the potential range for the different variable levels (e.g. as low as 6? for “middle”), and N in total? I have probably misunderstood this part, so please clarify in the text.

Line 202: I miss a short description of which parameters comes into play in the different models (except for TDR-position, all seems to be omitted from the text (expect from being included in the sup. tables). It is thus difficult so see the rationale for doing a model selection at all, if these variables are of no interest.

Figures: Are these the predicted response across the range of all other variables, or the for example the mean? Please clarify.

Reviewer #3: This manuscript presents an assessment of the sinking speed of floated-demersal longline to determine potential bycatch risks to seabirds. First of all, I would like to present my congratulations to the author by this interesting and hard piece of work. Generally, the manuscript is well written and easy to understand. The main claim of the manuscript is well defined and properly placed in the context of the reviewed literature. However, despite data is likely to support the claims, some more details on the results are needed, namely in terms of the modelling exercises, how final models fit each sample and which predictor variables were included. At the end it was hard to find any detail on predicted variables other than TDR position or water depth, i.e. wind speed, setting speed and number of hooks. Such lack of details prevents the assessment of the appropriate and rigorous performance of the statistical analysis. Some details on GAMM results might be available under the supplementary material which as reviewer I had no access. Please find bellow my detailed revision:

Introduction

Line 44 Consider to replace "Puffinus Gravis" by "Ardenna gravis"

Line 46 Replace "Celtic Sea" by "Celtic Seas"

Line 69 to 87 Consider to move this paragraph to a later section of your discussion. Here would be more interesting to introducing the reader to a narrow scope of your main topic/goal.

Methods

Line 139 and 140 I found this affirmation to important to be based in such weak peaces of information reflecting a notable lack of data. Remembering the Gran Sol bycatch figures are based in very low observation effort, from only one vessel. Despite, there are strong evidences of seabird bycatch to occur in the area, listing top bycatch species and bycatch numbers in the north east Atlantic floated demersal longline fishery based in such little information requires care. There is not much available information on how Great Shearwater bycatch rates were estimated for the Spanish fleet operating in Gran Sole, one important aspect to be into consideration is the presence of birds in the area during a narrow period of time. In the other hand, there are published evidences showing Northern Gannets mainly followed by Cory's Shearwater are being caught in serious numbers in longlines. Perhaps replace the assumption of "most bycaught species" by "two of the known bycaugh species" fits better the state of the art. Also, after reading the introduction and the methods I was expecting a need from the authors to discuss their findings in terms of any king of ecological or biological aspects of Fulmar or Great Shearwater. But, I understood those species are only used to justify the chosen distance bands. Perhaps, authors might be wider and justify their distance bands choice by the existence of different feeding strategies in seabird species. There are some classifications proposed by other authors, e.g. surface feeders, pelagic feeders, etc. And perhaps using Fulmar and Great Shearwater as an example of those groups, and add others if relevant.

Line 160 Please clarify the choice of using log number of hooks instead the raw number.

Line 161 Please detail the standardized method used for setting speed and water depth

Results

Line 211 Table 1 - to confirm within the journal guidelines about measurement units, knots are not an IS unit; Perhaps it is worth to remove "Species targeted", because all samples were targeting hake.

214 What about the results of the modelling? How much the predictor variables explain the sink rate variability? Which predictor variables were taking into account in the final model? Only the effect of position of TDR on the line on sink rates are given. Perhaps the results of the LOO CV step to retain or remove predictors are missing, but important to clarify the authors' choices. More details on those results are needed.

Line 237 As stating before, apart from the outputs of GAMM modelling, the results of the modelling exercise itself will help reader to understand the goodness of fit of the selected model.

Line 255 It is a very non-important detail, but as sink rate is expectable better when above 0.3m/s my suggestion is to illustrate it with a black line, and when bellow as a red line.

Line 262 Still this paragraph about sink rate or more about "Distance from stern for specific water depth ranges". Consider to add this "sub-chapter" here or follow the same approach used in the Methods - "Sink Rate and Distance from..." - for the entire sub-chapter

266 Authors use knots, please confirm guidelines for the need of using IS units

Discussion

Line 275 to 324 Depending on the results, it would be worth to discuss the possible effects of wind speed, setting speed and number of hooks (as proxy for size of a longline) on the sink rates and then on the bycatch risk.

Also, only one vessel was sampled in this study. Could such fact be a limitation for this study? Might be desired to discuss it in the light of the variability of longlines even within the same fishing fleet.

Lines 313 to 316 Could authors discuss the main challenges for fishers regarding changes in the longline weighting?

Reviewer #4: General issues:

Rouxel et al. present a method to assess the sink rate of a floated-demersal longline, and apply this method to a longline used to target European Hake in the offshore waters of Scotland, and find that the sink rate is slower than the rate recommended by the ACAP, in the top 2m of the water column in particular, and the distance astern of the vessel for hooks to sink beyond seabirds’ reach largely exceeds the optimum coverage. They indicate that hooks from floated-demersal longlines present a clear bycatch risk.

I found the experimental method adequately motivated, and I believe it will be useful in the following seabird bycatch analyses. I have some concerns about the sufficiency of an experiment on a single longline to achieve a general conclusion. Moreover, there is no observation on seabird bycatch during this experiment, and no analysis on the impacts of sink rate and other factors on seabird bycatch, so it is a hypothesis that the slow sink rate in this floated-demersal longline may pose high bycatch risk to seabirds, but to what extent, we don’t know. I imagine that this kind of fundamental experiment will be useful if a following analysis is conducted to relate seabird bycatch to the sink rate. At this point, I suggest to revise the title to avoid overstating this study and add more references in the discussion to emphasize the potential impacts of sink rate on seabird bycatch. All concerns should be addressed via major revision.

Specific issues:

Lines 41-43: Where were these numbers (160,000-320,000) from?

Line 76: Please specify the best practice mitigation measures.

Lines 125-129: Did weight keep constant?

Lines 158-160: Does other factors, such as wind direction, wave height, weight (if constant during this experimental period), influence the sink rate?

Lines 182-183: It was found that reducing setting speed reduced line tension and resulted in gear sinking closer to the vessel (Pierre et al. 2013). Is it the case in this study?

Lines 288-289: Need references.

Lines 312-324: There have been sufficient studies investigating the impacts of sink rate of longline fisheries on seabird bycatch (e.g. Cortés et al. 2018). Methods such as weighted lines, Chilean system, thawing the bait, smaller distance between consecutive weights have been adopted to increase sink rate in order to reduce seabird bycatch, although these methods have been found with some operational problems. For example, an increase of entanglements between the branch lines and hooks may happen during the setting operations when additional weights were used. More discussion on how to increase sink rate and the corresponding impacts on fish catch will improve this paper. A study found that weighted lines increased blackmouth catshark catches, possibly because catsharks forages in the near bottom layer and on the seabed (Anastasopoulou et al. 2013). An introduction on the life history of the target species European Hake would help to discuss the potential impacts of increased sink rate.

Figures 4 and 5: The four panels for each figure can be put in one page, using some R packages such as ggplot2.

References:

Pierre JP, Goad DW, Thompson FN, Abraham ER. Reducing seabird bycatch in bottom-longline fisheries. Final Report prepared for the Department of Conservation: Conservation Services Programme projects MIT2011-03 and MIT2012-01. Department of Conservation, Wellington. 2013.

Anastasopoulou A, Mytilineou C, Lefkaditou E, Dokos J, Smith C., Siapatis P, et al. Diet and feeding strategy of blackmouth catshark Galeus melastomus. J Fish Biol. 2013; 83: 1637–1655.

Cortés V, González-Solís J. Seabird bycatch mitigation trials in artisanal demersal longliners of the Western Mediterranean. PLoS One. 2018;13(5): e0196731.

Reviewer #5: This is an excellent manuscript and was a pleasure to review. I have a few minor suggestions for consideration:

- line 83. Provide a reference for 'Clean Catch UK'

- consider combining Figures 1 and 3

- lines 130-132. Consider rewriting as follows: Fishers were asked to deploy gear as they would in a normal fishing operation. Eight to ten TDRs were deployed on each experimental set, depending on fishing conditions.

- line 139. Delete 'reportedly the most bycaught species in the north east Atlantic floated demersal longline fishery', as this repeats infomation provided earlier.

- line 182. insert 'that' between stern and different, i.e., ..... from the stern that different positions ....

- line 255. Spelling - polygon.

- reference 41. Please provide more details, e.g., name of university.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

Reviewer #3: No

Reviewer #4: No

Reviewer #5: Yes: Barry Baker

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

18 Feb 2022

Thank you for giving us the opportunity to improve greatly our manuscript, based on constructive comments from the reviewers. We believe we have now answered to all the issues which required Major Revisions, in particular in relation to the statistical analysis and in light of the PLOS data policy & Funding information requirements. Response to all specific comments can be found in the submitted File "Response to Reviewers_Rouxel et al."

We hope you will find the revised version of the Manuscript and Figures meet the quality and publication requirements of the PLOS journal.

Submitted filename: Response to Reviewers_Rouxel et al..docx

Click here for additional data file.

18 Mar 2022

30 Mar 2022

All reviewers' comments were addressed in the attached "Response to Reviewers_Rouxel et al_2V" document.

Submitted filename: Response to Reviewers_Rouxel et al_2V.docx

Click here for additional data file.

4 Apr 2022

Slow sink rate in floated-demersal longline and implications for seabird bycatch risk

PONE-D-21-37923R2

Dear Dr. Rouxel,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Vitor Hugo Rodrigues Paiva, Ph.D.

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

11 Apr 2022

PONE-D-21-37923R2

Slow sink rate in floated-demersal longline and implications for seabird bycatch risk

Dear Dr. Rouxel:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Vitor Hugo Rodrigues Paiva

Academic Editor

PLOS ONE