All trials carried out during this study followed standard commercial fishing practices and snow crabs were not exposed to any additional harm. Therefore, this study did not require any permits from the authorities. Furthermore, the trials did not involve any endangered or protected species.

Sea trials were carried out on board the commercial fishing vessel “Northeastern” (overall length 55.2 m and 2250 HP) which operates 9000 pots and has a capacity of deploying and retrieving an average of 2000 pots per day. Comparative fishing experiments were conducted from 1^st to 24^th of May 2020 in the Barents Sea (latitude between N75°4262 and N76°1160; longitude between E33°1789 and E36°7520) (ICES area 1.b). The depth of the fishing grounds varied between 250 and 280 m.

During the sea trials, the same conical snow crab pots used in the commercial fishery operating in the Barents Sea and along the east coast of Canada were used [9, 20]. The dimensions of the pots were 130 cm in diameter at the bottom and 55 cm at the top, where the entrance to the pot is located. The height of the pots was 60 cm. The pot frame and the lower ring were made of 12-mm and 14-mm steel bars, respectively, yielding a pot weight of approximately 12.5 kg. Diamond mesh netting (mesh size 140 mm) made of Ø4 mm single braided polyethylene (PE) twine was used (mesh opening angle ranges from 64° to 89° [10]). The 140 mm netting has a CW50 of approx. 95 mm [8], meaning that 50% of legal-size snow crabs have the probability of escaping or being retained. The pots were attached to longlines by a quick link system with at least a 30 m distance between each pot.

Pots with and without lights, herein called test (T) and baseline (B) pots, respectively, were deployed on the same longline. After every second or third B pot there was a T pot.

Three series of experiments were carried out during the sea trials using lights. Pots in each of the series were linked on one separate longline resulting in three deployments of longlines:

Therefore, each of the series consisted of pots connected to one separate longline each. The pots of the three series were not fished simultaneously; however, they were deployed during the same fishing trials using the same snow crab fishing vessel. The experiments were conducted on board commercial snow crab fishing boat; therefore, the fishing area of the line deployments were slightly different and dependent on conditions of the vessel operation, weather conditions and drift ice. Therefore, it was not possible to keep similar soaking time conditions for all three experiments.

Each pot was baited with approximately 2 x 400 g of squid (Illex spp.). Half of the bait was placed in a bait container and the other half in a bait bag, which was mounted under the entrance of the pot (Fig 1). White and green PotLight LED fishing lights—manufactured by FishTekMarine [21]—were used in this experiment. The lights have a maximum operation depth of 1200 m and a battery life of approximately 500 hours.

The relative size-dependent catch efficiency between pots with and without LED-lights was independently estimated for each of the three experimental trials.

To assess the relative size-dependent catch efficiency derived from adding LED lights to the pots, we used unpaired catch comparison and catch ratio analyses [23]. We were interested in the size-dependent catch comparison rate values summed over the deployments of both T and B pots (deployment is equivalent to the number of baseline or treatment pots). During the experimental fishing trials, the catch data obtained for both types of pots were not collected in pairs, nor did the data correspond to the same total number of deployments, since each T pot was followed by two or three B pots on each longline. The catch data from the deployments of both pot types were summed and compared with the total deployment data by minimising Eq (1) to estimate the functional form of the summed catch comparison rate (the experimental rate being expressed by Eq (2) for T versus B pots). Eq (1) is given by: −∑w{∑i=1tq[ntwi×ln(CC(w,v))]+∑j=1bq[nbwj×ln(1.0−CC(w,v))]} (1) where v is a vector representing the parameters of the function describing the catch comparison curve defined by CC(w,v), and nt_wi and nb_wj are the numbers of snow crabs measured in each CW class w for T and B pots, respectively. In Eq (1), tq and bq are the number of deployments carried out with T and B pots, respectively, and the inner summations in the equation represent the sum of the total deployment data. The outer summation in the equation is the summation over the size classes w. Eq (1) is equivalent to maximising the likelihood for the observed data based on a maximum likelihood formulation for binominal data [29]. This maximization problem is turned to minimisation problem by putting a minus sign before the Eq (1). The experimental summed catch comparison rate, CC_w, where w denotes the snow crab CW, is given by Eq (2): CCw=∑i=1tqntwi∑i=1tqntwi+∑j=1bqnbwj (2)

When both the catch efficiency of T and B pots and the number of deployments are equal (i.e., tq = bq), the expected value for the summed catch comparison rate would be equal to 0.5. In this case the presence of unequal numbers for T and B pots in each series, tq/(tq + bq) would be used to judge whether or not there is a difference in catch efficiency between the pots. The experimental CC_w is often modelled by the function CC(w,v), which has the following form [30]: CC(w,v)=exp(f(w,v0,…,vk))1+exp(f(w,v0,…,vk)) (3)

In Eq 3, f is a polynomial of order k with coefficients from v₀ to v_k. We considered an f of up to an order of 4 with parameters v₀–v₄. Leaving out one or more of the parameters v₀…v₄ led to 31 additional models that were also considered as potential models for the catch comparison CC(w,v) between T and B pots. Among these models, estimations of the catch comparison rate were made using a multi-model inference to obtain a combined model [23, 31].

The ability of the combined model to describe the experimental data was evaluated based on the p-value, which quantifies the probability of obtaining by coincidence at least as big a discrepancy between the experimental data and the model as the one observed. Therefore, this p-value, which was calculated based on the model deviance and the degrees of freedom, should not be <0.05 for the combined model to describe the experimental data sufficiently well [32].

Based on the estimated catch comparison function CC(w, v) we obtained the relative catch efficiency (also named catch ratio) CR(w, v) between the pots with and without LED lights, and described by the following relationship [23]: CR(w,v)=qb×CC(w,v)qa×(1−CC(w,v)) (4)

The catch ratio is a value that represents the ratio between the catch efficiency of pots with and without LED lights. Thus, if the catch efficiency of both pot types is equal, CR(w,v) should always be 1.0. Thus, CR(w,v) = 1.5 would mean that the T pots with added LED lights are catching 50% more snow crab with size w than the B pots. In contrast, if CR(w,v) = 0.5, it would mean that the T pots are only catching 50% of snow crabs with size w that the B pots are retaining.

The confidence limits for the catch comparison curve and catch ratio curve were estimated using a double bootstrapping method [23]. The procedure accounted for the variation between pots in terms of snow crab availability and catch efficiency by selecting tq pots with replacement from the pool of T pots, and bq pots with replacement from the pool of B pots deployed during each bootstrap repetition. Uncertainty in the size structure of the catch data within the pots was accounted for by randomly selecting crabs with replacement from each of the selected pots separately. We performed 1000 bootstrap repetitions and calculated the Efron 95% [33] confidence limits for the catch comparison and catch ratio curve.

To identify the snow crab sizes associated with a significant difference in pot catch efficiency, we checked for size classes in which the confidence limits for the combined catch comparison curve did not contain tq/(tq + bq) and in which the confidence limits for the catch ratio curve did not contain the value 1.0.

Size-integrated average values (in percentage) for the catch ratio (CR_average) were estimated directly from the experimental catch data by the following equations: CRaverage−=100×qb×∑w<mw∑i=1tqntwiqa×∑w<mw∑j=1bqnbwjCRaverage+=100×qb×∑w≥mw∑i=1tqntwiqa×∑w≥mw∑j=1bqnbwj (5) where the outer summations include the size classes measured in the catch during the experimental fishing period under (for CR_average−) and over (for CR_average+) the minimum landing size (mw = 95 mm CW), respectively.

To investigate how well the size selectivity of the T and B pots matches with the size structure of the snow crab population present in the fishing ground, two fishing sustainability indicators (DiscardRatio_Test and DiscardRatio_Baseline) were estimated directly from the experimental catch data using the following equations: DiscardRatioTest=100×∑w<mw∑i=1tqntwi∑w∑i=1tqntwiDiscardRatioBaseline=100×∑w<mw∑j=1bqnbwj∑w∑j=1bqnbwj (6) where the outer summations include the size classes present in the catch during the experimental fishing period under (in nominator) and over (in denominator) all snow crab sizes. The DiscardRatio indicators quantify the ratio between undersized and target-size captured snow crabs (snow crab below and above mw, respectively). Ideally their value should be a as low as possible. DiscardRatio values are both affected by the size selectivity of the pots and by the size structure of the crab population present in the fishing grounds, and therefore provide an estimate that is specific for the targeted population and that cannot be extrapolated to other areas and seasons.

We used this equation to provide a size-averaged value for the effect of adding green or white LED lights on the catch efficiency of the pots. In contrast to the size-dependent evaluation of the catch ratio, CR_average is specific for the snow crab population structure encountered during the experimental sea trials, and it cannot be applied to other scenarios in which the size structure of the population may be different, depending on time period and fishing area [9].

Finally, size frequency distribution and cumulative size frequency distribution analyses were used. The method provides the proportion of the total catch up to a given size for the T and B pots. The analysis was carried out for T and B pots using the following equation: CDnW=∑i=1q∑w=0Wniw∑i=1q∑wniw (7) where n_wi is the number of snow crabs of size class w in pot i (T or B pot, respectively). Based on this information, the cumulative size frequency distribution CDn_W was obtained by the summations of i and w in Eq (7) over the q number of pots deployed during the experimental trials and CW size class w, respectively. CDn_W quantifies the proportion (in number of snow crabs) of a total catch reaching a given size class W. By incorporating the evaluation of Eq (7) in the double bootstrap described above we obtained the confidence intervals for CDn_W.