Methods of trap study

To investigate the research question ‘’ Which method of lionfish trapping is the most effective for targeting lionfish on the Saba Bank?’’, a lionfish trap study was conducted on the Saba Bank between March 2018 and September 2019. This study made use of two trap designs to test which one was most effective for targeting lionfish. The traditional chevron-shaped trap, which hardly required any alteration, and a modification of this design which has three extra funnels attached to its sides. Both designs made use of a plastic crate with holes cut in the sides to serve as a FAD, the specifics of which will be elaborated later. No live or organic bait was used. It was assumed that the structure of the trap in combination with the artificial reef would be enough to lure lionfish. The reason this type of FAD was chosen was to exploit the habit lionfish have to seek out small crevices and hiding spots (Hunt et al., 2019). The use of live bait has been shown before to attract more bycatch while not influencing lionfish catch rate (Pitt & Trott, 2015). The three extra funnels of the funnel trap design measure 15 centimetres in diameter and serve as an escape for smaller reef fish and thus aims to diminish bycatch.

This study used two data sets to investigate which trap design is most effective. Data from 2018 has been made available by the Saba Conservation Foundation and was used to get more data points.

Caution is required when comparing data since the traps used in the 2018 experiments contained a smaller mesh size then the traps in the current study.

Construction of the traps in 2018 happened in March. A total of 25 traps were made with a mesh size for both sides and top and bottom of 1,5 by 1,5-inch (38 mm). During the 2019 data collection the mesh size of the sides of the fish traps had to change from 1.5 by 1.5-inch to a bigger 2 by 2-inch (50 mm) mesh wire due to the high costs of importing 1,5 by 1,5-inch mesh wire from Atlantic and Gulf from Miami. 2 by 2-inch mesh wire can easily and relative inexpensively be imported from Minville Marine SXM from nearby Saint-Martin Unfortunately this supplier only alternative is 1,5 inch by 1-inch and according to the Fisheries Regulations BES 10-10-2010 mesh size cannot be smaller than 1,5-inch.

construction of traps happened in July and August. In total 28 traps were built, fourteen of each design. Both designs were constructed from the same materials using the same techniques and tools.

The basic shape and strength of the trap is provided by a rebar frame. The trap itself was constructed with two different types of vinyl-coated wire mesh. Due to a change in regulations regarding the wire mesh of fish traps, the top and bottom were made from 1,5 by 1,5-inch wire mesh and the sides from 2 by 2-inch wire mesh. In the 2018 data set, the sides of the traps

were also made from 1,5 by 1,5 wire mesh. The sides were attached to the top and bottom parts using galvanized steel hog rings. To match the design fishermen use, electrical wire was used to lock the cage within the rebar frame. Two zinc nodes were attached to the rebar frames to prevent corrosion (Nordic Galvanizers, N.D.). A door was put into the trap to empty the trap from its contents after having soaked for a designated period.

The hinges of the door are made from biodegradable rope to prevent ghost fishing in the case of a lost trap. Ropes were measured with a measuring tape and were attached on the two outer sides of the top of the cage using a clove hitch knot (see Figure 7).

Traps of both designs have the same general dimensions of 153 cm in length, 157 cm in width and 50 cm in height. The main funnel is the same size for both designs as well, with a length of 75 cm and a

Figure 7. Clove hitch knot used to attach the ropes to the trap

15

width of 20 cm. The entrance hole has a length of 18 cm. The door is 20 cm in length and 41 cm in Width. Diameter of the funnels is 15 cm.

A FAD was placed in each trap. As seen in Figure 8, in this study the FAD was represented by orange and blue coloured plastic crates used previously for storing and transporting bottles for retail purposes. In each crate a set of randomly sized rectangular holes have been cut. Originally, and in previous experiments conducted by the SCF, only orange coloured crates have been used. Due to the unavailability of the required amount of 28 of these crates, the choice has been made to use similarly shaped blue crates as well to match the required amount. The choice has been made to use an even amount of each type of crate, thus fourteen of each, divided evenly over each trap design. it was thought important to consider the implications of these different coloured and slightly differently shaped crates for the results. It was hypothesized that the colour difference would not matter since both the colours orange and blue would disappears on the depth on which the traps would be deployed (Chaplin, 2019). However, the slightly different shape and size could influence the results and must be considered. Table 1 emphasis the difference in dimensions between the two FAD’s.

Table 1. L x W x H dimensions of the two FAD’s in centimetres.

Orange FAD (in cm) Blue FAD (in cm)

Length 30 34

Width 45 48

Height 38,5 38

Figure 8. The front and side view of the two FAD colour designs used in the experiment during the summer of 2019. Notice the holes cut into the sides to mimic crevices and allow for lionfish to enter.

16 3.2.1 Data collection

From May 2018 until September 2019 modified lionfish traps have been deployed by the Saba Bank Management Unit and hauled on average about once a week. Between the 15^th of July and the 15^th of August 2019, 28 traps were deployed and hauled on the Saba Bank. It was attempted to deploy one of each trap design, thus four in total, on one line at one specific depth between 60 and 450 feet. This would result in a total of seven lines, of each four traps, over a gradient of increasing depth. This would allow for comparison of trap designs over a depth profile from 60 feet to 450 feet. In practise this method is difficult to apply because of the movement of the vessel during deployment. Graph 1 gives the depth profile on which the two designs have been hauled during the research. The number of traps hauled are not evenly distributed over the fishing depth. Traps have been hauled between a depth of 60 and 450 feet. Most traps were hauled in the range of 151 to 360 feet and less so between 60 and 150 feet and 361 and 450 feet. No traditional traps have been hauled in the depth category of 421-450 feet. Funnel traps have been hauled at every depth category.

Traps were deployed ensuring the top side remains up. Afterwards the rope was quickly uncoiled and continuously fed after the sinking trap. The three buoys on the end of the rope where thrown overboard at the end. Immediately after the deployment of a trap, the time and date of deployment, waypoint, coordinates, trap design, trap number and observers were recorded. In figure 9 a map with the sampling location of the 2019 data set can be found. Figure 10 shows a magnified map of the sampling location with the 7 different lines of traps from 2019. On each line 4 traps were deployed.

Hauling of a trap was done by hand and by a hydraulic winch on separate occasions, depending on the availability of a ship outfitted with such a mechanism. Hauling by hand was done on Queen Beatrix II (SCF’s ship). Hauling with winch was done on Bradley Johnson’s ship, a Saban fisherman. The hauling of a trap by hand is substantially more time and physically intensive than by using a hydraulic winch powered by the vessel’s engine. The technique used to haul a trap by hand is as follows. Once the

Graph 1. Depth distribution of all traditional and funnel traps hauled during the 2019 data collection session. Most traps have been hauled between the depths of 211-240 feet. In total 129 traditional traps and a total of 146 funnel traps have been hauled.

17

buoys of a specific trap were located, a large extra buoy was attached to the rope of the trap using a carabiner or shackle. By accelerating the boat for a short while, the buoy slowly slides back on the rope, bringing the trap up to the surface and allowing for a person to haul the slack of the rope in by hand. When the rope has been pulled in and the buoy has been brought back to the boat, the rope is tightly secured to the cleat and the processes of accelerating starts over again. The buoy moves backward, the engine is reversed, the rope unwounded and pulled in again. After a couple repetitions of this process the trap reached the surface and was hauled on board by two persons. The door was opened, and the catch unloaded in a tray. Lionfish and bycatch were recorded and measured. In total 75 traps were hauled in 2019. In 2018 a total of 183 traps were hauled.

After the measurements were taken and bycatch and lionfish where either thrown overboard or kept for further analysis on shore, the trap would be quickly but thoroughly inspected for defects. These could be missing zinc nodes, loosened cable zip ties, broken ropes for the door, missing tag numbers, and broken hog rings. After emptying the traps, they were deployed again at the same depth.

Risk analysis

Fishing on the Saba Bank is not without risks. The bank is known for its rough seas and rapidly changing currents and weather conditions. One must consider that it might not be possible or responsible to set out at any time and changes have been bade in scheduling fieldwork dates. This may have and has

Figure 9. Location of sampling Figure 10. Magnification of sampling location with the 7 different lines of traps. On each line 4 traps were deployed.

Map of sampling points Map of sampling location

18

had consequences for the research, as the soaking time of seven days which was aimed for may not always have been realized. Another problem which arises due to the Saba Bank’s characteristics is the potential to lose fish traps, either by simply not being able to find the buoys or because the trap has been set adrift by rough underwater currents and may be lost forever and will thus start contributing to ghost fishing. If traps couldn’t be found during a trip, it may have been found again on the next trip likely doubling the soaking time. The depth meter used during fieldwork on the Queen Beatrix II (research vessel) did not always function properly and gave inaccurate estimations. Traps may not always have been deployed on the depth which they were planned for. Most important was to write down the depth accurately each time the trap is hauled and make sure it was accounted for during statistics.

Another contributing factor during the data collection phase was that traps have been stolen by most probably by another fisherman or fishermen. The trap is hauled and emptied and thrown back again.

Evidence for trap stealing is the cable zip tie being cut off from the door or a different style of knotting the electrical wire into the frame to lock the door. Unfortunately, this happened a few times and measures to at least be able to observe evidence have been put in place only during the last phases of data collection. This made it unable to tell exactly how many traps have been stolen and to what extend the data may be flawed. It does not necessarily mean that the stolen trap is hauled empty as new batch of fish may have been caught in the trap in the meantime.

3.2.2 Data analysis trap study

After collection of data, the data was processed into Microsoft Excel 2016 data sheets. The data from the excel sheets was then transferred into IBM SPSS Statistics 26 for statistical testing. Several datasets have been produced. One for the sample session of 2019 and one which has bulked together the data for both 2018 and 2019, one for lionfish size data, one for bycatch species count and size data. The bycatch species count, and composition dataset has been filtered in such a way that all species which comprised less than 1% of the total caught individuals were left out. This way a more significant overview can be given of the catch variety and most important bycatch species. The reason datasets have been divided between 2019 and 2018-2019 was to take in to account any possible differences in results which might be due to the difference in mesh sizes or the colour of the FAD. The other two datasets are combined data from both 2018 and 2019. The variable ‘’depth’’ has been visually binned for all datasets to create categories to allow for better comparisons between frequencies over a depth gradient. Thirteen categories have been created with a split at every 30 feet. The first category

Table 1. Data and variables of the four datasets used in analysis. Top left is the 2019 dataset, top right the combined 2018-2019 dataset, bottom left the bycatch data and bottom right the lionfish size data.

Table 2. Overview of the different datasets and associated variables and their unit of measurement.

19

(shallowest) goes from 60-90 feet and the last category (deepest) goes from 420-450 feet. Table 2 gives an overview of the different variables for all datasets as they were used in analysis in IBM SPSS.

A Shapiro-Wilk test (Shapiro & Wilk, 1965) was used to investigate if the data was normally distributed for the 2019 and 2018-2019 datasets. Because the assumption of normal distribution was not met for the variables Number of lionfish (p = 0,000), Number of bycatch (p = 0,000), and number of bycatch species (p = 0,000) for the 2019 dataset, it was decided to use non-parametric tests to test for any significant differences. All test had an a-priori significance level of 0.05. Similar results came from the 2018-2019 dataset. The assumption of normal distribution was not met either for the variables Number of lionfish (p = 0,000), Number of bycatch (p = 0,000), and number of bycatch species (p = 0,000).

The new data from 2019 was analysed first. The significance of the FAD colour in relation to lionfish and bycatch rate was analysed at the start. Hypothesized was that the different colours would not influence the catch rate of lionfish or bycatch. Because no statistical significant differences could be found using a Mann-Whitney non-parametric test between FAD colour and the number of lionfish, the number of bycatch, and the number of species, it was decided to not take FAD colour into account and to combine the 2018 and 2019 data to get more data points.

The following analysis were carried out with the combined 2018-2019 dataset It is important to note that when this data was analysed the FAD colour was no longer considered and assumed to have no influence on the variables analysed. The decision was made to discard mesh size in the analysis all together and focus purely on the differences between the two to trap designs. If a difference can be found between the mesh size and the amount of lionfish, bycatch or bycatch individuals, it does not tell whether the difference is in fact due to the difference in mesh size. By combining the datasets and eliminating the mesh size variable the n-values increase significantly in size.

Mann-Whitney U non-parametric tests and Kruskal-Wallis H non-parametric tests (Mann & Whitney, 1947; Kruskal & Wallis, 1952) were performed to investigate possible differences between the lionfish catch rates and amount of bycatch for both trap designs and the influence of fishing depth on the amount of lionfish, bycatch and bycatch species caught. Significant results between lionfish catch rate and associated bycatch over trap designs does give a satisfactory answer as to which method works most desirable. Significant results about fishing depths in relation to amount of bycatch species and individuals and lionfish allow for recommendations in combination with the most effective trap design.

As a complementary tool, scoring system has been developed to rank all individual hauls based on their content of lionfish and amount of bycatch. A score from one to ten was given to each trap haul.

Ten being the highest score and one being the lowest. This score is based upon two separate scores, on based on the amount of lionfish and one on the amount of bycatch. A scale was created containing six categories lionfish and 5 categories for bycatch amount per single trap haul (see Table 3). Since the maximum amount of lionfish caught in a single trip in this study was nine, the category ‘’9-10’’ was given the highest score of 5. Since the maximum amount of bycatch in a single trap in the study was 50, the category ‘’41-50’’ was given the lowest score of 1. After each trap haul has been given two separate grades, they were added in to a singular 1-10 grade. In Table 3 an overview of the scoring system is given. For example, if a trap contained nine individual lionfish and eight individual bycatch fish, the trap will get a score of 10. When a trap would have 0 lionfish and 49 individual bycatch fish the score would be 1. A completely empty trap would still get a score of 5.

After the scores were calculated for every trap. A Shapiro-wilk tests was used to test if the variable

20

effectiveness score was normally distributed. The variable effectiveness score however was not normally distributed (P=0.000).

Table 3. Overview of the score categories for assessing the most effective lionfish trapping method.

Number of Lionfish Score Number of bycatches Score

0 0 0-10 5

Because the assumption of normality was not met, a Man-Whitney U nonparametric test was used to test for significant differences between the effectiveness score and the two trap designs. The means of the scores for both trap designs were compared to each other to get a grasp of the effectiveness of both methods. Based upon statistical testing of the different trap designs and their contents, and the results of the scoring system the most efficient trap can be determined.