Laboratory test of tagging methods
Four methods to externally tag green sea urchins with acoustic transmitters (V7 tags from VEMCO/Innovasea) were tested over a three-month lab experiment. Divers collected 600 + urchins in December 2017 from 8 to 12 m near Mont-Joli (Quebec), Canada. Urchins were fed ad libitum (Saccharina latissima and Fucus sp.) at the Maurice-Lamontagne Institute until tagging experiments started on February 6, 2018. The four methods tested were: Dyneema® fishing line (1 – FL1), nickel titanium fishing line (2 – FL2), T-bar tags (3 – TB), and nylon screws (4 – NS) (Fig. 1 and Supplementary Material: Details of tagging methods). The two fishing line methods (FL1 and FL2) consisted of threading a length of fishing line directly through the test to form a loop (sensu [23]). Two types of fishing line were tested: Dyneema® nylon fishing line (0.18 mm diameter) and a more durable but less flexible nickel titanium fishing line (0.254 mm diameter). The urchin’s test was pierced with a 0.69 mm diameter hypodermic needle, first on the aboral side about 1 cm from the mouth in an interambulacral plate to enter the test and then on the oral side to exit the test, again through an interambulacral plate. The fishing line was maintained in place as the needle was withdrawn, the two ends of line knotted together, and the tag fixed to the resulting loop using a combination of electrical tape and glue (Lepage UltraGel super glue). The nickel titanium fishing line was evaluated because of concerns that urchins might graze on the dyneema fishing line, thereby detaching tags over longer temporal scales and at higher urchin densities (Ebert, 1965).
The third method (TB) used commercially available plastic T-bar tags (sensu [24]). A tagging gun for clothing was used to pierce the test and insert a T-bar tag on the aboral side of the urchin about 1 cm from the anus in an interambulacral plate and the printed tag was attached to the stalk of the plastic tag using electrical tape and glue. The fourth method (NS) used a nylon screw cut to a length of 6mm (diameter: 3mm; sensu [25]). A hole slightly smaller in diameter than the gauge of the screw was made on the aboral side of the urchin using a dissection needle and the nylon screw carefully rotated into the hole. The tag was then glued on the head of the screw. To minimize costs, “tags” were printed with a 3D printer to have the same length (18mm), weight (0.7g) and volume in water as V7 tags.
A total of 288 urchins (80 each of FL1, TB and NS and 48 FL2) with test diameters > 35 mm were tagged from February 6–13. Before tagging, test diameter (mm) and condition (righting time sensu [26]) of all urchins were measured. Urchins were randomly distributed among eight separate 1m2 tanks (each with an independent supply of flow-through seawater to a depth of 50cm), with 10 individuals of each of three tagging methods (FL1, TB, and NS), 6 FL2 individuals, and 30 non-tagged individuals in each tank to match densities observed in the field trial (51 + 15 individuals·m-2). Tanks were located inside the aquaculture facility at MLI and seawater was at ambient conditions, drawn from the adjacent intake at 15m; temperature ranged from − 0.1 to 4.3℃ and salinity from 25.0 to 30.2 PSU over the course of the experience. Tanks were examined weekly to evaluate tag loss and mortality, and dead individuals removed when found.
Urchin health and condition were measured four times during the experiment (February 21, March 21, April 23, and May 22). Test diameter (mm), wet weight (g), righting time, and visible injuries were reported at each time for all individuals still retaining their tag, all urchins that had lost their tags (‘hole’ urchins), and ten control urchins from each tank. Righting time was measured by placing urchins individually, upside down, in a floating chamber large enough to ensure they were unable to touch the sides and measuring the time necessary for the urchin to flip itself back over. Trials were limited to 15 minutes and urchins not righting themselves in that period were assigned this value. Injuries were coded using four categories: 0 – no injury; 1 – small injury with minimal loss of spines and some discolouration; 2 – apparent injury with dark or discoloured plaque or significant loss of spines in an area more than 2 mm in diameter (necrotic, very dark borders); and 3 – very apparent injury with dark or discoloured plaque more than 5 mm in diameter (often with an apparent hole). The experiment finished at the end of May 2018 and all remaining tagged individuals, ‘hole’ urchins, plus 10 control urchins from each tank were dissected to determine gonad wet weight.
Field test of tagging methods
In August 2017, 30 urchins were tagged with V7 acoustic telemetry tags (VEMCO/Innovasea) using two methods: FL1 and TB. Urchins were collected by SCUBA divers near a salmon farm in Doctor’s Cove (southwest New Brunswick, Canada) where an existing acoustic receiver array was deployed (Fig. 2). The urchin population at the site was described by collecting all urchins from 5 quadrats (0.25 m2) from the tagging site.
Urchins (> 45 mm) were tagged on-site in two batches (16 at ‘a’ and 14 at ‘b’) with equal numbers of both tagging methods at each location (Fig. 3). Total manipulation time was limited, with divers replacing the urchins as soon as tagging was completed, and all tagging was done by the same person (KAM). At ‘a’, two markers separated by 280 cm were set up on the bottom and all 16 urchins released at known positions in relation to these markers. An unattached tag was also placed on the bottom at this release location to mimic a lost or detached tag. Two days later, SCUBA divers returned to ‘a’ and measured positions of all located tagged urchins to calculate 2-day net displacement. Divers revisited ‘a’ again on November 8 and recorded tag presence and status (attached or lost) for all tags found.
Data analysis
Laboratory
Tag retention and survival were analyzed with Cox proportional hazards models (right-censored data) using the R packages survival, MASS and nnet [27–29]. The nylon screw tagging method was dropped from the analysis of survival because most urchins lost their tag within the first 10 days, leaving few individuals for which mortality could be observed. Differences between tagging treatments were evaluated using log-rank statistics.
Urchins that lost their tags were easily separated from control urchins due to the presence, placement, and form of the injury resulting from tag loss. These were almost entirely NS urchins (only 3 TB and no FL1 or FL2 urchins lost their tags throughout the entire study, whereas 75 NS tags were lost: Fig. S1). These urchins were included in subsequent analyses as a separate group (‘hole’ urchins).
Condition was log-transformed because of heterogeneity of residuals and analyzed using a linear mixed effects model including three fixed effects (tagging method, days since tagging, and urchin diameter), the two-way interaction between tagging method and days since tagging and two random effects to account for repeated measures on the same individual and for urchins in the same tank (Table 2; Eq. 1). The fixed factor ‘Tag’ had 6 levels: control, FL1, FL2, TB, NS, and ‘hole’.
Diameter was analyzed using a linear mixed effects model, as in Eq. 1, but with (1|Tank) removed from the final model because it accounted for almost zero variance in the response, causing singularity issues during model fitting (Table 2; Eq. 2). One outlier observation was removed because the difference between measures was clearly beyond possibility (increase of > 1 cm in test diameter) and represented a recording error.
Differences in wet weight (final – initial) were analyzed using a linear mixed effects model with a single fixed effect and a random effect for urchins in the same tank (Table 2; Eq. 3). The fixed factor for tagging treatment had only three levels (FL1, FL2 and TB). The NS treatment was dropped from this analysis because only 4 individuals remained tagged at the end of the experiment; control and ‘hole’ treatments were excluded because data on individual urchins could not be calculated as they were not individually identifiable. Gonad wet weight at the end of the experiment was log-transformed because of heterogeneity of residuals and was analyzed using a linear model. The random effect to account for urchins in the same tank again accounted for almost zero variance in the response, causing singularity issues, and was removed from the model (Table 2; Eq. 4). The fixed factor ‘Tag’ had 5 levels: control, FL1, FL2, TB, and ‘hole’; the NS tagging method was excluded because two few urchins remained at the end of the experiment (n = 4) when dissections for gonad weight were done.
Severity of injury for tagged urchins was analyzed using ordinal logistic regression with a 5-level ordered factor response variable and a log-log link using the polr function from the MASS package (Table 2; Eq. 5). The fixed factor ‘Tag’ had 4 levels: FL1, FL2, TB, and NS. Both control and ‘hole’ treatments were dropped from this analysis because each showed no change for all urchins across all time points, causing problems with model fit (‘hole’ urchins always had very apparent injuries and control urchins had no injuries except for 2 individuals with small injuries). Probabilities of injury severity through time were predicted using the fit model.
Model fit was verified graphically in all linear and linear mixed effects models by plotting residuals versus fitted values, qqplots, histograms of residuals, and plots of residuals against all factors included in the model. Multicollinearity was verified by calculating variance inflation factors (< 4). Type II Wald F-tests with degrees of freedom calculated using the Kenward-Rogers method were used to evaluate significance of interactions and main effects because of the unbalanced nature of the experimental design due to tag losses and mortalities (see above), except in the analysis of injury severity, where Log-likelihood ratio Chi-squared type II tests were used. Tukey HSD post-hoc comparisons were used, where appropriate, to compare tagging treatments (function emmeans; [30]).
Table 2
Equations used in statistical analyses; see text for details.
Laboratory |
\(\text{log}\left(Condition\right)\tilde Tag+ Days since tagging + diameter + Tag:Days + \left(1\right|ID)+ (1\left|Tank\right)\) | Eqn 1 |
\(Diameter \tilde Tag+ Days since tagging + diameter + Tag:Days + \left(1\right|ID)\) | Eqn 2 |
\((Final-Intial weight) \tilde Tag+ \left(1\right|Tank)\) | Eqn 3 |
\(log\left(Gonad wet weight\right) \tilde Tag+ diameter\) | Eqn 4 |
\(Injury severity \tilde Tag+ Days since tagging + Tag:Days since tagging\) | Eqn 5 |
Field |
\(Net displacement \tilde Tag\) | Eqn 6 |
\(75\% MCP area \tilde Tag+month\) | Eqn 7 |
Field
Field – diver observations
Variation in observed net displacement explained by tagging method was analyzed using a linear model with a single fixed factor (tagging method, with two levels; Table 2; Eq. 6).
Field – acoustic telemetry
To evaluate if movement behaviour differed between tagging treatments, we calculated Minimum Convex Polygon (MCP) estimates of area occupied per month for each urchin; detections were first filtered by HPE < 200 to eliminate several extremely high-error positions and urchin-month combinations with fewer than 5 detections were discarded. Because HPE values for this array were inconsistently related to error in distance due to interference and reflections from the aquaculture gear in the water, the pattern of increase in MCP area as a function of the percentage of detections included was used to choose a cut-off percentage of points to discount high-error detections but capture movement behaviour (Fig. S2). A threshold value of 75% was identified that excluded the steep increase in area associated with including the few points far from the majority and this was assumed to represent an area with a high probability of actual occupation. 75% MCPs were then calculated for each urchin and the detached reference tag for every month. Due to the overall low numbers of detections at release point ‘b’ and the fact that the receiver array was removed from the water in mid-November, only urchins from release point ‘a’ and only estimates for September and October were included in the analysis of MCP area. 75% MCP area was analyzed using a linear model to test for the effect of tagging method, month and their interaction. The interaction was not significant and was dropped from the final model (Table 2; Eq. 7). Model fit and multicollinearity were verified as above.