Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You and Your Team.

Learn More →

Under pressure: comparing in situ and boat tagging methods using time-to-event analyses

Under pressure: comparing in situ and boat tagging methods using time-to-event analyses Background: With the increase in telemetry studies over the past decade, improving understanding of how differ - ent tagging methods influence the probability of presence in a receiver array is important in maximizing the result - ing data. Disappearance from the array may be due to mortality from surgery complications, tag loss, predation, or emigration. Internally tagging fish on a boat can cause barotrauma injuries, increased stress from prolonged handling times, or predation after fish have been released back into the water. Conducting in situ internal acoustic tagging at depth of capture removes barotrauma stresses and simplifies the release method, which may improve fish survival and decrease risk of disappearance from the array. In this study, we used 8 years of acoustic tagging data to determine if the tagging method (in situ versus on the boat) influenced the likelihood of a fish being detected 4 and 6 days after the tagging event. Results: At 6 days after tagging, Kaplan–Meier survival curves revealed that the probability of presence for fish tagged on the boat was 66% compared to 90% for fish tagged in situ. Tagging method was the only variable to sig- nificantly affect probability of presence based on Cox proportional hazards models, with fish tagged in situ ~ 75% less likely to disappear from the array compared to fish tagged on the boat at both 4 and 6 days after tagging. Examining tagging methods separately, handling time only marginally influenced probability of presence of boat-tagged fish and no variables had a significant effect on probability of presence of in situ tagged fish. Conclusions: In this study, tagging method was the only variable to significantly affect the probability of presence for internally tagged fish. Other factors had little to no influence, but correlation of variables limited what factors could be included in the models. Implanting internal acoustic tags in situ is not a practical method for every species and for every environment, but given the increased probability of presence demonstrated here, we strongly suggest it be considered where applicable. Keywords: In situ tagging, Acoustic telemetry, Time-to-event analysis Background wide variety of biological and ecological questions. Acoustic telemetry has become a widely accepted For example, acoustic telemetry data have provided method for collecting animal movement data in answers to broad ecological questions, such as reveal- the marine environment [16]. Data collected from ing previously unknown migration patterns for cer- acoustically tagged fish can be used to investigate a tain species [12, 28] and have also been used to answer more localized questions, such as patterns of habitat use in a specific location [15, 18]. Although acoustic *Correspondence: jessica.keller@myfwc.com 1 telemetry studies have been rapidly increasing in num- Florida Fish and Wildlife Conservation Commission, Fish and Wildlife Research Institute, 2796 Overseas Highway, Suite 119, Marathon, FL ber over the past 20 years [6], there are far fewer stud- 33050, USA ies that examine the influence of tagging methodology Full list of author information is available at the end of the article © The Author(s) 2021. This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creat iveco mmons .org/publi cdoma in/ zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Keller et al. Anim Biotelemetry (2021) 9:5 Page 2 of 15 on acoustic telemetry results [9]. This information is minimize the risk of these events by shortening the sur- needed because the physical act of tagging a fish can gery recovery period and reducing the risk of both preda- affect the outcome of the tagging event, potentially tion and emigration by releasing the fish directly into its change the behavior of that fish, and even influence the natural sheltering habitat. interpretation of results. We suggest consideration of conducting in  situ inter- In traditional mark/recapture studies, external tags are nal acoustic tagging of fish at depth of capture to remove inserted into the dorsal musculature. Traditional mark/ barotrauma stressors and simplify the release method. recapture studies are a cost-effective method of tagging Tagging fish in situ with acoustic transmitters at the same animals, but these studies may provide only limited spa- depth as capture and release has become more popular tial information, require the recapture of tagged fish, and as studies have shown the negative effects of barotrauma often have low success rates [25]. To overcome some of on fish survival [31, 32]. Externally tagging fish in situ has these limitations, researchers use more advanced tags, been reported for multiple studies [1, 2, 34], and inter- such as acoustic tags. While externally attaching acous- nally tagging fish in situ has become more common. For tic tags has been accomplished through a wide array of example, Starr et al. [38] captured deep-sea green spotted techniques, these tags can be quickly shed, causing a rockfish ( Sebastes chlorostictus) at depths between 100 truncation of data collection [17]. An alternative method and 200  m, brought them up to 20  m where they were of acoustic tagging that increases tag retention is inter- internally tagged, then lowered them back down to depth nally inserting the tag, usually through surgery [45]. and released them. Lindholm et  al. [22] used saturation However, internally tagging a fish on a boat may increase diving from the Aquarius underwater laboratory in the the chances of fish mortality post-release due to stress Florida Keys to internally tag fish, and Tuohy et  al. [43] from barotrauma, changes in water and air temperature, used closed circuit rebreathers. Recently, Feeley et  al. prolonged fish handling, or the amount of time the fish [12], McCallister et  al. [26], Bryan et  al. [4], and Keller spends at the surface [46]. Even with survival from the et al. [18] all describe internally tagging reef fish in situ. tagging surgery, there is still the potential for physical While there has been comparison of discard mortality trauma (bruising, bleeding, or acute damage) or physi- between in  situ and boat-tagged fish using external tags ological disturbance (changes affecting gasses, blood, and [31], to our knowledge there has been no study directly pH levels) associated with stress from capture or mishan- comparing the results of internally tagging fish on a boat dling [13, 36, 47]. versus in situ. In this study, we use 8 years of acoustic tag- Additionally, newly tagged fish may be more suscepti - ging data and our personal experiences with surgically ble to predation, especially if the fish must travel through implanting transmitters in fish to examine (1) the influence the water column before reaching a protective habi- of tagging method (in situ versus on the boat) on detecta- tat [27]. Acoustic tagging release method experiments bility of a fish post-surgery and (2) the effects of other tag - have demonstrated that fish descender devices, which ging variables (e.g., handling time and surgeon experience) return fish to depth, can increase fish survivorship [3], that might influence detection probability after release. but devices that both return a fish to depth and protect against predators (e.g., cages) are best for increased fish Methods survivorship [27, 46]. However, a review of barotrauma Fish tagging treatment (venting and descending devices) in catch- Fourteen species of groupers (Serranidae) and snappers and-release studies revealed inconsistency about which (Lutjanidae) were internally tagged with acoustic trans- method is most beneficial in reducing or treating baro - mitters (tag life: 417–1825  days) from 2008 to 2016 by trauma, likely because of differences in environmental or Florida Fish and Wildlife Conservation Commission’s Fish species-specific physiological variables and differences in and Wildlife Research Institute (FWRI) staff in the Florida assessment methods [11]. Keys and Dry Tortugas, Florida (Fig. 1). During the 8 years All tagging methods have the potential to cause health of tagging examined in this study, fish were captured using issues for the fish, so decreasing the risk of health issues either hook and line or underwater baited traps. Hook- should be a priority when planning a tagging project. In and-line captured fish were brought to the boat where the case of acoustic telemetry, it is often difficult to deter - they were surgically implanted with acoustic transmitters mine the fate of fish when they are no longer detected. and returned to depth by divers. Fish caught in underwa- While effects of the tagging procedure can cause mortal - ter traps (times between checking and rebaiting traps were ity through physical trauma, the fish may also experience approximately 1–12 h) had their tags inserted in situ by a mortality via predation or fishing pressure, lose the trans - team of divers. A detailed description of boat and in  situ mitter, or immediately emigrate from the tagging area. A tagging and acoustic array designs can be found in Feeley tagging procedure that minimizes stress to fish may also et al. [12], Herbig et al. [15], and Keller et al. [18]. Keller  et al. Anim Biotelemetry (2021) 9:5 Page 3 of 15 Fig. 1 Fish tagging locations and acoustic receiver arrays in the Lower Florida Keys and Dry Tortugas. Fish and Wildlife Research Institute receivers are represented as circles and Integrated Tracking of Aquatic Animals in the Gulf of Mexico receivers are represented as triangles. Bathymetry data in the subplots shows depth as a gradient from light (shallower) to dark (deeper) blue (0–132 m). Receiver location depths ranged from 3 to 43 m. Receivers depicted on this map were deployed during different times from 2008 to 2018, depending on the study. Details regarding receiver timing can be found in [12, 15, 18]. Bathymetry data is a part of the Florida Keys National Marine Sanctuary Digital Atlas [40] Tagging variables The surgical procedures during boat and in  situ tag - During both tagging methods, the following tagging vari- ging were similar with the exception that anesthesia ables were recorded: species, total length of fish (TL), (AQUI-S, 50% isoeugenol; aqui-s.com) was not used depth of capture, tagging location, handling time, tagging in  situ. We noticed that fish were calm once they were time, and name of the person performing the surgery. turned ventral side up and their behavior was simi- Tagging time was the total time of the tag implantation lar to that of fish that had been anesthetized. Tonic surgery, while handling time was either the total time the immobility is commonly used as an alternative to fish was on the boat or total time the fish was handled chemical-based anesthesia for surgical implantation in underwater. Handling times on the boat included the elasmobranchs [19, 37] and after a few initial efforts, we time it took for fish to show effects of anesthesia before realized that use of chemical-based anesthesia for tag- surgery and time for the fish to regain its equilibrium ging fish in situ was not necessary or beneficial and did and display other signs of recovery before being released not allow a quick and immediate release of the fish after by divers. “Fight” times for fish caught via hook and surgery was completed. Acoustic tagging procedures in line were not recorded, so analyses examined only the this study were approved by Florida Fish and Wildlife amount of time the fish was on the boat. Tagging loca - Conservation Commission’s Fish and Wildlife Research tion was the general location of the tagging, either the Institute and every effort was made to follow accepted Lower Florida Keys or the Dry Tortugas (Fig. 1), and was standards and ensure the ethical treatment of captured included to account for any environmental differences fish. Keller et al. Anim Biotelemetry (2021) 9:5 Page 4 of 15 between locations. Habitat types within the two tagging only one detection were excluded because a single detec- locations were similar and contained low- to mid-relief tion could be caused by transmission  collisions or noise continuous coral reefs and patch reefs. and is not considered reliable [35]. Detections from fish The family name of fish, either Serranidae or Lutjani - that were detected only on one receiver were examined dae (n = 2), was also included as a covariate to examine to determine whether detections were the result of nor- differences in detectability based on physiological differ - mal fish activity or from a tag loss or mortality event. ences between taxonomic groups while avoiding overpa- Consistent detections regardless of hour or day or a pat- rameterization of the models because of the high number tern of higher detections during a particular diel period of species (n = 14). Depth of capture was also recorded, could be due to a tag sitting on the seafloor with changes though for fish tagged on the boat, the depth recorded in environmental conditions throughout the day influ - was that of the seafloor as it was not possible to deter - encing transmission success. If detections were suspected mine the exact depth in the water column the fish was to be due to a tag sitting on the seafloor, those detections first caught on hook and line. Total length of fish at time were removed. of capture was transformed into a two-factor variable called “fish size” to compare small and large fish without Time‑to‑event analysis the confounding factor of fish family. Mean total lengths We used time-to-event analysis (also called survival were determined separately for each family and used as analysis) to compare the risk of tagged fish disappearing a threshold value. Fish were designated as “small” if their from the receiver array between the two tagging meth- length was below this value for the respective family and ods. The Cox proportional hazards model is a common “large” if their length was equal to or greater than this analytical method for assessing what variables cause the value. most risk to survival in discard mortality and delayed The variable month was used to examine whether mortality studies [3, 7, 33]. We used the Cox proportional time of year influenced the risk of disappearance. Water hazards model to examine which tagging variables most temperature varies throughout the year and could affect influenced the risk of fish disappearing from the receiver a fish’s health during surgery and after release. Fish array. In this study, acoustic telemetry data could only behavior (e.g., courtship/spawning behaviors) also var- provide presence/absence data, making it  difficult to ies throughout the year and between families, which can determine mortality (i.e., an end of detections could be influence the likelihood of a fish emigrating from the due to emigration, predation, harvest, etc.). As presence/ area. absence data is unable to determine mortality, time-to- To determine if the tagging experience of the surgeon event analyses examined the risk of fish disappearing had an influence on the risk of disappearance, a vari - from the receiver array for the final time (i.e., the date of able called “surgery count” was created. This variable last detection) rather than risk of mortality. was determined by ordering the tagging data by date and Before covariates were included in the model, they named surgeon (regardless of tagging method) for all fish were tested for covariance with other variables via a cor- tagged in this study that had surgeon name recorded. relation matrix and Pearson’s test of correlation. All cat- This information was used to assign each fish a surgery egorical variables were two-factored and were changed count number based on how many surgeries the surgeon to binary outputs to allow for point-biserial correlation. had previously completed. For example, surgery count Variables that had statistically significant correlation with had a value of one if it was the first fish the surgeon had each other (p < 0.05) were not included in the same haz- ever tagged or a value of 17 if it was the 17th fish a sur - ards model. geon had tagged. The data included in this study includes The influence of tagging procedure variables on the risk all, or nearly all, of each surgeon’s internal tagging experi- of disappearance from the array was examined at two dif- ence until the end of the study period or until the person ferent time periods, 4 and 6  days after surgery. The two left the agency. A few surgeons had some prior experi- time periods were chosen based on literature suggesting ence internally tagging fish, but were unable to accurately 4–6  days post-surgery to be a time frame when events quantify their experience, so all surgery counts in this (mortality, tag loss, or emigration) are most likely due study started with a value of 1. to tagging artifacts [27, 42, 43]. Any shorter time frame may miss delayed effects from tagging and longer time Data validation frames increase the chance of mistaking behavioral events Before inclusion in analyses, fish that did not have for those related to the tagging procedure. We decided complete tagging procedure data were excluded from to examine risk of disappearance at both 4 and 6  days the dataset. Detection data was then validated for fish after tagging to determine if there were any differences included in analyses. To validate detection data, fish with between these two time frames. For each s fi h, the last day Keller  et al. Anim Biotelemetry (2021) 9:5 Page 5 of 15 of its detection was used to calculate the number of days disappearance as the value of the covariate increases by it was present before it disappeared from the array. For 1 unit. A hazard ratio greater than 1 is the proportional example, if a fish was last detected 5 days after the tagging increase in risk of a disappearance as the value of the procedure, it was marked as having an event at day 5. covariate increases by 1 unit. The farther away a hazard If a fish was detected past the 4- or 6-day mark, the ratio is from 1, the greater the influence of that covariate. number of days it was considered present was truncated A hazard ratio near 1 means that covariate has a marginal to 4 or 6 days, depending on which time frame was used effect on the risk of disappearance (and thus the prob - for analysis. For example, if a fish was not detected on ability of presence)  while a hazard ratio at 1, or a confi - day 6, but was detected more than once on day 50, it dence interval that crosses 1, has no effect on the risk of was considered alive and present on day 6. Long gaps in disappearance [20]. detections may have occurred as fish often left receiver Independence between residuals and time is an coverage area before returning, but all fish included in assumption of the proportional hazards model [20] and this study were regularly detected and did not display was checked in all models by testing the correlation suspicious data such as getting detected the first few days between scaled Schoenfeld residuals and time using the after tagging and only having one subsequent detection a “cox.zph” function in the “survival” package. A statisti- year later. cally significant correlation of Schoenfeld residuals and To increase the chances of determining the most accu- time rejects the null hypothesis and violates the assump- rate activity level of a tagged fish, detection data from tion of the proportional hazards model. acoustic receivers deployed throughout the Florida Keys during 2008–2018 (not just the receivers in the array Results where the fish was tagged), were included to determine Of the fish tagged from 2008 to 2016, 194 were internally a fish’s status. Detection data came from receivers owned tagged grouper and snapper. Fish with incomplete tag- by FWRI and the Integrated Tracking of Aquatic Animals ging information (n = 87) were excluded and two addi- of the Gulf of Mexico (iTAG) network ([24]; Fig. 1). tional fish were excluded from analyses because they had To visually compare the probability of presence and only a single detection. Data from fish detected at only number of disappearances between the two tagging one location were examined, but no fish had continu - methods, Kaplan–Meier survival curves were created ous, consistent detection patterns suspected of result- using the “survival” package [41] in R [29]. To examine ing from a discarded tag. Receivers in both the Lower what factors were most influential on risk of disappear - Keys and Dry Tortugas arrays were often spaced farther ance, Cox proportional hazards models were calculated apart to try to maximize coverage area and therefore at both 4 and 6 days after tagging also using the “survival” most receivers did not have overlapping coverage. Since package. Model selection started with the full model most fish detected at a single location were groupers or using all covariates that were not correlated with tag- small snappers that were not expected to make long-dis- ging method (the main variable of interest) or each other. tance movements, smaller movements could have been The best-fitting model was chosen through a backwards missed because of spacing between receivers. Therefore, stepwise search, starting with the full model, and model since examined data did not display suspicious patterns, selection was  based on lowest Akaike information crite- no fish detected at only a single location was excluded rion with a correction for small sample size (AICc) using from analyses. In total, 105 fish from 14 different species the “MASS” package [44]. Beyond the initial Cox propor- were included in analyses (35 boat tagged and 70 in  situ tional hazards model examining the influence of tagging tagged; Table 1). method on risk of disappearance, two secondary hazards The mean total length was 57.0 ± 1.4  cm (mean ± SE) models were performed to assess which variables influ - for Lutjanidae and 64.6 ± 1.7  cm for Serranidae. These enced risk of disappearance for (1) fish tagged on the values were used to determine whether the variable of boat and (2) fish tagged in situ. “fish size” was “small” or “large”. Mean handling time and Hazard ratios from the hazards model provide infor- depth of capture varied between tagging method, but mation on the covariates’ effects on risk of disappearance. mean tagging time was similar (Table 2). Hazard ratios above a value of 1 indicate an increase in In situ tagged fish had a lower probability of disappear - risk of disappearance compared to a reference point, ing from the array compared to fish tagged on the boat, whereas a value below 1 indicates a reduction in risk. For based on Kaplan–Meier survival curves (Fig. 2). Six days categorical covariates, a hazard ratio less than 1 is the after tagging, the probability of presence of fish tagged on proportional reduction in risk compared to the reference the boat was only 66% while the probability of presence level and vice versa. For continuous covariates, a hazard of fish tagged in  situ was 90%. The null hypothesis that ratio less than 1 is the proportional reduction in risk of a there was no difference between the two methods in the Keller et al. Anim Biotelemetry (2021) 9:5 Page 6 of 15 Table 1 Number of  fish in  time-to-event analyses + Boat + In situ Tagging Method by species, tagging method, and size 1.00 Tagging method and fish species Total fish Small Large + 0.75 Boat tagged 35 19 16 0.50 Black grouper, Mycteroperca bonaci 8 5 3 Gray snapper, Lutjanus griseus 1 1 – 0.25 p = 0.002 Mutton snapper, Lutjanus analis 14 2 12 0.00 Red grouper, Epinephelus morio 1 1 – Days after tagging Yellowfin grouper, Mycteroperca venenosa 1 – 1 Number at risk: n (%) Yellowtail snapper, Ocyurus chrysurus 10 10 – Boat 35 (100)27 (77) 25 (71) 23 (66) In situ tagged 70 28 42 In situ 70 (100)66 (94) 65 (93) 63 (90) Black grouper, Mycteroperca bonaci 17 6 11 Days after tagging Dog snapper, Lutjanus jocu 6 – 6 Fig. 2 Kaplan–Meier survival curves showing probability of presence for fish tagged in situ and on the boat from day of tagging to 6 days Gag, Mycteroperca microlepis 2 – 2 afterward. Shaded areas indicate 95% confidence interval for the Gray snapper, Lutjanus griseus 8 8 – survival curves. A risk table shows the number of individual fish and Mutton snapper, Lutjanus analis 13 – 13 percentage of total fish still detected in 2-day bins for both tagging Nassau grouper, Epinephelus striatus 8 6 2 methods Red Hind, Epinephelus guttatus 1 1 – Rock Hind, Epinephelus adscensionis 1 1 – Scamp, Mycteroperca phenax 4 3 1 Schoolmaster, Lutjanus apodus 1 1 – Yellowfin grouper, Mycteroperca venenosa 8 1 7 Yellowmouth grouper, Mycteroperca inter- 1 1 – stitialis Total 105 47 58 Table 2 Mean values ± SE for handling time, tagging time, and depth between boat and in situ tagging methods Handling time Tagging time (min) Depth (m) (min) Boat tagged 72.17 ± 6.02 6.37 ± 0.35 13.29 ± 2.14 In situ tagged 9.30 ± 0.31 6.96 ± 0.27 24.61 ± 0.94 probability of presence from day of tagging to 6 days after tagging was rejected by a log-rank test (p = 0.002). Fish size, tagging time, and surgery count were the Fig. 3 Pearson correlation coefficients for all covariates for all fish only variables not significantly correlated with tagging tagged. Size and color of dots indicate strength and direction of method (the main variable of interest) or each other correlation. Boxes without dots indicate no statistically significant correlation (p > 0.05). Boxes with a thick black outline indicate the and therefore were the only variables included in the variables included in the full model initial Cox proportional hazards models (Fig.  3). Vari- ables that were significantly correlated with tagging method or with each other could not be included in the 6-day mark, tagging method was the only variable in hazards model without compromising model results. the best-fitting model (Table 3 ). Because of correlation among variables, model selec- Hazards model results for the best-fitting models tion using backward stepwise selection started with a indicate that fish internally tagged in  situ were 76% less full model including only tagging method, tagging time, likely (one minus the hazard ratio) to disappear from surgery count,  and fish size. Backward stepwise model the array 4  days after tagging and 75% less likely to dis- selection found the best-fitting hazards model (based appear 6  days after tagging compared to fish tagged on on lowest AICc) at the 4-day mark to include tagging the boat (Table  4). Tagging time was also included in method and tagging time as covariates, while at the Tagging Method Probability of Presence Keller  et al. Anim Biotelemetry (2021) 9:5 Page 7 of 15 Table 3 Cox proportional hazards models with AICc stepwise model ranking for both 4 and 6 days after tagging Days after tagging Hazards model AICc ΔAICc χ df p 4 Method + tagging time 131.19 0.00 11.11 2 0.004 Method 131.47 0.28 8.98 1 0.003 Method + tagging time + surgery count 131.72 1.53 11.92 3 0.008 Method + tagging time + fish size + surgery count 134.83 3.64 11.12 3 0.010 Null 137.52 6.33 6 Method 166.68 0.00 9.78 1 0.002 Method + tagging time 167.58 0.90 10.82 2 0.004 Method + tagging time + fish size 169.68 3.00 10.85 3 0.010 Method + tagging time + fish size + surgery count 171.83 5.15 10.87 4 0.030 Null 173.38 6.70 All models include probability of presence as the dependent variable and list independent variables included in each individual model. All fish were included in these models (n = 105). p-values are from the log-rank test Table 4 Cox proportional hazards models results for the best-fitting model at 4 and 6 days after tagging Days after tagging Factor Coefficient SE Hazard ratio 95% CI p 4 Tagging method: boat Reference – – – – Tagging method: in situ − 1.43 0.55 0.24 0.08–0.70 0.009* Tagging time − 0.19 0.12 0.83 0.65–1.05 0.120 6 Tagging method: boat Reference – – – – Tagging method: in situ − 1.38 0.48 0.25 0.10–0.64 0.004* For the categorical covariates, the hazard ratio is the proportion of risk compared to the reference covariate. For the continuous covariates, the hazard ratio is the proportion of risk with an increase of 1 unit (e.g., 1 min). A variable with a hazard ratio close to a value of 1 has a marginal effect and a confidence interval (CI) spanning 1 has no effect on hazard. All models include probability of presence as the dependent variable and list independent variables included in the best-fitting models. p-values are from the log-rank test and asterisks indicate statistical significance the best-fitting 4-day model, though the 95% confidence correlated with month (Fig.  5b). Three full models were interval spanned 1, indicating there was no effect. For all created for in situ tagged fish: (1) one with handling time, covariates in both best-fitting models, the proportional depth, and all other variables not correlated with each hazard assumption was supported by a nonsignificant other, and (2) one with handling time, fish size, and all relationship between scaled Schoenfeld residuals and other variables not correlated with each other, and (3) time. To visually compare each covariate’s risk to fish sur - one with month and all non-correlated variables. vival, hazard ratio forest plots were produced from the Handling time was the only covariate included in the full and best-fitting models, and plots for 4 days after tag - best-fitting model for boat-tagged fish at 4  days after ging are shown as an example (Fig. 4a, b). tagging (Table  5). At 6  days after tagging, both handling Pearson correlation coefficients were also calculated for time and family were included in the best-fitting model. all covariates when splitting the data into fish tagged on Handling time was significant in both best-fitting models; the boat and fish tagged in situ. Handling time and depth however, the hazard ratio was very close to 1, meaning were the two main covariates of interest. However, since the effect of handling time on hazard of disappearance these two covariates were correlated for fish tagged on was marginal (Table  6). A hazard ratio of 2.55 for fam- the boat (Fig. 5a), they could not be included in the same ily Serranidae in the 6-day model indicated an increase hazards model. As a result of these correlations, two full in risk compared to Lutjanidae. However, this covariate models were created for fish tagged on the boat: (1) one had a wide confidence interval that crossed 1, indicating with depth and all variables not correlated with each no significant effect of hazard and a somewhat unreliable other, and (2) one with handling time and all variables point estimate [20]. For all covariates in both best-fitting not correlated with each other. models, the proportional hazard assumption was sup- For in  situ tagged fish, handling time and depth were ported by a nonsignificant relationship between scaled not correlated with each other, but depth was correlated Schoenfeld residuals and time. with fish size and both depth and handling time were Keller et al. Anim Biotelemetry (2021) 9:5 Page 8 of 15 0.10.2 0.51 2 Fig. 4 Forest plots showing covariate hazard ratios at 4 days after tagging for the a full model and b best-fitting Cox proportional hazard model. Hazard ratios and 95% confidence intervals in comparison to reference points are shown for each covariate with p-values listed on the right-hand side of the plots (asterisks indicating statistical significance). Hazard ratio values less than 1 (represented by the dotted line) indicate a reduction in risk of disappearance from the array while values greater than 1 indicate increased risk of disappearance. Covariates with hazard ratios close to 1 have a marginal effect on hazard and confidence intervals crossing 1 indicate no significant effect For in  situ tagged fish, month was the only covariate effect. Additionally, the particularly wide confidence included in the best-fitting model at 4  days after tag - interval for fish size in the best-fitting 6-day model ging and both month and fish size were in the best-fit - indicated the point estimate was unreliable (Table  8). ting model at 6  days after tagging (Table  7). However, For all covariates in both best-fitting models, the pro - the global log-rank tests examining significance of the portional hazard assumption was supported by a non- models overall were nonsignificant in both cases. The significant relationship between scaled Schoenfeld hazard ratio confidence intervals crossed one for all residuals and time. covariates in the best-fitting models, indicating no Keller  et al. Anim Biotelemetry (2021) 9:5 Page 9 of 15 1 1 Depth Depth Depth 0.8 0.8 0.8 Family Family 0.6 Family 0.6 0.6 0.4 0.4 0.4 Fish.Size Fish.Size Fish.Size 0.2 0.2 0.2 Handling.Time Handling.Time Handling.Time 0 0 Location Location Location −0.2 −0.2 −0.2 Month −0.4 Month Month −0.4 −0.4 −0.6 Surgery −0.6 −0.6 Surgery.Count Surgery.Count −0.8 Tagging.Time −0.8 −0.8 Tagging.Time Tagging.Time −1 −1 −1 a b Fig. 5 Pearson correlation coefficients for all covariates for a boat-tagged and b in situ tagged fish. Size and color of dots indicate strength and direction of correlation. Boxes without dots indicate no statistically significant correlation (p > 0.05) Discussion seafloor via divers after they recovered from anesthesia to Using an historic dataset of 8  years of fish tagging by ensure a safe return to the reef. Nevertheless, the tagged Florida Fish and Wildlife Conservation Commission sci- fish may still have suffered from barotrauma, increased entists in the Florida Keys, we examined factors a poste- stress from being caught by hook and line, increased riori that influence probability of presence of internally handling times (which could also increase the impacts tagged fish. At 6  days after tagging, Kaplan–Meier sur - of barotrauma), or long-lasting effects of anesthesia. All vival curves demonstrated that fish tagged in  situ had a these factors could increase risk of boat-tagged fish dis - 25% higher probability of detection within the acoustic appearing from the array. array (i.e., 75% lower probability of disappearing) com- At the start of in  situ tagging, we decided not to use pared to fish tagged on the boat. Tagging method was anesthesia at depth because we discovered that fish the only factor affecting the risk of disappearance when held ventral side up with their eyes covered were just as we examined the initial Cox proportional hazards model calm as anesthetized fish. We also deemed it advanta - assessing the probability of presence of all fish. geous to avoid unnecessarily impairing the tagged fish’s In this study, there were several variables that could not response to predators once the fish was released into be included in the initial hazards model because of cor- sheltering habitat after surgery. Using anesthesia dur- relation, but their inclusion and limited influence in the ing fish surgeries may not always be the best course of subsequent boat-tagged and in  situ tagged models rein- action in every situation, and all potential effects on the forced that the tagging method was the most important fish must be considered [8 , 23]. We believe the benefits factor in risk of disappearance from the array. Acoustic of reduced total handling time for fish tagged in  situ tracking data of in  situ surgically tagged fish from two versus fish tagged on the boat, which usually included other telemetry studies also demonstrate the benefit of a recovery period (mean ± SE: In  situ = 9.30 ± 0.31  min, tagging method, indicated by high survival rates (97– Boat = 72.17 ± 6.02  min), coupled with immediate 100%) post-surgery [4, 43]. The decrease in probability of release into protective habitat and the lack of pressure presence of boat-tagged fish could be due to several fac - and temperature changes, explain the increased prob- tors. As described in Feeley et  al. [12] and Herbig et  al. ability of presence within the acoustic array for in  situ [15], all fish tagged on the boat were returned to the tagged fish. The increased probability of presence of fish Depth Family Fish.Size Handling.Time Location Month Surgery.Count Tagging.Time Depth Depth Family Family Fish.Size Fish.Size Handling.Time Handling.Time Locatio Locationn Month Month Surgery Surgery.Count Tagging.Time Tagging.Time Keller et al. Anim Biotelemetry (2021) 9:5 Page 10 of 15 Table 5 Cox proportional hazards models examining probability of  presence for  fish tagged on  the  boat (n = 35) at both 4 and 6 days after tagging Days Hazards model AICc ΔAICc χ df p after tagging 4 Handling time 66.00 0.00 4.17 1 0.04 Handling time + family 67.08 1.08 5.76 2 0.06 Tagging time 68.13 2.13 2.37 1 0.10 Null 68.26 2.26 Handling time + tagging time + family 68.60 2.60 6.96 3 0.07 Tagging time + surgery count 69.12 3.12 3.79 2 0.20 Tagging time + surgery count + family 70.64 4.64 5.06 3 0.20 Handling time + tagging time + month + family 70.67 4.67 7.30 4 0.10 Tagging time + surgery count + month + family 72.31 6.31 5.84 4 0.20 Handling time + tagging time + surgery count + month + family 72.91 6.91 8.34 5 0.10 Depth + tagging time + surgery count + month + family 74.85 8.85 5.86 5 0.30 Depth + tagging time + surgery count + month + fish size + family 77.79 11.79 5.90 6 0.40 6 Handling time + family 77.51 0.00 8.28 2 0.02 Handling time 77.55 0.04 5.43 1 0.02 Handling time + month + family 78.13 0.62 10.32 3 0.02 Month 79.66 2.15 3.22 1 0.07 Handling time + tagging time + month + family 79.75 2.24 11.22 4 0.02 Family + month 79.85 2.34 5.43 2 0.07 Tagging time + month + family 80.73 3.22 6.96 3 0.07 Null 81.06 3.55 Tagging time + fish size + month + family 81.61 4.10 9.98 4 0.06 Handling time + tagging time + surgery count + month + family 82.28 4.77 11.73 5 0.04 Tagging time + surgery count + month + family 84.22 6.71 9.33 5 0.10 Depth + tagging time + surgery count + month + family 87.07 9.56 9.36 6 0.20 Models are ranked by AICc values. All models include probability of presence as the dependent variable and list independent variables included in each individual model. p-values are from the log-rank test Table 6 Cox proportional hazards models results for  best-fitting models (lowest AICc) of  boat-tagged fish at  both  4 and 6 days after tagging Days after tagging Covariates Coefficient SE hazard ratio 95% CI p 4 Handling time − 0.02 0.01 0.98 0.96–0.99 0.048* 6 Handling time − 0.02 0.01 0.98 0.96–0.99 0.016* Family: Lutjanidae Reference – – – – Family: Serranidae 0.94 0.60 2.55 0.79–8.23 0.116 All models include probability of presence as the dependent variable and list independent variables included in each individual model. p-values are from the log-rank test and asterisks indicate statistical significance tagged in situ compared to fish tagged on the boat indi - that additional handling time would increase the risk cates that any negative effects due to the lack of anes - of a tagged fish having an event. The models, however, thesia are minimal and do not outweigh the benefits of suggested an increase in 1  min of handling time slightly in situ tagging. decreased the risk of disappearance for boat-tagged fish, There were several surprising results presented from though the confidence intervals neared or crossed 1 in the analysis of this data. Some of the variables that were all cases, indicating little to no effect. Handling time may originally believed to impact the probability of detec- not have been very influential in risk of disappearance, tion were either not influential in the hazards models or but it is still surprising that increased time may have had the opposite effect. For example, we hypothesized reduced risk, especially as handling times for fish tagged Keller  et al. Anim Biotelemetry (2021) 9:5 Page 11 of 15 Table 7 Cox proportional hazards models examining fish survival for  fish tagged in  situ (n = 70) at  both  4 and  6  days after tagging Days after tagging Hazards model AICc ΔAICc χ df p 4 Month 40.38 0.00 3.26 1 0.07 Fish size + month 41.49 1.11 4.51 2 0.10 Handling time 41.65 1.27 2.34 1 0.10 Depth + month 41.99 1.61 3.96 2 0.10 Null 42.19 1.81 Handling time + fish size 42.30 1.92 3.78 2 0.20 Handling time + family 42.74 2.36 3.25 2 0.20 Fish size + family + month 43.24 2.86 5.05 3 0.20 Depth + family + month 43.76 3.38 4.46 3 0.20 Handling time + fish size + family 43.79 3.41 4.45 3 0.20 Handling time + depth + family 44.77 4.39 3.41 3 0.30 Tagging time + fish size + family + month 45.21 4.83 5.09 4 0.30 Handling time + fish size + surgery count + family 45.57 5.19 4.52 4 0.30 Tagging time + depth + family + month 45.92 5.54 4.49 4 0.30 Handling time + depth + surgery count + family 46.97 6.59 3.44 4 0.50 Tagging time + fish size + surgery count + family + month 47.53 7.15 5.11 5 0.40 6 Fish size + month 57.43 0.00 5.69 2 0.06 Fish size 57.84 0.41 3.17 1 0.08 Handling time + fish size + surgery count 57.87 0.44 5.41 3 0.10 Fish size + surgery count 58.37 0.94 4.45 2 0.10 Month 58.70 1.27 2.03 1 0.20 Null 58.86 1.43 Fish size + surgery count + month 58.90 1.47 6.48 3 0.09 Depth + month 59.52 2.09 3.5 2 0.20 Handling time + fish size + surgery count + family 59.54 2.11 5.67 4 0.20 Surgery count 60.17 2.74 1.15 1 0.30 Depth + surgery count 60.92 3.49 1.72 2 0.40 Fish size + surgery count + family + month 60.95 3.52 6.68 4 0.20 Depth + surgery count + month 61.02 3.59 3.94 3 0.30 Handling time + depth + surgery count 62.10 4.67 2.32 3 0.50 Tagging time + depth + surgery count + month 62.88 5.45 4.62 4 0.30 Tagging time + fish size + surgery count + family + month 63.19 5.76 7.5 5 0.20 Handling time + depth + surgery count + family 64.08 6.65 2.64 4 0.60 Tagging time + depth + surgery count + family + month 64.97 7.54 4.87 5 0.40 Models are ranked by AICc values. All models include probability of presence as the dependent variable and list independent variables included in each individual model. p-values are from the log-rank test Table 8 Cox proportional hazards models results for best-fitting models of in situ tagged fish Days after tagging Covariates Coefficient SE Hazard ratio 95% CI p‑ value 4 Month 0.59 0.36 1.82 0.89–3.69 0.10 6 Month 0.34 0.24 1.41 0.89–2.24 0.14 Fish size: Small 1.44 0.84 4.23 0.82–21.84 0.08 All models include probability of presence as the dependent variable and list independent variables included in each individual model. p-values are from the log-rank test Keller et al. Anim Biotelemetry (2021) 9:5 Page 12 of 15 on the boat were high compared to other tagging studies of year and water temperature, and thus linked to the [3, 39] and minimizing handling of fish has been reported variable month. Although time of year was not a sig- as one of the most important considerations in implan- nificant in this study it should  also be considered when tation surgeries [30]. For in  situ tagging, handling time deciding how to maximize fish survival and probability of was not included in best-fitting models, but that may be presence when setting up a tagging project. because there was little variance in the handling times of Experience of the surgeon has been shown to influence fish tagged in situ. tag retention and survivorship [5, 10], but this was not Fish size was not a significant factor in the best-fitting seen in our study. Surgery count was not in any best-fitting models. Fish tagged in this study varied between 32 and model, but was a covariate in the third best-fitting model 107 cm TL, a range large enough that differences in prob - for both the initial model at 4  days after tagging and the ability of presence because of size were expected. We in  situ model at 6  days after tagging. Although the effect hypothesized that smaller fish would be at a higher risk was not statistically significant, both models indicated of disappearance because they might not have the same an increase in surgery count decreased risk of fish disap - energy resources to aid in recovery as larger fish or they pearance. Consistent training among surgeons may have may be more susceptible to predation. However, this was reduced the effect of surgery count enough so that other not supported by the models. “Fish size” was included in variables had larger influences. Before the start of a tagging the best-fitting model for in situ tagged fish at 6 days after project, surgeons practiced the tagging procedure with tagging but was not significant, and a large spread in the guidance from an experienced team member, and initial hazard ratio confidence interval (0.82–21.84) represented training was provided by a professional plastic surgeon high uncertainty in the hazard ratio estimate. While there (D. Hawtof, personal communication). Additionally, less are likely species-specific differences (e.g., body type, experienced surgeons were always paired with an expe- stress tolerance) that influence probability of presence of rienced surgeon during the tagging procedure to provide different sized fish, in our study fish size did not signifi - guidance during both boat and in  situ tagging. The train - cantly affect probability of presence post-tagging. Also, ing and oversight throughout the tagging procedure may fish less than 20  cm FL have been shown to have high have helped to reduce the effect of surgeon experience on survival from internal tagging [21], suggesting that sur- risk of disappearance. gery and implantation of appropriately sized tags for the Hazards models were analyzed at both 4 and 6  days body size of the fish may not impair survival, regardless after tagging because there was discrepancy in the lit- of fish size. erature about how long after surgery to examine survival Fish family was also not a significant covariate, related to the tagging procedure. Results between these although differences in tolerance to barotrauma and two time periods were similar (e.g., hazard ratios for tag- handling were seen among species. The high number ging method in the initial model differed by only 0.015), of species in this study (which originally led to overpa- but we felt it was informative to present both cases. rameterization in models) and the unevenness of species Although differences between the results at both time between tagging methods resulted in grouping species by periods were minimal, they are still interesting to note family. There are physiological differences between spe - because they show how the dynamics of fish survival may cies of the same family, and grouping them by family may subtly shift over time. The maximum time after surgery have masked these differences and may help explain why that a fish was still detected was just over 4  years (data family did not affect probability of presence. Although an from this fish was analyzed in [12]), and a possible future in-depth look at the effect of tagging method on species direction of work could include examining longer term was not possible in this study, a planned, paired study differences in detection patterns between different tag - with fewer species may find that fish species influences ging methods, particularly if the goal is to maximize data probability of presence. For any tagging project, factors acquisition. Analyzing probability of presence on a longer such as body type, susceptibility to barotrauma, and time frame may find different factors which are more stress tolerance should be considered when deciding on influential than those found in this study, but considera - the appropriate tagging method. tion should be given to whether disappearances from the Water temperatures and differences between water array past 6 days post-surgery are a result of the tagging and air temperatures vary throughout the year and could procedure or from a natural behavioral event. affect the risk of a fish disappearing from the array, but This study was not originally designed to rigorously month as a proxy for temperature did not have a sig- examine all factors affecting boat and in  situ tagged fish. nificant influence in either boat-tagged or in  situ tagged Therefore, some variables may not have proven as influen - models. Life history traits of the fish families (e.g., the tial as they would in a more systematic study. Depth was time of year that they reproduce) are also linked to time expected to have more of an influence on hazard because Keller  et al. Anim Biotelemetry (2021) 9:5 Page 13 of 15 detrimental effects of barotrauma have been shown to record). However, despite these caveats, we believe that increase with depth of capture [14]. In this study, depth this historic dataset using over 100 internally tagged fish may not have been the best representation of the sever- representing 14 different species surgically tagged by 11 ity of barotrauma. Boat-tagged fish were generally caught different scientists was robust enough to demonstrate using hook and line and were not necessarily caught from that tagging method had an effect on risk of fish disap - the seafloor, which was the depth recorded, and further - pearance. Telemetry studies are expensive to conduct, more, mean depth of capture of boat-tagged fish was both in terms of effort and money, and often outcomes significantly shallower compared to in  situ tagged fish are not known until months later when the data can (Wilcoxon two-sample rank test, p < 0.05). Fish that were be downloaded. Therefore, it is critical to have a better near the surface or mid-water column when caught on understanding of how tagging method influences prob - hook and line would have a greater depth recorded than ability of presence (which can be considered a proxy for where they were physically caught in the water column, fish survivorship). By examining the effects of tagging meaning that they may have exhibited fewer signs of baro- methods, we can inform future telemetry studies, thereby trauma than if they had actually been caught at the depth increasing the usefulness of telemetry data. recorded. Fish tagged in situ were caught via baited traps This study has shown that there were clear benefits to on the seafloor, so the depth of capture was the depth of tagging fish in situ, though due to high correlation among the trap and the actual depth of the fish when it entered variables we are unable to pinpoint the source behind the trap. Additionally, after the start of in  situ tagging, decreased risk of disappearance compared to boat-tagged this method was preferentially chosen over tagging on fish. The lack of barotrauma, shorter handling time, lack the boat at deep sites and only a few fish were captured, of anesthesia, and use of baited traps instead of hook and tagged, and released from the boat in deep water (> 30 m) line may all have played a role in increasing an in  situ and in  situ tagged fish had a higher overall sample size tagged fish’s probability of presence. We understand that compared to boat-tagged fish. The method of tagging this in  situ method may not always be practical. In  situ was based on what was deemed best for survival of the tagging can increase the amount of effort spent to tag fish and logistically made sense for each tagging event. If each fish because of the logistics required and the time boat tagging data had included more fish captured from limitations when tagging at deeper depths. Some target deep water, resulting in no discrepancy in mean depth per species may be located outside recreational diving limits, tagging method, we hypothesize depth would have had which can increase effort and cost for specialized training a larger effect on survival. Additionally, comparing risk and dive gear. Also, in this study we used baited traps to of disappearance between tagging methods where boat- capture fish in situ, but other methods should be consid - tagged fish were either released at the surface or brought ered if the species of interest is unlikely to be captured in back down to depth for release (the method used for all this manner. Furthermore, studies that rely on anglers to boat-tagged fish in this study) would further explore the supply the tagging subjects or focus on species found in effects of depth, barotrauma, and release method. very shallow water may not find in situ tagging feasible or Fight time, location of hook placement on fish, whether worthwhile. However, we suggest that all aspects of the the fish exhibited signs of barotrauma, whether a fish tagging procedure, from tagging method (e.g., on a boat with barotrauma was vented before its return to depth, vs. in  situ) to use of anesthesia, method of capture, and and overall health of the fish at time of release were not method of release be carefully considered for each target often recorded and thus were not included as covariates species to reduce stress to the fish, and maximize success in the hazards models. The amount a time a fish spent of the study. in a trap could influence the fish’s health and probability of presence after tagging, but the time a trap was baited Conclusions does not equal the time a fish was in a trap (a trap may In this study, fish internally tagged in  situ with acous - have been baited and left for 12  h, but a fish may have tic tags were ~ 75% less likely to disappear from the entered the trap at hour 11). Use of cameras for identi- receiver array (i.e., mortality, emigration, or tag loss fying how long fish were in traps and recording specifics event) shortly after tagging compared to fish brought on condition upon release would provide more details on to the surface and tagged on a boat. Other factors had why risk of disappearance was higher in boat-tagged fish. little to no influence on probability of presence after This study demonstrated some of the limitations of a tagging, but correlation of variables limited what could posteriori-designed study including correlated variables be included in hazards models. Despite the limitations and missing data (particularly during the initial stages of due to correlation, fish had a higher probability of being conducting acoustic telemetry research when it was not present and a lower probability of disappearance when known what supplementary data would be beneficial to tagged with the in  situ methods presented. Therefore, Keller et al. Anim Biotelemetry (2021) 9:5 Page 14 of 15 References when designing a telemetry study, serious considera- 1. Atkins JL, Morris JA, Green SJ. In situ tagging technique for fishes pro - tion of tagging method (in situ versus on a boat) and vides insight into growth and movement of invasive lionfish. Ecol Evol. other variables (capture method, release method, use of 2014;4(19):3768–77. 2. Balázs G, Lewarne B, Herczeg G. In situ underwater tagging of aquatic anesthesia, etc.) should be evaluated because they play organisms: a test using the cave-dwelling olm Proteus anguinus. Ann Zool a critical role in the overall health of fish and success Fennici. 2015;52(3):160–6. https ://doi.org/10.5735/086.052.0303. of the study. Implanting internal acoustic tags in situ is 3. Bohaboy EC, Guttridge TL, Hammerschlag N, Van Zinnicq Bergmann MP, Patterson WF III. Application of three-dimensional acoustic telemetry to not a practical method for every species of interest and assess the effects of rapid recompression on reef fish discard mortality. for every environment, but given the increased prob- ICES J Mar Sci. 2019;77(1):83–96. ability of presence demonstrated here, we suggest it be 4. Bryan DR, Feeley MW, Nemeth RS, Pollock C, Ault JS. Home range and spawning migration patterns of queen triggerfish Balistes vetula in considered for fish species where applicable. StCroix US Virgin Islands. Mar Ecol Prog. 2019;616:123–39. https ://doi. org/10.3354/meps1 2944. Acknowledgements 5. Cooke SJ, Graeb BDS, Suski CD, Ostrand KG. Eec ff ts of suture material We thank B Binder, A Collins, M Dancy, D DeMaria, D Eaken, D Hawtof, J Javech, on incision healing, growth and survival of juvenile largemouth bass A Johnson, M McCallister, J Renchen, A Tobin, and S Walters-Burnsed for assis- implanted with miniature radio transmitters: case study of a novice and tance in the field. We also thank S Barbieri for sharing data and Dry Tortugas experienced fish surgeon. J Fish Biol. 2003;62:1366–80. National Park Staff and the crews of the M/V Fort Jefferson, M/V Spree, M/V 6. Crossin GT, Heupel MR, Holbrook CM, Hussey NE, Lowerre-Barbieri SK, Peter Gladding, and NOAA ship Nancy Foster for helping with logistics. Thanks Nguyen VM, Raby GD, Cooke SJ. Acoustic telemetry and fisheries man- to E Ault, K Parsons, C Shea, and two anonymous reviewers for valuable manu- agement. Ecol Appl. 2017;27(4):1031–49. script comments and D Fields and C Ames for assistance with bathymetry 7. Curtis JM, Johnson MW, Diamond SL, Stunz GW. Quantifying delayed data. In memoriam Jeff Simonds, FWC staff (April 24th 1978–February 18th, mortality from barotrauma impairment in discarded red snapper using 2010), and David Hawtof, FWC volunteer (September 5th, 1933–February 5th, acoustic telemetry. Mar Coast Fish. 2015;7:434–49. 2019), who contributed to this study. 8. Damon-Randall K, Bohl R, Bolden S, et al. Atlantic sturgeon research techniques NOAA technical memorandum NMFSNE-215. Woods Hole: Authors’ contributions National Marine Fisheries Service Northeast; 2010. p. 19. JK analyzed the data. JK, DM, JH, MF, and AA contributed to the interpretation 9. Dance MA, Moulton DL, Furey NB, Rooker JR. Does transmitter placement of results and JK, DM, JH, AA, MF, and PB contributed to the writing of the or species affect detection efficiency of tagged animals in biotelemetry manuscript. PB, MF, DM, JK, AA, and JH performed fieldwork. AA, DM, and MF research? Fish Res. 2016;183:80–5. secured funding. All authors read and approved the manuscript. 10. Deters KA, Brown RS, Carter KM, Boyd JW, Eppard MB, Seaburg AG. Perfor- mance assessment of suture type, water temperature, and surgeon skill Funding in juvenile Chinook salmon surgically implanted with acoustic transmit- Research was supported by National Park Service CESU contract ters. Trans Am Fish Soc. 2010;139(3):888–99. H5000065040-J5281060165, USGS State Partnership Program grant 11. Eberts RL, Somers CM. Venting and descending provide equivocal ben- 08ERAG00070, NOAA NMFS Coral Reef Conservation Program efits for catch-and-release survival: study design influences effectiveness grants NA17RJ1226 and NA08OAR4320892, and NOS Agreement # more than barotrauma relief method. N Am J Fish Manag. 2017;37:612– MOA-21010–026/8081. 23. https ://doi.org/10.1080/02755 947.2017.13072 92. 12. Feeley MW, Morley D, Acosta A, Barbera P, Hunt J, Switzer T, Burton Availability of data and materials M. Spawning migration movements of Mutton Snapper in Tortugas, The datasets generated during and/or analyzed during the current study are Florida: spatial dynamics within a marine reserve network. Fish Res. available by request in the FWCDL repository (https ://f5000 6a.eos-intl.net/ 2018;204:209–23. F5000 6A/OPAC/Detai ls/Recor d.aspx?BibCo de=53861 79). 13. Gallagher AJ, Serafy JE, Cooke SJ, Hammerschlag N. Physiological stress response, reflex impairment, and survival of five sympatric shark Ethics approval and consent to participate species following experimental capture and release. Mar Ecol Prog Ser. Acoustic tagging procedures in this study were approved by Florida Fish 2014;496:207–18. and Wildlife Conservation Commission’s Fish and Wildlife Research Insti- 14. Hannah RW, Matteson KM. Behavior of nine species of Pacific rockfish tute and every effort was made to follow accepted standards and ensure after hook-and-line capture, recompression, and release. Trans Am Fish the ethical treatment of captured fish. This research was conducted under Soc. 2007;136:24–33. scientific research permits issued by the Florida Keys National Marine 15. Herbig JL, Keller JA, Morley D, Walter K, Barbera P, Acosta A. Examining Sanctuary (FKNMS-2008-015, FKNMS-2013-040, FKNMS-2013-040-A2, movement patterns of yellowtail snapper, Ocyurus chrysurus, in the Dry FKNMS-2018-047) and the Dry Tortugas National Park (DRTO-2008-SCI0003, Tortugas. Florida Bull Mar Sci. 2019;95(1):45–67. DRTO-2010-SCI-0006). 16. Hussey NE, Kessel ST, Aarestrup K, Cooke SJ, Cowley PD, Fisk AT, et al. Aquatic animal telemetry: a panoramic window into the underwater Consent for publication world. Science. 2015;348(6240):1255642. Not applicable. 17. Jepsen N, Thorstad EB, Havn T, Lucas MC. The use of external electronic tags on fish: an evaluation of tag retention and tagging effects. Anim Competing interests Biotelemetry. 2015;3:49. The authors declare that they have no competing interests. 18. Keller JA, Herbig JL, Morley D, Wile A, Barbera P, Acosta A. Grouper tales: use of acoustic telemetry to evaluate grouper movements at Western Dry Author details Rocks in the Florida Keys. Mar Coast Fish. 2020;12:290–307. Florida Fish and Wildlife Conservation Commission, Fish and Wildlife Research 19. Kessel ST, Hussey NE. Tonic immobility as an anaesthetic for elasmo- Institute, 2796 Overseas Highway, Suite 119, Marathon, FL 33050, USA. branchs during surgical implantation procedures. Can J Fish Aquat Sci. National Park Service, South Florida/Caribbean Inventory and Monitoring 2015;72(9):1287–91. Network, 18001 Old Cutler Rd Suite 419, Palmetto Bay, FL 33157, USA. 20. Kleinbaum DG, Klein M. Survival analysis. New York: Springer; 2012. p. 3. 21. Klinard NV, Halfyard EA, Fisk AT, Stewart TJ, Johnson TB. Eec ff ts of Received: 7 August 2020 Accepted: 5 January 2021 surgically implanted acoustic tags on body condition, growth, and survival in a small, laterally compressed forage fish. Trans Am Fish Soc. 2018;147:749–57. Keller  et al. Anim Biotelemetry (2021) 9:5 Page 15 of 15 22. Lindholm J, Fangman S, Kaufman L, Miller S. In situ tagging and tracking pulse position modulation acoustic telemetry monitoring equipment. of coral reef fishes from the Aquarius Undersea Laboratory. Mar Technol Anim Biotelemetry. 2015;3(1):55. Soc J. 2005;39:68–73. 36. Skomal GB. Evaluating the physiological and physical consequences of 23. Lowerre-Barbieri SK, Villegas-Ríos D, Walters-Burnsed S, Bickford J, Cooper capture on post-release survivorship in large pelagic fishes. Fish Manag W, Muller R, Trotter A. Spawning site selection and contingent behavior in Ecol. 2007;14:81–9. Common Snook, Centropomus undecimal. PLoS ONE. 2014;9(7):e101809. 37. Sloan KA, Bouyoucos IA, Brooks EJ, Sneddon LU. Ethical considerations in 24. Lowerre-Barbieri S, Burnsed S, Bickford J, Boucek R, Staley H. Report on fish research. J Fish Biol. 2019;94:556–77. the Integrated Tracking of Aquatic Animals in the Gulf of Mexico (iTAG) 38. Starr RM, Heine JN, Johnson KA. Techniques for tagging and tracking 2017 workshop: building network capacity. St. Petersburg: Florida Fish deepwater rockfishes. N Am J Fish Manag. 2000;20:597–609. and Wildlife Conservation Commission; 2017. 39. Starr RM, Sala E, Ballesteros E, Zabala M. Spatial dynamics of the Nassau 25. Lucas MC, Baras E. Methods for studying spatial behavior of freshwater grouper Epinephelus striatus in a Caribbean atoll. Mar Ecol Prog Ser. fishes in the natural environment. Fish Fish. 2000;1:283–316. 2007;343:239–49. 26. McCallister M, Renchen J, Binder B, Acosta A. Diel activity patterns and 40. Taylor C, Field D, Vander Pluym J, Viehman S, Ebert E. NCCOS Mapping: movement of invasive lionfish (Pterois volitans/P. miles) in the Florida Keys seafloor mapping products for the Florida Keys National Marine Sanctu- identified using acoustic telemetry. Gulf Caribb Res. 2018;29(1):27–40. ary from 2004-01-01 to 2016-12-31 (NCEI Accession 0170215). NOAA 27. Piraino MN, Szedlmayer ST. Fine-scale movements and home ranges of National Centers for Environmental Information. Dataset. 2018. https :// Red Snapper around artificial reefs in the Northern Gulf of Mexico. Trans doi.org/10.7289/v55b0 0rz. Accessed 22 Nov 2020 Am Fish Soc. 2014;143(4):988–98. 41. Therneau T. A package for survival analysis in R. 2020. R package version 28. Pratt HL, Pratt TC, Morley D, Lowerre-Barbieri S, Collins A, Carrier JC, Hart 3.2-7. https ://CRAN.R-proje ct.org/packa ge=survi val. KH, Whitney NM. Partial migration of the nurse shark Ginglymostoma 42. Topping DT, Szedlmayer ST. Site fidelity, residence time and movements cirratum (Bonnaterre), from the Dry Tortugas Islands. Environ Biol fish of red snapper Lutjanus campechanus estimated with long-term acoustic 2018;101:515–30. monitoring. Mar Ecol Prog Ser. 2011;437:183–200. 29. R Core Team. R: a language and environment for statistical computing. 43. Tuohy E, Nemeth MI, Bejarano I, Schärer MT, Appeldoorn RS. In situ Vienna: R Foundation for Statistical Computing; 2019. https ://www.R- tagging of Nassau Grouper Epinephelus striatus using closed-circuit proje ct.org/. rebreathers at a spawning aggregation in Puerto Rico. Mar Technol Soc J. 30. Rub AMW, Jepson N, Liedtke TL, Moser ML, Weber EPS III. Surgical inser- 2015;49:115–23. tion of transmitters and telemetry methods in fisheries research. Am J Vet 44. Venables WN, Ripley BD. Modern applied statistics with S-Plus. 4th ed. Res. 2014;75(4):402–16. New York: Springer; 2002. 31. Rudershausen PJ, Buckel JA, Hightower JE. Estimating reef fish discard 45. Wagner GN, Cooke S, Brown RS, Deters KA. Surgical implantation tech- mortality using surface and bottom tagging: effects of hook injury niques for electronic tags in fish. Rev Fish Biol Fish. 2011;21:71–81. and barotrauma. Can J Fish Aquat Sci. 2014;71:514–20. https ://doi. 46. Williams LJ, Herbig JL, Szedlmayer ST. A cage release method to improve org/10.1139/cjfas -2013-0337. fish tagging studies. Fish Res. 2015;172:125–9. 32. Rummer JL, Bennett WA. Physiological effects of swim bladder overex - 47. Wilson SM, Raby GD, Burnett NJ, Hinch SG, Cooke SJ. Looking beyond the pansion and catastrophic decompression on Red Snapper. Trans Am Fish mortality of bycatch: sublethal effects of incidental capture on marine Soc. 2005;134:1457–70. animals. Biolog Conserv. 2014;2014(171):61–72. 33. Runde BJ, Rudershausen PJ, Sauls B, Mikles CS, Buckel JA. Low discard survival of gray triggerfish in the southeastern US hook-and-line fishery. Publisher’s Note Fish Res. 2019;219:105313. Springer Nature remains neutral with regard to jurisdictional claims in pub- 34. Sigurdsson T, Thorsteinsson V, Gústafsson L. In situ tagging of deep-sea lished maps and institutional affiliations. redfish: application of an underwater, fish-tagging system. ICES J Mar Sci. 2006;63:523–31. 35. Simpfendorfer CA, Huveneers C, Steckenreuter A, Tattersall K, Hoenner X, Harcourt R, Heupel MR. Ghosts in the data: false detections in VEMCO Re Read ady y to to submit y submit your our re researc search h ? Choose BMC and benefit fr ? Choose BMC and benefit from om: : fast, convenient online submission thorough peer review by experienced researchers in your field rapid publication on acceptance support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year At BMC, research is always in progress. Learn more biomedcentral.com/submissions http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Animal Biotelemetry Springer Journals

Under pressure: comparing in situ and boat tagging methods using time-to-event analyses

Loading next page...
 
/lp/springer-journals/under-pressure-comparing-in-situ-and-boat-tagging-methods-using-time-pyZHKrePM6
Publisher
Springer Journals
Copyright
Copyright © The Author(s) 2021
eISSN
2050-3385
DOI
10.1186/s40317-021-00229-z
Publisher site
See Article on Publisher Site

Abstract

Background: With the increase in telemetry studies over the past decade, improving understanding of how differ - ent tagging methods influence the probability of presence in a receiver array is important in maximizing the result - ing data. Disappearance from the array may be due to mortality from surgery complications, tag loss, predation, or emigration. Internally tagging fish on a boat can cause barotrauma injuries, increased stress from prolonged handling times, or predation after fish have been released back into the water. Conducting in situ internal acoustic tagging at depth of capture removes barotrauma stresses and simplifies the release method, which may improve fish survival and decrease risk of disappearance from the array. In this study, we used 8 years of acoustic tagging data to determine if the tagging method (in situ versus on the boat) influenced the likelihood of a fish being detected 4 and 6 days after the tagging event. Results: At 6 days after tagging, Kaplan–Meier survival curves revealed that the probability of presence for fish tagged on the boat was 66% compared to 90% for fish tagged in situ. Tagging method was the only variable to sig- nificantly affect probability of presence based on Cox proportional hazards models, with fish tagged in situ ~ 75% less likely to disappear from the array compared to fish tagged on the boat at both 4 and 6 days after tagging. Examining tagging methods separately, handling time only marginally influenced probability of presence of boat-tagged fish and no variables had a significant effect on probability of presence of in situ tagged fish. Conclusions: In this study, tagging method was the only variable to significantly affect the probability of presence for internally tagged fish. Other factors had little to no influence, but correlation of variables limited what factors could be included in the models. Implanting internal acoustic tags in situ is not a practical method for every species and for every environment, but given the increased probability of presence demonstrated here, we strongly suggest it be considered where applicable. Keywords: In situ tagging, Acoustic telemetry, Time-to-event analysis Background wide variety of biological and ecological questions. Acoustic telemetry has become a widely accepted For example, acoustic telemetry data have provided method for collecting animal movement data in answers to broad ecological questions, such as reveal- the marine environment [16]. Data collected from ing previously unknown migration patterns for cer- acoustically tagged fish can be used to investigate a tain species [12, 28] and have also been used to answer more localized questions, such as patterns of habitat use in a specific location [15, 18]. Although acoustic *Correspondence: jessica.keller@myfwc.com 1 telemetry studies have been rapidly increasing in num- Florida Fish and Wildlife Conservation Commission, Fish and Wildlife Research Institute, 2796 Overseas Highway, Suite 119, Marathon, FL ber over the past 20 years [6], there are far fewer stud- 33050, USA ies that examine the influence of tagging methodology Full list of author information is available at the end of the article © The Author(s) 2021. This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creat iveco mmons .org/publi cdoma in/ zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Keller et al. Anim Biotelemetry (2021) 9:5 Page 2 of 15 on acoustic telemetry results [9]. This information is minimize the risk of these events by shortening the sur- needed because the physical act of tagging a fish can gery recovery period and reducing the risk of both preda- affect the outcome of the tagging event, potentially tion and emigration by releasing the fish directly into its change the behavior of that fish, and even influence the natural sheltering habitat. interpretation of results. We suggest consideration of conducting in  situ inter- In traditional mark/recapture studies, external tags are nal acoustic tagging of fish at depth of capture to remove inserted into the dorsal musculature. Traditional mark/ barotrauma stressors and simplify the release method. recapture studies are a cost-effective method of tagging Tagging fish in situ with acoustic transmitters at the same animals, but these studies may provide only limited spa- depth as capture and release has become more popular tial information, require the recapture of tagged fish, and as studies have shown the negative effects of barotrauma often have low success rates [25]. To overcome some of on fish survival [31, 32]. Externally tagging fish in situ has these limitations, researchers use more advanced tags, been reported for multiple studies [1, 2, 34], and inter- such as acoustic tags. While externally attaching acous- nally tagging fish in situ has become more common. For tic tags has been accomplished through a wide array of example, Starr et al. [38] captured deep-sea green spotted techniques, these tags can be quickly shed, causing a rockfish ( Sebastes chlorostictus) at depths between 100 truncation of data collection [17]. An alternative method and 200  m, brought them up to 20  m where they were of acoustic tagging that increases tag retention is inter- internally tagged, then lowered them back down to depth nally inserting the tag, usually through surgery [45]. and released them. Lindholm et  al. [22] used saturation However, internally tagging a fish on a boat may increase diving from the Aquarius underwater laboratory in the the chances of fish mortality post-release due to stress Florida Keys to internally tag fish, and Tuohy et  al. [43] from barotrauma, changes in water and air temperature, used closed circuit rebreathers. Recently, Feeley et  al. prolonged fish handling, or the amount of time the fish [12], McCallister et  al. [26], Bryan et  al. [4], and Keller spends at the surface [46]. Even with survival from the et al. [18] all describe internally tagging reef fish in situ. tagging surgery, there is still the potential for physical While there has been comparison of discard mortality trauma (bruising, bleeding, or acute damage) or physi- between in  situ and boat-tagged fish using external tags ological disturbance (changes affecting gasses, blood, and [31], to our knowledge there has been no study directly pH levels) associated with stress from capture or mishan- comparing the results of internally tagging fish on a boat dling [13, 36, 47]. versus in situ. In this study, we use 8 years of acoustic tag- Additionally, newly tagged fish may be more suscepti - ging data and our personal experiences with surgically ble to predation, especially if the fish must travel through implanting transmitters in fish to examine (1) the influence the water column before reaching a protective habi- of tagging method (in situ versus on the boat) on detecta- tat [27]. Acoustic tagging release method experiments bility of a fish post-surgery and (2) the effects of other tag - have demonstrated that fish descender devices, which ging variables (e.g., handling time and surgeon experience) return fish to depth, can increase fish survivorship [3], that might influence detection probability after release. but devices that both return a fish to depth and protect against predators (e.g., cages) are best for increased fish Methods survivorship [27, 46]. However, a review of barotrauma Fish tagging treatment (venting and descending devices) in catch- Fourteen species of groupers (Serranidae) and snappers and-release studies revealed inconsistency about which (Lutjanidae) were internally tagged with acoustic trans- method is most beneficial in reducing or treating baro - mitters (tag life: 417–1825  days) from 2008 to 2016 by trauma, likely because of differences in environmental or Florida Fish and Wildlife Conservation Commission’s Fish species-specific physiological variables and differences in and Wildlife Research Institute (FWRI) staff in the Florida assessment methods [11]. Keys and Dry Tortugas, Florida (Fig. 1). During the 8 years All tagging methods have the potential to cause health of tagging examined in this study, fish were captured using issues for the fish, so decreasing the risk of health issues either hook and line or underwater baited traps. Hook- should be a priority when planning a tagging project. In and-line captured fish were brought to the boat where the case of acoustic telemetry, it is often difficult to deter - they were surgically implanted with acoustic transmitters mine the fate of fish when they are no longer detected. and returned to depth by divers. Fish caught in underwa- While effects of the tagging procedure can cause mortal - ter traps (times between checking and rebaiting traps were ity through physical trauma, the fish may also experience approximately 1–12 h) had their tags inserted in situ by a mortality via predation or fishing pressure, lose the trans - team of divers. A detailed description of boat and in  situ mitter, or immediately emigrate from the tagging area. A tagging and acoustic array designs can be found in Feeley tagging procedure that minimizes stress to fish may also et al. [12], Herbig et al. [15], and Keller et al. [18]. Keller  et al. Anim Biotelemetry (2021) 9:5 Page 3 of 15 Fig. 1 Fish tagging locations and acoustic receiver arrays in the Lower Florida Keys and Dry Tortugas. Fish and Wildlife Research Institute receivers are represented as circles and Integrated Tracking of Aquatic Animals in the Gulf of Mexico receivers are represented as triangles. Bathymetry data in the subplots shows depth as a gradient from light (shallower) to dark (deeper) blue (0–132 m). Receiver location depths ranged from 3 to 43 m. Receivers depicted on this map were deployed during different times from 2008 to 2018, depending on the study. Details regarding receiver timing can be found in [12, 15, 18]. Bathymetry data is a part of the Florida Keys National Marine Sanctuary Digital Atlas [40] Tagging variables The surgical procedures during boat and in  situ tag - During both tagging methods, the following tagging vari- ging were similar with the exception that anesthesia ables were recorded: species, total length of fish (TL), (AQUI-S, 50% isoeugenol; aqui-s.com) was not used depth of capture, tagging location, handling time, tagging in  situ. We noticed that fish were calm once they were time, and name of the person performing the surgery. turned ventral side up and their behavior was simi- Tagging time was the total time of the tag implantation lar to that of fish that had been anesthetized. Tonic surgery, while handling time was either the total time the immobility is commonly used as an alternative to fish was on the boat or total time the fish was handled chemical-based anesthesia for surgical implantation in underwater. Handling times on the boat included the elasmobranchs [19, 37] and after a few initial efforts, we time it took for fish to show effects of anesthesia before realized that use of chemical-based anesthesia for tag- surgery and time for the fish to regain its equilibrium ging fish in situ was not necessary or beneficial and did and display other signs of recovery before being released not allow a quick and immediate release of the fish after by divers. “Fight” times for fish caught via hook and surgery was completed. Acoustic tagging procedures in line were not recorded, so analyses examined only the this study were approved by Florida Fish and Wildlife amount of time the fish was on the boat. Tagging loca - Conservation Commission’s Fish and Wildlife Research tion was the general location of the tagging, either the Institute and every effort was made to follow accepted Lower Florida Keys or the Dry Tortugas (Fig. 1), and was standards and ensure the ethical treatment of captured included to account for any environmental differences fish. Keller et al. Anim Biotelemetry (2021) 9:5 Page 4 of 15 between locations. Habitat types within the two tagging only one detection were excluded because a single detec- locations were similar and contained low- to mid-relief tion could be caused by transmission  collisions or noise continuous coral reefs and patch reefs. and is not considered reliable [35]. Detections from fish The family name of fish, either Serranidae or Lutjani - that were detected only on one receiver were examined dae (n = 2), was also included as a covariate to examine to determine whether detections were the result of nor- differences in detectability based on physiological differ - mal fish activity or from a tag loss or mortality event. ences between taxonomic groups while avoiding overpa- Consistent detections regardless of hour or day or a pat- rameterization of the models because of the high number tern of higher detections during a particular diel period of species (n = 14). Depth of capture was also recorded, could be due to a tag sitting on the seafloor with changes though for fish tagged on the boat, the depth recorded in environmental conditions throughout the day influ - was that of the seafloor as it was not possible to deter - encing transmission success. If detections were suspected mine the exact depth in the water column the fish was to be due to a tag sitting on the seafloor, those detections first caught on hook and line. Total length of fish at time were removed. of capture was transformed into a two-factor variable called “fish size” to compare small and large fish without Time‑to‑event analysis the confounding factor of fish family. Mean total lengths We used time-to-event analysis (also called survival were determined separately for each family and used as analysis) to compare the risk of tagged fish disappearing a threshold value. Fish were designated as “small” if their from the receiver array between the two tagging meth- length was below this value for the respective family and ods. The Cox proportional hazards model is a common “large” if their length was equal to or greater than this analytical method for assessing what variables cause the value. most risk to survival in discard mortality and delayed The variable month was used to examine whether mortality studies [3, 7, 33]. We used the Cox proportional time of year influenced the risk of disappearance. Water hazards model to examine which tagging variables most temperature varies throughout the year and could affect influenced the risk of fish disappearing from the receiver a fish’s health during surgery and after release. Fish array. In this study, acoustic telemetry data could only behavior (e.g., courtship/spawning behaviors) also var- provide presence/absence data, making it  difficult to ies throughout the year and between families, which can determine mortality (i.e., an end of detections could be influence the likelihood of a fish emigrating from the due to emigration, predation, harvest, etc.). As presence/ area. absence data is unable to determine mortality, time-to- To determine if the tagging experience of the surgeon event analyses examined the risk of fish disappearing had an influence on the risk of disappearance, a vari - from the receiver array for the final time (i.e., the date of able called “surgery count” was created. This variable last detection) rather than risk of mortality. was determined by ordering the tagging data by date and Before covariates were included in the model, they named surgeon (regardless of tagging method) for all fish were tested for covariance with other variables via a cor- tagged in this study that had surgeon name recorded. relation matrix and Pearson’s test of correlation. All cat- This information was used to assign each fish a surgery egorical variables were two-factored and were changed count number based on how many surgeries the surgeon to binary outputs to allow for point-biserial correlation. had previously completed. For example, surgery count Variables that had statistically significant correlation with had a value of one if it was the first fish the surgeon had each other (p < 0.05) were not included in the same haz- ever tagged or a value of 17 if it was the 17th fish a sur - ards model. geon had tagged. The data included in this study includes The influence of tagging procedure variables on the risk all, or nearly all, of each surgeon’s internal tagging experi- of disappearance from the array was examined at two dif- ence until the end of the study period or until the person ferent time periods, 4 and 6  days after surgery. The two left the agency. A few surgeons had some prior experi- time periods were chosen based on literature suggesting ence internally tagging fish, but were unable to accurately 4–6  days post-surgery to be a time frame when events quantify their experience, so all surgery counts in this (mortality, tag loss, or emigration) are most likely due study started with a value of 1. to tagging artifacts [27, 42, 43]. Any shorter time frame may miss delayed effects from tagging and longer time Data validation frames increase the chance of mistaking behavioral events Before inclusion in analyses, fish that did not have for those related to the tagging procedure. We decided complete tagging procedure data were excluded from to examine risk of disappearance at both 4 and 6  days the dataset. Detection data was then validated for fish after tagging to determine if there were any differences included in analyses. To validate detection data, fish with between these two time frames. For each s fi h, the last day Keller  et al. Anim Biotelemetry (2021) 9:5 Page 5 of 15 of its detection was used to calculate the number of days disappearance as the value of the covariate increases by it was present before it disappeared from the array. For 1 unit. A hazard ratio greater than 1 is the proportional example, if a fish was last detected 5 days after the tagging increase in risk of a disappearance as the value of the procedure, it was marked as having an event at day 5. covariate increases by 1 unit. The farther away a hazard If a fish was detected past the 4- or 6-day mark, the ratio is from 1, the greater the influence of that covariate. number of days it was considered present was truncated A hazard ratio near 1 means that covariate has a marginal to 4 or 6 days, depending on which time frame was used effect on the risk of disappearance (and thus the prob - for analysis. For example, if a fish was not detected on ability of presence)  while a hazard ratio at 1, or a confi - day 6, but was detected more than once on day 50, it dence interval that crosses 1, has no effect on the risk of was considered alive and present on day 6. Long gaps in disappearance [20]. detections may have occurred as fish often left receiver Independence between residuals and time is an coverage area before returning, but all fish included in assumption of the proportional hazards model [20] and this study were regularly detected and did not display was checked in all models by testing the correlation suspicious data such as getting detected the first few days between scaled Schoenfeld residuals and time using the after tagging and only having one subsequent detection a “cox.zph” function in the “survival” package. A statisti- year later. cally significant correlation of Schoenfeld residuals and To increase the chances of determining the most accu- time rejects the null hypothesis and violates the assump- rate activity level of a tagged fish, detection data from tion of the proportional hazards model. acoustic receivers deployed throughout the Florida Keys during 2008–2018 (not just the receivers in the array Results where the fish was tagged), were included to determine Of the fish tagged from 2008 to 2016, 194 were internally a fish’s status. Detection data came from receivers owned tagged grouper and snapper. Fish with incomplete tag- by FWRI and the Integrated Tracking of Aquatic Animals ging information (n = 87) were excluded and two addi- of the Gulf of Mexico (iTAG) network ([24]; Fig. 1). tional fish were excluded from analyses because they had To visually compare the probability of presence and only a single detection. Data from fish detected at only number of disappearances between the two tagging one location were examined, but no fish had continu - methods, Kaplan–Meier survival curves were created ous, consistent detection patterns suspected of result- using the “survival” package [41] in R [29]. To examine ing from a discarded tag. Receivers in both the Lower what factors were most influential on risk of disappear - Keys and Dry Tortugas arrays were often spaced farther ance, Cox proportional hazards models were calculated apart to try to maximize coverage area and therefore at both 4 and 6 days after tagging also using the “survival” most receivers did not have overlapping coverage. Since package. Model selection started with the full model most fish detected at a single location were groupers or using all covariates that were not correlated with tag- small snappers that were not expected to make long-dis- ging method (the main variable of interest) or each other. tance movements, smaller movements could have been The best-fitting model was chosen through a backwards missed because of spacing between receivers. Therefore, stepwise search, starting with the full model, and model since examined data did not display suspicious patterns, selection was  based on lowest Akaike information crite- no fish detected at only a single location was excluded rion with a correction for small sample size (AICc) using from analyses. In total, 105 fish from 14 different species the “MASS” package [44]. Beyond the initial Cox propor- were included in analyses (35 boat tagged and 70 in  situ tional hazards model examining the influence of tagging tagged; Table 1). method on risk of disappearance, two secondary hazards The mean total length was 57.0 ± 1.4  cm (mean ± SE) models were performed to assess which variables influ - for Lutjanidae and 64.6 ± 1.7  cm for Serranidae. These enced risk of disappearance for (1) fish tagged on the values were used to determine whether the variable of boat and (2) fish tagged in situ. “fish size” was “small” or “large”. Mean handling time and Hazard ratios from the hazards model provide infor- depth of capture varied between tagging method, but mation on the covariates’ effects on risk of disappearance. mean tagging time was similar (Table 2). Hazard ratios above a value of 1 indicate an increase in In situ tagged fish had a lower probability of disappear - risk of disappearance compared to a reference point, ing from the array compared to fish tagged on the boat, whereas a value below 1 indicates a reduction in risk. For based on Kaplan–Meier survival curves (Fig. 2). Six days categorical covariates, a hazard ratio less than 1 is the after tagging, the probability of presence of fish tagged on proportional reduction in risk compared to the reference the boat was only 66% while the probability of presence level and vice versa. For continuous covariates, a hazard of fish tagged in  situ was 90%. The null hypothesis that ratio less than 1 is the proportional reduction in risk of a there was no difference between the two methods in the Keller et al. Anim Biotelemetry (2021) 9:5 Page 6 of 15 Table 1 Number of  fish in  time-to-event analyses + Boat + In situ Tagging Method by species, tagging method, and size 1.00 Tagging method and fish species Total fish Small Large + 0.75 Boat tagged 35 19 16 0.50 Black grouper, Mycteroperca bonaci 8 5 3 Gray snapper, Lutjanus griseus 1 1 – 0.25 p = 0.002 Mutton snapper, Lutjanus analis 14 2 12 0.00 Red grouper, Epinephelus morio 1 1 – Days after tagging Yellowfin grouper, Mycteroperca venenosa 1 – 1 Number at risk: n (%) Yellowtail snapper, Ocyurus chrysurus 10 10 – Boat 35 (100)27 (77) 25 (71) 23 (66) In situ tagged 70 28 42 In situ 70 (100)66 (94) 65 (93) 63 (90) Black grouper, Mycteroperca bonaci 17 6 11 Days after tagging Dog snapper, Lutjanus jocu 6 – 6 Fig. 2 Kaplan–Meier survival curves showing probability of presence for fish tagged in situ and on the boat from day of tagging to 6 days Gag, Mycteroperca microlepis 2 – 2 afterward. Shaded areas indicate 95% confidence interval for the Gray snapper, Lutjanus griseus 8 8 – survival curves. A risk table shows the number of individual fish and Mutton snapper, Lutjanus analis 13 – 13 percentage of total fish still detected in 2-day bins for both tagging Nassau grouper, Epinephelus striatus 8 6 2 methods Red Hind, Epinephelus guttatus 1 1 – Rock Hind, Epinephelus adscensionis 1 1 – Scamp, Mycteroperca phenax 4 3 1 Schoolmaster, Lutjanus apodus 1 1 – Yellowfin grouper, Mycteroperca venenosa 8 1 7 Yellowmouth grouper, Mycteroperca inter- 1 1 – stitialis Total 105 47 58 Table 2 Mean values ± SE for handling time, tagging time, and depth between boat and in situ tagging methods Handling time Tagging time (min) Depth (m) (min) Boat tagged 72.17 ± 6.02 6.37 ± 0.35 13.29 ± 2.14 In situ tagged 9.30 ± 0.31 6.96 ± 0.27 24.61 ± 0.94 probability of presence from day of tagging to 6 days after tagging was rejected by a log-rank test (p = 0.002). Fish size, tagging time, and surgery count were the Fig. 3 Pearson correlation coefficients for all covariates for all fish only variables not significantly correlated with tagging tagged. Size and color of dots indicate strength and direction of method (the main variable of interest) or each other correlation. Boxes without dots indicate no statistically significant correlation (p > 0.05). Boxes with a thick black outline indicate the and therefore were the only variables included in the variables included in the full model initial Cox proportional hazards models (Fig.  3). Vari- ables that were significantly correlated with tagging method or with each other could not be included in the 6-day mark, tagging method was the only variable in hazards model without compromising model results. the best-fitting model (Table 3 ). Because of correlation among variables, model selec- Hazards model results for the best-fitting models tion using backward stepwise selection started with a indicate that fish internally tagged in  situ were 76% less full model including only tagging method, tagging time, likely (one minus the hazard ratio) to disappear from surgery count,  and fish size. Backward stepwise model the array 4  days after tagging and 75% less likely to dis- selection found the best-fitting hazards model (based appear 6  days after tagging compared to fish tagged on on lowest AICc) at the 4-day mark to include tagging the boat (Table  4). Tagging time was also included in method and tagging time as covariates, while at the Tagging Method Probability of Presence Keller  et al. Anim Biotelemetry (2021) 9:5 Page 7 of 15 Table 3 Cox proportional hazards models with AICc stepwise model ranking for both 4 and 6 days after tagging Days after tagging Hazards model AICc ΔAICc χ df p 4 Method + tagging time 131.19 0.00 11.11 2 0.004 Method 131.47 0.28 8.98 1 0.003 Method + tagging time + surgery count 131.72 1.53 11.92 3 0.008 Method + tagging time + fish size + surgery count 134.83 3.64 11.12 3 0.010 Null 137.52 6.33 6 Method 166.68 0.00 9.78 1 0.002 Method + tagging time 167.58 0.90 10.82 2 0.004 Method + tagging time + fish size 169.68 3.00 10.85 3 0.010 Method + tagging time + fish size + surgery count 171.83 5.15 10.87 4 0.030 Null 173.38 6.70 All models include probability of presence as the dependent variable and list independent variables included in each individual model. All fish were included in these models (n = 105). p-values are from the log-rank test Table 4 Cox proportional hazards models results for the best-fitting model at 4 and 6 days after tagging Days after tagging Factor Coefficient SE Hazard ratio 95% CI p 4 Tagging method: boat Reference – – – – Tagging method: in situ − 1.43 0.55 0.24 0.08–0.70 0.009* Tagging time − 0.19 0.12 0.83 0.65–1.05 0.120 6 Tagging method: boat Reference – – – – Tagging method: in situ − 1.38 0.48 0.25 0.10–0.64 0.004* For the categorical covariates, the hazard ratio is the proportion of risk compared to the reference covariate. For the continuous covariates, the hazard ratio is the proportion of risk with an increase of 1 unit (e.g., 1 min). A variable with a hazard ratio close to a value of 1 has a marginal effect and a confidence interval (CI) spanning 1 has no effect on hazard. All models include probability of presence as the dependent variable and list independent variables included in the best-fitting models. p-values are from the log-rank test and asterisks indicate statistical significance the best-fitting 4-day model, though the 95% confidence correlated with month (Fig.  5b). Three full models were interval spanned 1, indicating there was no effect. For all created for in situ tagged fish: (1) one with handling time, covariates in both best-fitting models, the proportional depth, and all other variables not correlated with each hazard assumption was supported by a nonsignificant other, and (2) one with handling time, fish size, and all relationship between scaled Schoenfeld residuals and other variables not correlated with each other, and (3) time. To visually compare each covariate’s risk to fish sur - one with month and all non-correlated variables. vival, hazard ratio forest plots were produced from the Handling time was the only covariate included in the full and best-fitting models, and plots for 4 days after tag - best-fitting model for boat-tagged fish at 4  days after ging are shown as an example (Fig. 4a, b). tagging (Table  5). At 6  days after tagging, both handling Pearson correlation coefficients were also calculated for time and family were included in the best-fitting model. all covariates when splitting the data into fish tagged on Handling time was significant in both best-fitting models; the boat and fish tagged in situ. Handling time and depth however, the hazard ratio was very close to 1, meaning were the two main covariates of interest. However, since the effect of handling time on hazard of disappearance these two covariates were correlated for fish tagged on was marginal (Table  6). A hazard ratio of 2.55 for fam- the boat (Fig. 5a), they could not be included in the same ily Serranidae in the 6-day model indicated an increase hazards model. As a result of these correlations, two full in risk compared to Lutjanidae. However, this covariate models were created for fish tagged on the boat: (1) one had a wide confidence interval that crossed 1, indicating with depth and all variables not correlated with each no significant effect of hazard and a somewhat unreliable other, and (2) one with handling time and all variables point estimate [20]. For all covariates in both best-fitting not correlated with each other. models, the proportional hazard assumption was sup- For in  situ tagged fish, handling time and depth were ported by a nonsignificant relationship between scaled not correlated with each other, but depth was correlated Schoenfeld residuals and time. with fish size and both depth and handling time were Keller et al. Anim Biotelemetry (2021) 9:5 Page 8 of 15 0.10.2 0.51 2 Fig. 4 Forest plots showing covariate hazard ratios at 4 days after tagging for the a full model and b best-fitting Cox proportional hazard model. Hazard ratios and 95% confidence intervals in comparison to reference points are shown for each covariate with p-values listed on the right-hand side of the plots (asterisks indicating statistical significance). Hazard ratio values less than 1 (represented by the dotted line) indicate a reduction in risk of disappearance from the array while values greater than 1 indicate increased risk of disappearance. Covariates with hazard ratios close to 1 have a marginal effect on hazard and confidence intervals crossing 1 indicate no significant effect For in  situ tagged fish, month was the only covariate effect. Additionally, the particularly wide confidence included in the best-fitting model at 4  days after tag - interval for fish size in the best-fitting 6-day model ging and both month and fish size were in the best-fit - indicated the point estimate was unreliable (Table  8). ting model at 6  days after tagging (Table  7). However, For all covariates in both best-fitting models, the pro - the global log-rank tests examining significance of the portional hazard assumption was supported by a non- models overall were nonsignificant in both cases. The significant relationship between scaled Schoenfeld hazard ratio confidence intervals crossed one for all residuals and time. covariates in the best-fitting models, indicating no Keller  et al. Anim Biotelemetry (2021) 9:5 Page 9 of 15 1 1 Depth Depth Depth 0.8 0.8 0.8 Family Family 0.6 Family 0.6 0.6 0.4 0.4 0.4 Fish.Size Fish.Size Fish.Size 0.2 0.2 0.2 Handling.Time Handling.Time Handling.Time 0 0 Location Location Location −0.2 −0.2 −0.2 Month −0.4 Month Month −0.4 −0.4 −0.6 Surgery −0.6 −0.6 Surgery.Count Surgery.Count −0.8 Tagging.Time −0.8 −0.8 Tagging.Time Tagging.Time −1 −1 −1 a b Fig. 5 Pearson correlation coefficients for all covariates for a boat-tagged and b in situ tagged fish. Size and color of dots indicate strength and direction of correlation. Boxes without dots indicate no statistically significant correlation (p > 0.05) Discussion seafloor via divers after they recovered from anesthesia to Using an historic dataset of 8  years of fish tagging by ensure a safe return to the reef. Nevertheless, the tagged Florida Fish and Wildlife Conservation Commission sci- fish may still have suffered from barotrauma, increased entists in the Florida Keys, we examined factors a poste- stress from being caught by hook and line, increased riori that influence probability of presence of internally handling times (which could also increase the impacts tagged fish. At 6  days after tagging, Kaplan–Meier sur - of barotrauma), or long-lasting effects of anesthesia. All vival curves demonstrated that fish tagged in  situ had a these factors could increase risk of boat-tagged fish dis - 25% higher probability of detection within the acoustic appearing from the array. array (i.e., 75% lower probability of disappearing) com- At the start of in  situ tagging, we decided not to use pared to fish tagged on the boat. Tagging method was anesthesia at depth because we discovered that fish the only factor affecting the risk of disappearance when held ventral side up with their eyes covered were just as we examined the initial Cox proportional hazards model calm as anesthetized fish. We also deemed it advanta - assessing the probability of presence of all fish. geous to avoid unnecessarily impairing the tagged fish’s In this study, there were several variables that could not response to predators once the fish was released into be included in the initial hazards model because of cor- sheltering habitat after surgery. Using anesthesia dur- relation, but their inclusion and limited influence in the ing fish surgeries may not always be the best course of subsequent boat-tagged and in  situ tagged models rein- action in every situation, and all potential effects on the forced that the tagging method was the most important fish must be considered [8 , 23]. We believe the benefits factor in risk of disappearance from the array. Acoustic of reduced total handling time for fish tagged in  situ tracking data of in  situ surgically tagged fish from two versus fish tagged on the boat, which usually included other telemetry studies also demonstrate the benefit of a recovery period (mean ± SE: In  situ = 9.30 ± 0.31  min, tagging method, indicated by high survival rates (97– Boat = 72.17 ± 6.02  min), coupled with immediate 100%) post-surgery [4, 43]. The decrease in probability of release into protective habitat and the lack of pressure presence of boat-tagged fish could be due to several fac - and temperature changes, explain the increased prob- tors. As described in Feeley et  al. [12] and Herbig et  al. ability of presence within the acoustic array for in  situ [15], all fish tagged on the boat were returned to the tagged fish. The increased probability of presence of fish Depth Family Fish.Size Handling.Time Location Month Surgery.Count Tagging.Time Depth Depth Family Family Fish.Size Fish.Size Handling.Time Handling.Time Locatio Locationn Month Month Surgery Surgery.Count Tagging.Time Tagging.Time Keller et al. Anim Biotelemetry (2021) 9:5 Page 10 of 15 Table 5 Cox proportional hazards models examining probability of  presence for  fish tagged on  the  boat (n = 35) at both 4 and 6 days after tagging Days Hazards model AICc ΔAICc χ df p after tagging 4 Handling time 66.00 0.00 4.17 1 0.04 Handling time + family 67.08 1.08 5.76 2 0.06 Tagging time 68.13 2.13 2.37 1 0.10 Null 68.26 2.26 Handling time + tagging time + family 68.60 2.60 6.96 3 0.07 Tagging time + surgery count 69.12 3.12 3.79 2 0.20 Tagging time + surgery count + family 70.64 4.64 5.06 3 0.20 Handling time + tagging time + month + family 70.67 4.67 7.30 4 0.10 Tagging time + surgery count + month + family 72.31 6.31 5.84 4 0.20 Handling time + tagging time + surgery count + month + family 72.91 6.91 8.34 5 0.10 Depth + tagging time + surgery count + month + family 74.85 8.85 5.86 5 0.30 Depth + tagging time + surgery count + month + fish size + family 77.79 11.79 5.90 6 0.40 6 Handling time + family 77.51 0.00 8.28 2 0.02 Handling time 77.55 0.04 5.43 1 0.02 Handling time + month + family 78.13 0.62 10.32 3 0.02 Month 79.66 2.15 3.22 1 0.07 Handling time + tagging time + month + family 79.75 2.24 11.22 4 0.02 Family + month 79.85 2.34 5.43 2 0.07 Tagging time + month + family 80.73 3.22 6.96 3 0.07 Null 81.06 3.55 Tagging time + fish size + month + family 81.61 4.10 9.98 4 0.06 Handling time + tagging time + surgery count + month + family 82.28 4.77 11.73 5 0.04 Tagging time + surgery count + month + family 84.22 6.71 9.33 5 0.10 Depth + tagging time + surgery count + month + family 87.07 9.56 9.36 6 0.20 Models are ranked by AICc values. All models include probability of presence as the dependent variable and list independent variables included in each individual model. p-values are from the log-rank test Table 6 Cox proportional hazards models results for  best-fitting models (lowest AICc) of  boat-tagged fish at  both  4 and 6 days after tagging Days after tagging Covariates Coefficient SE hazard ratio 95% CI p 4 Handling time − 0.02 0.01 0.98 0.96–0.99 0.048* 6 Handling time − 0.02 0.01 0.98 0.96–0.99 0.016* Family: Lutjanidae Reference – – – – Family: Serranidae 0.94 0.60 2.55 0.79–8.23 0.116 All models include probability of presence as the dependent variable and list independent variables included in each individual model. p-values are from the log-rank test and asterisks indicate statistical significance tagged in situ compared to fish tagged on the boat indi - that additional handling time would increase the risk cates that any negative effects due to the lack of anes - of a tagged fish having an event. The models, however, thesia are minimal and do not outweigh the benefits of suggested an increase in 1  min of handling time slightly in situ tagging. decreased the risk of disappearance for boat-tagged fish, There were several surprising results presented from though the confidence intervals neared or crossed 1 in the analysis of this data. Some of the variables that were all cases, indicating little to no effect. Handling time may originally believed to impact the probability of detec- not have been very influential in risk of disappearance, tion were either not influential in the hazards models or but it is still surprising that increased time may have had the opposite effect. For example, we hypothesized reduced risk, especially as handling times for fish tagged Keller  et al. Anim Biotelemetry (2021) 9:5 Page 11 of 15 Table 7 Cox proportional hazards models examining fish survival for  fish tagged in  situ (n = 70) at  both  4 and  6  days after tagging Days after tagging Hazards model AICc ΔAICc χ df p 4 Month 40.38 0.00 3.26 1 0.07 Fish size + month 41.49 1.11 4.51 2 0.10 Handling time 41.65 1.27 2.34 1 0.10 Depth + month 41.99 1.61 3.96 2 0.10 Null 42.19 1.81 Handling time + fish size 42.30 1.92 3.78 2 0.20 Handling time + family 42.74 2.36 3.25 2 0.20 Fish size + family + month 43.24 2.86 5.05 3 0.20 Depth + family + month 43.76 3.38 4.46 3 0.20 Handling time + fish size + family 43.79 3.41 4.45 3 0.20 Handling time + depth + family 44.77 4.39 3.41 3 0.30 Tagging time + fish size + family + month 45.21 4.83 5.09 4 0.30 Handling time + fish size + surgery count + family 45.57 5.19 4.52 4 0.30 Tagging time + depth + family + month 45.92 5.54 4.49 4 0.30 Handling time + depth + surgery count + family 46.97 6.59 3.44 4 0.50 Tagging time + fish size + surgery count + family + month 47.53 7.15 5.11 5 0.40 6 Fish size + month 57.43 0.00 5.69 2 0.06 Fish size 57.84 0.41 3.17 1 0.08 Handling time + fish size + surgery count 57.87 0.44 5.41 3 0.10 Fish size + surgery count 58.37 0.94 4.45 2 0.10 Month 58.70 1.27 2.03 1 0.20 Null 58.86 1.43 Fish size + surgery count + month 58.90 1.47 6.48 3 0.09 Depth + month 59.52 2.09 3.5 2 0.20 Handling time + fish size + surgery count + family 59.54 2.11 5.67 4 0.20 Surgery count 60.17 2.74 1.15 1 0.30 Depth + surgery count 60.92 3.49 1.72 2 0.40 Fish size + surgery count + family + month 60.95 3.52 6.68 4 0.20 Depth + surgery count + month 61.02 3.59 3.94 3 0.30 Handling time + depth + surgery count 62.10 4.67 2.32 3 0.50 Tagging time + depth + surgery count + month 62.88 5.45 4.62 4 0.30 Tagging time + fish size + surgery count + family + month 63.19 5.76 7.5 5 0.20 Handling time + depth + surgery count + family 64.08 6.65 2.64 4 0.60 Tagging time + depth + surgery count + family + month 64.97 7.54 4.87 5 0.40 Models are ranked by AICc values. All models include probability of presence as the dependent variable and list independent variables included in each individual model. p-values are from the log-rank test Table 8 Cox proportional hazards models results for best-fitting models of in situ tagged fish Days after tagging Covariates Coefficient SE Hazard ratio 95% CI p‑ value 4 Month 0.59 0.36 1.82 0.89–3.69 0.10 6 Month 0.34 0.24 1.41 0.89–2.24 0.14 Fish size: Small 1.44 0.84 4.23 0.82–21.84 0.08 All models include probability of presence as the dependent variable and list independent variables included in each individual model. p-values are from the log-rank test Keller et al. Anim Biotelemetry (2021) 9:5 Page 12 of 15 on the boat were high compared to other tagging studies of year and water temperature, and thus linked to the [3, 39] and minimizing handling of fish has been reported variable month. Although time of year was not a sig- as one of the most important considerations in implan- nificant in this study it should  also be considered when tation surgeries [30]. For in  situ tagging, handling time deciding how to maximize fish survival and probability of was not included in best-fitting models, but that may be presence when setting up a tagging project. because there was little variance in the handling times of Experience of the surgeon has been shown to influence fish tagged in situ. tag retention and survivorship [5, 10], but this was not Fish size was not a significant factor in the best-fitting seen in our study. Surgery count was not in any best-fitting models. Fish tagged in this study varied between 32 and model, but was a covariate in the third best-fitting model 107 cm TL, a range large enough that differences in prob - for both the initial model at 4  days after tagging and the ability of presence because of size were expected. We in  situ model at 6  days after tagging. Although the effect hypothesized that smaller fish would be at a higher risk was not statistically significant, both models indicated of disappearance because they might not have the same an increase in surgery count decreased risk of fish disap - energy resources to aid in recovery as larger fish or they pearance. Consistent training among surgeons may have may be more susceptible to predation. However, this was reduced the effect of surgery count enough so that other not supported by the models. “Fish size” was included in variables had larger influences. Before the start of a tagging the best-fitting model for in situ tagged fish at 6 days after project, surgeons practiced the tagging procedure with tagging but was not significant, and a large spread in the guidance from an experienced team member, and initial hazard ratio confidence interval (0.82–21.84) represented training was provided by a professional plastic surgeon high uncertainty in the hazard ratio estimate. While there (D. Hawtof, personal communication). Additionally, less are likely species-specific differences (e.g., body type, experienced surgeons were always paired with an expe- stress tolerance) that influence probability of presence of rienced surgeon during the tagging procedure to provide different sized fish, in our study fish size did not signifi - guidance during both boat and in  situ tagging. The train - cantly affect probability of presence post-tagging. Also, ing and oversight throughout the tagging procedure may fish less than 20  cm FL have been shown to have high have helped to reduce the effect of surgeon experience on survival from internal tagging [21], suggesting that sur- risk of disappearance. gery and implantation of appropriately sized tags for the Hazards models were analyzed at both 4 and 6  days body size of the fish may not impair survival, regardless after tagging because there was discrepancy in the lit- of fish size. erature about how long after surgery to examine survival Fish family was also not a significant covariate, related to the tagging procedure. Results between these although differences in tolerance to barotrauma and two time periods were similar (e.g., hazard ratios for tag- handling were seen among species. The high number ging method in the initial model differed by only 0.015), of species in this study (which originally led to overpa- but we felt it was informative to present both cases. rameterization in models) and the unevenness of species Although differences between the results at both time between tagging methods resulted in grouping species by periods were minimal, they are still interesting to note family. There are physiological differences between spe - because they show how the dynamics of fish survival may cies of the same family, and grouping them by family may subtly shift over time. The maximum time after surgery have masked these differences and may help explain why that a fish was still detected was just over 4  years (data family did not affect probability of presence. Although an from this fish was analyzed in [12]), and a possible future in-depth look at the effect of tagging method on species direction of work could include examining longer term was not possible in this study, a planned, paired study differences in detection patterns between different tag - with fewer species may find that fish species influences ging methods, particularly if the goal is to maximize data probability of presence. For any tagging project, factors acquisition. Analyzing probability of presence on a longer such as body type, susceptibility to barotrauma, and time frame may find different factors which are more stress tolerance should be considered when deciding on influential than those found in this study, but considera - the appropriate tagging method. tion should be given to whether disappearances from the Water temperatures and differences between water array past 6 days post-surgery are a result of the tagging and air temperatures vary throughout the year and could procedure or from a natural behavioral event. affect the risk of a fish disappearing from the array, but This study was not originally designed to rigorously month as a proxy for temperature did not have a sig- examine all factors affecting boat and in  situ tagged fish. nificant influence in either boat-tagged or in  situ tagged Therefore, some variables may not have proven as influen - models. Life history traits of the fish families (e.g., the tial as they would in a more systematic study. Depth was time of year that they reproduce) are also linked to time expected to have more of an influence on hazard because Keller  et al. Anim Biotelemetry (2021) 9:5 Page 13 of 15 detrimental effects of barotrauma have been shown to record). However, despite these caveats, we believe that increase with depth of capture [14]. In this study, depth this historic dataset using over 100 internally tagged fish may not have been the best representation of the sever- representing 14 different species surgically tagged by 11 ity of barotrauma. Boat-tagged fish were generally caught different scientists was robust enough to demonstrate using hook and line and were not necessarily caught from that tagging method had an effect on risk of fish disap - the seafloor, which was the depth recorded, and further - pearance. Telemetry studies are expensive to conduct, more, mean depth of capture of boat-tagged fish was both in terms of effort and money, and often outcomes significantly shallower compared to in  situ tagged fish are not known until months later when the data can (Wilcoxon two-sample rank test, p < 0.05). Fish that were be downloaded. Therefore, it is critical to have a better near the surface or mid-water column when caught on understanding of how tagging method influences prob - hook and line would have a greater depth recorded than ability of presence (which can be considered a proxy for where they were physically caught in the water column, fish survivorship). By examining the effects of tagging meaning that they may have exhibited fewer signs of baro- methods, we can inform future telemetry studies, thereby trauma than if they had actually been caught at the depth increasing the usefulness of telemetry data. recorded. Fish tagged in situ were caught via baited traps This study has shown that there were clear benefits to on the seafloor, so the depth of capture was the depth of tagging fish in situ, though due to high correlation among the trap and the actual depth of the fish when it entered variables we are unable to pinpoint the source behind the trap. Additionally, after the start of in  situ tagging, decreased risk of disappearance compared to boat-tagged this method was preferentially chosen over tagging on fish. The lack of barotrauma, shorter handling time, lack the boat at deep sites and only a few fish were captured, of anesthesia, and use of baited traps instead of hook and tagged, and released from the boat in deep water (> 30 m) line may all have played a role in increasing an in  situ and in  situ tagged fish had a higher overall sample size tagged fish’s probability of presence. We understand that compared to boat-tagged fish. The method of tagging this in  situ method may not always be practical. In  situ was based on what was deemed best for survival of the tagging can increase the amount of effort spent to tag fish and logistically made sense for each tagging event. If each fish because of the logistics required and the time boat tagging data had included more fish captured from limitations when tagging at deeper depths. Some target deep water, resulting in no discrepancy in mean depth per species may be located outside recreational diving limits, tagging method, we hypothesize depth would have had which can increase effort and cost for specialized training a larger effect on survival. Additionally, comparing risk and dive gear. Also, in this study we used baited traps to of disappearance between tagging methods where boat- capture fish in situ, but other methods should be consid - tagged fish were either released at the surface or brought ered if the species of interest is unlikely to be captured in back down to depth for release (the method used for all this manner. Furthermore, studies that rely on anglers to boat-tagged fish in this study) would further explore the supply the tagging subjects or focus on species found in effects of depth, barotrauma, and release method. very shallow water may not find in situ tagging feasible or Fight time, location of hook placement on fish, whether worthwhile. However, we suggest that all aspects of the the fish exhibited signs of barotrauma, whether a fish tagging procedure, from tagging method (e.g., on a boat with barotrauma was vented before its return to depth, vs. in  situ) to use of anesthesia, method of capture, and and overall health of the fish at time of release were not method of release be carefully considered for each target often recorded and thus were not included as covariates species to reduce stress to the fish, and maximize success in the hazards models. The amount a time a fish spent of the study. in a trap could influence the fish’s health and probability of presence after tagging, but the time a trap was baited Conclusions does not equal the time a fish was in a trap (a trap may In this study, fish internally tagged in  situ with acous - have been baited and left for 12  h, but a fish may have tic tags were ~ 75% less likely to disappear from the entered the trap at hour 11). Use of cameras for identi- receiver array (i.e., mortality, emigration, or tag loss fying how long fish were in traps and recording specifics event) shortly after tagging compared to fish brought on condition upon release would provide more details on to the surface and tagged on a boat. Other factors had why risk of disappearance was higher in boat-tagged fish. little to no influence on probability of presence after This study demonstrated some of the limitations of a tagging, but correlation of variables limited what could posteriori-designed study including correlated variables be included in hazards models. Despite the limitations and missing data (particularly during the initial stages of due to correlation, fish had a higher probability of being conducting acoustic telemetry research when it was not present and a lower probability of disappearance when known what supplementary data would be beneficial to tagged with the in  situ methods presented. Therefore, Keller et al. Anim Biotelemetry (2021) 9:5 Page 14 of 15 References when designing a telemetry study, serious considera- 1. Atkins JL, Morris JA, Green SJ. In situ tagging technique for fishes pro - tion of tagging method (in situ versus on a boat) and vides insight into growth and movement of invasive lionfish. Ecol Evol. other variables (capture method, release method, use of 2014;4(19):3768–77. 2. Balázs G, Lewarne B, Herczeg G. In situ underwater tagging of aquatic anesthesia, etc.) should be evaluated because they play organisms: a test using the cave-dwelling olm Proteus anguinus. Ann Zool a critical role in the overall health of fish and success Fennici. 2015;52(3):160–6. https ://doi.org/10.5735/086.052.0303. of the study. Implanting internal acoustic tags in situ is 3. Bohaboy EC, Guttridge TL, Hammerschlag N, Van Zinnicq Bergmann MP, Patterson WF III. Application of three-dimensional acoustic telemetry to not a practical method for every species of interest and assess the effects of rapid recompression on reef fish discard mortality. for every environment, but given the increased prob- ICES J Mar Sci. 2019;77(1):83–96. ability of presence demonstrated here, we suggest it be 4. Bryan DR, Feeley MW, Nemeth RS, Pollock C, Ault JS. Home range and spawning migration patterns of queen triggerfish Balistes vetula in considered for fish species where applicable. StCroix US Virgin Islands. Mar Ecol Prog. 2019;616:123–39. https ://doi. org/10.3354/meps1 2944. Acknowledgements 5. Cooke SJ, Graeb BDS, Suski CD, Ostrand KG. Eec ff ts of suture material We thank B Binder, A Collins, M Dancy, D DeMaria, D Eaken, D Hawtof, J Javech, on incision healing, growth and survival of juvenile largemouth bass A Johnson, M McCallister, J Renchen, A Tobin, and S Walters-Burnsed for assis- implanted with miniature radio transmitters: case study of a novice and tance in the field. We also thank S Barbieri for sharing data and Dry Tortugas experienced fish surgeon. J Fish Biol. 2003;62:1366–80. National Park Staff and the crews of the M/V Fort Jefferson, M/V Spree, M/V 6. Crossin GT, Heupel MR, Holbrook CM, Hussey NE, Lowerre-Barbieri SK, Peter Gladding, and NOAA ship Nancy Foster for helping with logistics. Thanks Nguyen VM, Raby GD, Cooke SJ. Acoustic telemetry and fisheries man- to E Ault, K Parsons, C Shea, and two anonymous reviewers for valuable manu- agement. Ecol Appl. 2017;27(4):1031–49. script comments and D Fields and C Ames for assistance with bathymetry 7. Curtis JM, Johnson MW, Diamond SL, Stunz GW. Quantifying delayed data. In memoriam Jeff Simonds, FWC staff (April 24th 1978–February 18th, mortality from barotrauma impairment in discarded red snapper using 2010), and David Hawtof, FWC volunteer (September 5th, 1933–February 5th, acoustic telemetry. Mar Coast Fish. 2015;7:434–49. 2019), who contributed to this study. 8. Damon-Randall K, Bohl R, Bolden S, et al. Atlantic sturgeon research techniques NOAA technical memorandum NMFSNE-215. Woods Hole: Authors’ contributions National Marine Fisheries Service Northeast; 2010. p. 19. JK analyzed the data. JK, DM, JH, MF, and AA contributed to the interpretation 9. Dance MA, Moulton DL, Furey NB, Rooker JR. Does transmitter placement of results and JK, DM, JH, AA, MF, and PB contributed to the writing of the or species affect detection efficiency of tagged animals in biotelemetry manuscript. PB, MF, DM, JK, AA, and JH performed fieldwork. AA, DM, and MF research? Fish Res. 2016;183:80–5. secured funding. All authors read and approved the manuscript. 10. Deters KA, Brown RS, Carter KM, Boyd JW, Eppard MB, Seaburg AG. Perfor- mance assessment of suture type, water temperature, and surgeon skill Funding in juvenile Chinook salmon surgically implanted with acoustic transmit- Research was supported by National Park Service CESU contract ters. Trans Am Fish Soc. 2010;139(3):888–99. H5000065040-J5281060165, USGS State Partnership Program grant 11. Eberts RL, Somers CM. Venting and descending provide equivocal ben- 08ERAG00070, NOAA NMFS Coral Reef Conservation Program efits for catch-and-release survival: study design influences effectiveness grants NA17RJ1226 and NA08OAR4320892, and NOS Agreement # more than barotrauma relief method. N Am J Fish Manag. 2017;37:612– MOA-21010–026/8081. 23. https ://doi.org/10.1080/02755 947.2017.13072 92. 12. Feeley MW, Morley D, Acosta A, Barbera P, Hunt J, Switzer T, Burton Availability of data and materials M. Spawning migration movements of Mutton Snapper in Tortugas, The datasets generated during and/or analyzed during the current study are Florida: spatial dynamics within a marine reserve network. Fish Res. available by request in the FWCDL repository (https ://f5000 6a.eos-intl.net/ 2018;204:209–23. F5000 6A/OPAC/Detai ls/Recor d.aspx?BibCo de=53861 79). 13. Gallagher AJ, Serafy JE, Cooke SJ, Hammerschlag N. Physiological stress response, reflex impairment, and survival of five sympatric shark Ethics approval and consent to participate species following experimental capture and release. Mar Ecol Prog Ser. Acoustic tagging procedures in this study were approved by Florida Fish 2014;496:207–18. and Wildlife Conservation Commission’s Fish and Wildlife Research Insti- 14. Hannah RW, Matteson KM. Behavior of nine species of Pacific rockfish tute and every effort was made to follow accepted standards and ensure after hook-and-line capture, recompression, and release. Trans Am Fish the ethical treatment of captured fish. This research was conducted under Soc. 2007;136:24–33. scientific research permits issued by the Florida Keys National Marine 15. Herbig JL, Keller JA, Morley D, Walter K, Barbera P, Acosta A. Examining Sanctuary (FKNMS-2008-015, FKNMS-2013-040, FKNMS-2013-040-A2, movement patterns of yellowtail snapper, Ocyurus chrysurus, in the Dry FKNMS-2018-047) and the Dry Tortugas National Park (DRTO-2008-SCI0003, Tortugas. Florida Bull Mar Sci. 2019;95(1):45–67. DRTO-2010-SCI-0006). 16. Hussey NE, Kessel ST, Aarestrup K, Cooke SJ, Cowley PD, Fisk AT, et al. Aquatic animal telemetry: a panoramic window into the underwater Consent for publication world. Science. 2015;348(6240):1255642. Not applicable. 17. Jepsen N, Thorstad EB, Havn T, Lucas MC. The use of external electronic tags on fish: an evaluation of tag retention and tagging effects. Anim Competing interests Biotelemetry. 2015;3:49. The authors declare that they have no competing interests. 18. Keller JA, Herbig JL, Morley D, Wile A, Barbera P, Acosta A. Grouper tales: use of acoustic telemetry to evaluate grouper movements at Western Dry Author details Rocks in the Florida Keys. Mar Coast Fish. 2020;12:290–307. Florida Fish and Wildlife Conservation Commission, Fish and Wildlife Research 19. Kessel ST, Hussey NE. Tonic immobility as an anaesthetic for elasmo- Institute, 2796 Overseas Highway, Suite 119, Marathon, FL 33050, USA. branchs during surgical implantation procedures. Can J Fish Aquat Sci. National Park Service, South Florida/Caribbean Inventory and Monitoring 2015;72(9):1287–91. Network, 18001 Old Cutler Rd Suite 419, Palmetto Bay, FL 33157, USA. 20. Kleinbaum DG, Klein M. Survival analysis. New York: Springer; 2012. p. 3. 21. Klinard NV, Halfyard EA, Fisk AT, Stewart TJ, Johnson TB. Eec ff ts of Received: 7 August 2020 Accepted: 5 January 2021 surgically implanted acoustic tags on body condition, growth, and survival in a small, laterally compressed forage fish. Trans Am Fish Soc. 2018;147:749–57. Keller  et al. Anim Biotelemetry (2021) 9:5 Page 15 of 15 22. Lindholm J, Fangman S, Kaufman L, Miller S. In situ tagging and tracking pulse position modulation acoustic telemetry monitoring equipment. of coral reef fishes from the Aquarius Undersea Laboratory. Mar Technol Anim Biotelemetry. 2015;3(1):55. Soc J. 2005;39:68–73. 36. Skomal GB. Evaluating the physiological and physical consequences of 23. Lowerre-Barbieri SK, Villegas-Ríos D, Walters-Burnsed S, Bickford J, Cooper capture on post-release survivorship in large pelagic fishes. Fish Manag W, Muller R, Trotter A. Spawning site selection and contingent behavior in Ecol. 2007;14:81–9. Common Snook, Centropomus undecimal. PLoS ONE. 2014;9(7):e101809. 37. Sloan KA, Bouyoucos IA, Brooks EJ, Sneddon LU. Ethical considerations in 24. Lowerre-Barbieri S, Burnsed S, Bickford J, Boucek R, Staley H. Report on fish research. J Fish Biol. 2019;94:556–77. the Integrated Tracking of Aquatic Animals in the Gulf of Mexico (iTAG) 38. Starr RM, Heine JN, Johnson KA. Techniques for tagging and tracking 2017 workshop: building network capacity. St. Petersburg: Florida Fish deepwater rockfishes. N Am J Fish Manag. 2000;20:597–609. and Wildlife Conservation Commission; 2017. 39. Starr RM, Sala E, Ballesteros E, Zabala M. Spatial dynamics of the Nassau 25. Lucas MC, Baras E. Methods for studying spatial behavior of freshwater grouper Epinephelus striatus in a Caribbean atoll. Mar Ecol Prog Ser. fishes in the natural environment. Fish Fish. 2000;1:283–316. 2007;343:239–49. 26. McCallister M, Renchen J, Binder B, Acosta A. Diel activity patterns and 40. Taylor C, Field D, Vander Pluym J, Viehman S, Ebert E. NCCOS Mapping: movement of invasive lionfish (Pterois volitans/P. miles) in the Florida Keys seafloor mapping products for the Florida Keys National Marine Sanctu- identified using acoustic telemetry. Gulf Caribb Res. 2018;29(1):27–40. ary from 2004-01-01 to 2016-12-31 (NCEI Accession 0170215). NOAA 27. Piraino MN, Szedlmayer ST. Fine-scale movements and home ranges of National Centers for Environmental Information. Dataset. 2018. https :// Red Snapper around artificial reefs in the Northern Gulf of Mexico. Trans doi.org/10.7289/v55b0 0rz. Accessed 22 Nov 2020 Am Fish Soc. 2014;143(4):988–98. 41. Therneau T. A package for survival analysis in R. 2020. R package version 28. Pratt HL, Pratt TC, Morley D, Lowerre-Barbieri S, Collins A, Carrier JC, Hart 3.2-7. https ://CRAN.R-proje ct.org/packa ge=survi val. KH, Whitney NM. Partial migration of the nurse shark Ginglymostoma 42. Topping DT, Szedlmayer ST. Site fidelity, residence time and movements cirratum (Bonnaterre), from the Dry Tortugas Islands. Environ Biol fish of red snapper Lutjanus campechanus estimated with long-term acoustic 2018;101:515–30. monitoring. Mar Ecol Prog Ser. 2011;437:183–200. 29. R Core Team. R: a language and environment for statistical computing. 43. Tuohy E, Nemeth MI, Bejarano I, Schärer MT, Appeldoorn RS. In situ Vienna: R Foundation for Statistical Computing; 2019. https ://www.R- tagging of Nassau Grouper Epinephelus striatus using closed-circuit proje ct.org/. rebreathers at a spawning aggregation in Puerto Rico. Mar Technol Soc J. 30. Rub AMW, Jepson N, Liedtke TL, Moser ML, Weber EPS III. Surgical inser- 2015;49:115–23. tion of transmitters and telemetry methods in fisheries research. Am J Vet 44. Venables WN, Ripley BD. Modern applied statistics with S-Plus. 4th ed. Res. 2014;75(4):402–16. New York: Springer; 2002. 31. Rudershausen PJ, Buckel JA, Hightower JE. Estimating reef fish discard 45. Wagner GN, Cooke S, Brown RS, Deters KA. Surgical implantation tech- mortality using surface and bottom tagging: effects of hook injury niques for electronic tags in fish. Rev Fish Biol Fish. 2011;21:71–81. and barotrauma. Can J Fish Aquat Sci. 2014;71:514–20. https ://doi. 46. Williams LJ, Herbig JL, Szedlmayer ST. A cage release method to improve org/10.1139/cjfas -2013-0337. fish tagging studies. Fish Res. 2015;172:125–9. 32. Rummer JL, Bennett WA. Physiological effects of swim bladder overex - 47. Wilson SM, Raby GD, Burnett NJ, Hinch SG, Cooke SJ. Looking beyond the pansion and catastrophic decompression on Red Snapper. Trans Am Fish mortality of bycatch: sublethal effects of incidental capture on marine Soc. 2005;134:1457–70. animals. Biolog Conserv. 2014;2014(171):61–72. 33. Runde BJ, Rudershausen PJ, Sauls B, Mikles CS, Buckel JA. Low discard survival of gray triggerfish in the southeastern US hook-and-line fishery. Publisher’s Note Fish Res. 2019;219:105313. Springer Nature remains neutral with regard to jurisdictional claims in pub- 34. Sigurdsson T, Thorsteinsson V, Gústafsson L. In situ tagging of deep-sea lished maps and institutional affiliations. redfish: application of an underwater, fish-tagging system. ICES J Mar Sci. 2006;63:523–31. 35. Simpfendorfer CA, Huveneers C, Steckenreuter A, Tattersall K, Hoenner X, Harcourt R, Heupel MR. Ghosts in the data: false detections in VEMCO Re Read ady y to to submit y submit your our re researc search h ? Choose BMC and benefit fr ? Choose BMC and benefit from om: : fast, convenient online submission thorough peer review by experienced researchers in your field rapid publication on acceptance support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year At BMC, research is always in progress. Learn more biomedcentral.com/submissions

Journal

Animal BiotelemetrySpringer Journals

Published: Jan 27, 2021

There are no references for this article.