Study site
Brewers Bay is located on the western end of St. Thomas, U.S. Virgin Islands (18° 20’ 28” N, 64°58’40” W) and is bounded by a commercial airport runway and small lagoon on the south, a sandy beach on the north-eastern shore, and a rocky headland and smaller bay (Perseverance Bay) to the northwest (Fig. 1). Brewers Bay is 1.8 km2 in area, ranges in depth between 0- 33.1 m (Fig. 1), and has steep vertical slopes along the airport runway and around the rocky headland. The bay is composed of a variety of habitat types including sand, seagrass, patch reefs, fringing coral reefs, rocky reefs, and rubble and reinforced concrete blocks (dolosse) around the seaward slopes of the airport runway. The lagoon is mostly soft muddy bottom with scattered rocks and dead corals. It is partly enclosed by the airport runway with the remaining shoreline composed of rocky reef or soft sediments, and red mangroves (Rhizophora mangle).
Acoustic Array
The acoustic monitoring system consisted of 45 omnidirectional receivers (VR2W, 69kHz, Vemco Inc, Halifax, NS, Canada), moored, and spaced equally across Brewers Bay, in the adjoining Perseverance Bay, and along the seaward side of the airport runway. (Fig. 1). Range testing of receivers [24] across the study site was conducted over four days in June 2015, by placing receivers in depths ranging from 5 to 19 m over different substrate types including shallow and deep coral/rock and seagrass/sand [25]. Probabilities of transmission were tested using three A69-1601 Innovasea (previously Vemco) transmitters V9-2H (151dB), V13-1H (153dB) and V16-4H (158dB) that transmitted every 60 seconds. Transmitters were attached to mooring lines, connected to cinder blocks, and suspended 1 m above the bottom. A detection probability of 70% for V13-1H transmitters was selected providing high coverage throughout the study area with estimated detection ranges of 101 m in seagrass/sand and 120 m in coral/rock substrates (Fig. 1). Water temperature and dissolved oxygen (DO) were collected at several stations in Brewers Bay using Hobo temperature loggers (Onset Computer Corporation, Bourne, Massachusetts) and miniDot DO loggers (Precision Measurement Engineering Inc, Vista, California) that were attached to acoustic receiver moorings. Temperature loggers were deployed in August 2015 and O2 loggers were deployed in February 2016 and both recorded at 15-minute intervals (Fig. 1).
Fish capture
All capture and tagging methodology on all fish in Brewers Bay was approved by the University of the Virgin Islands Institutional Animal Care and Use Committee (IRB #747807-1). Juvenile Atlantic tarpon were caught using hook and line from a boat or dock between September 2015 and November 2016. As each fish was reeled in, it was guided alongside the boat or dock and into a floating cradle constructed of PVC pipe, plastic mesh and foam noodles for buoyancy. Once in the cradle, the fish was turned upside-down to induce tonic-immobility and hook was removed from mouth. Fish remained immersed in open seawater the entire time, so no general or local anesthetic was administered, which also allowed us to release fish shortly after completion of data collection and tagging. Each fish (n=14) was measured for fork length (FL) and total length (TL) to the nearest millimeter (mm). Acoustic transmitters (either V13 (13 mm x 36 mm; n=8) or V13P (13 mm x 46 mm; n=6) 69kHz, Innovasea Inc, Halifax, NS, Canada) were surgically implanted into the body cavity on the ventral side of the fish [26]. The V13P transmitters provided depth data of fish. The incision was closed with surgical staples and treated with antibacterial ointment. Fish was turned back over, faced into the current to increase ventilation, and after a few minutes of recovery, fish was released at its capture location (Fig 1).
Data processing
Detections were downloaded from receivers every three months and analyzed using R Version 3.4.3[27]. Detections for each individual tarpon by receiver were plotted through time to investigate the presence of dropped tags, dead individuals, and short-term residency. Of the 14 juvenile tarpon that were tagged, this exploratory method was used to identify four (n=4) individuals with adequate data to conduct spatial home range analysis, eight (n=8) tarpon that were in array two days or less and had insufficient detections for analyses, and two (n=2) tarpon that either died or shed their tags (Table 1). Thus, four juvenile tarpon were considered resident and included in most analyses, whereas the remaining ten juvenile tarpon, whose fate was unknown, were excluded. Three of four resident tarpon had detections for 344 d to 472 d and also had pressure transmitters, thus were used to analyze monthly and seasonal trends in rates of movement, activity space, and vertical distribution (Table 1).
Temporal data were examined for seasonal and diel patterns. Seasons were defined as Spring (March, April, May), Summer (June, July, August), Fall (September, October, November) and Winter (December, January, February). Crepuscular periods were calculated using astronomical twilight based on daily sunrise/sunset time charts for Charlotte Amalie, St. Thomas, USVI [28]. Specifically, dawn was defined as -1 hr before Astronomical morning and +1 hr after sunrise to account for seasonal changes in day length. Likewise, dusk was defined as -1 hr before astronomical twilight to +1 hr after sunset. Day and night periods were the remaining hours between bracketed dawn and dusk, respectively.
Data Analysis and Statistics
Residency index - For each resident tarpon, the total number of detections, first/last day detected, number of days between first and last day, total days, and residency index within Brewers Bay array were calculated. Residency Index was defined as the percentage of time spent within Brewers Bay and was calculated by dividing total days detected within the array by number of days between the first and last detection.
Center of Activity - The center of activity (COA) location for juvenile tarpon (n=4) was calculated every 30 minutes using mean position (latitude and longitude) of all detections during that time step [29]. Distance between COA relocation points and difference in time between each relocation point were calculated for each fish using ‘adehabitatLT’ package of R environment [30]. COA values were used to calculate rate of movement (ROM) and activity space for individual fish, and included minimum convex polygons (100% MCP) and kernel utilization distributions (50% and 95% KUD).
Rate of movement – ROM (m/s) was calculated by dividing the distance between consecutive COA position values by the time difference between these consecutive points. Kruskal-Wallis and a Tukey post hoc test were used to test differences in ROM between diel periods and a two-way ANOVA tested differences in diel ROM across seasons.
Activity space - MCP, 50% KUD and 95% KUD were calculated using the ‘move’ and ‘adehabitat’ package in R environment [30,31]. MCPs provided information on the extent of an individual’s range or area used and included all outlying points that might be the result of exploratory movement or periodic migration not part of their typical activity. KUDs highlight the density of positions of an individual within the activity space based on COAs (i.e. 50% KUD = high density, 95% KUD = low density), as well as estimated error around these positions [32,33]. When necessary, a ‘land’ barrier polygon was used to clip out the area of MCP and KUD polygons that fell on land (rgeos package, [34]). The calculated MCP and KUD (50% and 95%) activity spaces were plotted in ArcGIS 10.6 for annual, monthly, and diel periods. To calculate the degree of overlap in 50% and 95% KUD among individuals over diel and monthly time periods, a Home Range (HR) percent overlap analyses was applied using the ‘kerneloverlaphr’ function of the ‘adehabitatHR’ package [30,35,36]. The percent overlap HR method analyses of the kerneloverlaphr function calculates the proportion of animal a’s home range that is overlapped by animal b’s home range [30,35,36]. The data output matrix with value of indices of overlap of all pairs of animals [35,36]. Using the matrix output, average and ranges in fish overlap values were calculated. Repeated Measures Analyses of Variance (RM-ANOVA) was used to test for differences in KUD across monthly and diel periods. All monthly analyses used data from only three (n=3) tarpon that had average KUD activity space representing each month (Table 1). Individual tarpon were treated as random variables, and either monthly or diel periods were treated for autocorrelation effects (corAR1) using the ‘lme’ function of the nlme package for R [37,38]. To assess relationship between monthly ROM and 50% KUD size, a linear regression was applied.
Vertical distribution - For resident tarpon with depth-enabled transmitters (n=3), depth measurements were binned into hourly and monthly periods and boxplots applied to elucidate their vertical movement patterns. ANOVA and Tukey post hoc tested for differences in vertical movement across both diel and monthly periods.
Environmental conditions - To assess relationship between daily average number of detections of tarpon and average temperature and dissolved oxygen within the lagoon and waters along the airport runway, a linear regression was applied for study period (September 2015-February 2018).