Using conventional and pop-up satellite transmitting tags to assess the horizontal movements and habitat use of thorny skate (Amblyraja radiata) in the Gulf of Maine

Abstract

Thorny skate (Amblyraja radiata) have experienced decreasing abundance and range contraction in the Gulf of Maine (GOM) in recent decades. To better understand the extent to which population structure, environmental conditions, and movement ecology may play a role in these disruptions, 128 “mark-report” pop-up satellite tags (mrPATs) and 2195 conventional tags were deployed from 2002 to 2019. Data obtained from 84 mrPATs and 43 conventional tag recaptures [127 individuals: 55 males, 72 females; 32–104 cm total length (TL)] revealed minimum linear horizontal movements of 0.4–46.8 km in all cardinal directions over periods 22–3435 d. There was no relationship between days at liberty, TL, sex, depth, reporting season, or tag type and minimum linear displacement, and no broad seasonal movements were evident. Skates were observed at depths 27–201 m and in water temperatures 2.5–12.5°C, with fluctuations in both depth and temperature evident by season. Given their restricted movements, thorny skate may represent a single stock/population with metapopulation-like structure in the GOM. The pervasiveness of sedentary behaviour may also place the species at risk of localized depletion and climate change but also demonstrates the potential efficacy of spatial closures for promoting population recovery.

climate change, metapopulation, population structure, seasonal movements

Introduction

Skates (family Rajidae) are a speciose group of cartilaginous fishes that are important epibenthic mesopredators in marine ecosystems worldwide (Last et al., 2016). Like other elasmobranchs, skates typically have low biological productivity (slow growth rates, late maturity, low fecundity; Frisk, 2010) rendering them vulnerable to population decline (Frisk et al., 2001). Despite their high contribution to global elasmobranch fisheries catches, and the resulting concerns over the conservation status of many species (Dulvy et al., 2014), skates have received considerably little research and management attention. Studies on skate movement, migration, and habitat selection patterns are particularly lacking, but it is generally assumed that skates exhibit geographically restricted home ranges and limited capabilities for large-scale movements due to their dorso-ventrally compressed morphology, mode of swimming, benthic association, and oviparous reproductive mode (Flowers et al., 2016; Siskey et al., 2019). To date, most of the empirical information on skate movement has been derived from conventional tag-recapture experiments (Siskey et al., 2019). However, the use of electronic (i.e. acoustic, archival, or satellite) tag technology to characterize movement and habitat use patterns has recently increased (e.g. Le Port et al., 2008, Wearmouth and Sims, 2009; Peklova et al., 2014; Neat et al., 2015; Farrugia et al., 2016; Frisk et al., 2019; Siskey et al., 2019). Fishery managers rely on basic movement data for stock identification, evaluating fisheries susceptibility, and establishing spatial fisheries management and conservation efforts (e.g. marine protected areas, time-area closures, habitat closures). Thus, there is a clear need to expand our scientific understanding of skate movements and habitat use.

The thorny skate (Amblyraja radiata) is a deep-water elasmobranch that occurs throughout much of the North Atlantic Ocean (Kulka et al., 2009). Off the northeast United States (US), the species predominantly occurs in the Gulf of Maine (GOM), which represents the southern fringe of its western North Atlantic distribution (Packer et al., 2003). Historically, thorny skates were considered one of the most common (McEachran and Musick, 1975) and most plentiful (Bigelow and Schroeder, 1953) deep-water skates in the GOM, but they have primarily been landed only as incidental catch in fisheries targeting more valuable species. Despite the lack of a directed fishery, thorny skate abundance in the GOM has declined by 80–95% since the 1960s [National Marine Fisheries Service (NMFS), 2017] and continues to decline in certain areas of the GOM (Sosebee et al., 2016). In addition, there is evidence that their distribution in the GOM has constricted around historical centres of abundance over the past 40 years (Nye et al., 2009). Due to the thorny skate’s chronically poor population status in US waters, it has been a prohibited (zero-possession) species under the New England Fishery Management Council (NEFMC) Northeast Skate Complex Fishery Management Plan since 2003 [New England Fishery Management Council (NEFMC), 2003], is classified as a National Oceanic and Atmospheric Administration (NOAA) Species of Concern, and is considered “Critically Endangered” by the IUCN Red List (Kulka et al., 2009). However, in spite of extensive conservation concerns and restrictive management measures, the GOM population remains overfished and there has been little apparent recovery [New England Fishery Management Council (NEFMC), 2019].

The factors driving the continued population decline and range contraction of thorny skate within the US GOM remain largely unknown. Although many factors may be at play [e.g. discard mortality (Mandelman et al., 2013; Knotek et al., 2019); environmental change (Nye et al., 2009)], the spatial scale at which any factor impacts thorny skate in the GOM is difficult to determine given the absence of information on their movement patterns and population structure in this rapidly warming region (Belkin, 2009). For example, it is not known if individuals in the GOM represent a single discrete population, or if they freely exchange with conspecifics throughout their broader western North Atlantic range. Previous thorny skate conventional tagging studies conducted off Newfoundland, Canada (Templeman, 1984), and in the North Sea (Walker et al., 1997) indicate that the species is rather sedentary and generally exhibits horizontal movements on the order of 50–100 km from tagging sites over periods of 0.2–20 years. However, broader horizontal movements of 180–444 km were also observed over 0.2–11 years in small number of tagged individuals, thereby demonstrating the species’ capability of broad-scale movement. Thorny skate genetic diversity is also relatively high throughout the North Atlantic (Chevolot et al., 2007; Lynghammar et al., 2014), with three major genetic clusters existing in the northeast Atlantic, Greenland/Iceland region, and the northwest Atlantic (Lynghammar et al., 2016). Of these, the Greenland/Iceland cluster exhibits the highest level of genetic diversity and may represent an admixture zone between the northeast and northwest Atlantic clusters (Chevolot et al., 2007; Lynghammar et al., 2016). In addition, the large amount of genetic haplotypes within each cluster indicates that fine-scale genetic structure may occur within each discrete region of the North Atlantic. Given these findings, physical mixing (via horizontal movement or migration) may play a key role in thorny skate population structure in the broader North Atlantic (Lynghammar et al., 2016), as well as within each cluster.

Population structure, habitat use, and the degree to which thorny skate can move and disperse throughout their range have direct implications for their susceptibility to population decline. For example, sedentary behaviour and slow rates of movement (relative to spatial fishing intensity) have been attributed to the localized depletion of thorny skate in the northeastern GOM and Canadian Scotian Shelf (Shackell et al., 2005). In addition, a shift in thorny skate distribution in the southern Gulf of St. Lawrence, Canada, has been linked to cooling bottom water temperatures in their preferred habitat and resulted in the geographic constriction of individuals into sub-optimal foraging habitat (Swain and Benoît, 2006). This distributional shift may also be impeding population recovery (through decreased fitness) and even exacerbating population declines due to the increased susceptibility of thorny skate to fishing mortality and/or predation in the area to which they have contracted (Swain and Benoît, 2006).

Given the clear link between thorny skate movements, habitat use, population structure, and population decline and recovery, a thorough understanding of these biological characteristics is critical to the effective management and conservation of the species in the GOM. In particular, such information is needed for the establishment of appropriate US stock boundaries, an evaluation of the efficacy of existing domestic and international management policy (including spatial closures), an assessment of the species’ vulnerability to environmental change, and to understand the spatial scale over which thorny skate are affected by fishing pressure and environmental conditions in the GOM. To address these needs, the objective of this study was to use conventional tag-recapture analyses and pop-up satellite archival transmitting (PSAT) tags to investigate the horizontal movements, dispersal, and habitat use of thorny skate in the US GOM.

Methods

Sampling and tagging

Thorny skate were captured in the US portion of the GOM from 2002 to 2019 with bottom otter trawls, sink gillnets, and demersal longlines on several research platforms including for-hire trips on commercial fishing vessels (F/V Mystique Lady, Gloucester, MA; F/V Lady Victoria, Seabrook, NH; F/V Mary Elizabeth, Scituate, MA), the NOAA Northeast Fisheries Science Center (NEFSC) GOM Bottom Longline (BLL) survey (McElroy et al., 2019), and the Massachusetts Division of Marine Fisheries (MADMF) Atlantic Cod Industry Based Survey (IBS). Following capture, each skate was measured for total length (TL; from the tip of the rostrum to the tip of the tail; cm), sexed, and tagged. All sampling and tagging was conducted using the New England Aquarium approved Animal Care and Use Protocols 2015-24 and 2016-02.

The maturity status of each skate was also assessed by macroscopic visual examination during all tagging trips in which one or more of the authors were present. Males were considered mature if claspers were large and fully calcified (Sulikowski et al., 2006). For females, the assessment of maturity status was more difficult due to the lack of external secondary sex characteristics that are indicative of sexual maturity. However, females were assumed to be mature if they possessed a distended cloaca, which is evidence of recent mating and/or ovoposition. This technique has previously been used to assess the maturity of skates in the GOM (Sosebee, 2004) and has been validated by the authors as a reliable, non-invasive method for assessing female maturity in this species. Note that maturity could not be assessed using published lengths at maturity in the GOM (e.g. Sulikowski et al., 2006) due to the highly variable size at maturity that is evident in thorny skate throughout the study area (Sosebee, 2004; Sulikowski and Kneebone, unpublished data). Maturity was also not assessed for any skate on the MADMF Cod IBS and was only assessed for a portion of skates tagged on the NEFSC BLL survey following Sosebee (2004). Skates for which maturity status could not be assessed were considered “unknown”.

Conventional tagging

Thorny skate measuring >20 cm TL were tagged with a conventional dart tag (Floy^® FT-94, Floy Tag & Mfg., Inc., Seattle, WA, USA) at the base of the tail and immediately released. Capture date and release location (latitude and longitude) were noted for each fish. Movements of conventional tagged skates were determined by recapture and data reporting. Recapture date, recapture location (latitude and longitude), skate size (TL or weight), and the name of the reporter were noted for all recaptures; additional information on recapture depth and skate condition was also recorded, if provided. Small monetary rewards were provided to those who reported recaptured tags. All tagging and recapture data compiled through the end of 2019 were included in this analysis.

Satellite tagging

Due to low recapture rates of conventional tags, it was necessary to use technology that could examine the horizontal movements of thorny skate in a fishery-independent framework. Accordingly, a subset of 128 thorny skate measuring >50 cm TL were tagged with Wildlife Computers (Redmond, WA, USA) “mark-report” pop-up satellite tags (mrPATs) from 2015 to 2017. The mrPAT is considerably smaller and lighter (length: 121 mm, diameter: 23 mm, weight in air: 26 g) than previously available PSAT models. Preliminary trials with captive thorny skate at the University of New England Marine Science Center showed that animals >50 cm TL were capable of supporting an mrPAT without adversely affecting animal health or normal swimming mechanics. To avoid potentially confounding our results, smaller individuals (≤50 cm TL) were not considered for mrPAT monitoring.

MrPATs were programmed to automatically release from tagged skates after pre-programmed periods at which time they would float to the surface and transmit their location to Argos satellites, thereby providing a fishery-independent “recapture” location. In addition, each tag also transmitted daily minimum and maximum temperature data (sensor range: −20 to 50°C, sensor accuracy: ±0.1°C) and daily maximum change in tilt (“delta tilt”; recorded via an accelerometer sensor) over the 100-d period preceding pop-up (i.e. the last 100 d of the deployment period). MrPATs were attached to the skate’s pectoral wing near the midline of the body using a method modified from Wearmouth and Sims (2009), wherein a monofilament line and Petersen discs (diameter: 25 mm) and baffles (diameter: 10 mm) were crimped on either side of the skate (Knotek et al., 2019). Tags were programmed for deployments of 30 (n = 10), 100 (n = 68), 200 (n = 48), and 300 (n = 2) d (Table 1). The majority of tags (84/128; 78%) were treated with anti-fouling paint prior to deployment. All mrPAT-tagged skates were maintained in onboard live wells that were continuously fed with fresh, ambient seawater immediately after capture and only select skates that showed no signs of apparent morbidity, no overt physical trauma, and high relative level of vigour were tagged. In addition, to minimize the potential impact of sampling and tagging on post-release fate, individual skates were processed as quickly as possible (1–6 min) and released immediately.

Table 1.

Summary of metadata for skates tagged with Wildlife Computers mrPATs and conventional tags.

Tag type	Total tags	TL (cm)	Male		Female		Unknown sex
			n	TL (cm)	n	TL (cm)	n	TL (cm)
mrPATs (2016–2017)
Deployments	128	52–105 (71 ± 13)	63	52–105 (69 ± 13)	65	55–101 (74 ± 14)	–	–
Reporting tags	84	55–104 (73 ± 13)	38	56–104 (72 ± 13)	46	55–99 (74 ± 13)	–	–
Conventional tags (2002–2019)
Deployments	2 195	20–109 (54 ± 16)	950	20–109 (54 ± 16)	1 105	22–106 (54 ± 16)	140	34–99 (59 ± 15)
Recaptures	43a	32–90 (63 ± 16)	17	39–90 (62 ± 15)	26	32–86 (63 ± 16)	–	–
All reporting tags	127	32–104 (69 ± 15)	55	39–104 (69 ± 14)	72	32–99 (70 ± 15)	–	–

Data are presented for all deployed tags and for reporting mrPATs and recaptured conventional and mrPATs. Total length [TL; range (mean ± SD)] and the number of individuals (n) of each sex are presented.

a

Includes two mrPAT recaptures.

Data analysis

All analyses were performed in the R Statistical Environment (v3.6.1, R Core Team, 2019). Horizontal movement maps were created using QGIS (v3.12.1).

mrPAT reporting location

In order for an mrPAT pop-up location to represent an accurate relocation, the tag must have: (i) been attached to its bearer at the time release was initiated and (ii) reported a position within a reasonable amount of time following detachment. Although some PSAT models have the capability of transmitting when they prematurely detach from their bearer and float to the surface, the mrPATs that were used in this study did not have this capability. Thus, for each mrPAT, it was necessary to confirm that the tag was attached to the skate at the time of release; prematurely detached mrPATs did not report until the end of their scheduled deployment period, regardless of when they detached and surfaced, and therefore did not provide locations representative of a skate’s location. To determine if mrPATs were attached to the skate at the time of release, archived daily minimum and maximum water temperature and delta tilt values were scrutinized over the last 100 d of the tag’s deployment. If daily delta tilt values (a proxy for tag orientation) were markedly and systematically different than would be expected if the tag was attached to the animal (i.e. floating vertically) and daily water temperatures were inconsistent with ambient bottom temperature in the vicinity of the tag’s reporting location, the tag was determined to have prematurely detached from the skate at some point during the deployment (Figure 1). If detachment occurred within the last 100 d of the deployment (i.e. the only period for which delta tilt and temperature data were available), an estimated date of detachment was noted.

Figure 1.

Example of daily minimum temperature and delta tilt data used to identify the premature detachment of Wildlife Computers mrPAT from their bearer. The vertical line in each plot denotes the date that systematic differences in both the ambient temperature and delta tilt records began, and thus the estimated date of tag detachment.

For all tags that were attached to the skate at the time release was initiated, the first transmitted Argos location of quality 1 or higher (i.e. 1, 2, or 3) was considered to represent its bearer’s location. This location quality restriction ensured that Argos locations were accurate to within 1000 m of the skate’s true location. Argos locations that were received within 12 hr of tag detachment were considered “accurate” locations and were included in all subsequent statistical analyses of horizontal movement. Locations that were received 12–36 hr after detachment from the animal were considered “approximate” and were not included in any statistical analyses but were instead plotted to examine general movements. Due to the surface currents evident in the GOM (Pettigrew et al., 2005), transmitted locations obtained >36 hr after tag detachment were not reliable indicators of their bearer’s location and were excluded from further analysis.

Horizontal movement

Relocation data obtained from mrPATs that reported within 12 hr of release and conventional recaptures were analogous and, thus, were analysed in an identical manner. For each tag, the minimum linear displacement distance (km) and displacement angle (degrees) between tagging and reporting locations were calculated using the “raster” (Hijmans, 2017) and “maptools” (Bivand and Lewin-Koh, 2018) packages, respectively. Time at liberty was also calculated for each tag as the date of tagging to the date of recapture/reporting. To qualitatively examine the distance and direction of thorny skate displacement relative to the seasons of tagging and tag reporting/recapture, polar plots were constructed using the “plotrix” package (Lemon, 2006). Seasons were defined meteorologically as: winter (December–February), spring (March–May), summer (June–August), and fall (September–November).

To examine the influence of biotic (e.g. TL, sex) and abiotic (e.g. tag type, days at liberty, reporting/recapture season, depth at release) variables on thorny skate minimum linear displacement from tagging locations, generalized linear models were constructed using the “glm” function in R. Models were fit using a Gamma distribution with TL and days at liberty included as continuous variables and sex (male/female), tag type (mrPAT/conventional), reporting/recapture season (winter/spring/summer/fall), and depth class (shallow or deep; see below) included as categorical variables. Maturity status was not included in the models due to the inability to assign a maturity status to each skate. Temperature was also not included in any models due to the inability to confidently assign bottom temperature (at tagging and recapture) to all conventional tags throughout the time series. Correlation tests were performed prior to modelling, and no strong correlations were evident among any variables. A fully saturated model was then constructed with all variables as fixed effects and an interaction term between days at liberty and reporting/recapture season. Stepwise backward model selection was then performed by removing individual terms that were non-significant (p > 0.05). Residual plots were examined for all models, and nested models were ranked and compared by their corrected Akaike Information Criterion (AICc; Burnham and Anderson, 2002) scores and their deviance explained. Models with AICc scores within 2 units were compared with likelihood ratio tests to evaluate their relative fit to the data.

Depth and temperature

To examine depth distribution of thorny skate by season, the local water depth was extrapolated at tagging and at each mrPAT pop-up location or conventional recapture location using the NOAA ETOPO1 Ocean Relief Model (doi:10.7289/V5C8276M) at 1 min (latitude/longitude) resolution in the “marmap” package (Pante and Simon-Bouhet, 2013). Preliminary examination of depth data revealed a broad and bi-modal range of depths at both tagging and tag reporting/recapture with a clear division at ∼120 m. Thus, to examine seasonal changes in depth distribution of skates throughout the study area, tagging and reporting/recapture depths were classified as “shallow” (≤120 m) and “deep” (>120 m). Boxplots of the depth at tag reporting/recapture were created for each depth class. The absolute change in depth from tagging to reporting/recapture locations was also calculated, and basic summary statistics were calculated for each depth class.

Daily minimum and maximum temperature data were compiled for each day in which an mrPAT remained attached to its bearer. Data were obtained from tags that remained on the skate for the full programmed period and tags that detached prematurely (only up to the day of detachment, when determined), regardless of the interval from detachment to reporting via Argos. To evaluate the extent to which daily temperature profiles differed throughout the sampled area of the GOM, skates were grouped based on depth class at tagging (shallow = ≤120 m; deep = >120 m) and boxplots were created to examine temperatures experienced by thorny skate by month and season.

Results

mrPAT

Thorny skate tagged with mrPATs (n = 128; 63 males, 65 females) measured 52–105 (mean ± SD: 71 ± 13) cm TL, of which 58 were sexually mature (27 males, 31 females), 60 were immature (34 males, 26 females), and 10 (2 males, 8 females) were of unknown status (Table 1 and Figure 2). Argos data transmissions were received from 116 (91%) mrPATs, and 11 (9%) tags did not report. Two mrPAT-tagged skates were recaptured during the study, one of which was deceased at the time of capture (ID: 159975) and another that died after being re-released (ID: 159891); Argos data were only received for tag 159891. Since accurate relocation data for these two tags were available solely via recapture, they were analysed as conventional recaptures (see below). Of the remaining 115 reporting mrPATs, 86 were determined to have been attached to their bearer when the release sequence was initiated (67% of all deployments). Of these, 82 yielded “accurate” locations (i.e. reported 0.3–9.4 hr after detachment from the skate), 2 yielded “approximate” locations (i.e. reported 27.2–29.5 hr after detachment), and 2 yielded “inaccurate” locations 5 and 32 d after detachment. For the remaining 29 reporting mrPATs, 24 (19%) detached from their bearer prematurely and floated for variable periods before reporting, and 5 (4%) transmitted weak signals with no associated geographic location (purported antennae failures). No viable horizontal movement data were obtained from these 29 tags.

Figure 2.

Map of the study area and the tagging locations for the 2195 conventional tags and 128 Wildlife Computers mrPATs deployed in the GOM. Note that numerous tags were deployed in very close proximity and are therefore overlapping. NOAA ETOPO1 Ocean Relief Model bathymetry data are presented to illustrate continuous depth over the study region. The western GOM (short dashed line) and Cashes Ledge (long dashed line) closed areas and the US Exclusive Economic Zone boundary (solid red line) are also shown. Map inset depicts the study area (black rectangle) with the 100-m bathymetric contour.

Conventional tags

A total of 2195 individuals (950 males, 1105 females, 140 unknown sex; 288 mature, 933 immature, 974 unknown status) measuring 20–109 (55 ± 16) cm TL were tagged with conventional tags from 2002 to 2019 (Table 1 and Figure 2). From these, a total of 41 recaptures were reported representing deployment periods of 22–3435 (median = 349) d, a reporting rate of 1.9%. Tags were primarily reported by commercial fishermen (trawlers and gillnetters); one recapture occurred during a NEFSC research trip, one during the NEFSC BLL survey, and two during the MADMF Atlantic Cod IBS.

Horizontal movements

Viable horizontal movement data were obtained from 127 skates [32–104 (69 ± 15) cm TL] including 55 males (21 mature, 26 immature, 8 unknown status) and 72 females (30 mature, 28 immature, 14 unknown status) (Table 1 and Figure 3). Data were statistically analysed for 125 skates (55 males, 70 females) with accurate relocation data originating from 82 mrPATs, 41 conventional tags, and 2 recaptured mrPATs. Approximate relocations were also plotted for two additional mrPAT-tagged skates (77 and 91 cm TL mature females). Relocation locations for mrPATs and recaptures were recorded in every calendar month. In general, thorny skate remained rather sedentary, exhibiting minimum linear displacements 0.4–46.8 (11.7 ± 8.5) km over periods 22–3435 (median = 268) d (Table 2). Results of generalized linear models indicated that there was no relationship between the number of days at liberty, TL, sex, tag type, recapture/reporting season, or depth and minimum linear displacement (Table 3). Horizontal movements occurred in all cardinal directions at 3–354° (149 ± 89°) relative to tagging locations and were similar for males and females, as well as mature, immature, and unknown maturity status individuals (Table 2 and Figure 3). Visual inspection of horizontal movement data revealed movements to be more random and not indicative of movement towards a discrete location/area during a given season, or to differ markedly by sex or maturity status (Figures 3 and 4).

Figure 3.

Seasonal horizontal movements of 127 thorny skates derived from 84 Wildlife Computers mrPATs with accurate or approximate locations and 43 conventional recaptures (including two mrPATs). The head of each arrow represents the pop-up (mrPAT) or recapture (conventional) location. NOAA ETOPO1 Ocean Relief Model bathymetry data are presented to illustrate continuous depth over the study region. The western GOM (short dashed line) and Cashes Ledge (long dashed line) closed areas and the US Exclusive Economic Zone boundary (solid red line) are also shown. Map inset depicts movements in relation to the Cashes Ledge closed area.

Figure 4.

Polar plots representing the angular displacement (degrees) of 127 tagged thorny skates with accurate or approximate locations relative to its tagging location by the season of deployment (different plots) and reporting/recapture (colors). Circular grids within the plots represent the minimum linear displacement distance travelled.

Table 2.

Summary of minimum linear displacements (km) and angular displacements (degrees relative to tagging location) exhibited by 125 thorny skates with accurate pop-up or recapture locations.

Movement metric	Mature	Immature	Unknown	All skates
Minimum linear displacement (km)
Male	1.2–32.8 (10.3 ± 8.7); 21	2.9–46.8 (15.7 ± 10.7); 26	1.2–19.0 (10.4 ± 5.9); 8	1.2–46.8 (12.9 ± 9.6); 55
Female	2.1–25.5 (10.3 ± 6.2); 28	1.2–25.7 (11.6 ±7.6); 28	0.4–36.2 (10.0 ± 9.5); 14	0.4–36.2 (10.8 ± 7.4); 70
Overall	1.2–32.8 (10.3 ± 7.3); 49	1.2–46.8 (13.6 ± 9.4); 54	0.4–36.2 (10.1 ± 8.2); 22	0.4–46.8 (11.7 ± 8.5); 125
Angular displacement (°)
Male	21–348 (161 ± 100)	19–353 (150 ± 95)	38–275 (160 ± 79)	19–353 (156 ± 93)
Female	34–352 (150 ± 91)	3–354 (140 ± 91)	36–324 (136 ± 72)	3–354 (143 ± 86)
Overall	21–352 (155 ± 94)	3–354 (145 ± 92)	36–324 (145 ± 74)	3–354 (149 ± 89)

Data are presented for individuals of each sex that were classified as sexually mature, immature, and unknown maturity status. The range (mean ± SD) and sample size are presented for each category.

Table 3.

Backward model selection, AICc ranking, and goodness of fit for generalized linear models fit to MLD from tagging location for 125 thorny skates with accurate relocations.

Model structure	Residual df	Deviance explained (%)	AICc	ΔAICc
MLD ∼ 1 + TL + sex + tag type + depth + DAL × Rseason	113	8.97	858.20	12.37
MLD ∼ 1 + TL + sex + tag type + depth	120	2.53	850.94	5.11
MLD ∼ 1 + TL + sex + tag type	121	2.53	848.73	2.90
MLD ∼ 1 + TL + sex	122	2.44	846.70	0.86
MLD ∼ 1 + TL	123	1.21	846.27	0.44
MLD ∼ 1	124	0.00	845.83	0.00

The final and most parsimonious model is presented in italics. df, degrees of freedom; DAL, days at liberty; MLD, minimum linear displacement; Rseason, reporting/recapture season; TL, total length.

Depth and temperature

At the time of mrPAT reporting or conventional tag recapture, thorny skate were observed in local depths of 27–201 (114 ± 50) m. Both males [27–201 (115 ± 50) m] and females [27–188 (112 ± 49) m] and mature [29–201 (107 ± 51) m], immature [27–185 (116 ± 47) m], and unknown maturity status [69–200 (143 ± 52) m] skates were observed in similar depths. Skates monitored west of 70°W longitude were generally observed at shallower depths [n = 83; 23–155 (80 ± 27) m] than those tagged to the east of this longitude [n = 42; 118–203 (164 ± 20) m]. Skates tagged in shallower waters ≤120 m (n = 83) were observed in similar mean depths over all seasons but were present in slightly deeper waters in the winter and shallower waters in the fall (Figure 5a). Skates tagged at deeper depths >120 m (n = 42) were observed in similar mean depths during the winter, summer, and fall but generally occupied shallower waters in the spring, which is the period when the coldest water temperature are evident at deeper depths (Figure 5b). From tagging to reporting/recapture locations, skates exhibited changes in depth from −76 to 80 m (negative values indicate a move to deeper depths), with similar changes in depth evident for skates tagged in shallow [−76 to 65 (median = 4) m] and deep [−59 to 80 (median = 3) m] waters. The vast majority of skates (94%) remained with the same depth class from tagging to reporting/recapture, with the exception of six skates that moved from shallow to deep waters and two skates that moved from deep to shallow waters.

Figure 5.

Boxplots of the interpolated depth at the mrPAT reporting or conventional tag-recapture location by season for 125 skates with accurate relocations (a and b) and the mean daily temperature by month and season experienced by 100 thorny skates tagged with Wildlife Computers mrPAT (c and d) for each depth class. The number of skates observed in each month or season is provided. Note the different y-axis scales among (a) and (b) and (c) and (d).

A total of 8563 d of minimum and maximum temperature measurements were available from 100 mrPATs, including 83 full-term tags and 17 that detached prematurely. During the study period, thorny skate inhabited water temperatures from 2.5 to 12.5°C in the GOM, with skates tagged at shallower depths ≤120 m and west of 70°W experiencing a broader temperature range (n = 53; 2.5–12.5°C) than those tagged at deeper depths >120 m and east of 70°W (n = 47; 4.0–10.0°C; Figure 5c and d). A seasonal temperature profile was evident in each depth class throughout the year, consistent with seasonal environmental temperature fluctuations in the GOM region.

Discussion

Here, we report on the results of a 17-year cooperative study that used conventional and mrPATs to examine the horizontal movements and habitat use of thorny skate within the rapidly warming (Pershing et al., 2015, Saba et al., 2016) GOM region. By deploying two tag types over a broad geographic area, we were able to provide novel insight into the short- (months) and long- (years) term movement patterns of thorny skate and make inferences about their population structure in the region. In addition, we were able to measure depth and temperature profiles experienced by thorny skate over all seasons and comment on how temperature may be affecting movements of individuals and the broader population. Collectively, this information advances our understanding of thorny skate life history and spatial ecology in the US GOM, will assist with future management and rebuilding of thorny skate, and foster an improved understanding of their susceptibility to localized depletion and climate change.

Horizontal movements

The results of this study corroborate previous tagging data in other regions that thorny skate are relatively sedentary and primarily exhibit small-scale, localized movements of <100 km in both the short and long terms. In the US GOM, individuals did not exhibit movements in excess of ∼50 km and there was no apparent relationship between time at liberty and minimum linear distance travelled. By comparison, Templeman (1984) reported that “usual” movements of tagged thorny skate were within 111 km of release locations off the coast of Newfoundland, Canada, while 85% of tagged thorny skates exhibited movements of <92 km in the North Sea (Walker et al., 1997). In addition, although both Templeman (1984) and Walker et al. (1997) documented more extensive horizontal movements than those observed in the GOM (i.e. 444 and 180 km, respectively), neither identified a relationship between displacement distance and time at liberty. For example, Templeman (1984) reported that “several” individuals were recaptured “very close” to tagging areas after 14–16 years at liberty while “some” others moved 180–200 km over periods of 0.2–0.6 years (∼70–220 d). It is unclear why thorny skate do not undergo more extensive movements in the GOM; however, data from Templeman (1984) and Walker et al. (1997) indicate that individuals were physically capable of undertaking broad-scale movements >50 km during the period over which they were monitored by this study (i.e. 70–3435 d).

Thorny skate monitored in the GOM also showed no evidence of large-scale (i.e. >100 km) seasonal movements, which have been documented or postulated to occur in conspecifics in Canadian waters (Macdonald et al., 1984; Clay, 1991; Simpson et al., 2012). There was also no evidence that movement occurred as a function of sex, suggesting that males and females exhibit similar movements. These findings align with Templeman (1984) who reported no evidence of seasonal movements in relation to spawning/reproduction in the vicinity of Newfoundland, Canada. It should be noted, however, that the nature of the conventional and mrPATs used in this study did not permit the continuous monitoring of individuals or groups of individuals. Thus, it is possible that thorny skate exhibited movements to distant areas during their time at liberty and returned back to near their location of tagging prior to recapture or mrPAT reporting. However, given the consistency between movements observed over periods of 30–300 d and published movement rates (0.002–0.091 km d⁻¹; Templeman, 1984) such movement activity seems unlikely. Nonetheless, additional continuous, fine-scale monitoring of individual skate over multiple years is necessary to reconcile how individual movements may change from year to year in response to temperature or reproduction (Frisk et al., 2019, Siskey et al., 2019). While horizontal movements may generally be limited in skates, they retain the physical capability for larger-scale movements and migration, although the drivers of such movements require more study (Frisk et al., 2019).

The lack of movement and connectivity among thorny skate within the US GOM provides some evidence that the species may display metapopulation-like structure in this region. Data from the NEFSC bottom trawl survey indicate that thorny skate have historically been distributed continuously throughout the GOM (Hogan et al., 2013); however, several distinct areas of localized abundance are evident (see Figure 2 in Curtis and Sosebee, 2015). Based on the extent of movements in the GOM, there is evidence that limited exchange may occur between these aggregations, at least on an annual or bi-annual basis. If true, this lack of connectivity, at least in the short term, suggests that thorny skate in a given area may be somewhat reproductively isolated from other areas. Thorny skate reproduce year round in the GOM (Sulikowski et al., 2005, Kneebone et al., 2007); thus, the perpetual existence of individuals in a discrete area may restrict mating with other individuals that occupy habitat within or adjacent to that area. The lack of seasonal movements also suggests that individuals from distant areas are not moving/migrating to a common area to mate, which, when combined with the depletion of the population throughout the GOM, may promote reproductive isolation of these aggregations.

Previous genetic studies on thorny skate have reported that a high proportion of physical mixing (i.e. gene flow) may occur over large spatial scales between the eastern and western North Atlantic (Lynghammar et al., 2016) and that the species’ migratory range is much greater than previously acknowledged (Chevolot et al., 2007). Curiously, there is also evidence of low haplotype diversity with discrete areas (e.g. North Sea; Chevolot et al., 2007) throughout the North Atlantic (Lynghammar et al., 2016), which suggests that local adaptation and genetic differentiation has occurred in the species on a smaller geographic scale. Considering these findings in the spatial and temporal context of our study, it is possible that thorny skate maintain (or have maintained) gene flow by undergoing broader movements a generational scale (once in a lifetime). However, the prevalence of restricted movements in all of the tagging data collected for the species in the North Atlantic (i.e. this study; Templeman, 1984; Walker et al., 1997) provides considerable evidence that long distance migration is rare in thorny skate, thereby promoting genetic substructure over broad areas (e.g. western North Atlantic; Lynghammar et al., 2016). Clearly, additional long-term, continuous monitoring both at the individual and aggregate levels is needed to better assess the connectivity of thorny skate movements in the GOM and to evaluate the extent to which distinct, reproductively isolated (or semi-isolated) groups exist throughout the broader region (Siskey et al., 2019). To this end, additional genetic testing of multiple markers in large number of skates sampled throughout the greater North Atlantic is required to reconcile the species’ population structure and to determine if the GOM supports a discrete stock or subpopulation.

Habitat use

Daily minimum and maximum temperate data obtained from mrPATs revealed that thorny skate occupy a wide thermal range in the GOM (2.5–12.5°C), which is largely consistent with published results using other methods. Previous fisheries-dependent data have reported thorny skate captures at temperatures ranging from −1.4 to 14°C from Nova Scotia to Cape Hatteras (McEachran and Musick, 1975) and from 2 to 13°C in the GOM (Packer et al., 2003), with most individuals observed between 4.5 and 9.5°C and markedly fewer in temperatures warmer than 10°C (Sosebee et al., 2016). However, thorny skate monitored in the GOM had broader thermal ranges than reported for the species in the Gulf of St. Lawrence, Canada (∼−1 to 6°C; Swain and Benoît, 2006), on the Grand Banks (−0.3 to 5°C; Pennino et al., 2019), and off the east coast of Greenland (0.1–5.1°C; Jørgensen et al., 2015).

There was some evidence that thorny skate exhibit seasonal shifts in depth distribution, perhaps in response to temperature (e.g. Figure 5). However, in general, thorny skate appear to experience natural temperature fluctuations throughout the year rather than moving to seek out a narrow (preferred) temperature range. For example, the differences in annual temperature range experienced by skates tagged in shallow (≤120 m) and deep (>120 m) waters reflect the variability in available thermal habitats across the GOM, with the deeper eastern portion of the study region (roughly east of 70°W) generally having more stable seasonal temperatures. Some thorny skates that remained in waters ≤120 m experienced strong seasonal temperature gradients as high as 10°C, which is markedly higher than that previously reported for the species in the GOM region (e.g. 2.34°C; Hogan et al., 2013). However, it is comparable to the seasonal thermal range of other more mobile skate species based on NEFSC spring and fall trawl survey capture records (e.g. winter skate, Leucoraja ocellata: 7.84°C; little skate, Leucoraja erinacea: 7.87°C; clearnose skate, Raja eglanteria: 9.84°C; Hogan et al., 2013). The lack of selection for a narrow temperature range is at odds with previous reports that thorny skate change their distribution to remain within their preferred thermal habitat (−0.5 to 3°C) on the Grand Banks (Pennino et al., 2019). Functionally, it is also possible that the relatively wide temperature range evident in the GOM region obviates the need for thorny skate to undergo broad-scale seasonal migrations to avoid unfavourably low ambient water temperatures (<2°C) as has been theorized in the Gulf of St. Lawrence and on the Grand Banks (Clay, 1991; Darbyson and Benoît, 2003; Swain and Benoît, 2006).

Differences in the temperatures experienced by thorny skate in the GOM also calls into question the effect that temperature and depth may be having on the abundance and distribution of the species in the region. Warming bottom temperatures in the GOM have been linked to a significant range contraction and shift of thorny skate to deeper, cooler depths from 1968 to 2007, although there was no concomitant evidence of a temperature-influenced shift in the centre of biomass (Nye et al., 2009). Visual inspection of NEFSC trawl survey data over this same time period corroborates a contraction of thorny skate biomass into deeper waters but also indicates a recent contraction into the western GOM (Sosebee et al., 2016), which is a shallower area that (according to our tag data) exhibits the warmest bottom water temperatures experienced by the skates during the summer and fall. Interestingly, thermal habitat models suggest that this area of the western GOM (i.e. where thorny skate have contracted) is less suitable habitat than exists in deeper areas of the central and eastern GOM (Kleisner et al., 2017). In addition, no obvious thermal barriers exist that would limit the species’ ability to move northward (e.g. to higher latitudes) in response to warming temperatures in the GOM. As a result, it seems logical that other factors, not just temperature, may be driving the range contraction of thorny skate in the GOM. For example, recent work indicates that benthic habitat type (e.g. soft vs. hard bottom) is important to the scale and spatial pattern of thorny skate distribution and population trends in US waters (Sosebee et al., 2016), including under various climate scenarios (Grieve, unpublished data). Thorny skate have also been demonstrated to prefer habitats where certain prey species, such as snow crab (Chionoecetes opilio), are available (Pennino et al., 2019), thereby suggesting that prey distribution may be an influential factor.

Low population biomass may be another factor influencing thorny skate range contractions in the GOM. Previous studies have demonstrated weak correlations between thorny skate biomass and water temperature in the Gulf of St. Lawrence, Canada, wherein skates tended to occupy warmer temperatures (3–6°C) and deeper depths (∼200 m) when their biomass was low (Swain and Benoît, 2006; Tamdrari et al., 2015). A similar shift to warm waters at low population biomass has also been observed in Atlantic cod (Gadus morhua) in the southern Gulf of St. Lawrence (Swain and Kramer, 1995) and for populations of American plaice (Hippoglossoides platessoides) in waters off Newfoundland (Swain and Morgan, 2001). More recently, the shift of thorny skate into deeper, warmer waters in the Gulf of St. Lawrence was found to be more closely linked to predation risk from grey seals (Halichoerus grypus) than shifting temperature regimes (Swain et al., 2015). In the GOM, it is unlikely that similar predation risk is a key driver of the contraction since skates have shifted towards the shallower area of the western GOM where such predation risk would theoretically be higher. The contraction of thorny skate into the western GOM is similar to Atlantic cod, which are speculated to have shifted their distribution southwestward in the GOM due to a range of factors including prey availability (Ames and Lichter, 2013), population biomass (Guan et al., 2017a), habitat suitability (Guan et al., 2017a), and metapopulation dynamics (Guan et al., 2017b). As sympatric, benthic species in the North Atlantic that possess metapopulation-like characteristics, it is possible that thorny skate biomass has also contracted into the GOM due to similar factors, such as habitat availability (Sosebee et al., 2016).

Previous studies suggest that thorny skate are metabolically sensitive to acute increases in temperature (Schwieterman et al., 2019) and are highly susceptible to climate change (Hare et al., 2016). Indeed, there is some evidence that the species has already undergone slight shifts in distribution and depth in the GOM in response to warming temperatures (Nye et al., 2009). However, given the aforementioned inconsistencies with respect to thorny skate temperature preferences, the true impacts of climate change on thorny skate distribution and abundance in the GOM remain relatively unknown [National Marine Fisheries Service (NMFS), 2017]. The results of our study suggest that thorny skate have a broad temperature tolerance (at least 2.5–12.5°C) and that ambient water temperatures have not yet warmed to a level that would elicit broad-scale movement out of the region. However, warming temperatures in the GOM may already be contributing to decreased population productivity. Additional research is necessary to establish the thermal tolerance of thorny skate and to quantify the effects of temperature on fecundity, embryo development, hatching rates, and adult and offspring fitness (e.g. Di Santo, 2016).

Satellite tag performance

The activities of this study represent one of the largest PSAT tagging efforts ever exerted on an elasmobranch species and reinforce utility of this technology as an effective tool for investigating the spatial ecology of batoids. The success of this study also demonstrates the ability of mrPATs to rapidly generate tag-recapture data on species with low recapture and/or conventional tag reporting rates, and for species that may be too small to endure protracted deployments of traditionally sized PSATs. For example, the size of the mrPAT permitted its attachment to thorny skate measuring as small as 52 cm TL, which, to our knowledge, represents the smallest batoids ever tagged with PSAT tags. This application of mrPATs to smaller individuals permitted the monitoring of both mature and immature individuals, and the investigation of ontogenetic changes in movement. A priori laboratory trials conducted on captive thorny skate measuring 50–55 cm TL also provided no evidence that swimming was impaired by the attachment of an mrPAT to the pectoral wing for up to 60 d. Furthermore, mrPAT-tagged skates exhibited similar horizontal movements to those of conventional tagged individuals, and skates fit with larger pop-up satellite tags (Lotek PSATLife, Lotek Inc.) for up to 28 d also exhibited extensive vertical movement activity (Knotek et al., 2019; Knotek, unpublished data). Taken together, there is strong evidence that the mrPATs likely had minimal impact on swimming behaviour and horizontal displacement in the long term. Lastly, the relatively low unit cost of the mrPATs (∼1500 USD) allowed us to achieve a large sample size, which is typically a major limitation of satellite tagging studies (Siskey et al., 2019).

Premature detachment of mrPATs was a notable problem encountered during this study and is one limitation of external electronic tags in general (Siskey et al., 2019). Of the 128 mrPATs deployed, 24 (20%) detached from their bearer prior to the end of the programmed deployment period and floated at the surface for variable amounts of time before transmitting. The rate of premature detachment also scaled with deployment period as a higher rate of non-reporting was evident in mrPATs programmed for longer (e.g. 200 d) deployments. The cause(s) of these premature detachments is/are unknown, but it is possible that the skates suffered mortality (and the tag detached as the skate decomposed), the tag simply was shed due to a failure in the tether, or the tag was ripped off the skate during a bycatch interaction (i.e. fisheries recapture event) or interaction with a predator. Regardless, although the relatively high rate of mrPAT detachment limited data collection, the experience gained in this study will permit the refinement of tag attachment techniques that may promote better retention of PSAT tags on skates and other dorsal-ventrally flattened species.

Management implications

The results of this study may inform the management of thorny skate in the GOM, as well as the broader North Atlantic. Given the extent of movements within the GOM, there is indication that thorny skate may rarely, if ever, exhibit broad-scale movements to adjacent regions (e.g. Scotian shelf, Grand Banks) and therefore may constitute a single stock/population in the GOM. The potential lack of connectivity between the GOM and the rest of the species’ North Atlantic range also indicates that population dynamics within the GOM are largely driven by local factors, rather than broader regional dynamics (e.g. recruitment, mortality, movement). In the absence of broad population connectivity, the current prohibition on landings in US waters appears to be an effective strategy for reducing directed fishing mortality on individuals inhabiting the GOM, although it does not constrain bycatch interactions or discard mortality. Additional research on the population structure of thorny skate in the GOM and broader North Atlantic is necessary to fully evaluate the efficacy of this management approach and assess the extent to which individuals in the GOM are connected to conspecifics in the broader population. Management decisions made by the NEFMC and NOAA Fisheries, which have the primary authority over fisheries management of the US portion of the GOM, will likely have a disproportionate impact on the future conservation outlook for this putative stock.

The sedentary behaviour of thorny skate has both positive and negative implications for population decline and recovery in the GOM. Due to the lack of horizontal movement throughout the GOM, thorny skate may be at risk of localized depletion due to discard mortality associated with commercial fisheries, particularly in areas of high fishing activity. Discard mortality rates of thorny skate captured in the GOM bottom trawl fishery, the fishery with the highest thorny skate discards (Sosebee et al., 2016), have been estimated at 16.5–24.5% (Mandelman et al., 2013; Knotek et al., 2019) and result in estimated annual bycatch mortality of ∼79 mt under recent levels of fishing effort. Although this annual mortality represents only ∼1% of the existing biomass in the GOM (Sosebee et al., 2016; Knotek et al., 2019), even small levels of mortality in the population can work to impair recovery or promote fragmentation. To complicate matters, thorny skate biomass is contracting towards the western GOM into an area that supports a high level of commercial fishing effort, which may increase fishery interactions. Indeed, the majority of skates tagged west of the western GOM closure (Figure 2) remained in areas that support relatively high levels of commercial fishing activity (e.g. trawl, gillnet, scallop dredge), and at least four individuals were recaptured in this area, including one skate that was recaptured twice. Of these recaptures, two skates died as a result of the capture process and two survived after being re-released.

In contrast, the thorny skate’s sedentary behaviour could prove beneficial to its conservation. Closed areas (e.g. western GOM closure, Cashes Ledge Closure) may function as sanctuaries in which thorny skate can escape incidental mortality occurring as a result of bycatch interactions. For example, 90% of the skates tagged with mrPATs in the Cashes Ledge Closure, the smallest permanent area closure in the US GOM, likely remained within the closed area over periods of 100–300 d (Figure 3). Given this high residency and the general prevalence of localized movements in the region, there is evidence that even small, well-placed area closures or gear-restricted areas may be beneficial to thorny skate population recovery. Functionally, closed areas similar to those that have been implemented for more economically valuable groundfish species may be particularly important for mitigating incidental mortality in areas that support high levels of commercial fishing activity, such as the western GOM. Efforts to identify thorny skate bycatch hotspots could provide a basis for future consideration of spatial management in the GOM to reduce this source of mortality. However, additional research is also still needed to disentangle the effects of environmental change, bycatch mortality, and life history in the regional decline in thorny skates and more clearly identify the factors that may best contribute to their recovery.

Acknowledgements

The authors are grateful to commercial fishermen Captain Phil Lynch (F/V Mary Elizabeth), Captain Eric Hesse (F/V Tenacious II), Captain Kevin Norton (F/V Miss Emily), and Captain Charles Felch (F/V Lady Victoria) and their crew members who provided use of their vessels for tagging. We also thank all of the commercial fishermen and fisheries observers who reported recaptured tags and Ben Church (Northeast Fisheries Observer Program) for providing detailed recapture information. Thanks also to Nick Buchan, William Hoffman, and Trevor Coble of the MADMF for deploying conventional tags on the Atlantic Cod IBS and Dominique St. Amand, Giovanni Gianesin, and Calvin Alexander and other staff of the NEFSC Cooperative Research Branch for deploying tags on the NEFSC BLL survey. We are also grateful to the past and present members of the Sulikowski and Mandelman laboratories who assisted with tagging over the course of the study and to Megan Winton (Atlantic White Shark Conservancy) and the two anonymous reviewers whose comments greatly improved this manuscript.

Author contributions

JK, JS, JM, and TC conceived the idea for the study and all authors contributed to tagging efforts. JK led the analysis and the composition of the manuscript with input and assistance from JS, RK, WDM, BG, TC, and JM.

Funding

Funding for this work was provided by the Northeast Consortium, the 2015 NOAA Bycatch Reduction and Engineering Program (Award #: NA15NMF4720373), and the 2015 NOAA Saltonstall-Kennedy Program (Award #: NA15NMF4270283).

Data availability statement

The data underlying this article will be shared on a reasonable request to the corresponding author.