Content uploaded by James Locascio
Author content
All content in this area was uploaded by James Locascio on Apr 06, 2021
Content may be subject to copyright.
MARINE ECOLOGY PROGRESS SERIES
Mar Ecol Prog Ser
Vol. 663: 157–177, 2021
https://doi.org/10.3354/meps13637 Published March 31
1. INTRODUCTION
There has been a call for unified approaches to
studying animal movement ecology (Nathan et al. 2008)
and using movement to understand ecosystem change
(Hazen et al. 2019, Lowerre-Barbieri et al. 2019b) and
im prove fisheries management (Link et al. 2020).
Movement affects vulnerability to fishing and spatially
© Inter-Research 2021 · www.int-res.com
*Corresponding author: elasmophile@gmail.com
#These authors contributed equally to this work
Regional-scale variability in the movement ecology
of marine fishes revealed by an integrative acoustic
tracking network
Claudia Friess1,*,# , Susan K. Lowerre-Barbieri1, 2,#, Gregg R. Poulakis3, Neil Hammerschlag4,
Jayne M. Gardiner5, Andrea M. Kroetz6, Kim Bassos-Hull7, Joel Bickford1, Erin C. Bohaboy8,
Robert D. Ellis1, Hayden Menendez1, William F. Patterson III2, Melissa E. Price9,
Jennifer S. Rehage10, Colin P. Shea1, Matthew J. Smukall11, Sarah Walters Burnsed1,
Krystan A. Wilkinson7,12, Joy Young13, Angela B. Collins1,14, Breanna C. DeGroot15,
Cheston T. Peterson16, Caleb Purtlebaugh17, Michael Randall9, Rachel M. Scharer3,
Ryan W. Schloesser7, Tonya R. Wiley18, Gina A. Alvarez19, Andy J. Danylchuk20,
Adam G. Fox19, R. Dean Grubbs21, Ashley Hill22, James V. Locascio7, Patrick M. O’Donnell23,
Gregory B. Skomal24, Fred G. Whoriskey25, Lucas P. Griffin20
1Fish and Wildlife Research Institute, Florida Fish and Wildlife Conservation Commission, St. Petersburg, FL 33701, USA
Full author addresses are given in the Appendix
ABSTRACT: Marine fish movement plays a critical role in ecosystem functioning and is increas-
ingly studied with acoustic telemetry. Traditionally, this research has focused on single species
and small spatial scales. However, integrated tracking networks, such as the Integrated Tracking
of Aquatic Animals in the Gulf of Mexico (iTAG) network, are building the capacity to monitor
multiple species over larger spatial scales. We conducted a synthesis of passive acoustic monitor-
ing data for 29 species (889 transmitters), ranging from large top predators to small consumers,
monitored along the west coast of Florida, USA, over 3 yr (2016−2018). Space use was highly vari-
able, with some groups using all monitored areas and others using only the area where they were
tagged. The most extensive space use was found for Atlantic tarpon Megalops atlanticus and bull
sharks Carcharhinus leucas. Individual detection patterns clustered into 4 groups, ranging from occa-
sionally detected long-distance movers to frequently detected juvenile or adult residents. Synchro-
nized, alongshore, long-distance movements were found for Atlantic tarpon, cobia Rachycentron
canadum, and several elasmobranch species. These movements were predominantly northbound
in spring and southbound in fall. Detections of top predators were highest in summer, except for
nearshore Tampa Bay where the most detections occurred in fall, coinciding with large red drum
Sciaenops ocellatus spawning aggregations. We discuss the future of collaborative telemetry re -
search, including current limitations and potential solutions to maximize its impact for understand-
ing movement ecology, conducting ecosystem monitoring, and supporting fisheries management.
KEY WORDS: Acoustic monitoring · Movement ecology · Ecosystem monitoring · Integrated
Tracking of Aquatic Animals in the Gulf of Mexico · iTAG · Collaboration
Resale or republication not permitted without written consent of the publisher
Mar Ecol Prog Ser 663: 157– 177, 2021
explicit stressors (Lowerre-Barbieri et al. 2019b), and
variation in migration, movement, or location can re-
sult in perceived changes in marine populations of in-
terest to managers (Link et al. 2020). In particular, a
better understanding of top predator spatiotemporal
abundance and movement patterns is needed because
they can serve as climate and ecosystem sentinels for
which monitored attributes (including movement) in-
dicate ecosystem change (Hays et al. 2016, Hazen et
al. 2019). Additionally, habitat use by top predators
can directly affect abundance and behavior of lower
trophic levels (Hammerschlag et al. 2012, Shoji et al.
2017), an important consideration in fisheries manage-
ment, as many top predator populations are under
threat from fisheries (Queiroz et al. 2019), while others
are showing signs of recovery from overfishing (Peter-
son et al. 2017). A seasonal influx of predators to an
area could lead to seasonal predation mortality patterns
and, if coinciding with a high-discard rate fishing sea-
son, higher-than expected discard mortality levels.
Acoustic telemetry is a valuable tool for studying
movement dynamics, migration, or centers of abun-
dance of aquatic species (Abecasis et al. 2018) and
has been widely used in marine and freshwater
environments (Donaldson et al. 2014, Crossin et al.
2017). Acoustic telemetry uses underwater hydro -
phones (here after referred to as receivers), typically
fixed in place and arranged in space and time within
a specific ‘array’ of receivers according to research
objectives (Brownscombe et al. 2019). Aquatic ani-
mals outfitted with acoustic transmitters are detected
by receivers when they come within detection range,
usually less than 500 m (Collins et al. 2008, Kessel et
al. 2014b, Mathies et al. 2014). Research applications
using acoustic telemetry have included studying life
history aspects, such as timing and location of spawn-
ing (Lowerre-Barbieri et al. 2016, Brownscombe et al.
2020); assessing levels of discard mortality (Bohaboy
et al. 2020); studying the effects of artificial reefs on
site fidelity and habitat connectivity (Keller et al.
2017); examining the effects of ecotourism on behav-
ior (Hammerschlag et al. 2017); monitoring compli-
ance with no-fishing zones (Tickler et al. 2019); and
evaluating the design of protected areas (Lea et al.
2016, Grin et al. 2020).
Acoustic tags can be detected on any receiver that
records within the frequencies transmitted by the
tags. Given the mobility of many aquatic species and
the connectivity of aquatic systems, acoustic tags are
often opportunistically detected on outside receiver
arrays (i.e. those deployed in other areas by re -
searchers tracking a different set of animals). To
facilitate the exchange of data between taggers and
acoustic array owners, several regional tracking net-
works have formed, including the Australian Inte-
grated Marine Observing System Animal Tracking
Facility (IMOS ATF), Atlantic Cooperative Telemetry
(ACT), Florida Atlantic Coast Telemetry (FACT; in -
cluding arrays from the Carolinas to the Bahamas),
and Integrated Tracking of Aquatic Animals in the
Gulf of Mexico (iTAG) networks. These networks ex -
pand the geographic area over which tagged animals
can be tracked, thereby widening the scope of indi-
vidual telemetry studies. Concurrently, conglomerates
such as the Ocean Tracking Network (OTN) serve as
data repositories and facilitators for the various track-
ing networks and telemetry studies. However, there
is a need to better leverage the strength of these net-
works to address the challenges facing our ocean eco-
systems (McGowan et al. 2017, Abecasis et al. 2018).
A number of tools exist that facilitate such retrospec-
tive analyses (Udyawer et al. 2018), but there are often
large differences in array design and transmitter
settings that cannot be fully accounted for during
standardization for data analysis and limit the scope
of the questions that can be asked of these data.
The goal of this study was to evaluate how an inte-
grative tracking approach can provide multi-species
movement data to improve our understanding of
movement ecology and ecosystem processes, with a
specific focus on the seasonal movements of predators
off the west coast of Florida (WCF), USA. We analyzed
3 years of data (2016−2018) from 21 acoustic telemetry
arrays within the iTAG network in the eastern Gulf of
Mexico (Gulf) to investigate the following 4 hypotheses:
(1) array coverage needed to track a given species
varies based on movements and space use by that
species, (2) movements vary due to external factors,
motion capacity, and navigation capacity (Nathan et
al. 2008); thus species, tagging location, and life stage
affect observed movement patterns, (3) there is com-
monality among species in seasonality and direction-
ality of movement, indicating similar underlying bio-
physical movement drivers, and (4) top predator
detection patterns show seasonal and spatial trends.
Multiple analytical approaches were used to address
these hypo theses, including quantification of detection
metrics, clustering analysis, and predictive modeling.
2. MATERIALS AND METHODS
2.1. Study areas
Data from 21 acoustic receiver arrays belonging to
the iTAG regional tracking network in the eastern
158
Friess et al.: Multi-species movement dynamics 159
Gulf were used in this analysis (details about the
individual iTAG arrays can be found in Supplement 1
and Table S1.1 at www. int-res. com/ articles/ suppl/
m663 p157_ supp/). These iTAG arrays, deployed on
the WCF during the study period (2016−2018), all
consisted of Vemco receivers capable of detecting
69 kHz acoustic transmitters. Their locations cov-
ered the range of the entire WCF, but they were
not evenly distributed. Because iTAG arrays were
developed to address individual study-scale objec-
tives, they exhibited a wide range of designs, vary-
ing in receiver number (3−60) and distribution (e.g.
gate, grid), with the finest spatial resolution coming
from arrays set up as Vemco Positioning Systems
(VPS).
It was necessary to regroup the receivers of some
iTAG arrays to form spatially distinct units for an -
alysis, resulting in 22 meta-arrays (referred to here-
after as arrays) (Fig. 1; Table S2.1 in Supplement 2
at www. int-res. com/ articles/ suppl/ m663 p157 _ supp/).
These arrays were further aggregated into nodes
for some analyses to reduce the spatial bias created
by heterogeneity in array distribution (Fig. 1). In the
present study, we referred to the arrays using the fol-
lowing 3-character naming system: sub-region (N =
north Florida, T = Tampa Bay area, C = Charlotte
Harbor area, S = south Florida), sequential number
within sub-region, and habitat (offshore = o, near-
shore = n, estuarine = e, riverine = r, where we define
‘offshore’ as being located in federal waters, greater
than 9 nautical miles away from shore, and 'near-
shore' as locations within state waters). For example,
array T3o is an offshore array in the Tampa Bay sub-
region, and it is also part of the Tampa Bay array
node that includes 6 arrays in close proximity in and
around the estuary (Fig. 1). Lastly, although not part
of the WCF, receivers in the Florida Keys (Fig. S2.1)
were included in the movement analysis portion of
the study to capture movements into and out of the
Gulf; the Keys array was considered part of the south
Florida (SFL) array group.
2.2. Detection data
Transmitter-owner information from iTAG and the
neighboring ACT and FACT telemetry network
databases were used to identify transmitters. Un -
identified transmitters detected on at
least 2 iTAG arrays were sent to Vemco
to help identify owners and species,
and transmitters were in cluded in this
study only after receiving owner per-
mission. For fish tagged in the WCF
area, each individual was assigned to
a tagging group based on a unique
combination of species, tagging loca-
tion, and life stage (juvenile or adult at
the time of tagging; assigned by trans-
mitter owner a priori). This was done
to address species which demonstrated
residency as juveniles and large-scale
movements as adults. Large juvenile
(2.0–3.4 m stretch total length, STL)
smalltooth sawfish Pristis pectinata
(hereafter referred to as sawfish) were
treated as their own tagging group,
given differences in movement ecol-
ogy from smaller juveniles (Brame et
al. 2019). Life stages were not distin-
guished for species tagged outside the
WCF region, as their detections within
the Gulf were dependent on large-
scale movements.
Individual tracking data were ag -
gregated at the array spatial scale and
date temporal scale (i.e. 24 h). This
Fig. 1. Florida, USA, with west coast array locations indicated by circles. Sym-
bol sizes are proportional to the number of receivers in each array (ranging
from 3 to 60). Also shown are the state−federal waters boundary (thin black
line), path of the Gulfstream gas pipeline (dotted line), and 200 m isobath
(thick black line). Arrays grouped into the same node due to spatial proximity
are within boxes. See Table S1.1 in Supplement 1 for corresponding iTAG array
numbers and Section 2.1 for explanation of the 3-character naming system
Mar Ecol Prog Ser 663: 157– 177, 2021
allowed us to: (1) control for differences in study
design (e.g. different transmitters and transmitter
delay programming; different array designs), (2) align
with the scope of this study to assess movement
across the entire WCF rather than at small spatial
scales, and (3) avoid overlap with ongoing and future
analyses at the species-specific study scale. Animals
with a known fate of shed transmitters, or mortality
(as evidenced by lack of vertical or lateral move-
ment or change in movement signature) were
removed prior to analysis, as were any animals with
less than a 10 d detection period (defined as the
period from tagging date or study start date, which -
ever came first, until last detection date on the WCF
or in the Florida Keys). Two detection filters, based
on R package ‘glatos’ functions (Binder et al. 2018),
were used to remove potentially spurious detections
before analysis: for a detection to be considered
valid, there had to be at least 2 detections within a
node in a 24 h period; or for VPS arrays, at least 2
detections on a single receiver within a 24 h period.
This stricter validation for VPS arrays was chosen to
avoid including spurious detections which were
more likely to occur with overlapping receiver
ranges and large numbers of high site fidelity ani-
mals tagged near receivers. Detection day (DD) was
defined as a transmitter detected within an array on
a calendar day. If a transmitter was detected at differ-
ent arrays on the same day, multiple DDs were
assigned. DD data were summarized and visualized
using the ‘tidyverse’ R package collection (Wickham
et al. 2019).
2.3. Movement patterns
We used clustering to analyze movement patterns.
Clustering was done on individual-based movement
variables (see below) created from the networked
telemetry data, which were first filtered for fish with
potential detection periods of at least 12 mo to evalu-
ate the detection period for potential seasonal effects.
Clustering was performed using the fuzzy C-means
clustering algorithm of Bezdek (1981) implemented
in the R package ‘ppclust’ (Cebeci 2019). Two cluster
validity indices were used to determine optimum
cluster size for a given set of variables: the fuzzy sil-
houette index and the modified partition coefficient
index computed with the R package ‘fclust’ (Ferraro
et al. 2019). The optimum number of clusters is that
for which the index takes on the largest value. Clus-
tering was done with different sets of exploratory
variables thought to capture the detection pattern
variability among existing groups, and the final
movement variables in the analysis were chosen
such that both cluster validity indices agreed on opti-
mal cluster size (Table S2.2). The 4-cluster solution
provided the clearest interpretability and was chosen
due to the a priori ex pectation of 4 movement types
ranging from highly resident to roaming or nomadic,
similar to what has been described in the literature
(Abrahms et al. 2017, Brodie et al. 2018). The result-
ing clusters were as signed names a posteriori based
on movement variable distributions.
The 5 movement variables used in the analysis
were: 1 distance-related measure (the 99th quantile
of distance traveled between successive detections),
2 detection frequency variables (the residence index
and the 99th quantile of days between successive
DDs on the WCF), 1 seasonality indicator variable (a
seasonality index), and 1 detection consistency index
(the gap ratio defined as the 99th to 75th quantiles of
days between successive DDs; see Table S2.3 for
variable summary statistics). Following Brodie et al.
(2018), we used the 99th quantiles rather than 100th
quantiles to provide better metrics of the movement
data distribution. Residence index (RI) was the num-
ber of days an individual was detected on the WCF
divided by the detection period. The seasonality
index was calculated using time series decomposi-
tion of the number of DDs mo−1 over the detection
period (see details in Supplement 2). The gap ratio is
low for individuals lacking variation in temporal
detection patterns and high for those characterized
by periods of both increased and decreased numbers
of DDs, regardless of whether or not these follow a
seasonal trend. For each tagging group, the propor-
tion of individuals in each movement group was cal-
culated, and within-tagging group variability in
movement group was estimated by calculating the
deviation from the mode, which ranges from 0 (no
variability) to 1 (equal proportions).
2.4. Movement pathways
Seasonality and directionality in observed move-
ment pathways were examined for species exhibiting
long-distance movements to and from the Florida
Keys. Where species-specific data were insufficient,
groupings of species with similar life history, move-
ment ecology, and shared taxonomy were created.
This resulted in a ‘coastal sharks’ group consisting of
great hammerhead Sphyrna mokarran, tiger Galeo-
cerdo cuvier, lemon Negaprion brevirostris, and
sandbar Carcharhinus plumbeus sharks. Movements
160
Friess et al.: Multi-species movement dynamics
were analyzed at the relatively coarse scale of calen-
dar season (winter = December−February, spring =
March−May, summer = June−August, fall = Septem-
ber−November) and node. Even though movements
were not expected to coincide perfectly with calen-
dar season, these time bins allowed for comparisons
of intra-annual patterns across species. Directed sea-
sonal movement networks were created, and move-
ments were classified according to alongshore direc-
tionality (northbound or southbound). To ensure the
validity of seasonal comparisons, 2 successive obser-
vations were only counted as a movement if they
occurred within a specific time period. This differed
among species and was based on visual inspection of
the time between DD quantiles for each group (see
details in Supplement 2 and Fig. S2.2). Resulting
cut-off values ranged from 57 d for cobia to 80 d for
Atlantic tarpon (hereafter referred to as tarpon). Sea-
sonal movement networks were constructed and
visualized using the R packages ‘igraph’ (Csardi &
Nepusz 2006), ‘ggplot2’ (Wickham et al. 2019), and
‘ggraph’ (Pedersen 2020).
Generalized linear models (GLMs) were used to
detect differences in the number of movement path-
ways (i.e. network edges) observed by movement
direction and season. For each species group, models
with and without an interaction between season and
movement direction were fitted. The response vari-
able was edge weight, which was a count of the num-
ber of times a potential movement path (between 2
different nodes) was used. It was assumed to follow a
negative binomial distribution. Since not all possible
movement paths would be expected to be used by all
species, a potential movement path was defined as a
path that was observed to be traveled by that species,
in either direction, during at least 1 season. Zero
counts were assigned to unused potential movement
paths. All models were fitted in the R package ‘rstan-
arm’ (Goodrich et al. 2020) which uses Stan (Carpen-
ter et al. 2017) for back-end estimation. Some combi-
nations of season and movement path direction had
very low or no positive observations, causing separa-
tion in the data that led to estimation problems with
standard GLMs using maximum likelihood. Therefore,
we chose Bayesian inference with weakly informa-
tive priors which can help obtain stable regression
coefficients and standard error estimates when sepa-
ration is present in the data (Gelman et al. 2008). All
models used 4 Markov chains with 2000 iterations
each, discarding 1000 as ‘burn-in,’ and all priors were
the default priors provided by ‘rstanarm’ (weakly
informative, normal priors with mean 0 and standard
deviation 2.5). We assessed convergence by calculat-
ing the potential scale reduction R
ˆstatistic (ensuring
that it was at most 1.1), inspecting trace plots, and
ensuring effective sample sizes of at least 1000 for all
parameters. Model fit was assessed using leave-one-
out cross-validation functionality provided by the R
package ‘loo’ (Vehtari et al. 2017), and the model with
the higher weight was used for inference. Model fits
were inspected graphically by conducting posterior
predictive checks using the R packages ‘bayesplot’
(Gabry & Mahr 2020) and ‘shinystan’ (Gabry 2018).
Marginal mean effects were computed and con-
trasted using the R package ‘emmeans’ (Lenth 2019)
to look for evidence of directional movement within
season (pairwise contrast) and whether directional
movements differed between seasons (i.e. comparing
each season to the average over all other seasons).
Hypothesis testing was done in the R package
‘bayestestR’ (Makowski et al. 2019a) by evaluating
evidence for existence and significance of effects.
Effect existence was assessed with the probability of
direction metric, the probability that a parameter is
strictly positive or negative, which is the Bayesian
equivalent of the frequentist p-value (Makowski et al.
2019b). Any probability of direction estimates above
97.5% were treated as strong evidence for effect
existence. Effect significance was assessed by calcu-
lating the portion of the full posterior density that fell
within the region of practical equivalence (ROPE; the
range of parameter values that is equivalent to 0).
The ROPE range was set from −0.18 to + 0.18, as is
recommended for parameters expressed in log odds
ratios, and values less than 5% in ROPE were consid-
ered significant (Makowski et al. 2019b). Overall, we
considered an effect important if there was evidence
for both effect existence and significance. We report
observed trends in the data, and all explicitly stated
comparisons constitute important effects.
2.5. Top predator hotspots
To test if top predator detections differed signifi-
cantly by season or location, we fitted 2 GLMs for
great hammerheads, bull Carcharhinus leucas, tiger,
sandbar, lemon, and white Carcharodon carcharias
sharks (individuals tagged as juveniles on the WCF
were excluded to omit nursery habitat use from the
analysis). The first model aimed to address whether
total top predator detection days varied by area
(definition below) and season (DD model). The sec-
ond model ad dressed whether the total number of
unique individuals detected varied by area and sea-
son (nind model). For both models, we were particu-
161
Mar Ecol Prog Ser 663: 157– 177, 2021
larly interested in the interaction effect between
area and season. Only a few arrays had sufficient
data to be included in this analysis and some
needed to be combined, resulting in 4 areas of com-
parison for this analysis: nearshore Charlotte Harbor
(the C1n array), the northern shelf (arrays N1o and
N2o), nearshore Tampa Bay (arrays T4n and T5n),
and offshore Tampa Bay (arrays T2o and T3o). The
response variable for the DD model was a daily
count of the number of individuals de tected by area
for each calendar day during the 3 yr study period.
The re sponse variable for the nind model was a
count of the number of unique individuals detected
mo−1. Both were assumed to follow a Poisson distri-
bution. The predictors for both models were area,
season, number of transmitters available for detec-
tion, and study year (defined as December through
November so as to not split winter across multiple
years). Study year was included as a predictor to
account for temporal changes in telemetry array
configuration (most notably, the C1n array was
largely re moved in 2018) and ecological effects (most
notably, the exceptionally strong and long-lasting
red tide event that affected coastal Tampa Bay [TB]
and Charlotte Harbor [CH] areas in 2018). Number
of available transmitters was included because
some individuals were tagged after this study
began (nstart = 24, nend = 54). The models included
interactions between area and season as well as
area and study year, an offset for the number of
available transmitters, and, for the DD model, a
nested random effect for month within year to
account for temporal autocorrelation patterns in the
data. Specifying available transmitters as an offset
variable results in modeling the response variable
as rates rather than counts (i.e. number of animals
detected per available transmitter). The models can
be written as follows, where irepresents calendar
day for the DD model and month for the nind model:
yi~Poisson(μi)
E(yi) = μi(1)
log(μ
i
) = Area
i
× Season
i
+ Area
i
× StudyYear
i
+ log(Tags
i
)
+ (1|Year
i
/Month
i
)
(DD model only)
Yeari~N(0, σyear
2)
Month:Yeari~N(0, σ2
month:year)
where yiis number of individuals observed d−1 for
the DD model and number of unique individuals ob -
served mo−1 for the nind model, μiis the expected
count, and log(Tagsi)is the offset term for number of
available transmitters. Models were fitted in the R
package ‘glmmTMB’ (Brooks et al. 2017), which uses
Laplace approximations to the likelihood via Tem-
plate Model Builder (Kristensen et al. 2015). Tem-
poral autocorrelation was checked visually using the
R package ‘forecast’ (Hyndman & Khandakar 2008).
Models were validated by simulating and testing
residuals from the fitted models using the R package
‘DHARMa’ (Hartig 2019). Post-hoc analyses were
conducted using the R package ‘emmeans,’ where
marginal ef fects for the variables of interest (i.e. area
and season) were calculated and contrasted to test
for significance of season and study year effects
within and among areas.
3. RESULTS
Detection data represented 889 fish from 29 species
(Table 1). These species range in terms of manage-
ment concerns from threatened and endangered
species (Gulf sturgeon Acipenser oxyrinchus desotoi
and sawfish, respectively) to unmanaged species
(e.g. hardhead Ariopsis felis and gafftopsail Bagre
marinus catfish). Habitat use was similarly wide-
ranging, from freshwater to offshore, with correspon-
ding management responsibility divided between
state and federal agencies. The following list typifies
the range from freshwater to marine life cycles: the
freshwater largemouth bass Micropterus salmoides,
the diadromous common snook Centropomus undec-
imalis (hereafter referred to as snook), the primarily
estuarine southern kingfish Menticirrhus ameri-
canus, estuarine-dependent species (e.g. tarpon and
red drum), reef fishes and elasmobranchs with estu-
arine nurseries (e.g. gray snapper Lutjanus griseus
and blacktip shark Carcharhinus limbatus), to off-
shore species such as red snapper L. campechanus
and white shark. The mean number of tagged fish
species−1 was 31 but ranged from 1 (3 species) to 163
individuals for sawfish (Table 1). Tagging dates var-
ied over the study period, contributing to a range of
detection periods from 1 to 899 d, with a relatively
short mean detection period for all species (235 d).
The tracking network on the WCF varies in broad
spatial acoustic monitoring coverage, array size (i.e.
number of receivers), and habitat being monitored:
riverine (n = 4), estuarine (n = 9), nearshore (n = 4),
and offshore (n = 5) arrays (Fig. 1). Only 15 % of the
individuals were observed in more than 1 node, but
these fish represented a fairly wide range of species:
great hammerhead, blacktip, bull, lemon, sandbar,
tiger, and white sharks, tarpon, cobia, snook, goliath
grouper Epinephelus itajara, greater amberjack Seri-
162
Friess et al.: Multi-species movement dynamics
ola dumerili, Gulf sturgeon, red drum, sawfish, and
whitespotted eagle ray Aetobatus narinari (hereafter
referred to as eagle ray).
3.1. Large-scale space use
We detected 55 unique tagging groups on the
WCF (Fig. 2). Species with multiple tagging groups
included tarpon, bull shark, gag grouper Myctero -
perca microlepis, goliath grouper, Gulf sturgeon, red
drum, red snapper, sawfish, snook, blacktip shark,
and eagle ray. Many tagging groups (49%) repre-
sented fish tagged within the WCF and detected on
multiple arrays. Another 31% of the tagging groups
were detected only in their study arrays, a pattern
driven by both site fidelity and proximity of a study
array to other arrays. These species included most
reef fishes, the catfishes, southern kingfish, sheeps -
head Archosargus probatocephalus, largemouth bass,
and bonnethead Sphyrna tiburo (Fig. 2). Lastly, 18%
of tagging groups were tagged outside the WCF
region, highlighting the role integrative tracking net-
works play for these species, which included a nurse
shark Ginglymostoma cirratum as well as a number
of top predators (great hammerhead, bull, lemon,
sandbar, tiger, and white sharks), which prey on many
of the resident species. The most expansive space
use on the WCF was seen for adult tarpon tagging
groups and bull sharks tagged in the Atlantic or CH
area (Fig. 2).
3.2. Movement patterns
The 4 groups generated by clustering of movement
variables for 554 individuals were characterized a
posteriori as: long-distance movers that were detected
infrequently (‘movers;’ n = 84), high-detection resi-
dents (‘HD residents;’ n = 191), low-detection residents
(‘LD residents;’ n = 168), and ‘seasonals’ (n = 111).
Both resident groups traveled short maximal dis-
tances between DDs (LD residents: mean ± SE = 7.4
± 1.45 km; HD residents: 0.45 ± 0.36 km), but they
differed in temporal detection patterns (Fig. 3). HD
residents (represented best by red snapper, red
163
Common name Scientific name No. of transmitters Total DD Mean DD Mean DP
Atlantic tarpon Megalops atlanticus 34 2101 62 274
Blacktip shark Carcharhinus limbatus 17 1431 84 245
Bonnethead Sphyrna tiburo 4782063
Bull shark Carcharhinus leucas 40 1351 34 471
Cobia Rachycentron canadum 18 84 5 202
Common snook Centropomus undecimalis 126 17264 137 316
Gafftopsail catfish Bagre marinus 12 413 34 117
Gag grouper Mycteroperca microlepis 29 2686 93 119
Goliath grouper Epinephelus itajara 14 951 68 106
Gray snapper Lutjanus griseus 44 3948 90 106
Great hammerhead Sphyrna mokarran 5 50 10 255
Greater amberjack Seriola dumerili 17 1363 80 134
Grey triggerfish Balistes capriscus 13 1749 135 136
Gulf sturgeon Acipenser oxyrinchus desotoi 82 7341 90 400
Hardhead catfish Ariopsis felis 8 84 11 100
Largemouth bass Micropterus salmoides 45 3830 85 284
Lemon shark Negaprion brevirostris 2 48 24 809
Nurse shark Ginglymostoma cirratum 1111
Red drum Sciaenops ocellatus 44 1704 39 303
Red grouper Epinephelus morio 26 11238 432 499
Red snapper Lutjanus campechanus 91 13672 150 156
Sandbar shark Carcharhinus plumbeus 210525
Scamp Mycteroperca phenax 1 106 106 106
Sheepshead Archosargus probatocephalus 1 262 262 274
Smalltooth sawfish Pristis pectinata 163 18164 111 210
Southern kingfish Menticirrhus americanus 3 152 51 111
Tiger shark Galeocerdo cuvier 3 27 9 440
White shark Carcharodon carcharias 11 40 4 113
Whitespotted eagle ray Aetobatus narinari 33 3067 93 428
Table 1. Species detection summary. Detection day metrics are transmitter-based. DD: detection days, DP: detection period (d)
Mar Ecol Prog Ser 663: 157– 177, 2021
164
Fig. 2. Overview of tagging groups detected on west coast of Florida acoustic telemetry arrays between 2016 and 2018. Species
is indicated on the left (with down arrows indicating the same species as the one above the arrow), and tagging location and
life stage, if not adult, are identified on the right. The number of detected transmitters in each tagging group is shown in paren-
theses. Box color indicates proportion of detection days (min = 9 × 10−5 = white; max = 1 = dark red). Boxes with bold black bor-
ders indicate the study array for that tagging group; general tagging locations are shown with hashes. Arrays are ordered on
the x-axis by geographic location, with the northwesternmost array on the far left and the southernmost on the far right. CH:
Charlotte Harbor, NFL: north Florida, SFL: south Florida, TB: Tampa Bay, ATL: Atlantic, MS: Madison-Swanson, SR: Suwannee
River, DT: Dry Tortugas, PL: Pipeline, WGOM: western Gulf of Mexico, EGOM: eastern Gulf of Mexico, Juv: juvenile, lg: large
Friess et al.: Multi-species movement dynamics
grouper Epinephelus morio, and grey triggerfish Bal-
istes capriscus) were detected consistently in moni-
tored areas (gap ratio mean: 1.93 ± 0.10; RI mean:
0.91 ± 0.01; maximal days between DD mean: 2.0 ±
0.11 d) whereas LD residents (represented by, e.g.,
some snook and largemouth bass) had less consistent
temporal detections (gap ratio mean: 17.0 ± 1.4;
RI mean: 0.36 ± 0.02; maximal days between DD
mean: 31.9 ± 2.64 d; Fig. 3; Fig. S3.1 in Supplement 3
at www. int-res. com/ articles/ suppl/ m663 p157 _ supp/).
Seasonals (represented best by eagle rays, some Gulf
sturgeon, and TB red drum) had the largest seasonal-
ity index (mean 0.51 ± 0.02) and gap ratio (mean 49.6
± 4.26). Movers (represented best by Atlantic-tagged
sharks and cobia) traveled the greatest maximal dis-
tances between successive DDs (mean 369 ± 25.2 km),
had the smallest RI (mean 0.04 ± 0.005) and the sec-
ond-highest seasonality index (mean 0.07 ± 0.01).
Both movers and seasonals went long maximal peri-
ods without being detected on the WCF (means 136 ±
15.2 and 123 ± 10.1 d, respectively), but seasonals
had periods of high detection frequencies in moni-
tored areas, unlike the movers (Fig. 3; Fig. S3.1).
Intraspecific, large-scale movement patterns dif-
fered for some tagging groups (i.e. fish tagged in
different locations or in different life stages) but not
for others. There were differences between life
stages for tarpon and red drum, with the juveniles
clustering as LD and HD residents while adults
clustered predominantly as movers (tarpon), season-
als (TB red drum), and LD residents (CH red drum;
Fig. 4). In contrast, juvenile eagle ray movement
patterns were like adults; both groups predomi-
nantly clustered as non-residents. However, sample
size for juveniles was low (n = 2). The strongest
intraspecific differences among tagging groups were
seen for sawfish. This difference was primarily be -
tween individuals tagged in SFL and those tagged
in the CH area. SFL large juveniles (n = 3) and
adults (n = 7) clustered exclusively as non-residents,
while CH large juveniles (n = 13) were primarily
residents. Small juveniles tagged in SFL (n = 6)
clustered as seasonals and LD residents, whereas
those tagged in the CH area (n = 77) clustered
exclusively as LD or HD residents (Fig. 4). Addi-
tional species differences between tagging locations
were seen for bull sharks, where all individuals
tagged in the Atlantic (n = 22) but only 50%
tagged off the central shelf (TB and CH, n = 4)
clustered as movers. No clear differences between
tagging locations were observed for red snapper or
snook. Mild differences were seen for Gulf sturgeon
and gag. Gulf sturgeon tagged in the Suwannee
River (SR) clustered predominantly as seasonals
and LD residents, while those tagged further west,
near Apalachicola Bay, also clustered as movers.
Gag tagged in the southern offshore TB array,
where re ceivers were more densely arranged, clus-
tered predominantly as HD residents, while those
tagged in the northern offshore TB array, where
receivers were more spread out, were evenly split
between the 2 resident groups.
165
Residence index Seasonality index Detection
consistency index Max distance Max time
MLRHR MLRHR MLRHR MLRHR MLRHR
0
200
400
600
0
200
400
600
800
0
100
200
300
0.00
0.25
0.50
0.75
0.00
0.25
0.50
0.75
1.00
Group
Value
SSSSS
Fig. 3. Distribution of covariates for movement pattern clustering analysis. The horizontal line is the median, upper and lower
hinges show the 25th and 75th percentiles, and whiskers extend from the hinge to the smallest (lower) and largest (upper) value
no further than 150% of the interquartile range; values outside that range are shown as black dots. Groups are M: low-detection,
long distance movers, S: seasonals, LR: low-detection residents, HR: high-detection residents. Max time is the 99th quantile of
days between successive detection days, distance is the 99th quantile of kilometers between successive detection days, and the
detection consistency index is the ratio of 99th to 75th quantiles of days between detection days
Mar Ecol Prog Ser 663: 157– 177, 2021
Within-tagging group variability for the 37 tagging
groups in the analysis ranged from 0 (for 6 roamer
and 3 resident groups) to 0.833 for juvenile blacktip
sharks. Median variability among tagging groups
was 0.444. Four tagging groups (SFL snook, SR Gulf
sturgeon, TB red drum, and CH large juvenile saw-
fish) clustered in all 4 movement groups.
3.3. Movement pathways
The number of potential movement paths, number
of movements, and number of individuals contribut-
ing to those movements differed among groups
(Table S3.1). Number of movement paths ranged
from 8 for white and juvenile blacktip sharks to 38 for
bull sharks; number of movements ranged from 10
for white sharks to 182 for eagle rays; and the num-
ber of individuals in the analysis was lowest for white
sharks (n = 7) and highest for bull sharks (n = 32).
Predictions from the fitted models generally captured
trends in the observed data (Fig. S3.2). The effect of
movement direction on the number of observed
movements differed among seasons (i.e. the season ×
movement direction interaction model was favored
over the additive model) for all groups except juve-
nile blacktip sharks, eagle rays, and white sharks
(see Tables S3.2−S3.4 for full model parameters and
post-hoc test results). The overall pattern was that
northbound movements dominated in spring, south-
bound movements dominated in fall, winter move-
ments were low, except for blacktip sharks, and sum-
mer patterns were more variable across species groups
(Figs. 5 & 6). Within season, cobia, bull sharks, and
the coastal sharks group (great hammerhead, lemon,
tiger, and sandbar sharks) had more northbound
than southbound movements in spring, and tarpon,
cobia, and bull sharks had more southbound than
northbound movements in fall (Table S3.4; note that
throughout this section, comparative language, e.g.
‘more,’ ‘higher,’ ‘fewer,’ indicates statistically impor-
tant effects, whereas adjectives or superlatives, e.g.
‘high,’ ‘low,’ ‘most,’ state an observed pattern that
was not statistically important with respect to the 2
types of comparisons that were made, i.e. directional
differences within season and seasonal differences
within direction). For tarpon, movements up the
coast occurred later in the year compared to cobia,
sharks, and sawfish: summer, not spring, movements
differed by direction, and northbound movements
were higher in summer than in other seasons. Within
movement direction, northbound movements were
higher than in other seasons in spring for cobia, bull
shark, and sawfish and lower than in other seasons in
fall for coastal sharks and sawfish (Table S3.3). More
southbound movements were observed in fall than
other seasons for tarpon, cobia, and bull sharks but
also in summer for coastal sharks and cobia (Fig. 5).
166
Fig. 4. Movement pattern clustering results by tagging
group, showing the proportion of animals (within tagging
group) in each movement group. The scale ranges from 0
(white) to 1 (black). Species is indicated on the left axis (with
down arrows indicating the same species as the one above
the arrow), and tagging location and life stage, if not adult,
are identified on the right. Groups are M: low-detection,
long distance movers, S: seasonals, LR: low-detection resi-
dents, HR: high-detection residents. Numbers in parenthe-
ses indicate number of transmitters included in the analy-
sis for each group. Of the 889 animals included in this
study, 554 were included in the movement network analy-
sis; most of the filtering occurred due to insufficient poten-
tial detection periods (≤12 mo). Only tagging groups repre-
sented by at least 2 animals are shown. Site abbreviations as
in Fig. 2
Friess et al.: Multi-species movement dynamics 167
Generally, northbound movements in fall were short
distance (<120 km; i.e. movement within central,
north, or south Florida; Figs. 5 & 6), while those in
spring were predominantly long distance for bull
sharks, coastal sharks, cobia, and tarpon and short
distance for the other species.
The species for which the models did not support a
difference in movement direction by season were those
with the fewest potential movement paths (Table S3.1).
Blacktip sharks had more movements in fall and win-
ter, and fewer in summer than in other seasons, while
more movements for white sharks were observed in
spring than other seasons (Fig. 5; Table S3.3). No sea-
sonal effects were supported for eagle rays.
While there was some commonality in spring and
fall movement direction, there were also group-
specific differences in space use that are apparent in
individual movement networks (Fig. 6). For example,
movement among SFL nodes occurred primarily for
tarpon and sawfish, and movement to and from off-
Fig. 5. Mean + SD by species or species group, season, and movement direction, relative to the maximum for each group. (a)
Juvenile and adult bull sharks, (b) juvenile and adult coastal sharks (great hammerhead, sandbar, lemon, and tiger sharks),
(c) cobia, (d) adult Atlantic tarpon, (e) large juvenile and adult smalltooth sawfish, (f) juvenile and adult white sharks, (g) juvenile
blacktip sharks, (h) juvenile and adult whitespotted eagle rays. np: number of unique potential movement paths. Generalized
linear models were fitted to number of movements (see Section 2 for details) for each group, and results from post-hoc com-
parisons of marginal means are indicated where there was strong evidence for both existence (probability of direction
>97.5%) and significance (< 5% in region of practical equivalence) of effects (see Tables S3.2−S3.4 in Supplement 3 for full
model results): panels highlighted with grey backgrounds indicate seasons within which the marginal means between north-
bound (N) and southbound (S) movements differed, and asterisks mark the season for which a marginal mean for the indicated
movement direction (green = south, red = north) differed from the mean over the other seasons. Bicolored asterisks were used to
note seasons that differed for those models where the data did not support direction-specific seasonal effects. Every group was
observed on west coast of Florida arrays in every season, even though movements, as defined in this study, were not observed
for every season−group combination
Mar Ecol Prog Ser 663: 157– 177, 2021
shore nodes was seen primarily for bull
sharks, coastal sharks, and cobia (also
for white sharks; not shown). Further-
more, there was variation in move-
ments among species within the coastal
sharks group: the only fall (southbound)
movements ob served were for great
hammerheads (Fig. 6); southbound
movements for lemon and tiger sharks
occurred in summer (not shown).
3.4. Top predator hotspots
There were significant area and sea-
sonal differences in top predator de -
tections on the WCF. Seasonal trends
were consistent across study years,
while area trends differed among years.
DDs were highest in summer in north
Florida (NFL), CH, and offshore TB,
and highest in fall in nearshore TB
(Fig. 7). Overall, the central shelf (TB
and CH) had higher DDs than NFL, but
inter-annual variation was high, with
2018 being the lowest year for the
central shelf and 2017 being the low-
est year for NFL (Tables S3.7 & S3.8).
The overall number of unique individ-
uals detected was consistently highest
in the offshore TB area in summer
(Fig. 8). Within areas, significantly more
168
Fig. 6. Spring and fall movement networks
for groups with season-specific movement
direction differences. (a) Juvenile and adult
bull sharks, (b) juvenile and adult coastal
sharks (great hammerhead, sandbar, lemon,
and tiger sharks), (c) cobia, (d) adult Atlantic
tarpon, (e) large juvenile and adult small-
tooth sawfish. Arrays in nodes were gr ouped
to focus on longer-distance movements.
South bound movements are drawn in
straight green lines and northbound move-
ments in curved red lines. Node color is
indicative of network degree, with darker
shades indicating higher degree (degree
calculations in cluded consecutive detec-
tions days at the same node, which are not
shown). Line width corresponds to edge
weight (i.e. number of times a path was
used). Species contributing to the spring
movement paths for the sharks group
were great hammerhead, tiger, and lemon
sharks, while only great hammerheads
were detected moving in fall
Friess et al.: Multi-species movement dynamics
unique individuals were detected in spring and sum-
mer offshore of TB than in other seasons, while in
nearshore TB, spring and fall had more unique indi-
viduals. In both TB areas, there were significantly
fewer unique individuals detected in winter (Table
S3.11). No significant seasonal effects were found for
CH or NFL (Fig. 8). The highest number of DDs
across years and areas occurred in summer 2017 for
the CH area, but no similar spike in the number of
unique individuals was seen in that year and season
(Figs. 7 & 8), suggesting repeat detections of the
same individuals created the DD effect. In contrast,
the DD data indicated that more unique individuals
visited the offshore TB area each summer, with fewer
DDs ind.−1, either because they spend less time there
or are present but not detected as frequently (see
Tables S3.5−S3.12 for model parameters, diagnostics,
and marginal means comparison results).
4. DISCUSSION
This study used collaborative acoustic telemetry
data from the iTAG network to show that (1) space
use of tracked species on the WCF was highly vari-
able, with some groups using all monitored areas and
169
Fig. 7. Observed (grey bars) and predicted (boxplots) number of top predators (great hammerheads, bull, white, tiger, sand-
bar, and lemon sharks; excluding juveniles tagged on the west coast of Florida) detected per day, summed by season. Within
area, seasons that had significantly (p ≤ 0.05) lower detection days are indicated by blue boxplots, those with significantly
higher estimates are red, and significantly higher or lower study years are highlighted with > and <, respectively (<< = p ≤ 0.01,
< = p ≤ 0.05). Representations of mid-line, hinges, and whiskers are as in Fig. 3
Mar Ecol Prog Ser 663: 157– 177, 2021
others using only the area where they were tagged;
(2) telemetry-derived movement types differed among
tagging groups (life stage and tagging location) for
some but not all species, but differences between
tagging locations cannot be conclusively attributed
to biological differences due the confoundment of
observation and process effects; (3) there was com-
monality in seasonal movement directionality for tar-
pon, cobia, bull sharks, coastal sharks, and sawfish
moving primarily northward in the spring and south-
ward in the fall; and (4) top predator detections
showed consistent spatiotemporal patterns that dif-
fered between season and area.
4.1. Large-scale space use
It was expected and confirmed in the present study
that the iTAG network was a valuable source for
monitoring highly migratory species. Additionally, we
showed that the tracking network helps to fill data
gaps in movement information for species monitored
in a specific area for part of their annual migration or
during early life stages. This includes species with
strong seasonal patterns such as eagle rays, red drum,
and Gulf sturgeon; movers such as tarpon; and juve-
nile elasmobranchs such as blacktips, bull sharks, and
sawfish as they leave their nursery areas and transi-
170
Fig. 8. Observed (grey points) and predicted (boxplots) number of unique top predator individuals (great hammerheads, bull,
white, tiger, sandbar, and lemon sharks; excluding juveniles tagged on the west coast of Florida) detected per month, aver-
aged by season. Within area, seasons that had significantly (p ≤ 0.05) lower unique individuals detected are indicated by blue
boxplots, those with significantly higher estimates are red, and significantly higher or lower study years are highlighted with
> and <, respectively (<< = p ≤ 0.01, < = p ≤ 0.05). Representations of mid-line, hinges, and whiskers are as in Fig. 3
Friess et al.: Multi-species movement dynamics
tion from residents to a different movement pattern.
The network allows re searchers studying these ani-
mals to ask new questions they would not have other-
wise been able to (Grin et al. 2018). Tracking net-
works benefit not only researchers studying highly
mobile animals but also those focused on resident
fishes. For example, collaborative work with network
taggers can provide insights into predation on resident
species by migratory predators (Bohaboy et al. 2020).
Additionally, many resident fishes exhibit spawning
movements which could result in detections on other
network arrays, and tracking networks allow for the
potential to discover previously unknown transient
behavior or shifts in space use over time.
Individuals tagged outside the WCF that have
observations in this data set were almost exclusively
tagged in the Atlantic (including the east coast of
Florida, The Bahamas, and the northeastern USA).
The only individual tagged in the western Gulf was a
sandbar shark. This is probably due in part to the
greater acoustic tagging effort in the Atlantic than
the western Gulf, but also the observed pattern of a
biogeographical break between the eastern and
western Gulf (Chen 2017), with many fish in the
western Gulf migrating south to Mexico rather than
east toward the WCF (Rooker et al. 2019).
There was a somewhat surprising lack of reef fish
detections, particularly red snapper, among arrays
located near the Gulfstream pipeline. Pipeline con-
struction created artificial hard bottom habitat on
and near the pipeline as part of the damage miti -
gation process from pipeline construction. It was
hypothesized that the pipeline and these artificial
hardbottom spots could contribute to the expansion
of red snapper into the eastern Gulf by serving as
steppingstones (Cowan et al. 2011). Red snapper
were tagged on 3 offshore reefs near the pipeline (i.e.
arrays N1o, N2o, and T1o), but none of the over 300
tagged fish were detected anywhere but on their
study arrays. Perhaps arrays in closer proximity to
each other along the pipeline artificial reefs can help
resolve the question of whether red snapper do use
them as steppingstones for range re-expansion to
areas occupied prior to intense fishing, or perhaps
the 3 yr time period of this synthesis was insufficient
to detect such movement.
4.2. Movement patterns
Multi-species clustering of movement patterns
would not be possible with data from only a small
number of arrays. The results of the clustering analy-
sis were dependent on the spectrum of movement
ecologies represented in the sample of tagged ani-
mals as well as the observation system, and the vari-
ables analyzed. Results were sensitive to the choice
of clustering variables, a result also reported by
Brodie et al. (2018) for Australian telemetry arrays.
Even though there were a number of differences in
our movement type clustering analysis compared to
theirs (e.g. different systems, different movement
variables, shorter study period, fewer species and
tagged individuals), 3 of the 4 groups generated in
this study were equivalent to those reported in the
Australian study (‘HD residents’ ≈ ‘residents,’ ‘LD
residents’ ≈ ‘occasionals,’ and ‘movers’ ≈ ‘roamers’).
Our ‘seasonals’ group was not previously reported,
which is not surprising given that we used a season-
ality index variable specifically to distinguish that
group. It should be noted here that many individuals
or entire groups that clustered as movers in our
analysis are known to undertake seasonal migrations
to and from the Gulf (Biesiot et al. 1994, Reyier et al.
2014, Skomal et al. 2017), but detections were so
infrequent that they could not be distinguished from
more nomadic movement patterns. Our analysis
identified individuals that spent a lot of time in areas
with acoustic monitoring coverage (e.g. eagle rays)
when seasonally present on the WCF, whereas
movers seasonally often travel even further into the
Gulf and spend less time in monitored areas, perhaps
also using habitats in deeper waters without acoustic
monitoring coverage.
Networked telemetry data extend the spatial scope
of observation but at the cost of disparate observation
capacity between monitored regions. Changes to the
telemetry infrastructure, especially the kinds that
would allow more detections along migratory routes,
could change the set of variables needed to discrimi-
nate amongst movement groups. Thus, movement
type clustering is a snapshot in time and results must
be interpreted with care, as apparent intraspecific
variability in movement patterns may be due to ob -
servation error rather than true movement patterns,
especially in species using habitats with low receiver
coverage. For example, receiver density is likely
what was driving the differences between move-
ment types (as high vs. low detection residents) for
gag tagged in 2 offshore TB areas. Similarly, the
observed differences in movement patterns between
tagging locations for sawfish are likely due to a
combination of ontogenetic changes in habitat use,
sample size, habitat complexity, and receiver density.
Most (84%) sawfish tagged in the CH estuarine sys-
tem (n = 89) were small juveniles (< 2 m STL), which
171
Mar Ecol Prog Ser 663: 157– 177, 2021
are known to be primarily resident within their natal
estuarine nurseries, some of which include extensive
creek and canal habitats (Poulakis et al. 2013, 2016,
Scharer et al. 2017). As individuals exceed 2 m STL,
they begin leaving the nurseries and moving to and
from SFL (Graham et al. 2021) where fewer fish (n =
16) were tagged and included in the clustering
analysis, and most (n = 10, 62.5%) were >2 m. Con-
sequently, within the CH area, where there were 2
dense arrays of receivers compared to SFL, some
small juveniles were almost constantly within
receiver range and clustered as HD residents, while
other small juveniles as well as large juveniles, went
undetected for longer periods and clustered as LD
residents. These apparent differences in movement
ecology by tagging location highlight the limitations
of the multi-species clustering approach and show
that detailed knowledge of local arrays and species-
specific re search is needed to address nuances in the
data (e.g. habitat complexity), to validate the results
and fully understand complex life histories that
encompass the entire eastern Gulf and beyond.
4.3. Movement pathways
The seasonal large-scale movement patterns re -
ported here are congruent with existing literature.
Tarpon generally move north in spring and summer,
and south in fall (Luo et al. 2020), and cobia move
from the Florida Keys into the northern Gulf in spring
(Franks et al. 1999). Large juvenile and adult sawfish
undergo seasonal migrations, consisting of spring
and summer northward and fall and winter south-
bound movements (Graham et al. 2021), and seasonal,
temperature-related residence patterns for sharks
have been described off southeast Florida (Kessel et
al. 2014a, Hammerschlag et al. 2015, Guttridge et al.
2017). Large sharks are found in deeper waters in fall
and winter (Ajemian et al. 2020), which is consistent
with the reduced movements we found in those sea-
sons, as deep-water sites are poorly monitored.
Our analysis failed to detect statistically relevant
differences in movement direction by season for
juvenile blacktip and white sharks. This was sur-
prising given that previous research revealed sea-
sonal movements into the Gulf in winter and spring
for white sharks (Skomal et al. 2017), and previous
tag−recapture data also suggested a pattern of sea-
sonal movements for WCF juvenile blacktip sharks
(Hueter et al. 2005). Our results are most likely attrib-
utable to low sample sizes, suggesting that the WCF
telemetry network did not adequately monitor long-
distance migrations for those species or that not
enough tagged individuals were available for detec-
tion during our study period. Unlike cobia, which had
an equal ratio of south- to northbound movements in
the data, blacktip and white sharks were predomi-
nantly observed moving in 1 direction (south for black-
tips and north for white sharks). It is unclear whether
this skew is an artifact of low sample size or represents
a real trend of systematically failing to detect direc-
tional movements for these species. Juvenile blacktip
sharks are vulnerable to predation and fishing mor-
tality in the nursery (Heupel & Simpfen dorfer 2002).
Mortality rates on their migratory routes may also be
high, which might be partially responsible for more
observed movements leaving the nursery and head-
ing south. White sharks might use deeper waters with
little re ceiver coverage when migrating from the
Gulf back to the Atlantic resulting in fewer records of
those movements.
Additional factors that could lead to failure to
detect interaction effects are (1) individual variation
in timing of migrations that could, at the population
level, give the appearance of bidirectional move-
ments in the same season, and (2) inclusion of shorter-
distance, within-season movements (particularly be -
tween the TB and CH areas) that may or may not be
part of long-distance migration tracks. Those factors
likely contributed to finding no significant movement
direction effects for eagle rays. Eagle rays occur off
the WCF in spring, summer, and fall, and are hypo -
thesized to migrate to offshore and southern areas
when water temperatures decrease (Bassos-Hull et
al. 2014, DeGroot et al. 2021). There was a lot of indi-
vidual variability in eagle ray movement direction,
but inspection of seasonal eagle ray movement net-
works revealed patterns that the GLM was not set up
to detect: a latitudinal progression of movement activ-
ity, from the southern part of the coast in winter to
the northern part in summer (Fig. S3.3).
The commonality in movement directionality over
coarse spatiotemporal scales observed for tarpon,
cobia, and most elasmobranchs supports the exis-
tence of shared biophysical movement drivers. Al -
though identifying the precise drivers is beyond the
scope of this study, some likely contributors are tem-
perature, which is a major factor for ectothermic
organisms (Lear et al. 2019b), reproduction (i.e.
movement to and from spawning, mating, and nurs-
ery areas), foraging (Lear et al. 2019a), and preda-
tion. Some sharks likely follow the migration routes
of their prey, a phenomenon called migratory cou-
pling (Furey et al. 2018), others change their move-
ments in response to reef fish spawning aggrega-
172
Friess et al.: Multi-species movement dynamics
tions (Pickard et al. 2016, Rhodes et al. 2019), and,
while most potential shark prey species prefer to
avoid their predators, some, such as cobia, are known
to associate with large elasmobranchs (Shaer &
Nakamura 1989).
4.4. Top predator hotspots
We found seasonal trends of top predator detec-
tions that differed by area and were consistent across
study years. Top predator DDs were highest in most
analyzed areas in the summer, which is consistent
with the finding of movement from the Florida Keys
into the Gulf in spring. Nearshore TB was the excep-
tion to the pattern in that fall was the season of high-
est detections. This could be driven by the large red
drum spawning aggregations that form in fall at the
mouth of TB (Lowerre-Barbieri et al. 2019a) which
also attract smaller shark species such as the black -
nose shark Carcharhinus acronotus (J. Bickford pers.
obs.). A seasonal influx of predators into the Gulf could
mean seasonally fluctuating predation rates, result-
ing in high predation levels in high-discard recre-
ational fisheries, such as red snapper. The federal
recreational red snapper season is in the summer,
coinciding with highest shark detections on the WCF.
While we have provided evidence for predictable
spatiotemporal fluctuations in predator presence on
the WCF, quantifying any potential predation effect
to be useful for management would require further
study and the use of additional tools and data sources
(Hammerschlag 2019). For example, Bohaboy et al.
(2020) used fine-scale movement monitoring in a
high-resolution acoustic telemetry array to estimate
that 83% of red snapper and 100% of grey trigger-
fish discard mortality was due to predation by large
pelagic predators. Predator−prey interactions could
also be studied with predation transmitters (Halfyard
et al. 2017) or Vemco Mobile Transceivers (Haulsee
et al. 2016). In addition, there could be other areas on
the WCF that are important shark hotspots but are
currently not acoustically monitored, particularly in
deeper waters. Spatial fisheries-dependent and inde-
pendent data could be evaluated to determine poten-
tial locations for additional arrays to expand top
predator monitoring capabilities.
Long-term monitoring of inter-annual differences
in movements and space use is needed to understand
ecosystem health. To make temporal comparisons
from networked telemetry data, consistency in tele -
metry infrastructure over time is needed. Without this
consistency, process and observation effects become
confounded in the data. We explicitly considered
year effects in analyzing spatiotemporal top predator
detection patterns, and there are process as well as
observation factors explaining the strong inter-
annual differences we observed. Of the 3 years ana-
lyzed, 2018 stood out as having lower DDs in all cen-
tral Florida areas. In this year, an abnormally strong
and long-lasting red tide event affected nearshore
central Florida waters. Unfortunately, the removal of
receivers from the nearshore CH array and offshore
TB arrays in 2018 made it impossible to attribute this
effect to red tide in those areas. The nearshore TB
array, however, has been maintained since 2012.
Thus, the reduction in DDs and number of unique
individuals detected here in 2018 should not be due
to changes in observation capacity, making it likely
that this was a signal from the red tide event.
One noteworthy caveat of the movement paths and
predator hotspot GLMs we fitted is that the data con-
sisted of repeated observations of the same individu-
als, thereby violating independence assumptions.
Re peated observations of the same individuals could
give the appearance of strong population trends that
may or may not hold if sample size was increased.
5. CONCLUSIONS
Fisheries science, like other sciences, is assessing
how best to use the emerging field of ‘technoecology’
(Allan et al. 2018) and incorporate non-extractive
sampling into standard monitoring schemes. Teleme-
try networks collect extensive information about the
movements of tagged marine animals, but the value
of networked telemetry data synthesis studies to
practical fisheries management is currently limited,
for 2 reasons. First, changes in detectability over time
cannot currently be separated from changes in be -
havior due to frequent changes in array configuration.
Unlike the Australian IMOS ATF, the WCF currently
does not have any state, federal, or consortium-
funded permanent receiver arrays. A network of
strategically placed, permanent receivers would en -
able temporal comparisons of movement patterns
and space use without the confounding influences of
changing observation capacity. Second, the fisheries
assessment and management process is currently not
capable of accepting outputs from telemetry studies,
much less telemetry syntheses, unless these outputs
come packaged in the form of a standard stock as -
sessment parameter such as natural mortality. Chang-
ing this will likely require the system to move beyond
management based on maximum sustainable yield
173
Mar Ecol Prog Ser 663: 157– 177, 2021
and its analogues, and there are currently no opera-
tional alternatives.
Telemetry synthesis studies have potential value
for ecology that is yet to be fully realized, although
they may be most valuable for exploratory data
analysis and hypothesis development. For example,
future research questions inspired by our work
include: (1) Are spawning aggregations the drivers of
a seasonal predator influx to the WCF? (2) Are there
seasonal, spatially specific fluctuations in predation
mortality of WCF resident fishes? (3) Are the differ-
ences between high-and-low site fidelity residents
observed in this study artifacts of the observation
system or do they reflect true behavioral differences
within populations? (4) How can observation effects
(e.g. differences in spatiotemporal detection proba-
bility over time) be formally incorporated into infer-
ence from networked acoustic telemetry data? For
iTAG to move beyond opportunistic data and fully
realize its potential for hypothesis-driven ecological
inquiry, it will require long-term funding to support
permanent monitoring infrastructure, coordinated
multi-species tagging, a Gulf-wide database, and the
personnel needed to oversee membership, database
management, workshops, and the website (https://
itagscience.com).
Acknowledgements. We thank all researchers, technicians,
and institutions that have contributed to collecting the
data used in this manuscript by tagging fish or maintaining
acoustic receiver arrays. We thank the following people
for giving permission to use their transmitter data in this
study: Debra Abercrombie, Judd Curtis, Tobey Curtis,
Keith Dunton, Joe Heublein, Adam Kaeser, Matt Kendall,
Frank Parauka, and Wes Pratt. We are grateful to Ray
Simpson and Lindsay Henderson for providing the fish
artwork, and to J. Walter and 2 anonymous reviewers
whose comments strengthened the manuscript. The Ocean
Tracking Network and Vemco provided receivers for some
iTAG arrays. S.K.L.-B. was funded by NOAA CIMAS grant
NA15AR4320064. Grant F-59 from the US Fish and
Wildlife Service Sport Fish Restoration program helped to
fund iTAG, iTAG workshops, and personnel to oversee the
running of iTAG. The views and conclusions are those of
the authors and do not necessarily reflect the opinions or
policies of the US government or any of its agencies. Any
use of trade, firm, or product names is for descriptive
purposes only and does not imply endorsement by the US
government.
LITERATURE CITED
Abecasis D, Steckenreuter A, Reubens J, Aarestrup K and
others (2018) A review of acoustic telemetry in Europe
and the need for a regional aquatic telemetry network.
Anim Biotelem 6: 12
Abrahms B, Seidel DP, Dougherty E, Hazen EL and others
(2017) Suite of simple metrics reveals common move-
ment syndromes across vertebrate taxa. Mov Ecol 5: 12
Ajemian MJ, Drymon JM, Hammerschlag N, Wells RJD and
others (2020) Movement patterns and habitat use of tiger
sharks (Galeocerdo cuvier) across ontogeny in the Gulf
of Mexico. PLOS ONE 15: e0234868
Allan BM, Nimmo DG, Ierodiaconou D, Vanderwal J, Koh
LP, Ritchie EG (2018) Futurecasting ecological research:
the rise of technoecology. Ecosphere 9: e02163
Bassos-Hull K, Wilkinson KA, Hull PT, Dougherty DA and
others (2014) Life history and seasonal occurrence of the
spotted eagle ray, Aetobatus narinari, in the eastern Gulf
of Mexico. Environ Biol Fishes 97: 1039−1056
Bezdek JC (1981) Pattern recognition with fuzzy objective
function algorithms. Plenum Press, New York, NY
Biesiot PM, Caylor RE, Franks JS (1994) Biochemical and
histological changes during ovarian development of
cobia, Rachycentron canadum, from the northern Gulf of
Mexico. Fish Bull 92: 686−696
Binder TR, Hayden TA, Holbrook C (2018) glatos: an R
package for the Great Lakes Acoustic Telemetry Obser-
vation System. R package version 2.0. https: // gitlab.
oceantrack.org/GreatLakes/glatos
Bohaboy EC, Guttridge TL, Hammerschlag N, Van Zinnicq
Bergmann MP, Patterson WF III (2020) Application of
three-dimensional acoustic telemetry to assess the effects
of rapid recompression on reef fish discard mortality.
ICES J Mar Sci 77: 83−96
Brame AB, Wiley TR, Carlson JK, Fordham SV and others
(2019) Biology, ecology, and status of the smalltooth sawfish
Pristis pectinata in the USA. Endang Species Res 39: 9−23
Brodie S, Lédée EJI, Heupel MR, Babcock RC and others (2018)
Continental-scale animal tracking reveals functional move-
ment classes across marine taxa. Sci Rep 8: 3717
Brooks ME, Kristensen K, van Benthem KJ, Magnusson A
and others (2017) glmmTMB balances speed and flexi-
bility among packages for zero-inflated generalized
linear mixed modeling. R J 9: 378−400
Brownscombe JW, Lédée EJI, Raby GD, Struthers DP and
others (2019) Conducting and interpreting fish telemetry
studies: considerations for researchers and resource
managers. Rev Fish Biol Fish 29: 369−400
Brownscombe JW, Griffin LP, Morley D, Acosta A and others
(2020) Seasonal occupancy and connectivity amongst
nearshore flats and reef habitats by permit Trachinotus
falcatus: considerations for fisheries management. J Fish
Biol 96: 469−479
Carpenter B, Gelman A, Hoffman MD, Lee D and others
(2017) Stan: a probabilistic programming language.
J Stat Softw 76: 1−32
Cebeci Z (2019) Comparison of internal validity indices for
fuzzy clustering. J Agric Bioinform 10: 1−14
Chen Y (2017) Fish resources of the Gulf of Mexico. In: Ward
C (ed) Habitats and biota of the Gulf of Mexico: before
the Deepwater Horizon oil spill (Vol 2). Springer, New
York, NY, p 869−1038
Collins AB, Heupel MR, Simpfendorfer CA (2008) Spatial
distribution and long-term movement patterns of cow -
nose rays Rhinoptera bonasus within an estuarine river.
Estuaries Coasts 31: 1174−1183
Cowan JH Jr, Grimes CB, Patterson WF III, Walters CJ and
others (2011) Red snapper management in the Gulf of
Mexico: science- or faith-based? Rev Fish Biol Fish 21:
187−204
Crossin GT, Heupel MR, Holbrook CM, Hussey NE and oth-
ers (2017) Acoustic telemetry and fisheries management.
Ecol Appl 27: 1031−1049
174
Friess et al.: Multi-species movement dynamics
Csardi G, Nepusz T (2006) The igraph software package for
complex network research. InterJournal Complex Syst
1695: 1−9
DeGroot BC, Bassos-Hull K, Wilkinson KA, Lowerri-Barbieri
S, Poulakis GR, Ajemian MJ (2021) Variable migration
patterns of whitespotted eagle rays Aetobatus narinari
along Florida’s coastlines. Mar Biol 168: 18
Donaldson MR, Hinch SG, Suski CD, Fisk AT, Heupel MR,
Cooke SJ (2014) Making connections in aquatic eco -
systems with acoustic telemetry monitoring. Front Ecol
Environ 12: 565−573
Ferraro MB, Giordani P, Serafini A (2019) Fclust: an R pack-
age for fuzzy clustering. R J 11: 198−210
Franks JS, Warren JR, Buchanan MV (1999) Age and
growth of cobia, Rachycentron canadum, from the north-
eastern Gulf of Mexico. Fish Bull 97: 459−471
Furey NB, Armstrong JB, Beauchamp DA, Hinch SG (2018)
Migratory coupling between predators and prey. Nat
Ecol Evol 2: 1846−1853
Gabry J (2018) shinystan: interactive visual and numerical
diagnostics and posterior analysis for Bayesian models.
R package version 2.5.0. https: //CRAN.R-project.org/
package= shinystan
Gabry J, Mahr T (2020) bayesplot: plotting for Bayesian
models. R package version 1.7.2. https: //mc-stan.org/
bayesplot
Gelman A, Jakulin A, Pittau MG, Su YS (2008) A weakly
informative default prior distribution for logistic and
other regression models. Ann Appl Stat 2: 1360−1383
Goodrich B, Gabry J, Ali I, Brilleman S (2020) rstanarm:
Bayesian applied regression modeling via Stan. R pack-
age version 2.19.3. https: //mc-stan.org/rstanarm
Graham J, Kroetz AM, Poulakis GR, Scharer RM and others
(2021) Large-scale space use of large juvenile and adult
smalltooth sawfish Pristis pectinata: implications for
management. Endang Species Res 44: 45−59
Griffin LP, Brownscombe JW, Adams AJ, Boucek RE and
others (2018) Keeping up with the silver king: using
cooperative acoustic telemetry networks to quantify the
movements of Atlantic tarpon (Megalops atlanticus) in
the coastal waters of the southeastern United States. Fish
Res 205: 65−76
Griffin LP, Smith BJ, Cherkiss MS, Crowder AG and others
(2020) Space use and relative habitat selection for imma-
ture green turtles within a Caribbean marine protected
area. Anim Biotelem 8: 22
Guttridge TL, Van Zinnicq Bergmann MPM, Bolte C,
Howey LA and others (2017) Philopatry and regional
con nectivity of the great hammerhead shark, Sphyrna
mokarran in the US and Bahamas. Front Mar Sci 4: 3
Halfyard EA, Webber D, Del Papa J, Leadley T, Kessel ST,
Colborne SF, Fisk AT (2017) Evaluation of an acoustic
telemetry transmitter designed to identify predation
events. Methods Ecol Evol 8: 1063−1071
Hammerschlag N (2019) Quantifying shark predation effects
on prey: dietary data limitations and study approaches.
Endang Species Res 38: 147−151
Hammerschlag N, Luo J, Irschick DJ, Ault JS (2012) A com-
parison of spatial and movement patterns between sym-
patric predators: bull sharks (Carcharhinus leucas) and
Atlantic tarpon (Megalops atlanticus). PLOS ONE 7:
e45958
Hammerschlag N, Broderick AC, Coker JW, Coyne MS
and others (2015) Evaluating the landscape of fear
between apex predatory sharks and mobile sea tur -
tles across a large dynamic seascape. Ecology 96:
2117−2126
Hammerschlag N, Gutowsky LFG, Gallagher AJ, Matich P,
Cooke SJ (2017) Diel habitat use patterns of a marine
apex predator (tiger shark, Galeocerdo cuvier) at a high
use area exposed to dive tourism. J Exp Mar Biol Ecol
495: 24−34
Hartig F (2019) DHARMa: residual diagnostics for hier -
archical (multi-level/mixed) regression models. R pack-
age version 0.2.4. https: //CRAN.R-project.org/ package=
DHARMa
Haulsee DE, Fox DA, Breece MW, Brown LM, Kneebone J,
Skomal GB, Oliver MJ (2016) Social network analysis
reveals potential fission−fusion behavior in a shark. Sci
Rep 6: 34087
Hays GC, Ferreira LC, Sequeira AM, Meekan MG and oth-
ers (2016) Key questions in marine megafauna move-
ment ecology. Trends Ecol Evol 31: 463−475
Hazen EL, Abrahms B, Brodie S, Carroll G and others (2019)
Marine top predators as climate and ecosystem sentinels.
Front Ecol Environ 17: 565−574
Heupel MR, Simpfendorfer CA (2002) Estimation of mortal-
ity of juvenile blacktip sharks, Carcharhinus limbatus,
within a nursery area using telemetry data. Can J Fish
Aquat Sci 59: 624−632
Hueter RE, Heupel MR, Heist E, Keeney DB (2005) Evidence
of philopatry in sharks and implications for the manage-
ment of shark fisheries. J Northwest Atl Fish Sci 35:
239−247
Hyndman RJ, Khandakar Y (2008) Automatic time series for
forecasting: the forecast package for R. J Stat Softw 27:
1−22
Keller K, Smith JA, Lowry MB, Taylor MD, Suthers IM
(2017) Multispecies presence and connectivity around a
designed artificial reef. Mar Freshw Res 68: 1489−1500
Kessel ST, Chapman DD, Franks BR, Gedamke T and others
(2014a) Predictable temperature-regulated residency,
movement and migration in a large, highly mobile mar-
ine predator (Negaprion brevirostris). Mar Ecol Prog Ser
514: 175−190
Kessel ST, Cooke SJ, Heupel MR, Hussey NE, Simpfendor-
fer CA, Vagle S, Fisk AT (2014b) A review of detection
range testing in aquatic passive acoustic telemetry stud-
ies. Rev Fish Biol Fish 24: 199−218
Kristensen K, Nielsen A, Berg CW, Skaug H, Bell B (2015)
TMB: automatic differentiation and Laplace approxima-
tion. J Stat Softw 70: 1−21
Lea JS, Humphries NE, von Brandis RG, Clarke CR, Sims
DW (2016) Acoustic telemetry and network analysis
reveal the space use of multiple reef predators and
enhance marine protected area design. Proc R Soc B 283:
20160717
Lear KO, Poulakis GR, Scharer RM, Gleiss AC, Whitney NM
(2019a) Fine-scale behavior and habitat use of the
endangered smalltooth sawfish (Pristis pectinata): in -
sights from accelerometry. Fish Bull 117: 348−359
Lear KO, Whitney NM, Morgan DL, Brewster LR and others
(2019b) Thermal performance responses in free-ranging
elasmobranchs depend on habitat use and body size.
Oecologia 191: 829−842
Lenth R (2019) emmeans: stimated marginal means, aka
least-squares means. R package version 1.4.3.01. https: //
CRAN.R-project.org/package=emmeans
Link JS, Huse G, Gaichas S, Marshak AR (2020) Changing
how we approach fisheries: a first attempt at an opera-
175
Mar Ecol Prog Ser 663: 157– 177, 2021
tional framework for ecosystem approaches to fisheries
management. Fish Fish 21: 393−434
Lowerre-Barbieri SK, Burnsed SLW, Bickford JW (2016)
Assessing reproductive behavior important to fisheries
management: a case study with red drum, Sciaenops
ocellatus. Ecol Appl 26: 979−995
Lowerre-Barbieri SK, Tringali MD, Shea CP, Walters Burnsed
S, Bickford J, Murphy M, Porch C (2019a) Assessing red
drum spawning aggregations and abundance in the
eastern Gulf of Mexico: a multidisciplinary approach.
ICES J Mar Sci 76: 516−529
Lowerre-Barbieri SK, Kays R, Thorson JT, Wikelski M (2019b)
The ocean’s movescape: fisheries management in the bio-
logging decade (2018−2028). ICES J Mar Sci 76: 477−488
Luo J, Ault JS, Ungar BT, Smith SG and others (2020) Migra-
tions and movements of Atlantic tarpon revealed by two
decades of satellite tagging. Fish Fish 21: 290−318
Makowski D, Ben-Shachar MS, Chen SHA, Lüdecke D
(2019a) Indices of effect existence and significance in the
Bayesian framework. Front Psychol 10: 2767
Makowski D, Ben-Shachar MS, Lüdecke D (2019b)
BayestestR: describing effects and their uncertainty,
existence and significance within the Bayesian frame-
work. J Open Source Softw 4: 1541
Mathies NH, Ogburn MB, McFall G, Fangman S (2014)
Environmental interference factors affecting detection
range in acoustic telemetry studies using fixed receiver
arrays. Mar Ecol Prog Ser 495: 27−38
McGowan J, Beger M, Lewison RL, Harcourt R and others
(2017) Integrating research using animal-borne teleme-
try with the needs of conservation management. J Appl
Ecol 54: 423−429
Nathan R, Getz WM, Revilla E, Holyoak M, Kadmon R, Saltz
D, Smouse PE (2008) A movement ecology paradigm for
unifying organismal movement research. Proc Natl Acad
Sci USA 105: 19052−19059
Pedersen TL (2020) ggraph: an implementation of grammar
of graphics for graphs and networks. R package version
2.0.3. https: //CRAN.R-project.org/package=ggraph
Peterson CD, Belcher CN, Bethea DM, Driggers WB III, Frazier
BS, Latour RJ (2017) Preliminary recovery of coastal sharks
in the south-east United States. Fish Fish 18: 845−859
Pickard AE, Vaudo JJ, Wetherbee BM, Nemeth RS,
Blondeau JB, Kadison EA, Shivji MS (2016) Comparative
use of a Caribbean mesophotic coral ecosystem and as -
sociation with fish spawning aggregations by three spe-
cies of shark. PLOS ONE 11: e0151221
Poulakis GR, Stevens PW, Timmers AA, Stafford CJ, Simpfen -
dorfer CA (2013) Movements of juvenile endangered
smalltooth sawfish, Pristis pectinata, in an estuarine river
system: use of non-main-stem river habitats and lagged
responses to freshwater inflow-related changes. Environ
Biol Fishes 96: 763−778
Poulakis GR, Stevens PW, Timmers AA, Stafford CJ and oth-
ers (2016) Long-term site fidelity of endangered small-
tooth sawfish (Pristis pectinata) from different mothers.
Fish Bull 114: 461−475
Queiroz N, Humphries NE, Couto A, Vedor M and others
(2019) Global spatial risk assessment of sharks under the
footprint of fisheries. Nature 572: 461−466
Reyier EA, Franks BR, Chapman DD, Scheidt DM, Stolen
ED, Gruber SH (2014) Regional-scale migrations and
habitat use of juvenile lemon sharks (Negaprion brevi-
rostris) in the US South Atlantic. PLOS ONE 9: e88470
Rhodes KL, Baremore I, Graham RT (2019) Grouper (Epi-
nephelidae) spawning aggregations affect activity space
of grey reef sharks, Carcharhinus amblyrhynchos, in
Pohnpei, Micronesia. PLOS ONE 14: e0221589
Rooker JR, Dance MA, Wells RJD, Ajemian MJ and others
(2019) Population connectivity of pelagic megafauna in
the Cuba−Mexico−United States triangle. Sci Rep 9: 1663
Scharer RM, Stevens PW, Shea CP, Poulakis GR (2017) All
nurseries are not created equal: large-scale habitat use
patterns in two smalltooth sawfish nurseries. Endang
Species Res 34: 473−492
Shaffer RV, Nakamura EL (1989) Synopsis of biological data
on the cobia Rachycentron canadum (Pisces: Rachycen-
tridae). NOAA Tech Rep NMFS 82. FAO Fish Synop 153.
http: //www.fao.org/3/ap958e/ap958e.pdf
Shoji J, Mitamura H, Ichikawa K, Kinoshita H, Arai N (2017)
Increase in predation risk and trophic level induced by
nocturnal visits of piscivorous fishes in a temperate sea-
grass bed. Sci Rep 7: 3895
Skomal GB, Braun CD, Chisholm JH, Thorrold SR (2017)
Movements of the white shark Carcharodon carcharias in
the North Atlantic Ocean. Mar Ecol Prog Ser 580: 1−16
Tickler DM, Carlisle AB, Chapple TK, Curnick DJ, Dale JJ,
Schallert RJ, Block BA (2019) Potential detection of illegal
fishing by passive acoustic telemetry. Anim Biotelem 7: 1
Udyawer V, Dwyer RG, Hoenner X, Babcock RC and others
(2018) A standardised framework for analysing animal
detections from automated tracking arrays. Anim
Biotelem 6: 17
Vehtari A, Gelman A, Gabry J (2017) Practical Bayesian
model evaluation using leave-one-out cross-validation
and WAIC. Stat Comput 27: 1413−1432
Wickham H, Averick M, Bryan J, Chang W and others (2019)
Welcome to the Tidyverse. J Open Source Softw 4: 1686
176
Friess et al.: Multi-species movement dynamics 177
Editorial responsibility: Elliott Hazen,
Pacific Grove, California, USA
Reviewed by: J. Walter and 2 anonymous referees
Submitted: September 9, 2020
Accepted: January 11, 2021
Proofs received from author(s): March 17, 2021
Appendix. Full list of author addresses
Claudia Friess1,*,# , Susan K. Lowerre-Barbieri1, 2,#, Gregg R. Poulakis3,
Neil Hammerschlag4, Jayne M. Gardiner5, Andrea M. Kroetz6, Kim Bassos-Hull7,
Joel Bickford1, Erin C. Bohaboy8, Robert D. Ellis1, Hayden Menendez1,
William F. Patterson III2, Melissa E. Price9, Jennifer S. Rehage10, Colin P. Shea1,
Matthew J. Smukall11, Sarah Walters Burnsed1, Krystan A. Wilkinson7,12, Joy Young13,
Angela B. Collins1,14, Breanna C. DeGroot15, Cheston T. Peterson16,
Caleb Purtlebaugh17, Michael Randall9, Rachel M. Scharer3, Ryan W. Schloesser7,
Tonya R. Wiley18, Gina A. Alvarez19, Andy J. Danylchuk20, Adam G. Fox19,
R. Dean Grubbs21, Ashley Hill22, James V. Locascio7, Patrick M. O’Donnell23,
Gregory B. Skomal24, Fred G. Whoriskey25, Lucas P. Griffin20
1Fish and Wildlife Research Institute, Florida Fish and Wildlife Conservation Commission, St. Petersburg, FL 33701, USA
2Fisheries and Aquatic Sciences, School of Forest Resources and Conservation, University of Florida, Gainesville,
FL 32653, USA
3Charlotte Harbor Field Laboratory, Fish and Wildlife Research Institute, Florida Fish and Wildlife Conservation
Commission, Port Charlotte, FL 33954, USA
4Rosenstiel School of Marine and Atmospheric Science, University of Miami, Miami, FL 33149, USA
5Division of Natural Sciences, New College of Florida, Sarasota, FL 34243, USA
6Riverside Technology, Inc. for NOAA, National Marine Fisheries Service, Southeast Fisheries Science Center,
Panama City, FL 32408, USA
7Mote Marine Laboratory, Sarasota, FL 34236, USA
8National Marine Fisheries Service, Pacific Islands Fisheries Science Center, Honolulu, HI 96818, USA
9U.S. Geological Survey Wetland and Aquatic Research Center (USGS-WARC), Gainesville, FL 32653, USA
10Institute of Environment, Florida International University, Miami, FL 33199, USA
11Bimini Biological Field Station Foundation, South Bimini, Bahamas
12Chicago Zoological Society’s Sarasota Dolphin Research Program c/o Mote Marine Laboratory, Sarasota, FL 34236, USA
13Tequesta Field Laboratory, Fish and Wildlife Research Institute, Florida Fish and Wildlife Conservation Commission,
Tequesta, FL 33469, USA
14University of Florida IFAS Extension, Florida Sea Grant, Palmetto, FL 34221, USA
15Harbor Branch Oceanographic Institute, Florida Atlantic University, Fort Pierce, FL 34946, USA
16Florida State University, Tallahassee, FL 32306, USA
17Senator George Kirkpatrick Marine Laboratory, Fish and Wildlife Research Institute, Florida Fish and Wildlife
Conservation Commission, Cedar Key, FL 32625, USA
18Havenworth Coastal Conservation, Palmetto, FL 34221, USA
19Warnell School of Forestry and Natural Resources, University of Georgia, Athens, GA 30602, USA
20Department of Environmental Conservation, University of Massachusetts Amherst, Amherst, MA 01003, USA
21Florida State University Coastal and Marine Laboratory, St. Teresa, FL 32358, USA
22Lynker Technologies for NOAA, National Ocean Services, Office of Response and Restoration, Marine Debris
Division, Silver Spring, MD 20910, USA
23Rookery Bay National Estuarine Research Reserve, Naples, FL 34113, USA
24Massachusetts Division of Marine Fisheries, New Bedford, MA 02744, USA
25Ocean Tracking Network, Department of Biology, Dalhousie University, Halifax, Nova Scotia B3H 4R2, Canada
