May et al.: Fine-scale movements of Pristispectinata 
363 
receivers required time-synchronization tags, which were 
placed 1 m from each receiver. VR2Tx receivers did not 
require external time-synchronization tags. Moored spa¬ 
tial reference tags were used to improve accuracy of posi¬ 
tion estimates. Functionality of the VPS was tested by 
using a V9 tag attached to the bottom of a pole to mimic 
the location of a sawfish. 
Tide modeling 
To investigate the influence of tidal fluctuations on hab¬ 
itat use by smalltooth sawfish near the shallow oyster 
reefs, a 2-dimensional hydrodynamic model (MIKE 21/3 
Flow Model-FM, DHI Water and Environment, Hprsholm, 
Denmark) was calibrated to the study area (Dye et al., 
in press). This model simulated water transport when 
forced by variations in wind, temperature, salinity, fresh¬ 
water discharges, and tides (Allahdadi et al., 2017). For 
the initial model, bathymetric data were obtained from 
the South Florida Water Management District (100-ft 
digital elevation model for the Lower West Coast region, 
available from website) and the General Bathymetric 
Chart of the Oceans (GEBCO_2014 Grid, available from 
website). To calibrate the model to the study site, water 
depth measurements were taken along 5 transects per¬ 
pendicular to shore (6 or 7 measurements per transect), 
and latitude, longitude, and time of day were recorded 
to relate water depth to the tidal cycle. Tide data from 
the Tides and Currents website of NOAA’s National 
Ocean Service for 2 stations, Locust Point (available 
from website) and Port Boca Grande (available from 
website), were also used to validate the modeled tidal 
heights. The model was run for the entire study period 
(27 May-26 October 2016) to discern when and where 
the receivers may have been too shallow to function 
effectively or were exposed at low tide. 
For analyses, tidal phase was separated into two 6-h 
categories: ebbing and flooding. Position estimates for 
smalltooth sawfish were analyzed by using these 2 cate¬ 
gories and were standardized as counts of detections per 
hour during each period (O’Shea et al., 2010). One test 
compared flooding to ebbing. Two other tests were con¬ 
ducted in recognition of the mixed semidiurnal tides of 
the study area: first and second occurrences were com¬ 
pared for both flooding and ebbing. Slack water was not 
investigated because the model outputs were hourly. 
These analyses were conducted by using Statgraphics 
Centurion 18 (Statpoint Technologies Inc., The Plains, 
VA). Because variances were unequal and the normality 
assumption was violated, non-parametric Kruskal-Wallis 
tests were used to test for differences in number of posi¬ 
tion estimates by tidal stage. 
Analyses of movement and habitat use 
Acoustic data were collected from 27 May through 26 Octo¬ 
ber 2016 within the VPS. Raw detection data were sent to 
Vemco for determination of position estimates for individ¬ 
ual smalltooth sawfish. Before further analysis, position 
estimates were filtered on the basis of their horizontal 
positioning errors (HPEs) following Scheel and Bisson 
(2012); only position estimates with HPE values <20 were 
analyzed because these estimates were likely within 5 m of 
the actual location. 
Next, to determine habitat use within the VPS, poly¬ 
gons corresponding to 3 habitat types were created and 
defined as shallow (<1 m), deep (>1 m), and oyster reef (all 
oyster restoration areas combined) for grouping position 
estimates. One meter was used to separate shallow and 
deep habitats because shallow depths are important for 
small juveniles (Norton et al., 2012). A likelihood ratio chi- 
square test (Cressie and Read, 1989; Rogers and White, 
2007) was used to compare position estimates between 
shallow, deep, and oyster polygons (by using JMP 14, SAS 
Institute Inc., Cary, NC). This test was based on the area 
of the polygons (shallow: 0.12 km 2 ; deep: 1.74 km 2 ; and 
oyster: 0.002 km 2 ), with a significance level of 0.05, and 
compared the ratios of actual use of these habitats to what 
would be expected on the basis of random chance. Tests 
were run for all position estimates combined and then for 
position estimates for each individual sawfish separately 
because pooling position estimates would have eliminated 
the ability to identify individual variation. Bonferroni- 
adjusted confidence intervals (CIs) were also calculated to 
explore variation around the selection ratios and provide 
a conservative measure of reliability (Neu et al., 1974; 
Rogers and White, 2007). In general, ratios >1 indicate 
selection for a habitat type, and ratios <1 indicate selec¬ 
tion against a habitat type. 
By using river-wide data, contour maps were created 
with the software Surfer 13 (Golden Software LLC, 
Golden, CO) to visualize large-scale patterns of habitat 
use and movement, with regard to the nursery hotspot 
boundary (see inset in Figure 1). Raw data were reduced 
by selecting the first detection per hour per receiver for 
each individual, by using SAS 9.4 (SAS Institute Inc.), to 
eliminate the chance of data for 1 or 2 individuals bias¬ 
ing the results (Huston et al., 2017). These maps were 
used to visualize presence of smalltooth sawfish through¬ 
out the river by time period (day: 0900-1659; night: 
2100-0559; crepuscular periods were omitted following 
Poulakis et al., 2013) and by age class following Scha- 
rer et al. (2012). Age classes were based on length (YOY: 
<1500 mm STL; >age 1: >1500 mm STL). 
Results 
In 2016, 20 smalltooth sawfish (726-1770 mm STL) were 
caught and tagged in upper Charlotte Harbor during 
daytime sampling (Peace River: number of tagged fish 
[/j] = 19; Myakka River: n= 1; Table 1). Most individuals 
were caught along the northern shoreline of the Peace 
River within the nursery hotspot, and none were cap¬ 
tured along the southern shoreline portion of the hotspot 
near the VPS. Both age classes, YOY (n=16) and >age 1 
(n=4), were represented. Of the 20 tagged individuals, 
11 sawfish (55%) were detected within the VPS on the 
