352 
Fishery Bulletin 117(4) 
Table 1 
Details for the 8 recovered acceleration data loggers (.ADLs) and the smalltooth sawfish 
(Pristis pectinata) on which they were deployed between May 2014 and November 2015 in 
the Peace River, Florida. Individuals were classified as either young of the year (YOY) or 
individuals age 1 or older, on the basis of stretch total length (Scharer et al., 2012). In this 
study, ADLs were attached either directly to the first dorsal fin, requiring recapture for 
retrieval, or were tethered to pop-off float packages that were retrieved while floating on the 
surface by using a very high frequency transmitter signal (Fig. 2). The asterisks (*) indicate 
the same individual caught 5 months after it was initially tagged. M=male; F=female. 
Sawfish ID 
Date tagged 
Sex 
Length 
(mm) 
Age class 
Deployment 
duration (d) 
Retrieval 
method 
1 
9 May 2014 
M 
808 
YOY 
5.03 
Recapture 
2 
12 Jun 2014 
M 
1735 
>age 1 
5.37 
Recapture 
3 
12 Jun 2014 
F 
1575 
>age 1 
5.09 
Recapture 
4 
11 Jun 2015 
M 
980 
YOY 
5.06 
Recapture 
5* 
11 Jun 2015 
F 
1025 
YOY 
5.13 
Recapture 
6 
11 Jun 2015 
F 
1049 
YOY 
5.12 
Recapture 
7 
10 Nov 2015 
F 
1831 
>age 1 
5.23 
Pop-off 
8* 
20 Nov 2015 
F 
1531 
>age 1 
5.42 
Pop-off 
with the lowest AICc was determined to have the best fit. 
Following the formation of this model describing the burst 
probability, the hours during which no burst events were 
detected were removed from the analysis, and the number 
of bursts observed in the remaining hours was modeled 
by using a GAMM with a negative binomial distribution. 
This model included the same predictor variables used in 
the previously described models, and the best-fit combina¬ 
tion of fixed-effect predictor variables was determined by 
using AICc. 
In all GAMMs, normality of residuals was confirmed 
by using the gam.check function in the mcgv package in 
R, and autocorrelation was accounted for by using the 
CorARl function. Cyclic smoothers were used for time 
of day to reflect the circular nature of the 24-h clock. 
Fixed effects that were maintained in the chosen model 
were further examined with Tukey’s honestly significant 
difference (HSD) tests by using the multcomp package, 
vers. 1.4-10 (Hothorn et al., 2008). Additionally, to deter¬ 
mine whether there was a shift in depth distributions 
with size, the frequency distributions of hourly depths 
of YOY and individuals age 1 or older were compared by 
using a Kolmogorov-Smirnov test. 
Data processing and analyses: passive acoustic monitoring 
Acoustic transmitter data were downloaded from acoustic 
receivers and examined to provide location data concurrent 
with ADL-derived behavioral data. Positions derived from 
acoustic data were matched with those from ADL data by 
using time stamps, and a location was designated for each 
hour that acoustic transmissions were received. Receivers 
were deployed in 2 habitats: creeks, lined with red man¬ 
groves (Rhizophora mangle), that are generally shallow, 
enclosed habitats (with depths <1 m at most tides) and 
habitats within the main stem (or main channel) of the 
Peace River that are generally deeper (with depths of 1-3 m 
at most tides), more open areas with less habitat complex¬ 
ity (Fig. 1). If acoustic transmissions were received from 
multiple habitats during the same hour of deployment or if 
they were not received during an hour of deployment, data 
were not included in location analyses. The percentage of 
time smalltooth sawfish were active and the probability 
and frequency of burst events were compared between the 
2 types of habitats by using analysis of variance (ANOVA), 
with depth, temperature, time of day, tide, and age class 
included as fixed-effect predictors in the models and indi¬ 
vidual included as a random effect. The dredge function in 
the MuMIn package (vers. 0.12.0; Barton, 2009) in R was 
used to determine which factors should be included as 
informative predictors of activity and burst events. 
To visualize the overall area and habitat use by smalltooth 
sawfish while they carried ADLs, a contour map was created 
by using Surfer 13 (Golden Software LLC, Golden, CO) fol¬ 
lowing the methods of Huston et al. (2017). Age classes were 
combined because all individuals used the same area of the 
river regardless of size during the 5-d ADL deployments. 
Results 
Between May 2014 and November 2015, ADLs were 
deployed on 10 smalltooth sawfish (Table 1). Eight of these 
tags were recovered, and over 994 h of high-resolution 
behavioral data were collected from them (Fig. 4). Acoustic 
detections were received during 225 h of the ADL deploy¬ 
ments, providing concurrent position and behavior data for 
these hours. Burst events were identified for all sawfish, 
with 6-58 burst events observed per day for individuals. 
These burst events composed 2 main categories: repeated, 
