Tolotti et al.: Vertical movements of oceanic whitetip sharks (Carcharhinus longimanus) 
383 
profiles in a grid with a 0.5-m resolution, according to 
the method described in Bauer et al. (2015). The in- 
terpolated temperature profiles were then plotted by 
using a heat-color scheme and examined in compari- 
son with daily average depths and their standard de- 
viations (SDs). This method was chosen because it has 
been shown to provide an accurate representation of 
the thermal signature and, therefore, to be a viable 
alternative in the absence of temperature time-series 
data (Bauer et al., 2015). 
To identify potential environmental variables that 
could drive the vertical behavior of oceanic whitetip 
sharks, we applied generalized additive models (GAMs). 
This type of model has been used to model habitat pref- 
erences of a variety of oceanic species, including sharks 
(Zagaglia et al., 2004; Damalas and Megalofonou, 2010; 
Bustamante and Bennett, 2013; Lam et al., 2014). The 
daily SD of depth records was considered a proxy for 
the vertical variability (movement amplitude) of sharks 
and, therefore, was chosen as a response variable. The 
explanatory variables included environmental vari- 
ables related to data derived from tag records: geolo- 
cation estimates (longitude and latitude), SST, mixed- 
layer depth (MLD), and shark size. These variables 
were introduced as smoothing terms (thin-plate regres- 
sion splines). To assess temporal effects, “month” was 
included as a factorial variable. Because sharks were 
tagged in 2 ocean basins (Atlantic and Indian oceans), 
“ocean” was also included as a factorial variable. Mod- 
eling was conducted by using the gam function of the 
mgcv package, vers. 1.8-12, in R (Wood, 2006), with a 
Gaussian link function. All possible combinations be- 
tween variables and factors were tested and yielded 63 
models. Models also were run separately for each in- 
dividual shark to investigate individual variability. In 
this case, “ocean” and “shark size” were not considered, 
resulting in 15 models. Model selection was based on 
the Akaike information criterion and further evaluated 
with residual analysis. 
Results 
Vertical movement patterns 
Diel cycles Diel changes in vertical behavior were vis- 
ible across the depth time series of all tagged individu- 
als. However, different patterns also were observed 
within and between individuals. The strongest signal 
was observed during crepuscular hours, especially at 
dawn, when sharks swam at considerably shallower 
depths (Fig. 1). This pattern was consistent, and it 
frequently was observed in the time series of all in- 
dividuals. Circular statistics, applied to the average 
depth per hour of each individual, revealed a lack of 
uniformity over a 24-h cycle for the oceanic whitetip 
sharks (Rao’s spacing test: P>0.001). Test results high- 
light the consistency of this crepuscular pattern, which 
was present even when day and night differences were 
not observed. 
Figure 1 also shows a general pattern of shallower 
average depths during the day than during night, when 
tagged sharks appear to move to deeper waters. This 
pattern was well marked for sharks AOCS4, AOCS5, 
AOCS7, and IOCS1, for which statistically significant 
differences were observed between occupied depths 
during light and dark hours (Suppl. Table 1) (online 
only). For sharks AOCS3 and AOCS6, a difference be- 
tween day and night average depths was not evident or 
statistically significant. As opposed to the averages, SD 
values in depth records did not vary much across the 
24-h cycle (Fig. 1). In contrast, for sharks IOCS1 and 
AOCS4, SD values were higher during the night than 
during the day. With SD considered a proxy of vertical 
amplitude, these sharks appear to explore the water 
column more extensively during nighttime. 
The spectral analysis of high-resolution depth time 
series from the recovered tag of shark AOCS3 revealed 
2 distinct frequency peaks, one at 12 h and another at 
24 h (Fig. 2). The sharp 12-h peak might represent the 
crepuscular pattern described previously. The sharp- 
ness of this peak also indicates a high degree of con- 
sistency in this diel pattern, i.e., a shift in the vertical 
behavior frequently occurred around the same time 
of the day. The 24-h peak indicates that periodic be- 
havioral shifts also occur with daytime and nighttime 
depths. The broad base of this peak, however, indicates 
that the shifts in vertical behavior at this scale are 
less consistent. This result is interesting in that it did 
not appear when the depth readings were aggregated 
by hour. In Figure 1 differences between daytime and 
nighttime average depths of this individual are not 
presented. 
The visual assessment of each depth time series re- 
vealed the identification of 3 main types of day and 
night behavioral patterns. Type-I behavior was char- 
acterized by a preference for shallower waters and 
by some sporadic deep dives during the day and by a 
preference for deeper waters and regular up-and-down 
movements during the night. Type-II behavior featured 
an inverse pattern of that described as type I; sharks 
occupied deeper waters during the day, as opposed to 
night, and also made regular up-and-down movements. 
In contrast, type-III behavior did not show a clear dif- 
ference between daytime and nighttime depth pref- 
erences. Examples of each behavior type can be seen 
in Figure 3. All individuals exhibited the 3 described 
behavioral patterns during their monitoring periods, 
but the frequency of each behavior type varied largely 
among sharks (Fig. 4). Type II was the least frequent 
behavior type observed in all time series and occurred 
most often for shark AOCS3, representing 23.7% of 
the time series of this individual. Type I dominated 
the time series of sharks AOCS4 (41.0%) and IOCS1 
(61.2%), and type III dominated the time series of 
sharks AOCS5 (62.8%) and AOCS6 (50.5%). Note that 
because of gaps in the transmitted depth time-series 
data, not all 24-h periods could be observed and hence 
classified. 
The temporal distribution of behavior types did not 
