MARINE ECOLOGY PROGRESS SERIES
Mar Ecol Prog Ser
Vol. 328: 205–213, 2006 Published December 20
INTRODUCTION
Diel, tidal and lunar rhythms have been observed in
the reproductive, locomotor, feeding and moulting
behaviour of a range of marine organisms (e.g. Hays et
al. 2001, Queiroga & Blanton 2005, Skov et al. 2005).
For organisms at low trophic levels, there may be a
strong advantage in coordinating activity with a cer-
tain phase of an abiotic cycle where it offers a reduc-
tion in predation pressure. For instance, zooplankton
throughout the world’s oceans are known to conduct a
diel pattern of vertical migration (DVM); moving up in
the water column at dusk and returning to depth at
dawn, in order to feed whilst minimising the risk of
predation from visual predators (Hays et al. 2001).
Planktonic organisms can also be passively influenced
by the tidal cycle, being advected by tidal currents
(Cotte & Simard 2005), aggregated in tidal fronts or
concentrated against topographic features (Genin
2004).
Whilst higher pelagic predators are less likely to be
directly affected by these cycles, they are known to
respond to cyclical patterns in prey availability by
adjusting patterns of activity and habitat selection
accordingly (Croll et al. 1998, Baird et al. 2001). In this
way the effect of diel, tidal and lunar cycles can propa-
© Inter-Research 2006 · www.int-res.com*Corresponding author. Email: dws@mba.ac.uk
Diel and tidal rhythms in diving behaviour
of pelagic sharks identified by signal processing
of archival tagging data
Emily L. C. Shepard
1, 5
, Mohammed Z. Ahmed
2
, Emily J. Southall
1
, Matthew J. Witt
3
,
Julian D. Metcalfe
4
, David W. Sims
1,
*
1
Marine Biological Association of the United Kingdom, The Laboratory, Citadel Hill, Plymouth PL1 2PB, UK
2
School of Computing, Communications and Electronics, University of Plymouth, Drake Circus, Plymouth PL4 8AA, UK
3
Centre for Ecology and Conservation, University of Exeter in Cornwall, Tremough TR10 9EZ, UK
4
Centre for Environment, Fisheries and Aquaculture Sciences, Pakefield Road, Lowestoft NR33 0HT, UK
5
Present address: School of Biological Sciences, University of Wales Swansea, Singleton Park, Swansea SA2 8PP, UK
ABSTRACT: Patterns of vertical movement in pelagic predators can be highly complex, reflecting
behaviours such as foraging, thermoregulatory excursions and spawning. Here we used fast Fourier
analysis to identify periodicity in the vertical movements of 6 basking sharks Cetorhinus maximus
from archival tagging data that totalled 595 d. We analysed quantitatively fine-scale vertical
movements of basking sharks over seasonal scales (May to February) and detected predominant
periodicities related to the vertical movements of the sharks’ zooplankton prey. Normal and reverse
diel vertical migration (DVM) represented the main periodic dive behaviour, occurring for 11 to 72%
of individual track times. A tidal pattern of vertical movement, previously unreported for sharks, was
also identified. A possible mechanism for this behaviour appears related to the shark exploiting
tidally-induced aggregations of zooplankton prey at depth. The youngest shark tagged showed a
markedly different pattern of vertical behaviour. Long-term data sets of swimming depth are becom-
ing increasingly available for pelagic predators from pressure-sensitive data loggers. This study
demonstrates the utility of signal processing techniques in objectively identifying both expected and
unexpected periodicity in these continuous, high-resolution tracks.
KEY WORDS: Telemetry · Fourier analysis · Dive profile · Strategy · Fish · Whale · Seal
Resale or republication not permitted without written consent of the publisher

Mar Ecol Prog Ser 328: 205–213, 2006
gate up the food chain, leaving stereotyped signatures
in patterns of predator movement (e.g. Wilson et al.
1993, Cartamil & Lowe 2004).
High-resolution data on the vertical movements of
large marine organisms can be collected by attaching
archival tags, which can record depth frequently and
store data for up to many years (e.g. Arnold & Dewar
2001). However, the identification of rhythmical move-
ment patterns in these data can be difficult because
organisms’ responses can change rapidly with respect
to a changing environment, rhythms can be hard to
detect when they exert an influence over monthly or
seasonal cycles, and different signals may confound
each other where they occur simultaneously.
Signal processing techniques such as fast Fourier
analysis accurately summarise the relative importance
of periodic components within time-series, and are
well-suited to the analysis of archival tagging records
as they achieve rapid throughput of high-resolution
data (Graham et al. 2006). They also provide advan-
tages over techniques such as autocorrelation, which
are likely to require preliminary de-trending of the
depth record (Neat et al. 2005). Summary swimming
depth data in plankton-feeding basking sharks Ceto-
rhinus maximus has shown that they do not hibernate
but remain vertically active in the winter (Sims et al.
2003). DVM has also been identified subjectively in
short, isolated portions of 4 archival tracks to date
(Sims et al. 2005). However, fine-scale depth
movements for C. maximus have yet to be analysed ob-
jectively at a seasonal scale. In this study, fast Fourier
transforms were applied to swimming depth data for
6 basking sharks from archival tags. The aim was to
identify periodic components in the dive records, in or-
der to detect underlying influences on the complex
dive behaviour of a non-air breathing vertebrate and
examine how these changed across seasonal scales.
MATERIALS AND METHODS
Archival tagging. Basking sharks
were tagged from May to July in 2001,
2002 and 2004 (Table 1) with pop-up
archival transmitting (PAT) tags (PAT
Versions 2, 3 and 4; Wildlife Computers)
using methods given in Sims et al.
(2003). Tagging procedures conformed
to institutional and national ethical
guidelines. PAT tags incorporated an
Argos-certified transmitter with a data
logger that recorded pressure to
1000 m, water temperature (–40 to 60°C)
and light level (W cm
–2
at 550 nm wave-
length) at 1 min intervals for the duration
of tag deployment. All tags were programmed with the
same sampling attributes except the tag on Shark 1
which did not record depths greater than 160 m, how-
ever, only 1.26% of the total track time occurred below
160 m. All tags were pre-programmed to release from
the shark after data recording periods of between 7 and
229 d (mean 103.8 d; Table 1). The full archival data set
was only accessible upon recovery of the tags. Six tags
were retrieved (Table 1) representing 24% of the total
tags deployed to date (n = 25). Geolocations were calcu-
lated as described in Sims et al. (2003, 2006). Shark
length and sex were recorded where possible at the time
of tagging (Sims et al. 2003).
Dive analysis. The fast Fourier transform (FFT) oper-
ates by approximating a function with a sum of differ-
ent sine and cosine terms (Chatfield 1996). The influ-
ence of each periodic component is indicated by the
magnitude of the corresponding spectral peak in the
periodogram. The FFT is particularly well-suited to
analysing long-term, high-resolution data sets such as
those from archival tagging studies as the resolution
and range of detectable frequencies are directly
related to the sampling frequency and duration (Gra-
ham et al. 2006). Specifically, fast Fourier transforms
can identify periodicities up to the Nyquist frequency,
which is half the sampling rate, and in this study was
1 cycle 30 s
–1
(Chatfield 1996). Furthermore, there is
potential for extremely high spectral resolution, as the
FFT-generated spectrum contains N/2 distinct fre-
quency components and adjacent components are sep-
arated by ∆f, defined as the sampling frequency/N.
These traits also differentiate the FFT from autocorre-
lation and periodogram analyses that produce outputs
of lower spectral resolution.
Depth data were extracted from archival tag records
and routinely corrected for pressure-sensor drift. An
FFT was applied to the full depth record for each shark
and monthly sub-sections of the tracks using program-
ming routines in MATLAB (MathWorks)
1
. Due to the
206
1
Programming codes available on request
Table 1. Cetorhinus maximus. A summary of the archival tags retrieved from the
25 deployed on basking sharks during May to July 2001, 2002 and 2004. Sex (F =
female) is included where known. (j) the only shark known to be a juvenile;
days: no. of track days. Dates given as d/mo/yr
Shark Shark Sex Date Tag location ‘Pop-up’ location Days
no. length (m) tagged °N °W °N °W
1 4.5 F 24/5/01 50.38 4.11 56.42 7.26 77
2 6.0 25/5/01 50.32 4.13 49.87 2.42 197
3 6.5 31/7/01 55.87 5.39 55.59 5.12 52
4 2.5(j) 31/7/01 55.89 5.40 51.67 6.64 229
5 6.0 F 18/6/02 50.18 4.26 50.35 5.17 7
6 4.5 01/6/04 50.26 4.03 50.93 4.59 33