Journal of Mammalogy, 85(2):245–253, 2004
SEASONAL CHANGES AND ACTIVITY-DEPENDENT
VARIATION IN HEART RATE OF ROE DEER
PETER K. THEIL, ALBERT E. COUTANT,
AND
CARSTEN RIIS OLESEN *
Department of Wildlife Ecology and Biodiversity, National Environmental Research Institute, Grenåvej 12,
DK-8410 Rønde, Denmark
Three free-living female European roe deer (Capreolus capreolus L.) were fitted with a telemetry implant to
measure heart-rate. Observations of behavior of undisturbed deer were made during daylight from 12 February to
12 September 1996 in the Hestehave Forest (185 ha), Kalø, Denmark. Behaviors were classified in 14 different
categories. Heart rate (HR, in beats/min) in 10-s intervals was characterized by the mean HR, short-term, longterm, and overall HR variation within an observation period as well as the autocorrelation of mean HR and shortterm variability. A statistical model described 90.1% of the observed variation in HR and included 4 parameters:
category of behavior, Julian date, deer age, and wind speed. From 11 April to 4 August, HR is described by
a tandem cosine curve with a peak on 6 May. In contrast, HR variability (short-term, long-term, and overall
variability) was constant throughout the year. HR increased by 0.36 beats/min each time wind speed increased by
1 m/s, and HR variability was also positively affected by the wind speed. HR and HR variability was highest for
locomotive behaviors, medium for standing behaviors, and lowest for bedded behaviors. Autocorrelation of mean
HR and short-term variability was not significantly affected by any of the observed variables. Using mean HR
and HR variability, it was possible statistically to distinguish 10 categories of behavior ranging from bedded with
closed eyes (60.5 beats/min) to fleeing (254.8 beats/min).
Key words:
wind effect
behavior, Capreolus capreolus, energy, energy metabolism, heart rate variability, roe deer, season, telemetry,
HR and rate of oxygen consumption has been reported to be
approximately linear for birds and several small mammals
(Morhardt and Morhardt 1971; Owen 1969) and for elk
(Robbins et al. 1981), moose (Renecker and Hudson 1985),
mule deer (Kautz et al. 1981), reindeer (Nilssen et al. 1984a),
and white-tailed deer (Holter et al. 1976; Mautz and Fair
1980).
Until recently, available techniques for HR measurements
have been of a poor quality, resulting in inaccurate data,
primarily due to interference from muscle potentials and
electrode movement, especially when animals are performing
locomotive activities (MacArthur et al. 1979). Previously,
studies of HR and corresponding behavior have focused on
enclosed animals such as roe deer (Herbold et al. 1991), red
deer (Cervus elaphus—Price et al. 1993), elk (Lieb 1981),
white-tailed deer (Jacobsen 1979; Moen 1978), mule deer
(Weisenberger et al. 1996), and bighorn sheep (Ovis canadensis—Geist et al. 1985; MacArthur et al. 1979). HR data
have also been collected from free-living animals, such as mule
deer (Freddy 1984), moose and red deer (Langvatn and
Andersen 1991), bighorn sheep (MacArthur et al. 1979), and
reindeer (Nilssen et al. 1984a). HR is influenced by habitat
(Geist et al. 1985), distance to roads (MacArthur et al. 1979),
ambient temperature (Johnson and Gessaman 1973; Lieb
1981), time of day (Freddy 1984; MacArthur et al. 1979; Price
Measurements of heart rate (HR) are very useful when
constructing energy or time budgets for free-living ungulates
(Holter et al. 1976; Moen 1978) or when studying behavior
and how animals respond to disturbances (Geist et al. 1985;
Herbold et al. 1991). HR has been shown to be a reliable
predictor of the metabolic rate of elk (Cervus elaphus
nelsoni—Robbins et al. 1979), moose (Alces alces—Renecker
and Hudson 1985), mule deer (Odocoileus hemionus—Kautz
et al. 1981), reindeer (Rangifer tarandus platyrhynchus and R.
t. tarandus—Nilssen et al. 1984a), and white-tailed deer
(Odocoileus virginianus—Holter et al. 1976; Mautz and Fair
1980). Seasonal changes in metabolic rate, with low levels
during winter and high levels during summer, have been well
documented among ungulates of the northern temperate zone
(moose—Regelin et al. 1985; Renecker and Hudson 1986;
Schwartz et al. 1991; reindeer—Nilssen et al. 1984b; roe
deer, Capreolus capreolus—Ellenberg 1978; Mauget et al.
1997; Weiner 1977; sheep, Ovis aries—Blaxter and Boyne
1982; and white-tailed deer—Holter et al. 1976; Silver et al.
1969; Worden and Pekins 1995). The relationship between
* Correspondent: [email protected]
Ó 2004 American Society of Mammalogists
www.mammalogy.org
245
246
JOURNAL OF MAMMALOGY
FIG. 1.—Frequency distribution of ratios between 2 consecutive
RR-intervals (n ¼ 1.28 106) from four 24-h periods for each of
3 female roe deer (Capreolus capreolus). Ratios were natural logarithmic (ln)-transformed and values ,0.47 and .0.47 were regarded
as artifacts (,0.4% of the total recordings) and deleted before the RRintervals (intervals between adjacent R waves) were converted to heart
rate.
et al. 1993), body mass (Espmark and Langvatn 1985;
Jacobsen 1979), age, season, and behavior (MacArthur et al.
1979; Moen 1978; Price et al. 1993; Weisenberger et al. 1996).
Furthermore, oscillations of HR in humans are reported to be
due to respiratory rhythms or baroreceptor-dependent rhythms
(Mølgaard 1995), or they may be random (Porges 1985).
Moen (1978) described HR variations throughout the year
for white-tailed deer by a sine wave. The statistical model used
in that study was later revised to a nonsymmetrical tandem
cosine curve to describe annual variation of energy metabolism
of moose (Moen and Moen 1998). The aim of our study was to
present a model for seasonal variation of HR in free-living
female roe deer that would make it possible to study energy and
time budgets even when deer are out of sight. We used HR
telemetry technology that eliminated muscle potential interference even when deer were active.
MATERIALS AND METHODS
Observations of behavior of deer were carried out in Hestehave
Forest at Kalø, Denmark. The forest is 185 ha and consists mainly of
deciduous trees (65% of the area). Hestehave Forest is part of the Kalø
estate in eastern Jutland (568179N, 108289E). The spring population of
roe deer in the forest is about 100 individuals. The forest is managed
according to general Danish forestry practices. No hunting is allowed
except for selective roe deer hunting with rifle performed by
administrative forestry personnel. Three female deer were captured
within temporarily opened, fenced exclosures (0.5–3.0 ha), which
were in place to protect young deciduous tree shoots from deer
browsing. The deer were captured, transported to Aarhus University
Hospital in Skejby, and surgically fitted with a HR telemetry implant
on 21 January 1996. Surgery and implantation of the HR implant
lasted 1.5 h per deer, and all deer were released again in the forest
within 12–14 h. An electronic device was implanted subcutaneously in
the neck, from which a wire was inserted into the heart through the
jugular vein. Design of the telemeter and implantation procedures have
been described by Olesen et al. (1998). Age of deer was determined at
the time of surgery by evaluating the set of teeth and wear. One 2year-old deer and 2 older deer had initial body masses of 22.2, 24.0,
and 26.0 kg, respectively. The 2 older deer were each accompanied by
Vol. 85, No. 2
a single fawn at the beginning of the experiment. The 3 deer each gave
birth to 2 fawns in early summer 1996. All 3 deer were registered to
give birth to fawns the year after, indicating the HR implant did not
have any negative implications on the reproduction, and they were
shot in the following hunting season (fall 1997).
Time intervals (ms) between 2 heartbeats (RR-interval, i.e., interval
between adjacent R waves) were transmitted from the implanted
device to a more powerful 150-MHz radio transmitter in a neck collar
with different color code. Each collar was equipped with a multidirectional antenna, and RR-intervals were continuously transmitted at
a specific tuned frequency. A data logger (RX 900, Televilt,
Lindesberg, Sweden) capable of collecting HR data from a single
deer at a time was placed in the forest. The data logger also registered
date, time, transmitter signal frequency, and signal strength (dB) for
each measurement. Observations of behavior were made in daylight
from 12 February to 12 September 1996, without disturbing the 3 deer.
Only the behaviors fleeing and trotting were recorded when the
observer intentionally disturbed the animals.
Wind direction and speed were measured 2 m above ground level
twice per hour every 24 h throughout the experiment at the Danish
Meteorological Institute, Station 06070, located about 7 km from the
boundary of the forest. Measurements were made according to
international standards; mean speed (0.1 m/s accuracy) and mean
direction were measured for 10 min every 30 min.
Data processing.—Artifacts were sometimes registered in sequential
series of RR-intervals. They occurred either if a single heartbeat was
not recorded or if a disturbance of radio transmission erroneously
divided a single RR-interval into 2 artifacts. To eliminate artifacts, it
was necessary to evaluate the ratio between 2 consecutive RRintervals. The ratio is known from humans to be 1.35 (H. Mølgaard,
pers. comm.) even though no standardized filter algorithms have yet
been developed to eliminate RR-interval artifacts (Mølgaard 1995).
Correction of the upper limit of the ratios (1.35) for the HR range
difference between humans (60–200 beats/min) and roe deer (60–300
beats/min—this study) gives 1.60 for the upper limit [0.35 (300 – 60/
200 – 60) þ 1 ¼ 1.60] and 0.625 for the lower limit (reciprocal of
1.60) as valid ratios of consecutive RR intervals. The frequency
distribution of all ratios throughout four 24-h periods per deer shows
that it is reasonable to assume ratios ,0.625 or .1.60 to be artifacts
(,0.47 or .0.47 after natural logarithmic transformation; Fig. 1).
All artifacts were deleted, amounting to ,0.4% of all recordings. RRintervals were later converted to HR (beats/min).
Start and stop times for each observation of behavior were
registered, defining an observation period. For all categories of
behavior, 2 s were added to start time, and a single second was
subtracted from stop time to minimize influence on HR of previous
and subsequent behaviors. When standing followed walking, the 1st
minute of standing was discarded due to high influence of the previous
behavior on HR. These adjustments were justified because HR
increases nearly instantly but decreases in a curvilinear fashion. For all
bedded categories of behaviors, procedures were different; 2 min were
added to start time if previous behavior was of standing categories,
and 1 min was subtracted from stop time if subsequent behavior was
of standing categories. The reason for these adjustments was that HR
of bedded roe deer often rose 30–40 s before the deer changed position
to standing behavior. The elevated HR may be interpreted as
a response to no observed stimuli (emotionally conditioned HR
increment).
Each observation period was divided into 10-s intervals within
which the mean (HR10) and standard deviations (SD10) of HR were
calculated from RR-intervals. To obtain normal distributions, SD10
values were logarithmic transformed. An autoregressive process of
April 2004
THEIL ET AL.—VARIATION IN HEART RATE OF ROE DEER
order 1 described these data, indicating that mean (HR10) and standard
deviation of HR (SD10) in one 10-s interval were influenced by levels
of HR10 and SD10 in preceding 10-s intervals. Autocorrelation
coefficient, mean, and standard deviation (Table 1) sufficiently
describe an autoregressive process of order 1 (Box and Jenkins
1976). Furthermore, describing HR data as an autoregressive process
takes into account that the data comprise a time series and that data are
not independent. Only observation periods with 6 10-s intervals
were used in calculations of autocorrelation coefficients. Autocorrelation coefficients [R(HR10) and R(SD10)] were quantified as means,
weighted by the number of 10-s intervals in the observation period.
Standard deviation of mean HR [SD(HR10)] describes variation of HR
within minutes and is termed the long-term variability (SDANN10
[standard deviation of the 10-s averages of NN intervals in the
observation period]—Mølgaard 1995). Such long-term variation can
be influenced by circadian rhythms of HR. The mean of standard
(SD10)] describes variation between successive
deviations of HR [X
heartbeats and is termed the short-term variability (SDNN10 [standard
deviation of NN intervals in 10-s periods]—Mølgaard 1995). Short
term oscillations can be a result of respiratory-dependent (4-s
intervals) and baroreceptor-dependent (10-s intervals) rhythms. To
our knowledge, no physiological explanation has yet been given for
overall variability [SD(SD10)], but a possible explanation might be
a nervously conditioned response. To obtain normal distributions,
SDANN10 and SD(SD10) were logarithmic transformed. When
calculating R(SD10), SDNN10, and SD(SD10), the 1st 10-s interval in
an observation period was ignored because the 1st value of SD10 was
significantly higher than consecutive values, obviously because a 2-s
adjustment of start time was too small. This did not hold for mean HR
because mean HR was 1st averaged within 10-s intervals and
afterward among 10-s intervals within an observation period. There
were no significant differences in HR between the 1st 10-s interval and
the following intervals within an observation period.
Owing to depletion of battery power, transmission of HR data
ceased before a 1-year cycle had been completed. All statistics were
performed using SAS v. 6.11. (SAS Institute Inc. 1997). The level of
significance employed was P , 0.05. Parameters describing autocorrelation and HR (tandem cosine model) are reported as means 6
standard errors. Long-term, short-term, and overall variability of HR
are reported by the lower and upper 95% confidence limits due to the
logarithmic transformation of these data. The cosine model used to
describe HR is a linear normal model. However, in order to determine
all parameters at once (i.e., wavelength, phase, amplitude, and mean
level of HR) for each category of behavior, it was necessary to use
nonlinear regression (PROC NLIN—SAS Institute Inc. 1997). The
estimated HR for categories of behavior was compared pairwise using
Wald’s test (Armitage and Colton 1998). Short-term, long-term, and
overall HR variability were tested by unbalanced analysis of variance
(Sokal and Rohlf 1995) using the PDIFF statement of the MIXED
procedure (SAS Institute Inc. 1997).
Tests for factors affecting autocorrelation were performed by
describing continuous variables as discrete variables using the
EXACT statement of the FREQ procedure (SAS Institute Inc.
1997). Seasons were categorized into 10 periods (21 days each), and
wind speed was classified as ,7.5 or 7.5 m/s.
RESULTS
Autocorrelation.—The autocorrelation coefficient is a value
between –1 and 1 and describes the correlation between
consecutive observations in a time series. The autocorrelation
coefficient for mean HR and short-term variability of HR was
247
TABLE 1.—Definitions of terms used to describe heart rate (HR)
within an observation period. Recordings of heart rate in a given
observation period were divided into multiple 10-s periods, and mean
(HR10) and standard deviation (SD10) of HR was calculated within 10s intervals (Mølgaard 1995).
Term
Statistics within an observation period
) HR
Mean HR
(X
Long-term variability
SDANN10
Autocorrelation of HR
Ra(HR10)
Short-term variability
Overall variability
SDNN10
SD(SD10)
Autocorrelation of
short term variability
Ra(SD10)
a
Unit
Beats/min
Beats/min
Beats/min
Beats/min
Definition/calculation
of (HR10)
X
Standard deviation
of (HR10)
Autocorrelation
of (HR10)
of (SD10)
X
Standard deviation
of (SD10)
Autocorrelation
of (SD10)
R represents the Greek letter rho.
found to be constant and hence not significantly affected by
any of the observed variables (wind speed, age, season, or
behavior). The autocorrelation coefficient for mean HR
[R(HR10)] was 0.35 6 0.01 (n ¼ 715) regardless of wind
speed (v2 ¼ 1.03 , d.f. ¼ 16, P ¼ 1.00), age (v2 ¼ 2.02,
d.f.¼ 15, P . 0.99), season (v2 ¼ 2.21, d.f. ¼ 3, P . 0.53),
or category of behavior (v2 ¼ 1.00, d.f. ¼ 1, P . 0.31). The
autocorrelation coefficient for the standard deviation within
10-s intervals [R(SD10)] was 0.13 6 0.01 (n ¼ 618)
regardless of wind speed (v2 ¼ 0.53, d.f. ¼ 13, P ¼ 1.00),
age (v2 ¼ 1.70, d.f. ¼ 12, P . 0.99), season (v2 ¼ 1.49, d.f.
¼ 3, P . 0.68), or category of behavior (v2 ¼ 0.29, d.f. ¼ 1,
P . 0.58).
HR model.—We describe annual variation in mean HR of
adult female roe deer with a tandem cosine model during
summer and constant HR during winter. The tandem cosine
model consists of an ascending half wave (equation 1) and
a descending half wave (equation 2) as proposed by Moen and
Moen (1998). The ascending and the descending half waves
had different wavelengths and hence were not symmetrical.
The tandem cosine model assumed 2 cosine models with
identical bases, identical phases, identical amplitudes, but
different wavelengths. A standard cosine curve has the
maximal Y value when X ¼ 0; hence, the phase of the tandem
cosine model needs to be corrected. Phase correction is
parameterized as the day of the year, where HR is at maximum
(max). The model incorporated seasonal variation for all
categories of behavior with increasing HR during April and
early May (102 Julian day 127; equation 1) and
decreasing HR in May to early August (128 Julian day 218; equation 2):
FðseasonÞascend ðbeats=minÞ ¼ Aj fcos½2pðd maxÞ=w1 þ 1g;
ð1Þ
where Aj ¼ behavior-dependent amplitude (j ¼ 1, 2, 3, 4;
beats/min), d ¼ Julian date (days 102–127), max ¼ day with
maximum HR (for adjusting the tandem cosine curve to the
248
JOURNAL OF MAMMALOGY
TABLE 2.—Annual variation in basal heart rate (HR; beats/min) for
14 behavioral categories displayed by roe deer (Capreolus capreolus).
HR was determined by radiotelemetry. Effects of age and wind speed
on HR are included. n ¼ number of observation periods in which
a given behavior was recorded, and the amplitude (Aj) is the behaviordependent change in HR (contingent on the category of behavior, j)
used to describe annual HR variation (equations 1 and 2). Mean and
standard error of HR for a given category of behavior were determined
using a tandem cosine model to describe annual variation in HR.
Behavior
b
Fleeing
Trottingb
Walking
Foraging
Standingþalertc
Standingþgrooming
Standingc
Urinating
Beddedþgrooming
Beddedþalert
Standingþruminating
Beddedþruminating
Bedded
Beddedþclosed eyes or head down
n
111
11
319
653
170
78
429
8
16
19
17
79
156
29
Aj
0
0
16.2
22.7
22.7
22.7
22.7
22.7
7.3
7.3
22.7
7.3
7.3
7.3
HRa (beats/min)
254.8
183.9
104.0
80.8
80.5
79.7
78.1
77.7
67.9
67.5
65.1
64.4
63.8
60.5
a
b
c
d
d
de
de
de
e
ef
ef
ef
ef
f
SE
1.3
4.1
1.1
0.9
1.3
1.8
1.0
4.9
3.5
3.2
3.4
1.8
1.4
2.6
Age effect
2 years old
3 years old
645
1,450
3.3
0.0
0.9
Wind effect (bws)
2,095
0.36
0.15
a
Different letters indicate significant differences (P , 0.05) in basal HR among
behaviors.
b
In deer engaged in fleeing, forelegs and hind legs move independently, while one
foreleg and one hind leg move contemporarily during trotting.
c
Deer standing alert (opposed to deer standing) did not move their head at all, and both
ears and eyes were focused in a specific direction.
annual cycle), and w1 ¼ wavelength of ascending half wave
(days), and
FðseasonÞdescend ðbeats=minÞ ¼ Aj fcos½2pðd maxÞ=w2 þ 1g;
ð2Þ
where Aj and max are as in equation 1, d ¼ Julian date (days
128–218), and w2 ¼ wavelength of descending half wave
(days).
Finally, during winter (i.e., d , 102 or d . 218), HR was
assumed to be constant (F(season) ¼ 0).
The limits defining the ascending and the descending half
wave and the constant HR during winter was not fixed but
derived by means of the estimated values of phase (max) and
wavelengths of the cosine curves (w1 and w2). The boundary
between the winter HR level and the ascending half wave was
specified as max w1/2, while the boundary between the
descending half wave and the constant HR level during winter
was specified as max þ w2/2.
Season (expressed through Julian date) had a significant
effect on HR (F ¼ 579,502, d.f. ¼ 13, 2073, P , 0.001). The
tandem cosine model described rhythmic changes in HR from
11 April (day 102) to 4 August (day 217). Estimates of
wavelengths of the cosine curves (w1 and w2) were 50 6 4
days (F ¼ 174.7, d.f. ¼ 1, 2,073, P , 0.001) and 181 6 7 days
Vol. 85, No. 2
(F ¼ 670.7, d.f. ¼ 1, 2,073, P , 0.001) for the ascending and
the descending cosine half waves, respectively, and maximum
HR (max) was reached on 6 May (day 127 6 1; F ¼ 19,367,
d.f. ¼ 1, 2,073, P , 0.001). Amplitude and max, tested for
individual differences in a single step, were not significantly
different (F ¼ 0.11, d.f. ¼ 4, 2,073, P ¼ 0.98). Behavioral
observation data were divided into 4 groups according to
amplitude. Estimated amplitudes were 1.1 6 4.8 beats/min
for fleeing and trotting (F ¼ 0.05, d.f. ¼ 1, 2,073, P ¼ 0.82; in
equation 1, no amplitude was applied for these 2 categories;
Table 2), 16.2 6 1.9 beats/min (F ¼ 185.5, d.f. ¼ 1, 2,073, P ,
0.001) for walking, 22.7 6 0.6 beats/min (F ¼ 1,368, d.f. ¼ 1,
2,073, P , 0.001) for all nonlocomotive standing behaviors,
and 7.3 6 1.1 beats/min (F ¼ 43.2, d.f. ¼ 1, 2,073, P , 0.001)
for all bedding behaviors. HR data collected from other
individuals using the same HR device indicated that HR was
constant during the rest of the year (August–April).
The tandem cosine model included a unity term in order to
make all values positive because a normal cosine curve ranges
from 1 to þ1 (Edwards and Penney 1990), and oscillations
begin with a descending half wave followed by an ascending
half wave.
HR was significantly higher (3.3 6 0.9 beats/min) for the 2year-old deer than for the 2 older deer (F ¼ 14.40, d.f. ¼ 1,
2,073, P , 0.001; Table 2). For each behavior, except for
fleeing and trotting, HR varied with seasons (Figs. 2A–D; F ¼
579,502, d.f. ¼ 13, 2,073, P , 0.001). HR increased
significantly (0.36 6 0.15 beats/min; F ¼ 5.7, d.f. ¼ 1,
2,073, P , 0.05) when wind speed increased by 1 m/s. Effects
of behavior, season, wind speed, and deer age on mean HR
were incorporated into equation 3 and explained 90.1% of the
observed variation (adjusted R2 ¼ 0.901; d.f. ¼ 22; n ¼ 2,095):
HRðbeats=minÞ ¼ ai þ FðseasonÞ þbWS WS þ age effect; ð3Þ
where ai ¼ behavior-dependent mean HR (i ¼ 1–14; beats/
min), F(season) ¼ HR at a given date specified in equation 1 or
equation 2 or ¼ 0 during winter, bWS ¼ coefficient for wind
effect (beats min1 s m1), and WS ¼ wind speed (m/s). The
age effect is included to account for the higher HR commonly
observed in young animals.
Long-term, short-term, and overall variability.—Long-term,
short-term, and overall variability was highest for locomotive
behaviors, medium for standing behaviors, and lowest for
categories of bedded behaviors. The wind speed affected
positively all 3 measures of HR variability (although not
significantly in case of long-term variability). In contrast, only
short-term variability was significantly affected by deer age.
Long-term variability (SDANN10) varied significantly among
categories of behavior (F ¼ 6.40, d.f. ¼ 13, 1,153, P , 0.001;
Table 3) and was highest for locomotive behavior or when deer
were alert. Further, wind speed tended to affect long-term
variability (F ¼ 3.11, d.f. ¼ 1, 1,153, P ¼ 0.08), whereas age
had no effect (F ¼ 0.07, d.f. ¼ 1, 1,153, P ¼ 0.79). Effects of
behavior and wind speed explained 7.9% of the long-term
variability (adjusted R2 ¼ 0.079).
Short-term variability (SDNN10) varied significantly among
behavioral groups (F ¼ 22.40, d.f. ¼ 13, 1,345, P , 0.001;
April 2004
THEIL ET AL.—VARIATION IN HEART RATE OF ROE DEER
249
FIG. 2.—Effect of various behaviors on heart rate (HR) of roe deer (Capreolus capreolus) at various times of year, as shown by observed
(points) and predicted by a tandem cosine model (solid line). Figures show HR when engaged in A) fleeing (n ¼ 111), B) walking (n ¼ 319), and
C) foraging (n ¼ 653) behaviors or D) when bedded (n ¼ 156).
Table 3) and was highest for locomotive behavior. Wind speed
(F ¼ 19.00, d.f. ¼ 1, 1,345, P , 0.001) and age (F ¼ 17.20,
d.f. ¼ 1, 1,345, P , 0.001) both affected the short-term
variability. Effects of behavior, wind speed, and age explained
23.9% of the variation of short-term variability (adjusted R2 ¼
0.239). Short-term variability for foraging (mean ¼ 2.82) was
significantly lower than for standing (mean ¼ 3.41; F ¼ 9.35,
d.f. ¼ 1, 1,345, P , 0.01).
Overall HR variability [SD(SD10)] varied with category of
behavior (F ¼ 2.49, d.f. ¼ 13, 1,153, P ¼ 0.002) and wind
speed (F ¼ 6.21, d.f. ¼ 1, 1,153, P ¼ 0.013; Table 3), and
overall variability was highest for locomotive categories of
behavior. In contrast, there was no significant effect of age of
the deer (F ¼ 0.59, d.f. ¼ 1, 1,153, P ¼ 0.44). Behavior and
wind speed explained 3.6% of the variation of overall HR
variability (adjusted R2 ¼ 0.036).
DISCUSSION
Seasonal variation.—Moen’s (1978) sine-based model of
temporal variation in HR throughout the year for white-tailed
deer served as the basis for our development of the statistical
model used in this report. Annual rhythms of HR have been
quantified more recently using a tandem cosine curve (Moen
and Moen 1998). Our model differs from that one in 2 ways.
First, the tandem cosine model presented here covers only part
of the annual cycle (and therefore an additional parameter is
required). Second, in contrast to Moen and Moen (1998) and
Moen and Boomer (2000), we fit the tandem cosine model to
observed data. We found a pronounced annual variation in HR
from April to August (equations 1 and 2) with a maximum on 6
May. The annual pattern of HR was not symmetrical around
the peak. Mean birth date for roe deer at Kalø was 2 June
(Strandgaard 1972); thus, the elevated HR during April and
May is probably due to the increased energetic costs of
gestation and preparation for lactation. This corresponds well
with results of Mauget et al. (1997), who showed that energy
expenditure of pregnant deer increased during the last 2 months
of gestation. Gestation (and early lactation) is well synchronized with increased availability of highly digestible herbs.
Digestibility of organic matter is highly correlated with
digestibility of energy (Weisbjerg and Hvelplund 1993). Based
on chemical analyses of buds and seedlings normally selected
by roe deer, digestibility of organic material was estimated to
be 70% during May and June and 30–40% the rest of the year
(Olesen et al. 1998). The greater availability and digestibility of
food occurred simultaneously with an increased frequency of
foraging bouts during spring as reported by Jeppesen (1989).
Higher food intake during spring also affected energy
metabolism. Elevated availability and digestibility of food
selected by roe deer changed the energy balance from negative
to positive in May (Cederlund and Liberg 1995). Metabolic rate
is known to increase for a period after food intake (specific
dynamic action, diet-induced thermogenesis, or thermal effect
of food—Hill and Wyse 1989). Digestive processes at least
partly explain these energetic costs. Elevated HR during spring
is likely caused by the combination of decreased intervals
between foraging bouts due to higher digestibility of food and
increased availability of higher-quality food, resulting in
250
JOURNAL OF MAMMALOGY
TABLE 3.—Measures of heart rate (HR) variability among
observation periods for 14 roe deer behaviors. Long-term variability
(SDANN10) is expressed as standard deviation of mean HR measured
among 10-s intervals. Short-term variability (SDNN10) is expressed as
means of standard deviation of HR measured within 10-s intervals.
SD(SD10) is standard deviation of the standard deviations measured
within 10-s intervals. When calculating SDANN10 and SD(SD10), only
observation periods consisting of 2 10-s intervals are included.
Effects on HR variability of age and wind speed are included. The
intervals represents the lower and the upper 95% confidence limits for
short-term, long-term, and overall HR variability. Values within
a column with the same letter are not significantly different.
95% confidence interval
Behavior category
Fleeing
Trotting
Walking
Foraging
Standingþalert
Standingþgrooming
Standing
Urinating
Beddedþgrooming
Beddedþalert
Standingþruminating
Beddedþruminating
Bedded
Beddedþclosed eyes
or head down
SDANN10
2.794.99
1.348.51
2.944.08
2.092.62
2.954.10
2.274.19
2.453.22
1.157.63
1.834.51
3.095.92
1.653.44
1.902.87
2.343.21
a
abcd
a
c
a
abc
b
abcd
abc
a
bc
b
bc
0.981.96 d
SDNN10
5.487.20
4.858.52
4.205.01
2.713.13
3.193.89
3.774.95
3.083.79
2.805.27
2.503.99
2.313.66
2.213.59
3.003.85
2.563.17
SD(SD10)
a
0.3100.535 a
ab
0.1340.850 ab
b
0.2930.402 a
d
0.2610.324 b
c
0.2530.357 ab
b
0.2940.514 a
c
0.3210.412 a
bcd 0.0690.393 ab
cde
0.1710.449 ab
cde
0.2110.436 ab
cde
0.2380.472 ab
cd
0.2570.375 a
de
0.2860.380 ab
1.842.76 e
0.1540298 b
Age effect
2 years old
3 years old
Wind effect (per m/s)
0.080.07 ns
0
0
0.060.15
0.0270.012 ns
0
0.000.03 ns
0.010.03
0.0010.010
decreased foraging time per bout, faster turnover, and
ultimately a higher metabolic rate. Overall, this HR elevation
was pronounced for all categories of behavior except fleeing
and trotting.
Another key factor for metabolic rate is body mass. Mauget et
al. (1997) and Regelin et al. (1985) found that energy
expenditure is correlated with total body mass (including fetus).
It was not possible to record body mass of the deer during our
field study, but annual changes of body mass are accounted for
indirectly by the tandem cosine model and age effects.
After parturition body mass drops markedly, although
energy expenditure remains high (Figs. 2B–D), probably due
to lactation. The intensity of lactation peaks 3–5 weeks after
fawning in black-tailed deer (Sadleir 1982), elk (Robbins et al.
1981), and red deer (Loudon et al. 1983). The period after peak
lactation (July–September) is characterized by increasing body
mass (Ellenberg 1978; Sadleir 1982) as muscle tissue and fat
reserves are replenished. Decreasing lactation intensity and
increasing body mass can account for the sigmoid curve of HR
following the breeding season. However, the decreased HR in
June and July may also, in part, be explained by a lower dietinduced thermogenesis as the organic matter digestibility
decreased (Olesen et al. 1998) and frequency of foraging bouts
dropped (Jeppesen 1989).
Vol. 85, No. 2
Our HR model corresponds well to the seasonal variation of
resting metabolic rate found by Weiner (1977) in roe deer.
Weiner (1977) reported lower metabolism in December–
February and higher metabolism during June–August. The
ratio of maximum summer HR to winter HR when bedded was
1.23 (equation 1). If we assume a linear relationship between
resting metabolic rate and the HR when roe deer were bedded,
this HR ratio should equal the ratio between summer and
winter resting metabolism. It is in close agreement with the
metabolism ratios for roe deer of 1.28 in May and 1.09 in June–
August reported by Weiner (1977). In contrast, Mauget et al.
(1997) reported a summer-to-winter resting metabolism ratio of
0.97 for nonpregnant roe deer.
Moen (1978) presented a sine-based model for an annual
cycle of HR measurements from white-tailed deer. It included
season and behavior variables and explained 51–84% of the
variation, depending on categories of behavior. The model
showed peak HR on 15 August and minimum HR on 15
February. These dates differ from those in our model probably
because Moen’s data were based on values from 4 males and 2
females without separating individual effects, thus biasing
results in favor of a peak nearer male rutting season. Further,
Moen calculated mean HR on a monthly basis, leading to some
loss of information. Moen reported an increased amplitude of
HR with increasing level of activity, but we found the highest
amplitude for standing categories of behaviors and lower
amplitudes for bedding and locomotive activities. This difference might be due to the sampling of observations of fleeing.
Most of our observations of fleeing were made when deer
escaped human harassment, while Moen’s observations were of
enclosed, undisturbed deer. Therefore, Moen’s observations
likely were of submaximal levels of flight speed. We found the
highest amplitude of HR for standing behaviors, probably due,
in part, to extra energetic expenses of carrying fetuses. Our
results, which show no seasonal variation for highly locomotive
activities, seem reasonable because HR for roe deer and for
white-tailed deer reaches a physiological maximum when
fleeing in any season. Whether this is true also for trotting is
not known because few observations were recorded.
HR varies seasonally with day length in other species
(sheep—Baldock et al. 1988; elk—Lieb 1981). Our study did
not examine specifically the possible association between day
length and HR, but day length and temperature affect
reproduction and metabolic rate, both of which affect HR.
Even though these parameters are also affected by food
availability, studies on moose (Regelin et al. 1985) and caribou
(Fancy 1986) show that seasonal variation in HR persists even
under conditions of ad libitum food supplementation. Thus,
annual HR variation apparently is not contingent on food
availability or is adapted to the natural abundance of food.
Effects of age and wind speed.—HR decreases with
increasing body mass in a number of species of mammals
(Johnson and Gessaman 1973), including red deer (Espmark
and Langvatn 1985) and sheep (Geist et al. 1985). Our study
showed that HR in adult female roe deer followed a similar
pattern, although we attribute the individual effects to age
rather than directly to body mass.
April 2004
THEIL ET AL.—VARIATION IN HEART RATE OF ROE DEER
Estimates for wind speed were taken from a meteorological
station outside the forest and do not necessarily reflect the wind
speed experienced by the deer. It was impossible to measure
the wind speed actually experienced in the microhabitats
chosen by the deer; thus, true effects of wind on HR were hard
to assess. Nevertheless, we attempted to evaluate wind effects
on HR during several behaviors. The effect of wind on HR
during fleeing and trotting was regarded as 0 because HR of roe
deer engaged in fleeing and trotting did not vary through the
year (Table 2). The elevated HR of other behaviors during high
winds might be explained by an emotionally conditioned
increment of HR due to nervousness caused by decreased
olfactory and auditory sensing. This is supported by the
observed elevations of short-term, long-term, and overall HR
variability on windy days. Jeppesen (1987) found that roe deer
behaved more nervously during high winds, and it is known
that nervousness in humans can lead to an increased HR
(Blix et al. 1974). The HR difference between standing and
standingþalert may be an example of an emotionally
conditioned HR elevation because deer are standing upright
in both behaviors. Another explanation for the elevated HR
during high winds could be loss of heat by convection. Most of
the observations were made after leaves were on trees, with
mean air temperatures well above 08C. Lack of observations
due to loss of power in the radio transmitters made it impossible to evaluate whether the wind effect increased during
winter.
Behavioral effects.—The ranking order of behaviors and
their associated HR corresponded well to the order found for
white-tailed deer (Jacobsen 1979; Moen 1978), elk (Lieb
1981), bighorn sheep (MacArthur et al. 1979; Weisenberger
et al. 1996), mule deer (Freddy 1984; Weisenberger et al. 1996),
and red deer (Price et al. 1993). The 3 locomotive behaviors
(fleeing, trotting, walking) had the highest HR, and HR
increased with increased level of mobility (Table 2). Nilssen et
al. (1984a) also reported that metabolic rate increased linearly
with increasing speed of running. Herbold et al. (1991)
reported similar HR for the behaviors standingþalert (92
beats/min), foraging (83 beats/min), and bedded (78 beats/
min) for a single enclosed adult female roe deer. Ruminating
behavior raised HR by 0.6 beats/min when roe deer were
bedded, almost the same amount as reported for red deer by Price
et al. (1993). We found that locomotive categories of behavior
had higher levels of short-term, long-term, and overall
variability than nonlocomotive behaviors (Table 3). Increasing
variability in HR with increasing activity has also been found in
bighorn sheep (MacArthur et al. 1979; Weisenberger et al.
1996), white-tailed deer (Mautz and Fair 1980), roe deer
(Herbold et al. 1991), and mule deer (Freddy 1984; Weisenberger et al. 1996). Foraging had one of the lowest levels of
short-term variability, possibly indicating that roe deer forage
only when undisturbed. The short-term variability was lower for
foraging than for standing, which is important when discriminating between behaviors on the basis of HR data. Overall
variability was highest when deer were moving or eliciting alarm
reactions (beddedþalert, standingþalert) and lowest when deer
were foraging or bedded with closed eyes. These findings
251
suggest that the physiological interpretation of overall variability
may be that it is an indicator of an emotional conditioned
response (i.e., nervousness).
By means of discriminant function analysis, our HR model
can be used to predict female roe deer behavior, allowing
precise determination of time and energy budgets. For that
purpose, the short-term variability can be used to detect
changes of behavior in a given 10-s interval. Another
application of the HR model is quantification of the response
a deer elicits to, for example, human recreational activities. We
(Olesen et al. 1998) found that roe deer responded 2.5 times as
much when deer were chased by an unleashed dog (German
hunting terrier, 8 kg) compared to when a person rides a bike
on a nearby road in the forest. The response deer elicited to
a person walking off the roads in the forest was intermediate.
The results are useful when evaluating the energetic consequences of human-induced harassments throughout the year
and may be used as a tool for management.
ACKNOWLEDGMENTS
Aage W. Jensen’s foundation, the Danish Forest and Nature
Agency, and the Danish Research Council provided financial support.
The Danish Experimental Inspectorate, Copenhagen, Denmark,
approved the protocol for the care and use of animals, and the
protocol followed the guidelines for the capture, handling, and care of
mammals approved by the American Society of Mammalogists. We
are grateful to H. Mølgaard and H. R. Andersen, Aarhus University
Hospital, Skejby, for surgical assistance and to J. L. Jeppesen and M.
C. Forchhammer for critically reviewing an earlier version of the
manuscript. We thank P. Grønvald and U. Marnæs for software
development; P. Hartmann, DMU Kalø, for technical assistance with
the fieldwork; and last but not least A. H. Andersen, Aarhus
University, Institute of Mathematics, for untiring statistical advice.
LITERATURE CITED
ARMITAGE, P., AND T. COLTON. 1998. Encyclopedia of biostatistics.
John Wiley & Sons, Inc., New York.
BALDOCK, N. M., R. M. SIBLY, AND P. D. PENNING. 1988. Behavior and
seasonal variation in heart rate in domestic sheep, Ovis aries.
Animal Behaviour 36:35–43.
BLAXTER, K. L., AND A. M. BOYNE. 1982. Fasting and maintenance
metabolism of sheep. Journal of Agricultural Science 99:611–620.
BLIX, A. S., S. B. STROMME, AND H. URSIN. 1974. Additional heart
rate—an indicator of psychological activation. Aerospace Medicine
45:1219–1222.
BOX, G. E. P., AND G. M. JENKINS. 1976. Time series analysis:
forecasting and control. 2nd ed. HoldenDay, San Francisco,
California.
CEDERLUND, G., AND O. LIBERG. 1995. Rådjuret—viltet, ekologien och
jakten. Svenska Jägareförbundet, Uppsala, Sweden.
EDWARDS, C. H., AND D. E. PENNEY. 1990. Calculus and analytic
geometry. 3rd ed. Prentice Hall, Upper Saddle River, New Jersey.
ELLENBERG, H. 1978. Zur Populationsökologie des Rehes (Capreolus
capreolus L., Cervidae) in Mitteleuropa. Spixiana. Zeitschrift für
Zoologie (Suppl) 2:1–211.
ESPMARK, Y., AND R. LANGVATN. 1985. Development and habituation
of cardiac and behavioral responses in young red deer calves
(Cervus elaphus) exposed to alarm stimuli. Journal of Mammalogy
66:702–711.
252
JOURNAL OF MAMMALOGY
FANCY, S. G. 1986. Daily energy budgets of caribou: a simulation
approach. Ph.D. dissertation. University of Alaska, Fairbanks.
FREDDY, D. J. 1984. Heart rates for activities of mule deer at pasture.
Journal of Wildlife Management 48:962–969.
GEIST, V., R. E. STEMP, AND R. H. JOHNSTON. 1985. Heart rate
telemetry of bighorn sheep as a means to investigate disturbances.
Pp. 92–99 in The ecological impact of outdoor recreation on
mountain areas in Europe and North America, Recreational Ecology
Research Group Report No. 9 (N. G. Bayfield and G. C. Barrow,
eds.). Wye College, Kent, United Kingdom.
HERBOLD, H., F. SUCHENTRUK, S. WAGNER, R. WILLING, AND R. MAD.
1991. Behavioral and heart rate reactions in roe deer during
disturbance experiments—preliminary results. Pp. 566–571 in
Transactions 20th Congress of the International Union of Game
Biologists (S. Csányi and J. Ernhaft, eds.), Gödöll~o, Hungary, 21–
26 August 1991, University of Agricultural Sciences, Gödöll~
o,
Hungary.
HILL, R. W., AND G. A. WYSE. 1989. Animal physiology. Harper &
Row Publishers Inc., New York.
HOLTER, J. B., W. E. URBAN., H. H. HAYES, AND H. SILVER. 1976.
Predicting metabolic rate from telemetered heart rate in white-tailed
deer. Journal of Wildlife Management 40:626–629.
JACOBSEN, N. K. 1979. Changes in heart rate with growth and activity
of white-tailed deer fawns (Odocoileus virginianus). Comparative
Biochemistry and Physiology A 62:885–888.
JEPPESEN, J. L. 1987. The disturbing effects of orienteering and hunting
on roe deer (Capreolus capreolus). Danish Review of Game
Biology 13(3):1–24.
JEPPESEN, J. L. 1989. Activity patterns of free-ranging roe deer
(Capreolus capreolus) at Kalø. Danish Review of Game Biology
13(8):1–32.
JOHNSON, S. F., AND J. A. GESSAMAN. 1973. An evaluation of heart rate
as an indirect monitor of free-living energy metabolism. Pp. 44–54
in Ecological energetics of homeotherms. Volume 20. (J. A.
Gessaman, ed.). Utah State University Press, Logan.
KAUTZ, M. A., W. W. MAUTZ, AND L. H. CARPENTER. 1981. Heart rate
as a predictor of energy expenditure of mule deer. Journal of
Wildlife Management 45:715–720.
LANGVATN, R., AND R. ANDERSEN. 1991. Støy og forstyrrelser, -metodikk til registrering av hjortedyrs reaksjon på militær aktivitet.
NINA Oppdragsmelding 98:1–48.
LIEB, J. W. 1981. Activity, heart rate, and associated energy
expenditure of elk in western Montana. Ph.D. dissertation,
University of Montana, Missoula.
LOUDON, A. S. I., A. S. MCNEILLY, AND J. A. MILNE. 1983.
Nutrition and lactational control of fertility in red deer. Nature
302:145–147.
MACARTHUR, R. A., R. H. JOHNSTON, AND V. GEIST. 1979. Factors
influencing heart rate in free-ranging bighorn sheep: a physiological
approach to the study of wildlife harassment. Canadian Journal of
Zoology 57:2010–2021.
MAUGET, C., R. MAUGET, AND A. SEMPÉRÉ. 1997. Metabolic rate in
female European roe deer (Capreolus capreolus): incidence of
reproduction. Canadian Journal of Zoology 75:731–739.
MAUTZ, W. W., AND J. FAIR. 1980. Energy expenditure and heart rate
for activities of white-tailed deer. Journal of Wildlife Management
44:333–342.
MOEN, A. N. 1978. Seasonal changes in heart rates, activity,
metabolism, and forage intake of white-tailed deer. Journal of
Wildlife Management 42:715–738.
MOEN, A. N., AND R. A. MOEN. 1998. Metabolic ratios for estimating
energy metabolism in moose. Alces 34:181–187.
Vol. 85, No. 2
M OEN, A. N., AND G. S. BOOMER. 2000. A tandem algorithm for
modeling rhythmic changes. Ecological Modeling 134:275–
282.
MORHARDT, J. E., AND S. S. MORHARDT. 1971. Correlations between
heart rate and oxygen consumption in rodents. American Journal of
Physiology 221:1580–1586.
MØLGAARD, H. 1995. 24-h heart rate variability, methodology and
clinical aspects. Forlaget Møl, Aarhus, Denmark.
NILSSEN, K. J., H. K. JOHNSEN, A. ROGNMO, AND A. S. BLIX. 1984a.
Heart rate and energy expenditure in resting and running Svalbard
and Norwegian reindeer. American Journal of Physiology 246:
R963–R967.
NILSSEN, K. J., J. A. SUNDSFJORD, AND A. S. BLIX. 1984b. Regulation of
metabolic rate in Svalbard and Norwegian reindeer. American
Journal of Physiology 247:R837–R841.
OLESEN, C. R., P. K. THEIL, AND A. E. COUTANT. 1998. Råvildt og
forstyrrelser. Danmarks Miljøundersøgelser. DMU-rapport nr. 238
[Report No. 237 from the National Environmental Research
Institute, Denmark] (in Danish with English summary).
OWEN, R. B., JR. 1969. Heart rate, a measure of metabolism in
blue-winged teal. Comparative Biochemistry and Physiology
31:431–436.
PORGES, S. W. 1985. Spontaneous oscillations in heart rate: potential
index of stress. Pp. 97–112 in Animal stress (G. P. Moberg, ed.).
American Physiological Society, Bethesda, Maryland.
PRICE, S., R. M. SIBLY, AND M. H. DAVIES. 1993. Effects of behavior
and handling on heart rate in farmed red deer. Applied Animal
Behavior Science 37:111–123.
REGELIN, W. L, C. C. SCHWARTZ, AND A. W. FRANZMANN. 1985.
Seasonal energy metabolism of adult moose. Journal of Wildlife
Management 49:388–393.
RENECKER, L. A., AND R. J. HUDSON. 1985. Telemetered heart rate as
an index of energy expenditure in moose (Alces alces). Comparative
Biochemistry and Physiology 82A:161–165.
RENECKER, L. A., AND R. J. HUDSON. 1986. Seasonal energy
expenditures and thermoregulatory responses of moose. Canadian
Journal of Zoology 64:322–327.
ROBBINS, C. T., C. T. COHEN, AND B. B. DAVITT. 1979. Energy
expenditure by elk calves. Journal of Wildlife Management 43:
445–453.
ROBBINS, C. T., R. S. PODBIELANCIK-NORMAN, D. L. WILSON, AND E. D.
MOULD. 1981. Growth and nutrient consumption of elk calves
compared to other ungulate species. Journal of Wildlife Management 45:172–186.
SADLEIR, R. M. F. S. 1982. Energy consumption and subsequent
partitioning in lactating black-tailed deer. Canadian Journal of
Zoology 60:382–386.
SAS Institute Inc. 1997. SAS/STAT user’s guide. Version 6.11, 4th
ed., vol. 1–2. SAS Institute Inc., Cary, North Carolina.
SCHWARTZ, C. C., M. E. HUBBBERT, AND A. W. FRANZMANN. 1991.
Energy expenditure in moose calves. Journal of Wildlife Management 55:391–393.
SILVER, H., N. F. COLOVOS, J. B. HOLTER, AND H. H. HAYES. 1969.
Fasting metabolism of white-tailed deer. Journal of Wildlife
Management 33:490–498.
SOKAL, R. R., AND F. J. ROHLF. 1995. Biometry: the principles and
practice of statistics in biological research, W. H. Freeman and Co.,
New York.
S TRANDGAARD , H. 1972. An investigation of corpora lutea,
embryonic development, and time of birth of roe deer (Capreolus
capreolus) in Denmark. Danish Review of Game Biology 6(7):
1–22.
April 2004
THEIL ET AL.—VARIATION IN HEART RATE OF ROE DEER
WEINER, J. 1977. Energy metabolism of the roe deer. Acta
Theriologica 22:3–24.
WEISBJERG, M. R., AND T. HVELPLUND. 1993. Bestemmelse af
nettoenergiindhold (FEk) i foder til kvæg [Estimation of net energy
content (FU) in feeds for cattle]. Report No. 3/1993 from the
National Institute of Animal Science, Denmark: 1–39 (in Danish
with English summary).
WEISENBERGER, M. E., P. R. KRAUSMANN, M. C. WALLACE, D. W. DE
YOUNG, AND O. E. MAUGHAN. 1996. Effects of simulated jet aircraft
253
noise on heart rate and behavior of desert ungulates. Journal of
Wildlife Management 60:52–61.
WORDEN, K. A., AND P. J. PEKINS. 1995. Seasonal change in feed
intake, body composition, and metabolic rate of white-tailed deer.
Canadian Journal of Zoology 73:452–457.
Submitted 17 September 2002. Accepted 27 January 2003.
Associate Editor was Ronald D. Gettinger.