Campbell et al.
1 S9: Effects of rainfall on the behavior of the Eurasian beaver
2
3 Introduction
4
5 The negative effect of inter-annual variation in rainfall on the body-weight and reproductive
6 success of the Eurasian beavers in our Telemark study site may arise from a direct effect of
7 rainfall on foraging activity, rather than from any effect on the forage quality. For example,
8 beavers may not detect predators as easily in rain (Lima and Dill 1990; Hilton et al.
1999) and
9 are therefore less inclined to risk exposure to predation by foraging on land. Beavers in Poland
10 have been found to adjust their foraging patterns in response to the placement of predator odors
11 (Rosell and Czech 2000) showing that beavers will adapt their behavior in response to
12 perceived predation risk. The fur of terrestrial mammals can lose half its insulating properties if
13 wetted (Webb and King 1984). While beaver fur is adapted for a semi-aquatic life-style, which
14 will likely have improved its insulating properties when wet over that of terrestrial mammal
15 fur, the thermodynamics of evaporative cooling will still effect the outer, exposed fur layer
16 (Kimmel et al.
1991), leading to chilling if exposure persists. Also, beaver fur insulates when
17 diving by trapping air (which warms up), whereas air cannot be ‘trapped’ in this way in the
18 terrestrial environment. An alternative hypothesis, therefore, is that wet terrestrial foraging
19 conditions (rain) may increase the thermoregulatory burden on beavers.
20 Here we test these two hypotheses by examining the relationship between the behaviour
21 of beavers in the study area and rainfall. We predict that:
22
23
1.
If predation risk increases in rain, beavers will spend a smaller proportion of their time away from the safety of water with increasing rainfall.
24
25
26
2.
If the thermoregulatory burden increases in wet weather, beavers will spend a greater proportion of their time in the shelter of a lodge or bank-den with increasing rainfall.
1

Campbell et al.
1 Materials and methods
2
3 The impacts of rain on beaver behavior were analyzed from behavioral observation of 17
4 animals. Twelve were observed in 2000 (eight adult males and four adult females from nine
5 territories in Saua and Gvarv rivers) and five (three adult males and two two-year old males,
6 from four territories in Saua and Sauar rivers) in 2007. In 2000, beavers were implanted with
7 an Alterra TX30.3A1 intraperitoneal 30 MHz-radio transmitter (63g) equipped with a
8 temperature sensor and movement sensor (Alterra-DLO) (Ranheim et al.
2004; Campbell et al.
9 2005). In 2007, beavers were fitted with either tail-mounted VHF radio-tags (142 MHz,
10 Advanced Telemetry Systems) or tail-mounted proximity loggers with VHF transmitters (142
11 MHz, SirTrack). All procedures were licenced and complied with the ethical laws in Norway.
12 All beavers were tracked for a minimum of two complete Principal Activity Periods
13 (PAP). A PAP started when the focal animal left the lodge in the evening and ended when it
14 returned to the lodge the following morning and remained there for at least 20 minutes.
15 Following Martin and Bateson (1999), the sampling rule followed was focal sampling and the
16 recording rule followed was continuous recording , to a resolution of one minute (Martin and
17 Bateson 1999, p84-88). When >1 behaviors occurred within the same minute, the time
18 allocated to each was proportional to the number of behaviors observed within that minute.
19 Radio-transmitters were used to maintain surveillance of the subject. All observations were
20 conducted from a boat, binoculars were used throughout and a spotlight was used during dark
21 periods, with no detectable influence on the subject (Herr and Rosell 2004). For this study, we
22 were interested in the effect that rain might have on an animal’s willingness to be exposed to
23 the risk of predation while out of the water and also whether animals prefer to shelter out of the
24 rain. During observations, in addition to the behavior observed, focal animals were recorded as
25 being on land or in water. Time spent in a lodge or bank-den was also recorded. During night-
26 time observation sessions on the river, temperature and rainfall was recorded at regular
2

Campbell et al.
1 intervals. Rainfall was recorded as a binary variable and the variable ‘Rain’ expressed as the
2 percentage of observations with rain each night.
3
4 Data analysis
5
6 To examine the effects of rainfall on the proportion of their PAP that animals spent on land and
7 in a lodge, linear mixed-effect models were created using the nlme package in R 2.14.0 (R
8 Development Core Team 2011). A random intercept model was used, following Schielzeth and
9 Forstmeier (2009). Predictors were Year, Month and Rain without interaction terms and Beaver
10 was set as the random effect. For each behavior, the full model was compared, with and
11 without a first-order autoregressive (AR1) correlation structure to account for repeated
12 measures for individuals, and with and without individual-based weightings on variance
13 structure. The full model with the lowest Akaike’s information criterion (AICc, Burnham and
14 Anderson 2002) value was selected for further analysis. Following this, all combinations of the
15 three predictors were run, yielding eight (2
3
) models for each behaviour. Candidate models
16 were compared and the parameter estimates for each predictor averaged over all eight models
17 based on Akaike weights, using the R package MuMIn (v1.6.4, Bartoń 2011). Models were run
18 using the ‘maximum likelihood’ method to allow this AICc based multi-model inference.
19
20 Results
21
22 Proportion of time spent on land
23 The full model describing proportion of time animals spent on land most comprehensively
24 contained an AR1 correlation structure but no individual variance weightings. The top two
25 models, based on Akaike weights, both contained a positive effect of Rain and accounted for
26 64% of the model weights (Table 1). Overall, models containing Rain accounted for 66% of
3

Campbell et al.
5
6
7
1 model weights, compared with 55% for Year and 3% for Month. A weighted average across all
2 models found that for every 1% increase in Rain, beavers increased their time on land by
3 0.11% but this relationship was not significant (95% CI = −0.05 to +0.22%, Fig. 1).
4
Table S9.1: Parameter estimates and Akaike weights for models describing proportion of time spent by beavers on land or in a lodge or bank-den. Models are listed in order of greatest weight for each predictor.
Model
Time on land
Year, Rain
Rain
Year
Null
Month, Rain
Month
Year, Month, Rain
Year, Month
Intercept Rain df AICc ∆ AICc weight
0.4399 0.0011
0.4173 0.0011
0.4638
0.4416
0.5121 0.0011
0.5171
0.5145 0.0011
0.5192
10
9
11
10
6
5
5
4
-37.0
-36.7
-35.8
-35.3
-30.3
-29.2
-28.8
-27.7
0.352
0.39 0.289
1.28 0.185
1.75 0.146
6.70 0.012
7.82 0.007
8.21 0.006
9.34 0.003
Time in lodge
Null
Rain
Year
Year, Rain
Month
Month, Rain
Year, Month
Year, Month, Rain
8
9 Proportion of time spent in lodge
0.2149
0.1986 0.0008
0.2238
0.2075 0.0008
0.2781
0.2828 0.0009
0.2786
0.2834 0.0009
9
10
10
11
4
5
5
6
-13.8
-12.9
-11.7
-10.8
-7.0
-6.6
-4.4
-3.8
0.432
0.85 0.282
2.04 0.155
2.98 0.097
6.72 0.015
7.18 0.012
9.37 0.004
9.92 0.003
10 The full model describing proportion of activity spent in a lodge or bank-den most effectively
11 contained an AR1 correlation structure but no individual variance weightings. The top model,
12 based on Akaike weights was the null model, containing none of the predictors, which
13 accounted for 43% of the model weights (Table 1). The second most supported model showed
14 a positive effect of Rain and accounted for 28% of model weights. Overall, models containing
15 Rain accounted for 39% of model weights, compared with 26% for Year and 3% for Month. A
4

Campbell et al.
1 weighted average across all models provided evidence that for every 1% increase in Rain,
2 beavers increased their time spent in a lodge or bank-den by 0.08%, but this relationship was
3 not significant (95% CI = −0.05 to +0.20%, Fig. 2).
4
5 Discussion
6
7 One potential mechanism behind the negative effects of rain we observed, would be that
8 beavers may not detect predators as easily during rain (Lima and Dill 1990; Hilton et al.
1999)
9 and are therefore less inclined to risk exposure to predation by foraging on land. We found no
10 negative effect of rainfall on the proportion of time beavers spent on land however. Instead, we
11 found weak evidence for a positive effect of rainfall on time spent on land. The size of the
12 effect and the fact that the 95% confidence intervals include zero both suggest that this effect is
13 unlikely to be biologically significant. This result indicates that predator avoidance is not the
14 mechanism underlying the effect of rain on inter-annual variation in beaver body-weight and
15
16 reproductive success. Predation risk is low in the study population since the wolves ( Canis lupus
), the main predator of beavers where the species’ ranges overlap (Rosell and Czech
17 2000; Sidorovich et al.
2003; Andersone and Ozoliņš 2004), have not been recorded in the
18 study area for approximately 100 years. While a reduction, or almost complete extinction, of
19 fear of non- human predators has been recorded in populations where such predators are absent
20 (Berger 1998; Ward et al. 1996), experiments on the Telemark beaver population have shown a
21 significant reduction in time spent scent-marking in presence of both lynx and wolf scent
22 (Rosell and Sanda 2006). While perceived predation risk could therefore still influence the
23 foraging decisions of beavers, we found no evidence that rainfall influences their perception of
24 predation risk.
25 Alternatively, beavers may suffer a thermodynamic cost to having wet fur. The fur of
26 the semi-aquatic mink Mustela vison does not appear to suffer reduced thermal efficiency when
5

Campbell et al.
1 wetted in still air (Korhonen and Niemelä 2002). Beavers are also highly adapted to the aquatic
2 environment with a similarly dense under-fur (12,000 cm 2 dorsal density and 23,000 cm 2
3 ventral density cf.
16,600 cm
2
mean density over the whole winter pelt in mink, Korhonen
4 1988; Rosell 2002), a lower surface to volume ration than mink, and an extensive counter-
5 current blood vessel arrangement (Cutright and McKeen 1979). Therefore this hypothesis
6 seems improbable. While we found that the parameter estimate for the effect of rainfall on time
7 spent in shelter was positive, the size of the effect and the crossing of the 95% confidence
8 intervals over zero both suggest that this effect is unlikely to be biologically significant. Thus,
9 if there is a thermodynamic cost to foraging in rain, it does not prove substantial enough to
10 encourage beavers to take food under shelter to be consumed.
11
12
13
14
15
16
17
18
19
Figure S9.1: Plot of the percentage of the observation period with rain against the proportion of the observation period each animal spent a) on land and b) in a lodge or bank den. Each colored line represents an individual. Where the same percentage rain value reoccurs within an individual, the mean of time on land or in a lodge is presented.
6

Campbell et al.
1 References
2
3
Andersone Ž, Ozoliņš J (2004) Food habits of wolves
Canis lupus in Latvia. Acta
4
5
Theriologica , 49 , 357-367.
Bartoń K (2011) Package ‘MuMIn’: Multi-model inference. R package, version 1.6.4.
6 Berger J (1998) Future prey: Some consiquences of the loss and restoration of large carnivores.
7 In: Behavioral Ecology and Conservation Biology (ed Caro T) pp. 80-100. Oxford
8 University Press, USA.
9 Burnham KP, Anderson D (2002) Model Selection and Multi-Model Inference. Springer, 488
10
12 pp.
11 Campbell RD, Rosell F, Nolet B et al.
(2005) Territory and group sizes in Eurasian beavers
( Castor fiber ): echoes of settlement and reproduction? Behavioral Ecology and
13 Sociobiology 58 , 597-607.
14 Cutright WJ, McKean T (1979) Countercurrent blood vessel arrangement in beaver ( Castor
15
17 canadensis ). Journal of Morphology , 161 , 169-175.
16 Herr J, Rosell F, (2004). Use of space and movement patterns in monogamous adult Eurasian beavers ( Castor fiber ). Journal of Zoology , 262 , 257-264.
18 Hilton GM, Ruxton GD, Cresswell W, (1999) Choice of foraging area with respect to predation
19 risk in redshanks: The effects of weather and predator activity. Oikos , 87 , 295-302.
20 Kimmel E, Arkin H, Broday D, Berman A (1991) A model of evaporative cooling in wetted
21 hide. Journal of Agricultural & Engineering Research , 49 , 227-241.
22 Korhonen HT (1988) Seasonal comparison of body composition and hair coat structure
23 between mink and polecat. Comparative biochemistry and physiology. A. Comparative
24
25 physiology , 91 , 469-473.
Korhonen HT, Niemelä P (2002) Water absorption and the drying and cooling rates in mink
26 ( Mustela vison ) following simulated diving. Animal Science , 74 , 277-283.
7

Campbell et al.
1 Lima SL, Dill LM (1990) Behavioral decisions made under the risk of predation: A review and
2 prospectus . Canadian Journal of Zoology, 68 , 619-640.
3 Martin P, Bateson, P. 1999. Measuring behaviour: An introductory guide . Cambridge
4 University Press, 187 pp.
8
9
5 R Development Core Team. 2011. R: A language and environment for statistical computing. R
6 Foundation for Statistical Computing , Vienna, Austria.
7 Ranheim B, Rosell F, Haga HA et al.
(2004) Field anaesthetic and surgical techniques for implantation of intraperitoneal radio transmitters in Eurasian beavers Castor fiber .
Wildlife Biology , 10 , 11-15.
10 Rosell F (2002) Do eurasian beavers smear their pelage with castoreum and anal gland
11 secretion? Journal of Chemical Ecology , 28 , 1697-1701.
12 Rosell F, Czech A (2000) Responses of foraging Eurasian beavers Castor fiber to predator
13 odours. Wildlife Biology , 6 , 13-21.
14 Rosell F, Sanda J (2006) Potential risks of olfactory signaling: the effect of predators on scent
15
17 marking by beavers. Behavioral Ecology , 17 , 897-904.
16 Schielzeth H, Forstmeier W (2009) Conclusions beyond support: overconfident estimates in mixed models. Behavioral Ecology , 20 , 416-420.
18 Sidorovich VE, Tikhomirova LL, Jedrzejewska B (2003) Wolf Canis lupus numbers, diet and
19 damage to livestock in relation to hunting and ungulate abundance in northeastern
20 Belarus during 1990-2000. Wildlife Biology , 9 , 103-111.
21 Ward JF, Macdonald DW, Doncaster CP et al.
(1996) Physiological response of the European
22
24 hedgehog to predator and nonpredator odour. Physiology & Behavior , 60 , 1469-1472.
23 Webb DR, King JR (1984) Effects of wetting on insulation of bird and mammal coats. Journal of Thermal Biology , 9 , 189-191.
25
8