Snow Cover Constrains The Behavioural Flexibility of A Winter Adapted Bird - Original Article (Habitat Use)

Ibis (2023), 165, 1186–1200 doi: 10.1111/ibi.
13211
Snow cover constrains the behavioural flexibility of a

winter-adapted bird
AMY A. SHIPLEY*1,2 & BENJAMIN ZUCKERBERG1
1
Department of Forest and Wildlife Ecology, University of Wisconsin-Madison, 1630 Linden Dr., Madison, WI, 53706,
USA
2
Office of Applied Science, Wisconsin Department of Natural Resources, 2801 Progress Rd, Madison, WI, 53716,
USA
Behavioural flexibility is an important way in which animals respond to changing envi-

ronmental conditions. During winter, snow cover is an important seasonal refuge that
provides thermal insulation and protects overwintering species from predators. However,
snow depth and quality can be highly variable throughout winter and it is unclear how
species that use snow cover adjust their behaviour with changing winter conditions and
in complex landscapes. During winter months, Ruffed Grouse Bonasa umbellus spend a
large portion of time roosting in trees, understorey and, in some cases, subnivean envi-
ronments. Importantly, the ability to roost under snow has been associated with reduced
stress levels and increased overwinter survival. Across three winters, we studied plasticity
of roosting behaviour across a diversity of winter conditions and land-cover types. In line
with predictions, Grouse were more likely to use snow burrows when snow was deep
and powdery, and experienced higher temperatures in snow roosts compared to other
roost types. However, snow roosting behaviour did not vary strongly across land-cover
types, and Grouse were not more likely to use snow burrows at lower temperatures,
potentially because snow roosting may serve to protect them from predators regardless
of winter weather conditions. Both the snow conditions necessary for snow roosting and
the occurrence of snow roosting behaviour were relatively rare during our 3-year study
(only 3% of all roosts were snow burrows). Loss of winter refugia due to warming win-
ters and declining snow cover may limit the use of behavioural flexibility for winter-
adapted species to cope with environmental change.
Keywords: behavioural adaptation, land cover, microrefugia, Ruffed Grouse, weather, winter
climate.
Behavioural plasticity is a crucial pathway by which uncertain, as behaviour is reversible over short time
animals respond to changing environmental condi- periods (Chevin et al. 2010). For example, Desert
tions (Piersma & van Gils 2011) and can include Wood Rats Neotoma lepida alter when they leave
changes in phenology (e.g. altering the timing of their dens so that they are not exposed to ambient
migration, Kolecek et al. 2020), physiology (e.g. night temperatures above their lethal physiological
lowering energy expenditure, Humphries threshold (Murray & Smith 2012). Similarly,
et al. 2005) and morphology (e.g. increasing fat American Pikas Ochotona princeps, which typically
stores, Pokallus & Pauli 2016). Responding rely on cooler talus areas, are increasingly shifting
behaviourally to changing environmental conditions to non-talus habitat in forests, such as downed logs,
is particularly advantageous when the future is snags and slash piles, to avoid extreme daytime
temperatures (Beever et al. 2017).
*Corresponding author.
Seeking out microrefugia, or localized patches
Email: shipley2@wisc.edu of milder or more stable conditions that can buffer
Twitter: @amy_a_shipley animals from unsuitable environmental conditions,
© 2023 The Authors. Ibis published by John Wiley & Sons Ltd on behalf of British Ornithologists' Union.
This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and
distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.
1474919x, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ibi.13211 by Welsh Assembly Government, Wiley Online Library on [11/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Behavioural flexibility of wintering grouse 1187
is an important behavioural adaptation in a chang- and increase overwinter survival (Shipley

ing climate. The availability of microrefugia and et al. 2020). However, unlike land cover, snow is
the subsequent behavioural response that pro- ephemeral (Thompson et al. 2018), and the flexi-
motes the use of those refugia during extreme bility to access ephemeral microrefugia may be
events may be an important aspect of adaptive important in environments where more predict-
capacity (Beever et al. 2017), especially if trade- able and persistent microrefugia are uncommon or
offs exist with other competing behaviours, such do not exist. Temperature, snow depth and snow
as foraging (Murray & Smith 2012). Microrefugia density all change through the winter but it is
can buffer organisms from extreme conditions, and unclear how species that use snow cover as a
both terrestrial and aquatic animals actively seek microrefugium adjust their behaviour with chang-
out microhabitats such as rock interstices (Hall ing conditions. Importantly, the conditions that
et al. 2016) or tree cavities (Coombs et al. 2010). promote the stability of the subnivium are deterio-
For example, the prevalence of microrefugia dur- rating with modern climate change. Winter air
ing heat waves is one of the most important pre- temperatures in the Great Lakes Region have
dictors of the occurrence of American Pika (Hall increased, but winter soil temperatures have
et al. 2016). The use of microrefugia is similarly decreased (Isard et al. 2007), probably due to
important for reptiles in responding to increasing decreased snowpack and duration of snow cover
temperatures and maintaining body temperatures (Brown & DeGaetano 2011). Under a +5 °C
(Carter et al. 2012, Valdecantos et al. 2013). How- warming scenario, the subnivium in the Great
ever, increased use of microrefugia to escape Lakes Region is predicted to decline by 45% in
extreme conditions (e.g. heat, cold, increased pre- extent and by more than a month in duration
cipitation) has costs and may limit time spent on (Thompson et al. 2021). Despite increasing ambi-
other important behaviours, such as foraging ent temperatures, increased variability in winter
(Murray & Smith 2012). temperature and ‘rain on snow’ events that reduce
Microrefugia may be particularly important in snow depth and snow quality may lead to a func-
winter, which is a time of resource scarcity and tionally colder winter environment (Pauli
extreme environmental conditions. Reduced access et al. 2013). It is important to understand how the
to food and exposure to low temperatures results behaviour of winter-adapted species shifts with
in increased energy expenditure (Shipley changing winter conditions and how it might
et al. 2019) and exposure to predators as a result change in the future with warming winters and
of more time spent foraging (Lima 1986). In fewer days with suitable subnivean conditions.
northern forests, warmer winter microclimates are The Ruffed Grouse is a winter-adapted bird
often associated with higher tree basal area (i.e. species associated with North American temperate
cross-sectional area at 1.3 m above ground; Suggitt forests. In winter, grouse can spend up to 80% of
et al. 2011, Ashcroft & Gollan 2012, Ford their time roosting in trees, on the ground or on
et al. 2013), higher elevations due to cold air pool- top of the snow, or completely or partially below
ing in valleys (Sears et al. 2011) or more contigu- the snow in snow burrows or snow bowls (Bump
ous landscapes with less fragmentation (Latimer & et al. 1947). Snow burrows are thought to provide
Zuckerberg 2016). In addition to these topographi- superior insulation relative to other roost sites
cal and land cover features, however, snow cover (Thompson & Fritzell 1988b), allowing grouse to
provides important microrefugia that insulate and remain within their thermo-neutral temperature
protect many species from harsh winter weather range (Marjakangas et al. 1984), and are used by a
(Pauli et al. 2013). For several winter-adapted spe- number of galliform species, including Black
cies capable of accessing the environment beneath Grouse Lyrurus tetrix (Marjakangas et al. 1984),
snow cover (the subnivium), deep snow is impor- Western Capercaillie Tetrao urogallus (Marjakangas
tant for foraging (e.g. American Marten Martes et al. 1984), Grey Partridge Perdix perdix (Wester-
americana, Taylor & Buskirk 1994; Tundra Vole skov 1965), Willow Ptarmigan Lagopus lagopus
Microtos oeconomus, Korslund & Steen 2006) and (Irving 1960) and Sage Grouse Centrocercus uro-
reproduction (lemmings Lemmus spp., Duchesne phasianus (Back et al. 1987). Use of deep snow
et al. 2011). In Ruffed Grouse Bonasa umbellus, and roosting in snow burrows is an important
deep snow can reduce physiological stress associ- behavioural adaptation that reduces physiological
ated with low temperatures (Shipley et al. 2019) stress (Shipley et al. 2019) and influences habitat
1188 A. A. Shipley and B. Zuckerberg
selection (Shipley et al. 2020) of Ruffed Grouse. SWA is 9 10.0 °C (mean 1 sd;
Importantly, individual Ruffed Grouse that seek range 26.5 to 5.5 °C).
out the deepest snow and roost in snow burrows
increase their chances of surviving the winter, even
Grouse trapping, radiotelemetry and
after accounting for low temperatures that may be
roost site observations
associated with harsher winter weather conditions
(Shipley et al. 2020). Broad-scale regional shifts in From September to November (2015–2017), we
climate have been linked to the dampening of captured Ruffed Grouse in walk-in pens connected
Ruffed Grouse population cycles (Pomara & Zuck- to drift fences made from chicken wire (Gul-
erberg 2017), and warming winters, loss of snow lion 1965). Traps were checked twice each day
cover and limited access to microrefugia may be between 09:00 and 11:00 h and 17:00 and
playing a role. However, it is unclear how winter 19:00 h. Traps were closed during inclement
weather conditions and land cover interact to weather (e.g. heavy rain or snowfall). The mesh
influence the behavioural decisions of grouse to securing the top of each pen was covered with
use microrefugia at a given location and, in turn, leaves to provide a natural camouflage to prevent
how these behavioural decisions influence grouse predators from locating trapped grouse and to
exposure to low temperatures. Further, it is reduce grouse stress. We collected morphological
unknown how often the conditions that promote measurements (i.e. body mass and tarsus length),
snow burrowing occur currently, which is impor- identified the age and sex of each grouse (Hale
tant to understand in the context of future warm- et al. 1954), and tagged birds with 12-g necklace
ing and loss of snow cover. style radio-transmitters (Advanced Telemetry Sys-
We tested the hypotheses that winter tempera- tems, Inc.; expected battery life of ~ 1 year).
ture, snow depth, snow density and land cover influ- Transmitters did not exceed 5% of each individual
ence behavioural shifts towards the use of winter grouse’s bodyweight (Cochran 1980) and we did
microrefugia, and that behavioural flexibility would not observe evidence that transmitters impeded
mediate exposure to low winter temperatures. Spe- grouse locomotion (e.g. grouse appeared to walk
cifically, we predicted: (1) snow roosting behaviour and fly like untagged grouse). Research methods
would be more likely with increasing snow depth, for work with wild birds were approved by the
more powdery snow and low temperatures; (2) University of Wisconsin-Madison College of Agri-
snow roosting would be more likely in open habitat culture and Life Sciences Animal Care and Use
where there is less forest cover and fewer alternative Committee (protocol # A005198). Over three
microrefugia (such as tree roosts); and (3) grouse winters (1 December–31 March 2015–2018, here-
using snow roosts would experience reduced expo- after Years 1, 2 and 3), we found roost sites for
sure to low ambient temperatures. each tagged grouse once or twice each week,
between 09:00 and 17:00 h. To confirm roosting
behaviour visually, we used the transmitter signals
METHODS
to locate grouse, which usually involved flushing
birds from roosts. We categorized roosting behav-
Study site
iour as ‘snow burrow’ if the grouse was roosting
The Sandhill Wildlife Area (SWA; 44°180 N, under the snow with at least a partial roof of snow
90°070 W) is a 3700-ha state wildlife area in central (> 75% covered), ‘snow bowl’ if the roost was a
Wisconsin and is a study site for a number of long- bowl-shaped indentation in the snow without a
term monitoring programmes for winter-adapted roof, ‘tree roost’ if the grouse was found in a
species. Vegetation at SWA is a mosaic of forested deciduous or coniferous tree, and ‘surface roost’ if
Red and White Pine Pinus resinosa and Pinus stro- the roost was on the ground or on top of the snow
bus, Tag Alder Alnus serrulata, aspen Populus tre- without an indentation.
muloides and Populus gradidentata, and oak
Quercus spp., fragmented by open water and
Environmental predictors of roosting
marshes. There were on average 47 winter nights
behaviour
each year when snow depth exceeded 20 cm
(range 0–101 nights) between 2001 and 2015. We extracted land cover classes at each grouse
Mean winter (December–March) temperature at roost site using our land cover map. We began
with the WisFIRS (Wisconsin Forest Inventory and in Ruffed Grouse (Shipley et al. 2019) and thus is
Reporting System) land cover map produced by likely to influence roosting behaviour. Tempera-
the Wisconsin Department of Natural Resources, ture rasters (10-m resolution) were created using
which has 31 separate land cover classes and a 5- regression kriging (Hengl et al. 2007) of data col-
m resolution. These land cover classes were lected from temperature sensors (DS1922L iBut-
assigned according to dominant land cover. For tons; Maxim Integrated) spaced across SWA to
mature stands, the dominant land cover class capture differences in temperature according to
accounted for > 50% of basal area (i.e. cross- land cover types and distances to forest edge
sectional area at 1.3 m above ground), and for (Shipley et al. 2020). While wind could influence
young stands, for > 50% of total stems. We con- wind chill and behaviour of grouse, we were not
densed the 31 initial land cover classes into five able to instal wind gauges at iButton temperature
classes: Mature Forest, Aspen–Alder (hereafter, stations. Further, publicly available wind speed
‘Dense Cover’), Scrub Oak, Open (emergent vege- data from weather stations that are usually located
tation, lowland brush/willow and lowland grass) in open areas would probably not have been repre-
and Other (Wilson et al. 2018, Shipley sentative of wind speed in our Mature Forest and
et al. 2020). Across our study site, ‘Other’ con- Dense Cover land cover classes.
sisted primarily of marsh and open water, but also
included 16 additional less common land cover
Grouse temperature exposure
types, some of which grouse used occasionally. To
aid model convergence and restrict our analysis to To investigate how grouse roosting behaviour
only four land cover classes (Mature Forest, Dense impacts the temperatures to which grouse are
Cover, Scrub Oak and Open), if grouse roosted in exposed, we attached iButton temperature sensors
one of the land cover classes classified as ‘Other’, (DS1922L; Maxim Integrated; expected battery
we reclassified these less common land cover types life of at least 3 years) to the radio-transmitters of
as the most similar of the four land cover classes: a subset of our tagged grouse. We attached sensors
White Birch and Red Pine reclassified as Mature to the side of the transmitter facing away from the
Forest; low-growing shrubs, lowland brush, oak bird’s body using steel-reinforced epoxy (JB
aged less than 10 years, and oak aged 10–20 years Weld). Attaching iButtons to transmitters
reclassified as Dense Cover; and marsh, true increased the mass of the transmitter from 12 to
grasses and clearcut reclassified as Open. ~ 15.3 g, but the total weight of the transmitter
We measured snow depth at roost sites to the and iButton still did not exceed 5% of any bird’s
nearest half centimetre. We measured snow sink body mass. The iButtons recorded temperature
depth (an index of snow density) at roost sites by every half hour during winter, beginning on 1
dropping a 200-g cylinder (diameter: 8.5 cm) from December, with an accuracy of 0.5 °C and a res-
50 cm above the top of the snow and recording olution of 0.5 °C. The iButtons and temperature
the sinking depth to the nearest half centimetre. data were retrieved from grouse transmitters
Larger values indicate powdery, less dense snow. opportunistically when grouse either died or were
A value of zero indicates very hard snow. When randomly recaptured during trapping the following
snow was not present, snow density was recorded year.
as ‘N/A’; we did not include roost observations
when snow was absent in our data analyses. When
Statistical analyses
the cylinder was dropped and sank all the way to
the ground (i.e. snow depth was equal to snow We used a multinomial logistic model to model
sink depth), snow sink depth estimates were not the effects of land cover and weather variables on
reliable because the cylinder may have sunk fur- the probability of different Ruffed Grouse roosting
ther had the snow been deeper and hence we dis- behaviours (snow burrow, snow bowl, surface or
carded such observations from data analysis tree). We tested for the effect of minimum tem-
(Thompson et al. 2021). For each grouse roost perature, snow depth, snow sinking depth, land
location, we extracted the minimum temperature cover, date and time of day (minutes from mid-
from the previous day from daily temperature ras- night) on roosting behaviour. Predictor variables
ters. Minimum temperature from the previous day (other than land cover) were scaled using a z-
is an important predictor of corticosterone levels transformation to enable comparisons among
predictors (presented as b se). We developed a occurred during the 12 h prior to the roost obser-
list of a priori candidate models (Supporting Infor- vation. We used a generalized linear model with a
mation Table S1), which included combinations of Gaussian error distribution to model the effect of
between one and six of the predictors and the roosting behaviour on the difference in tempera-
two-way interactions. We ranked models based on ture between the median grouse iButton tempera-
Akaike’s Information Criterion corrected for small ture and the minimum ambient temperature.
sample sizes (Burnham & Anderson 2002), and Roosting behaviour was categorized as ‘snow roost’
models were considered competitive if they were (snow bowls and snow burrows were pooled due
within 2 ΔAICc of the best model. If more than to small sample sizes of observed snow burrows by
one model was competitive, we selected the model grouse with iButtons), ‘surface roost’ and ‘tree
with the greatest AICc weight (wi) for interpreta- roost’. Mean values are given sd. We visualized
tion. We performed a residual analysis to confirm the effects of the roosting behaviour on the tem-
that statistical assumptions were met. We imple- perature difference between the grouse iButtons
mented the multinomial logistic model using the and ambient temperature using ggplot2 (version
nnet package (Venables & Ripley 2002) version 3.1.1).
7.3.14 in the R statistical platform (R Core For each of the three analyses described above,
Team 2022, version 4.0.0 used for all analyses). we also ran each analysis as a mixed model,
To explore further the effects of land cover and including individual grouse ID as a random effect.
snow depth on the probability of using snow for Because analyses with and without a random effect
roosting (both snow burrows and snow bowls), we produced almost identical results (see below), we
used a generalized linear model with a logit link present the results of the analysis that resulted in a
function. We coded roosting behaviour as model with a lower AIC value to aid in
0 = ‘surface roosts’ and ‘tree roosts’ and 1 = ‘snow parsimony.
burrows’ and ‘snow bowls’ and included snow
depth and land cover as main effects, as well as an
RESULTS
interaction between snow depth and land cover.
For both the multinomial and binomial logistic From December to March 2015–2018, we
model, we visualized predictors with partial observed 55 grouse using 754 roosts (17 snow bur-
dependency plots using ggplot2 (version 3.1.1) to rows, 86 snow bowls, 565 surface roosts and 86
relate each predictor variable to the probability of tree roosts). We discarded 234 roost observations
roosting behaviours while keeping the other pre- when snow depth was equal to zero and 26 obser-
dictors conditional on their means. vations when snow depth was equal to snow sink
To quantify the effect of roosting behaviour on depth. Thirty-six roost observations were classified
temperatures experienced by grouse, we used tem- in habitat as ‘Other’. Of these, we reclassified six
perature data recorded by the iButtons attached to roosts in White Birch Betula papyrifera and one
grouse transmitters. For each roost observation of roost in Red Pine as Mature Forest, one roost in
a grouse with an iButton, we extracted the median low-growing shrubs, one roost in lowland brush,
temperature from the bird’s iButton during the five roosts in oak aged less than 10 years and two
previous 12 h, representing temperature experi- roosts in oak aged 10–20 years as Dense Cover;
enced by the roosting grouse. Galliforms usually and two roosts in marsh, one roost in true grasses
defecate between one and three solid faeces per and 14 roosts in clearcut as Open (see Methods).
hour (Baltic et al. 2005), and the number of faeces Our final dataset included 494 roost observa-
that we observed at Ruffed Grouse roosts ranged tions across the four land cover classes, represent-
between one and 22. Twelve hours is therefore ing 53 radiotagged grouse (Table 1, Fig. 1). The
likely to be representative of the time grouse mean number of roost observations per grouse was
spend at roosts. We then extracted the minimum 9.3 5.8 (mean 1 sd, range = 1–28). Mean
temperature from the previous day from our daily values ( sd) of snow depth, snow sink depth and
temperature rasters and calculated the difference minimum temperature observed at roost sites are
between the median grouse iButton temperature shown in Table 2.
and the minimum ambient temperature. As mini- We deployed iButtons on the transmitters of 25
mum temperature usually occurs in the early grouse and retrieved eight. The iButtons collected
morning hours, minimum temperature probably ambient temperature data for an average of
Table 1. Ruffed Grouse roosting behaviour by land cover temperature from the previous night from our
class. daily temperature rasters that were available
(models for some daily temperature rasters did not
Roosting Dense Scrub Mature
converge), our final dataset had 67 roost observa-
behaviour Open Cover Oak Forest
tions: 12 snow roosts (snow burrows and bowls
Snow burrow 2 5 3 7 pooled), 41 surface roosts and 14 tree roosts.
Snow bowl 10 32 11 33
Surface 36 105 41 132
Tree 2 7 18 50 Effects of land cover and winter
weather on roosting behaviour
Values indicate sample sizes of roost observations in each of
four land cover classes at Sandhill Wildlife Area in central Wis- We found strong support for the correlation of
consin, USA. snow depth, snow sink depth and land cover with
roosting behaviour in Ruffed Grouse. Model rank-
69 days (range 7–129 days). While the iButtons ings were nearly identical in the analysis with and
were collecting temperature data, we obtained 75 without inclusion of individual grouse ID as a ran-
roost observations. After extracting minimum dom effect (Supporting Information Tables S1 and
Figure 1. Ruffed Grouse roosting behaviour and land cover categories at Sandhill Wildlife Area in central Wisconsin, USA. Snow
burrows (a, purple), snow bowls (b, blue), surface roosts (c, black) and tree roosts (d, green). Surface roosts were most common, fol-
lowed by snow bowls, tree roosts and snow burrows.
Table 2. Snow conditions and minimum temperature observed at Ruffed Grouse roost sites.
Roosting behaviour Snow depth (cm) Snow sink depth (cm) Temperature (°C)
Snow burrow 23.9 6.6 (13–44) 9.5 3.2 (4–15) 13.6 6.7 (26.7 to 4.3)
Snow bowl 15.6 7.5 (3–40) 6.7 2.5 (1–17) 16.5 7.7 (29.4 to 1.0)
Surface 11.2 7.6 (0.5–39) 3.4 2.3 (0–13) 13.6 7.8 (29.2 to 0.8)
Tree 11.7 6.8 (1–31) 4.6 2.6 (0–17) 14.4 8.6 (27.9 to 1.1)
Values shown are means sd. The range is shown in parentheses.
S3). We present results of the analysis without a 1.02–2.95) relative to Dense Cover (Fig. 2c,
random effect because the AIC value of the top Table S3).
model (AIC = 810.58) was lower than the AIC We found evidence that land cover influenced
value of the identical top model with a random the relationship between roosting behaviour and
effect (AIC = 821.14; Table S1). We used the top snow depth. We used a binomial model of snow
model for interpretation, which included snow roosting behaviour that did not include a random
depth, snow sink depth and land cover (AICc effect of grouse ID (AIC = 476.10 for the model
weight = 0.47). An additional competitive model with a random effect; AIC = 474.29 for the model
was identical to the top model but included time without a random effect; Supporting Information
(Tables 3 and S2). As snow depth increased, Table S4). We found a significant main effect of
grouse were more likely to use snow burrows snow depth (b = 0.202, 95% CI: 0.08–0.36):
(b = 1.47, 95% confidence interval (95% CI): grouse were more likely to use snow roosts (bowls
0.81–2.13) and snow bowls (b = 0.36, 95% CI: or burrows) than other roosts (surface or tree
0.08–0.64) relative to surface roosts (Fig. 2a, roosts) as snow depth increased. Additionally, we
Table S3). As snow sink depth increased and snow found a significant interaction between snow depth
became powdery, grouse were more likely to use and land cover. As predicted, grouse were more
snow burrows (b = 2.27, 95% CI: 1.65–2.89), likely to use snow roosts in shallower snow when
snow bowls (b = 1.44, 95% CI: 1.09–1.79) and, to they were in Open habitat than when in Mature
a lesser extent, tree roosts (b = 0.69, 95% CI: Forest (b = 0.13, 95% CI: 0.30 to 0.00) and
0.36–1.03) relative to surface roosts (Fig. 2b, Scrub Oak (b = 0.13, 95% CI: 0.27 to 0.01;
Table S3). Although surface roosts were the most Fig. 3, Table S4).
common roosting behaviour across all land cover
classes, grouse were more likely to roost in trees
Effects of roosting behaviour on
when they were in Mature Forest (b = 1.78, 95%
temperature exposure
CI: 0.94–2.62) and Scrub Oak (b = 1.98, 95% CI:
We used a linear model of the effects of roosting
behaviour on temperature exposure that included
Table 3. Candidate models (ΔAICc ≤ 6) and the intercept null
a random effect of grouse ID (AIC = 436.12 for
model evaluating the effects of winter weather variables and
land cover on Ruffed Grouse roosting behaviour. the model with a random effect; AIC = 441.78 for
the model without a random effect). The differ-
Model K AICc ΔAICc wi ence between grouse temperature exposure and
ambient temperature was significantly greater
Snow Depth + Snow Sink 18 810.59 0.00 0.47 when grouse used snow roosts as compared with
Depth + Cover
Snow Depth + Snow Sink 21 812.30 1.72 0.20
surface roosts (b = 4.40, se = 2.07, 95% CI:
Depth + Cover + Time 8.42 to 0.38; Fig. 4, Supporting Information
Snow Depth*Snow Sink 21 813.25 2.67 0.12 Table S5). Thus, grouse experienced temperatures
Depth + Cover 4.4 °C higher in snow roosts than on surface
Snow Depth + Snow Sink 21 813.64 3.06 0.10 roosts. Grouse also tended to experience higher
Depth + Cover +
Temperature
temperatures in snow roosts than in tree roosts but
Snow Depth*Snow Sink 24 815.04 4.46 0.05 confidence intervals overlapped zero (b = 3.69,
Depth + Cover + Time se = 2.48, 95% CI: 8.51 to 1.13).
Snow Depth + Snow Sink 24 815.35 4.77 0.04
Depth + Cover +
Time + Temperature Frequency of snow roosting and suitable
NULL 3 992.11 181.53 0.00 snow roosting conditions
See Supporting Information for a full list of candidate models. Snow roosting and the minimum conditions
Snow sink depth represents the depth that a weighted cylinder required for this behaviour were uncommon dur-
sinks into the snow when dropped and is a proxy for snow ing the study (Fig. 5). To calculate the frequency
density; larger values indicate more powdery snow. Cover is a
categorical land cover variable and includes Mature Forest,
of snow roosting and suitable snow roosting condi-
Dense Cover, Open habitat and Scrub Oak. Temperature rep- tions, we used the entire dataset of roost observa-
resents minimum temperature at the roost location during the tions (754 roost observations, including
previous 24 h. Time is measured in minutes from midnight. observations when snow depth was equal to zero).
Figure 2. Partial dependency plots showing the effects of (a) snow depth (cm), (b) snow sink depth (cm; higher values indicate more
powdery snow) and (c) land cover classes on the probability of Ruffed Grouse using snow burrows (purple), snow bowls (blue), sur-
face roosts (black) or tree roosts (green). Lines and shaded bands (a, b) and points and error bars (c) represent means and 95%
confidence intervals. Vertical tick marks indicate observed snow depth (a) and observed snow sink depth (b) at roost sites. The prob-
ability of grouse using snow bowls increased steadily as snow depths increased from 0 to 35 cm, but the probability of snow burrow-
ing increased considerably only between 20 and 40 cm snow depth. The probability of using snow bowls quickly increased with
increasing snow sink depth, but grouse did not use snow burrows unless snow was considerably more powdery.
We defined the minimum conditions as the shal- of ephemeral winter microrefugia. Snow density
lowest snow depth (3 cm for snow bowls, 13 cm and snow depth were the most important factors
for snow burrows) and shallowest snow sink depth influencing roosting behaviour, and grouse rapidly
(1 cm for snow bowls, 4 cm for snow burrows) changed behaviour towards roosting in snow bowls
where we observed the behaviours, which we con- and burrows as snow became more powdery and
sider a conservative estimate. The snow depth and deeper. Temperature did not influence roosting
snow density conditions necessary for snow bur- behaviour. Although the likelihood of snow roost-
rowing were present on between ~ 15 and 35% of ing did not differ among land cover types, grouse
winter days, and only 0.3–5% of roosts were snow were more likely to use snow roosts in open habi-
burrows in any given year. Because we observed tat at shallower snow depths, potentially indicating
snow bowl behaviour with as little as 3 cm snow an increased importance of using snow roosts for
depth, the minimum conditions for snow bowls protection from predators in open habitat or a lack
occurred more frequently (present on ~ 58–75% of alternative roosts. Importantly, although snow
of winter days), but only ~ 9–13% of roosts were roosting resulted in grouse being less exposed to
snow bowls in any given year. lower temperatures, snow burrows and the winter
weather conditions that promote them were rela-
tively rare, and snow burrowing only accounted
DISCUSSION
for a very small proportion (0.3–5%) of observed
Our study demonstrates how snow conditions and winter roosting across the 3-year study (Fig. 5). As
land cover influence grouse behaviour and the use winters in temperate forests warm with climate
Figure 3. Partial dependency plot showing the effect of an

interaction between snow depth (cm) and land cover on the Figure 4. The effect of Ruffed Grouse roosting behaviour on
probability that Ruffed Grouse will use snow roosts (snow bur- grouse exposure to winter temperatures. Snow roosts repre-
rows and snow bowls pooled) over all other roost behaviours. sent pooled observations of snow burrows and snow bowls.
Open habitat (orange), Dense Cover (black), Mature Forest Temperature increase (°C) represents the difference between
(red) and Scrub Oak (blue). Grouse were more likely to use the median temperature recorded by iButtons attached to
snow roosts at shallower snow depths in Open habitat than in grouse radio-transmitters during the 12 h prior to the roost
Mature Forest or Scrub Oak. Solid lines represent means and observation and the minimum ambient temperature during the
shaded bands and dashed lines represent 95% confidence previous 24 h. Roost behaviours that do not share letters have
intervals. Vertical tick marks indicate observed snow depth at 95% confidence intervals that do not overlap zero. Grouse in
roost sites. There were 12 snow roosts and 38 other roosts in snow roosts experienced significantly higher temperatures
Open habitat, 40 snow roosts and 182 other roosts in Mature than grouse using surface roosts. Grey circles represent
Forest, 14 snow roosts and 59 other roosts in Scrub Oak, and observed temperature differences for each roosting behaviour.
37 snow roosts and 112 other roosts in Dense Cover. There were 12 snow roost observations, 41 surface roost
observations and 14 tree roost observations.
change, loss of snow cover and its insulating prop-

erties may paradoxically result in colder environ- 13 cm, yet we observed grouse using snow bowls
ments for species that seek shelter under the snow with as little as 3 cm of snow. Previous studies
(Pauli et al. 2013). have cited 15–20 cm as the minimum depth
required for Ruffed Grouse to use snow burrows
(Gullion 1970, Thompson & Fritzell 1988a,
Effects of winter weather and land
Blanchette et al. 2007, Heinrich 2017). Although
cover on roosting behaviour
grouse can roost in snow bowls when snow is shal-
Grouse were more likely to use snow bowls and low, roosting in shallower snow bowls with more
snow burrows than surface roosts as snow depth of the bird’s body exposed may not provide the
increased, but this behavioural flexibility was non- same benefits as deeper snow bowls or snow bur-
linear. The probability that grouse would use snow rows. In fact, we found that as snow depths
bowls increased steadily between 0 and 35 cm approached 15 cm, grouse avoided deeper snow
snow depth, but the probability that grouse would and only began selecting for deeper snow at
roost in snow burrows remained near 0% until depths > 15 cm (Shipley et al. 2020). Notably,
snow depths reached 20–25 cm, increasing rapidly exposure to deeper snow (close to 15 cm) is asso-
to nearly 50% when snow depths exceeded 40 cm. ciated with increased winter mortality (Shipley
This rapid increase in the likelihood of using snow et al. 2020) and may explain why the probability
burrows when snow depth exceeds 25 cm, as of grouse roosting in snow bowls only marginally
compared with the more gradual increase in the increases with deeper snow until a threshold is
likelihood of using snow bowls with increasing reached where snow burrowing becomes possible.
snow depth, is indicative of the benefits of snow When snow depth exceeds 15 cm, and snow bur-
burrows over snow bowls. The minimum snow rowing becomes possible, individual grouse that
depth in which grouse used snow burrows was seek out the deepest snow and use snow burrows
Figure 5. The proportion of winter days on which we observed snow roosting behaviour in Year 1 (2015–2016), Year 2 (2016–2017)
and Year 3 (2017–2018) that had the minimum conditions for snow bowls (blue) and snow burrows (purple) (a). The proportion of
roosts in each year that were tree roosts (green), surface roosts (black), snow bowls (blue) and snow burrows (purple) (b). Conditions
for snow burrowing and the frequency of snow burrowing behaviour were uncommon during the 3-year study.
have higher winter survival probability (Shipley that may be more difficult to move through and
et al. 2020). In Quebec, Blanchette et al. (2007) which increases stress levels and susceptibility to
found that the incidence of snow burrowing predation. As sink depths increase beyond 6 cm,
behaviour in Ruffed Grouse was as high as 36% snow burrowing and use of snow bowls become
between 2001 and 2003, during which snow more likely, and grouse are less likely to roost in
depths at the end of January were around 40 cm, trees. Even at the maximum snow sink depths that
which closely mirrors our predicted probability of we observed, snow burrowing was never a more
snow burrowing at 40 cm (Fig. 2a). Presumably if common roosting behaviour than use of snow
we had observed mean snow depths closer to bowls. This may indicate that deep snow bowls
40 cm, we would have documented a greater have similar benefits to snow burrows or that there
number of snow burrows. For species that seek may be trade-offs associated with using snow bur-
shelter under the snow, the depth of snow needed rows, such as the inability to observe surroundings.
to obtain the insulating benefits of snow probably Nonetheless, the ability to use snow bowls or snow
depends on body size. For example, Sage Grouse, burrows under similar conditions further points to
which are at least twice as large as Ruffed Grouse, the importance of behavioural flexibility.
use snow bowls and snow burrows when snow While minimum temperature is an important
depths are greater than 25 and 30 cm, respectively predictor of grouse stress (Shipley et al. 2019) and
(Back et al. 1987). roost site selection (Shipley et al. 2020), we did
Grouse were more likely to use snow bowls and not find an effect of ambient temperature on
snow burrows than surface roosts as snow sink roosting behaviour, although we were not able to
depth increased and snow became powdery, but test for an effect of wind chill. Under a different
the relative hardness of snow had a much stronger suite of weather conditions (e.g. deeper snow and
effect. Snow burrowing probability only exceeded or lower temperatures than we observed), or if we
0% when sink depths were greater than 6 cm, and were able to include data on wind speed, we may
then increased to about 50% at 15 cm sink depth. have detected an effect of temperature on the
Increased snow sink depth is associated with probability of snow roosting behaviour. Blanchette
increased physiological stress in grouse (Shipley et al. (2007) found that Ruffed Grouse were more
et al. 2019), probably due to increased energy likely to use snow burrows when air temperatures
required to move through more powdery snow. were lower, but at their Quebec study site, mean
When snow sink depths are moderate (around snow depths exceeded 40 cm, and a larger number
6 cm), grouse do not use snow burrows and are of snow burrow observations may have enabled
relatively unlikely to roost in snow bowls. At these detection of an effect of temperature on snow
moderate sink depths, grouse may spend more roosting behaviour. Alternatively, predator avoid-
time roosting in trees to avoid more powdery snow ance may also be an important benefit of snow
roosting. Snow roosting increases overwinter sur- burrows were up to 7 °C warmer than other roost
vival (Shipley et al. 2020) and thus grouse may sites. The lower critical temperature for Ruffed
continue to use snow roosts even when tempera- Grouse (below which they must increase their
tures are higher. Anecdotally, Prawdzik (1963) metabolic rate to balance environmental demands)
observed a Ruffed Grouse pursued by a Cooper’s has been estimated as 0.3 °C (Rasmussen &
Hawk Accipiter cooperii dive into and repeatedly Brander 1973) and as 1.5 °C (Thompson &
bury itself in the snow, even after being removed Fritzell 1988a). Given that the average minimum
from the burrow by the author, indicating that temperature at our study site was 9 °C, the use
Ruffed Grouse may use snow roosts as refuge from of snow burrows may allow grouse to remain
predators. Spidsø et al. (1997) attributed increased much closer to their lower critical temperature.
predation of Black Grouse by Northern Goshawks Snow burrows clearly are important for heat reten-
Accipiter gentilis to a winter with limited snow fall tion and energy savings. Most of our snow roost
that precluded snow roosting. The benefits of observations were snow bowls, which are open on
snow roosting could therefore be relevant regard- top, and thus our estimate of the heat retention of
less of temperature, and grouse may use snow snow roosts is likely to be an underestimate and
roosts as soon as minimum snow depth and snow would probably be greater with a higher propor-
sink depth conditions are met. tion of true snow burrows.
Land cover did not influence the probability
that grouse would use snow bowls or snow bur-
Deteriorating conditions for an adaptive
rows, and surface roosts were the most common
behavioural strategy
roosting behaviour across cover types. Grouse
were more likely to roost in trees in mature forest Seeking refugia below the snow is an important
and scrub oak, probably due to the increased avail- behavioural adaptation that allows for exposure to
ability of large trees for roosting as compared with higher temperatures, reduces stress (Shipley
open areas and dense cover. However, in open et al. 2019) and is associated with increased over-
habitat, as predicted, grouse were more likely to winter survival (Shipley et al. 2020). However, the
use snow roosts at shallower snow depths than in conditions that promote snow roosting (deep,
mature forest or scrub oak. Grouse stress increases powdery snow) are already relatively uncommon,
with the size of open habitat patches (Shipley particularly for snow burrows (Fig. 5), and are pre-
et al. 2022), and grouse avoid open areas (Shipley dicted to deteriorate further. Across Wisconsin,
et al. 2020). In this study, open habitat was not average snow depths currently peak at 20 cm in
associated with lower minimum temperatures. midwinter, and peak snow depths are projected to
Thus, the tendency to use snow roosts at shallower decline to 8–15 cm in the future (Notaro
depths in open habitat may reflect an increased et al. 2011, 2014), well below that needed for
perceived risk of predation and a greater need to snow burrowing to be more than a rare occur-
roost under the snow as a refuge from predators. rence. An increase in the number of freeze–thaw
events is also predicted (Mellander et al. 2007,
Sinha & Cherkauer 2010), resulting in a hardened
Effects of roosting behaviour on
snow layer that will further reduce snow quality
temperature exposure
for roosting. During our 3-year study, we observed
When grouse used snow roosts, they experienced progressively fewer days that met the minimum
temperatures 4.4 °C above ambient temperature conditions necessary for snow burrowing behav-
compared with when they used surface roosts. iour, and as few as 0.3% of all roosts were snow
The temperatures recorded by the sensors attached burrows. Although our study period was only a
to grouse transmitters were probably partially small snapshot in time, and another 3-year period
influenced by grouse body temperature, as temper- might have shown an increase in snow burrowing
atures recorded by sensors were considerably conditions, the pattern we observed is consistent
higher than ambient temperature recorded at sur- with the expected trend predicted with warming
face roosts. However, the effect of body heat winters in Wisconsin. In more northerly parts of
would presumably be the same regardless of roost- the range of Ruffed Grouse (e.g. Quebec; Blanch-
ing behaviour. Using a heated taxidermic mount, ette et al. 2007) deep snow conditions and snow
Thompson and Fritzell (1988b) found that snow burrowing behaviour are still common and can
effectively buffer grouse from low temperatures. and logistical assistance, and we are grateful to the
In more southerly parts of their range (e.g. the Ruffed Grouse Society for additional funding. The Uni-
Appalachian Mountains; Whitaker & Stauf- versity of Wisconsin-Madison Department of Forest and
Wildlife Ecology provided additional support. We also
fer 2003) winter temperatures may already be
thank L. Korslund and K. Cutting for reviewing an ear-
mild enough that inadequate snow cover for bur- lier version of the manuscript.
rowing is no longer detrimental. For species that
use snow as a refuge from harsh winter conditions
when snow cover is adequate, the limits of beha- AUTHOR CONTRIBUTIONS
vioural flexibility are being stretched as deep and Amy A. Shipley: Conceptualization; data curation;
powdery snow becomes less common. To lessen formal analysis; investigation; methodology; project
the impacts of snow cover losses for winter- administration; validation; visualization; writing –
adapted species, regional managers and conserva- original draft; writing – review and editing.
tionists can protect the subnivium when and Benjamin Zuckerberg: Conceptualization; data
where it exists by limiting the use of snowmobiles curation; formal analysis; funding acquisition;
and silvicultural practices that involve heavy investigation; methodology; project administration;
machinery which compacts the snow (Zuckerberg resources; software; supervision; validation; visuali-
& Pauli 2018). zation; writing – original draft; writing – review
Snow depth and snow density are important and editing.
limiting factors influencing grouse roosting behav-
iour and their exposure to harsh winter conditions.
Grouse rapidly shifted from roosting on top of the CONFLICT OF INTEREST
snow to roosting under the snow as snow became STATEMENT
deeper and more powdery. Behavioural flexibility We have no conflict of interest to declare.
and snow roosting behaviour have clear benefits
for heat retention, stress reduction and increased
survival via predator avoidance but the conditions ETHICAL NOTE
that promote the use of these microrefugia are None.
predicted to become rarer with modern climate
change. While using snow bowls is currently possi-
ble for a larger proportion of winter days (58–75% FUNDING
across years), roosting in shallower snow bowls
This study is based upon work funded by the
may not provide the same benefits as deeper snow
National Institute of Food and Agriculture, U.S.
bowls or snow burrows. Importantly, the loss of
Department of Agriculture, Hatch Projects
snow cover is truncating the range of behavioural
1006604 and 1003605; Ruffed Grouse Society;
flexibility for this winter-adapted bird. Given these
University of Wisconsin-Madison College of Agri-
non-linear behavioural responses to the use of
cultural and Life Sciences; Wisconsin Department
these important winter microrefugia, it is likely
of Natural Resources.
that a rapidly changing winter climate represents a
tipping point in this behavioural syndrome.
Although winters are expected to warm in the
Data Availability Statement
future across northern forests, loss of snow cover
that provides insulating properties may result in a Data used in this paper are available via Dryad
functionally colder winter environment for snow- (Shipley & Zuckerberg 2022; doi:10.5061/
adapted species where and when cold snaps occur dryad.2bvq83btr).
(Pauli et al. 2013, Zuckerberg & Pauli 2018) and
may reduce the effectiveness of evolved adapta-
tions and behavioural flexibility. REFERENCES
Ashcroft, M.B. & Gollan, J.R. 2012. Fine-resolution (25 m)
We thank B. Heindl, A. Walker, K. Kovach, T. Gettel- topoclimatic grids of near-surface (5 cm) extreme
man, A. Elzinga, J. Ostroski, A. Bradley, A. Wilkie and temperatures and humidities across various habitats in a
E. Leicht for many hours collecting data. We thank the large (200 x 300 km) and diverse region. Int. J. Climatol. 32:
Wisconsin Department of Natural Resources for funding 2134–2148.
Back, G.N., Barrington, M.R. & McAdoo, J.K. 1987. Sage Humphries, M.M., Boutin, S., Thomas, D.W., Ryan, J.D.,
grouse use of snow burrows in northeastern Nevada. Wilson Selman, C., McAdam, A.G., Berteaux, D. & Speakman,
Bull. 99: 488–490. J.R. 2005. Expenditure freeze: The metabolic response of
Baltic, M., Jenni-Eiermann, S., Arlettaz, R. & Palme, R. small mammals to cold environments. Ecol. Lett. 8: 1326–
2005. A noninvasive technique to evaluate human- 1333.
generated stress in the black grouse. Ann. N. Y. Acad. Sci. Irving, L. 1960. Birds of Anaktuvuk Pass, Kobuk, and Old
1046: 81–95. Crow: 1–409. Washington, D.C: United States National
Beever, E.A., Hall, L.E., Varner, J., Loosen, A.E., Dunham, Museum Bulletin.
J.B., Gahl, M.K., Smith, F.A. & Lawler, J.J. 2017. Isard, S.A., Schaetzl, R.J. & Andresen, J.A. 2007. Soils cool
Behavioural flexibility as a mechanism for coping with as climate warms in the Great Lakes region: 1951-2000.
climate change. Front. Ecol. Environ. 15: 299–308. Ann. Assoc. Am. Geogr. 97: 467–476.
Blanchette, P., Bourgeois, J.-C. & St-Onge, S. 2007. Winter Kolec ek, J., Adamık, P. & Reif, J. 2020. Shifts in migration
selection of roost sites by ruffed grouse during daytime in phenology under climate change: Temperature vs.
mixed nordic-temperate forests, Quebec, Canada. Can. J. abundance effects in birds. Clim. Change 159: 177–194.
Zool. 85: 4–504. Korslund, L. & Steen, H. 2006. Small rodent winter survival:
Brown, P.J. & DeGaetano, A.T. 2011. A paradox of cooling Snow conditions limit access to food resources. J. Anim.
winter soil surface temperatures in a warming northeastern Ecol. 75: 156–166.
United States. Agric. For. Meteorol. 151: 947–956. Latimer, C.E. & Zuckerberg, B. 2016. Forest fragmentation
Bump, G.R., Darrow, R.W., Edminster, F.C. & Crissey, W.F. alters winter microclimates and microrefugia in human-
1947. The Ruffed Grouse: Life History, Propagation, and modified landscapes. Ecography 39: 1–13.
Management. Buffalo, NY: New York State Conservation Lima, S.L. 1986. Predation risk and unpredictable feeding
Department. conditions - determinants of body-mass in birds. Ecology 67:
Burnham, K.P. & Anderson, D.R. 2002. Model Selection and 377–385.
Multimodel Inference. New York, NY: Springer. Marjakangas, A., Rintamaki, H. & Hissa, R. 1984. Thermal
Carter, A., Goldizen, A. & Heinsohn, R. 2012. Personality responses in the Capercaillie Tetrao urogallus and the black
and plasticity: Temporal behavioural reaction norms in a grouse Lyrurus tetrix roosting in the snow. Physiol. Zool. 57:
lizard, the Namibian rock agama. Anim. Behav. 84: 471– 99–104.
477. Mellander, P.E., Lofvenius, M.O. & Laudon, H. 2007.
Chevin, L.M., Lande, R. & Mace, G.M. 2010. Adaptation, Climate change impact on snow and soil temperature in
plasticity, and extinction in a changing environment: boreal scots pine stands. Clim. Change 85: 179–193.
Towards a predictive theory. PLoS Biol. 8: 8. Murray, I.W. & Smith, F.A. 2012. Estimating the influence of
Cochran, W.W. 1980. Wildlife telemetry. In Schemnitz, S.D. the thermal environment on activity patterns of the desert
(ed) Wildlife Management Techniques Manual: 507–520. woodrat (Neotoma lepida) using temperature chronologies.
Washington, DC: The Wildlife Society. Can. J. Zool. 90: 1171–1180.
Coombs, A.B., Bowman, J. & Garroway, C.J. 2010. Thermal Notaro, M., Lorenz, D.J., Vimont, D., Vavrus, S., Kucharik,
properties of tree cavities during winter in a northern C. & Franz, K. 2011. 21st century Wisconsin snow
hardwood forest. J. Wildl. Manag. 74: 1875–1881. projections based on an operational snow model driven by
Duchesne, D., Gauthier, G. & Berteaux, D. 2011. Habitat statistically downscaled climate data. Int. J. Climatol. 31:
selection, reproduction and predation of wintering lemmings 1615–1633.
in the Arctic. Oecologia 167: 967–980. Notaro, M., Lorenz, D., Hoving, C. & Schummer, M. 2014.
Ford, K.R., Ettinger, A.K., Lundquist, J.D., Raleigh, M.S. & Twenty-first-century projections of snowfall and winter
Lambers, J.H.R. 2013. Spatial heterogeneity in ecologically severity across central-eastern North America. J. Climate
important climate variables at coarse and fine scales in a 27: 6526–6550.
high-snow mountain landscape. PLoS One 8: 13. Pauli, J.N., Zuckerberg, B., Whiteman, J.P. & Porter, W.
Gullion, G.W. 1965. Improvements in methods for trapping 2013. The subnivium: A deteriorating seasonal refugium.
and marking ruffed grouse. J. Wildl. Manag. 29: 109–116. Front. Ecol. Environ. 11: 260–267.
Gullion, G.W. 1970. Factors affecting ruffed grouse Piersma, T. & van Gils, J.A. 2011. The Flexible Phenotype:
populatons in boreal forests of northern Minnesota, USA. A Body-Centered Integration of Ecology, Physiology, and
Finnish Game Res. 30: 103–117. Behaviour. Oxford: Oxford University Press.
Hale, J.B., Wendt, R.F. & Halazon, G.C. 1954. Sex and Age Pokallus, J.W. & Pauli, J.N. 2016. Predation shapes the
Criteria for Wisconsin Ruffed Grouse: 24. Madison, WI: movement of a well-defended species, the north American
Wisconsin Conservation Department Technical Bulletin. porcupine, even when nutritionally stressed. Behav. Ecol.
Hall, L.E., Chalfoun, A.D., Beever, E.A. & Loosen, A.E. 27: 470–475.
2016. Microrefuges and the occurrence of thermal Pomara, L.Y. & Zuckerberg, B. 2017. Climate variability
specialists: Implications for wildlife persistence amidst drives population cycling and synchrony. Divers. Distrib. 23:
changing temperatures. Clim. Change Responses 3: 1–12. 421–434.
Heinrich, B. 2017. Winter strategies of ruffed grouse in a Prawdzik, T.R. 1963. Ruffed grouse escaping a Cooper’s
mixed northern forest. Northeast. Nat. 24: B55–B71. hawk. J. Wildl. Manag. 27: 639–642.
Hengl, T., Heuvelink, G.B.M. & Rossiter, D.G. 2007. About R Core Team2022. R: A Language and Environment for
regression-kriging: From equations to case studies. Comput. Statistical Computing. Vienna: R Foundation for Statistical
Geosci. 33: 1301–1315. Computing. https://www.R-project.org/
Rasmussen, G. & Brander, R. 1973. Standard metabolic rate

Received 14 February 2022;
and lower critical temperature for ruffed grouse. Wilson Bull.
Revision 17 February 2023;
85: 223–229.
revision accepted 21 March 2023.
Sears, M.W., Raskin, E. & Angiletta, M.J.Jr. 2011. The world
Associate Editor: Chevonne Reynolds.
is not flat: Defining relevant thermal landscapes in the
context of climate change. Integr. Comp. Biol. 51: 666–675.
Shipley, A.A. & Zuckerberg, B. 2022. Ruffed Grouse
SUPPORTING INFORMATION
Roosting Behaviour Dataset, Dryad, Dataset. https://doi.org/
10.5061/dryad.2bvq83btr Additional supporting information may be found
Shipley, A.A., Sheriff, M.J., Pauli, J.N. & Zuckerberg, B.
online in the Supporting Information section at
2019. Snow roosting reduces temperature-associated stress
in a wintering bird. Oecologia 190: 309–321. the end of the article.
Shipley, A.A., Cruz, J. & Zuckerberg, B. 2020. Personality Table S1 Comparison of candidate model rank-
differences in the selection of dynamic refugia have ings based on a multinomial logistic regression
demographic consequences for a winter-adapted bird. Proc. evaluating the effects of winter weather variables
Royal Soc. B 287: 10.
and land cover on Ruffed Grouse roosting behav-
Shipley, A.A., Sheriff, M.J., Pauli, J.N. & Zuckerberg, B.
2022. Weather and land cover create a predictable "stress- iour. Models that did not include a random effect
scape" for a winter-adapted bird. Landsc. Ecol. 37: 779–793. of individual grouse ID were ranked based on AIC
Sinha, T. & Cherkauer, K.A. 2010. Impacts of Future Climate and are shown on the left side of the table. Models
Change on Soil Frost in the Midwestern United States. J. that did include a random effect of individual
Geophys. Res. 115: 1–16.
grouse ID were ranked based on AIC and are
Spidsø, T.K., Hjeljord, O. & Dokk, J.G. 1997. Seasonal
mortality of black grouse Tetrao tetrix during a year with little shown on the right side of the table. Model rank-
snow. Wildl. Biol. 3: 205–209. ings were nearly identical, but models without a
Suggitt, A.J., Gillingham, P.K., Hill, J.K., Huntley, B., random effect consistently had lower AIC values
Kunin, W.E., Roy, D.B. & Thomas, C.D. 2011. Habitat than those with a random effect. The set of candi-
microclimates drive fine-scale variation in extreme
date models without random effects were used for
temperatures. Oikos 120: 1–8.
Taylor, S.L. & Buskirk, S.W. 1994. Forest microenvironments further analysis.
and resting energetics of the American marten Martes Table S2. Candidate models and the intercept
americana. Ecography 17: 249–256. null model evaluating the effects of winter weather
Thompson, F.R. & Fritzell, E.K. 1988a. Ruffed grouse variables and land cover on Ruffed Grouse roosting
metabolic rate and temperature cycles. J. Wildl. Manag. 52:
behaviour.
450–453.
Thompson, F.R. & Fritzell, E.K. 1988b. Ruffed grouse winter Table S3. Relationships between Ruffed Grouse
roost site preference and influence on energy demands. J. roosting behaviour and winter weather variables
Wildl. Manag. 52: 454–460. and land cover. Only variables in the top model
Thompson, K.L., Zuckerberg, B., Porter, W.P. & Pauli, J.N. (based on AIC model selection of a list of multino-
2018. The phenology of the subnivium. Environ. Res. Lett.
mial logistic models) are presented here. For Snow
13: 12.
Thompson, K.L., Zuckerberg, B., Porter, W.P. & Pauli, J.N. Depth and Snow Sink Depth, beta estimates rep-
2021. The decline of a hidden and expansive microhabitat: resent the effect of Snow Depth and Snow Sink
The subnivium. Front. Ecol. Environ. 19: 268–273. Depth on the probability that a surface roost
Valdecantos, S., Martinez, V., Lobo, F. & Cruz, F.B. 2013. would change to a snow bowl, snow burrow or
Thermal biology of Liolaemus lizards from the high Andes:
tree roost. For land cover types, beta estimates
Being efficient despite adversity. J. Therm. Biol. 38: 126–
134. represent the effect of changing from Dense Cover
Venables, W.N. & Ripley, B.D. 2002. Modern Applied to Mature Forest, Open habitat or Scrub Oak on
Statistics with S. New York: Springer. the probability that a surface roost would change
Westerskov, K. 1965. Winter ecology of the partridge (Perdix to a snow bowl, snow burrow or tree roost. b, se
perdix) in the Canadian prairie. Proc. N Z Ecol. Soc. 12: 23–30.
and 95% CI marked with a * represent values
Whitaker, D.M. & Stauffer, D.F. 2003. Night roost selection
during winter by ruffed grouse in the Central Appalachians. from an identical model that also included individ-
Southeast. Nat. 2: 377–392. ual grouse ID as a random effect. The model that
Wilson, E.C., Shipley, A.A., Zuckerberg, B., Peery, M.Z. & did not include a random effect was used for
Pauli, J.N. 2018. An experimental translocation identifies inference.
habitat features that buffer camouflage mismatch in
Table S4. Relationships between Ruffed Grouse
snowshoe hares. Conserv. Lett. 12: 12614.
Zuckerberg, B. & Pauli, J.N. 2018. Conserving and roosting behaviour and snow depth and land cover
managing the subnivium. Conserv. Biol. 32: 774–781. based on a binomial logistic model of behaviour
where snow roosts (snow bowls and snow burrows linear model. Temperature increase (°C) repre-
pooled) = 1 and other roosts (surface roosts and sents the difference between the median tempera-
tree roosts pooled) = 0. b, se and 95% CI marked ture recorded by iButtons attached to grouse radio
with a * represent values from an identical model transmitters during the 12 h prior to the roost
that also included individual grouse ID as a ran- observation and the minimum ambient tempera-
dom effect. The model that did not include a ran- ture during the previous 24 h. b, se and 95% CI
dom effect was used for inference. marked with a * represent values from an identical
Table S5. Relationships between Ruffed Grouse model that also included individual grouse ID as a
exposure to winter temperatures and Ruffed random effect. The model that included a random
Grouse roosting behaviour based on a generalized effect was used for inference.

Snow Cover Constrains The Behavioural Flexibility of A Winter Adapted Bird - Original Article (Habitat Use)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Snow Cover Constrains The Behavioural Flexibility of A Winter Adapted Bird - Original Article (Habitat Use)

Uploaded by

Copyright:

Available Formats

Ibis (2023), 165, 1186–1200 doi: 10.1111/ibi.

Snow cover constrains the behavioural flexibility of a

Behavioural ﬂexibility is an important way in which animals respond to changing envi-

is an important behavioural adaptation in a chang- and increase overwinter survival (Shipley

Values shown are means sd. The range is shown in parentheses.

Figure 3. Partial dependency plot showing the effect of an

change, loss of snow cover and its insulating prop-

Rasmussen, G. & Brander, R. 1973. Standard metabolic rate

You might also like