938 ESTIMATING WOLF DENSITY IN FORESTED AREAS
Few reliable methods exist for estimating popu-
lation size of large terrestrial carnivores (Crête and
Messier 1987, Fuller and Snow 1988, Becker 1991,
Ballard et al. 1995, Miller et al. 1997, Becker et al.
1998). This is particularly true for forested areas
where sightability is low and when radiocollared
individuals are unavailable in the target population.
Although radiotelemetry might remain the best
technique for estimating wolf density associated
with intensive, relatively small study areas, it is
expensive and may not be logistically or socially
feasible in all areas (Crête and Messier 1987, Fuller
and Sampson 1988). These difficulties notwith-
standing, estimating population size remains cen-
Wildlife Society Bulletin 2004, 32(3):938–947 Peer refereed
Address for Brent R. Patterson: Ontario Ministry of Natural Resources, Wildlife Research and Development Section, 300 Water
Street, 3rd Floor N., Peterborough, ON, K9J 8M5 Canada; e-mail: brent.patterson@mnr.gov.on.ca. Address for Norman W. S.
Quinn: Algonquin Provincial Park, PO Box 219 Whitney, ON, K0J 2M0 Canada. Address for Earl F. Becker: Alaska Department of
Fish and Game, Division of Wildlife Conservation, 333 Raspberry Road, Anchorage, AK 99518, USA. Address for Derek B. Meier:
Department of Environmental Resource Studies, Trent University, Peterborough, ON, K9J 7B8 Canada.
Estimating wolf densities in forested
areas using network sampling of tracks
in snow
Brent R. Patterson, Norman W. S. Quinn, Earl F. Becker,
and Derek B. Meier
Abstract Few reliable methods exist for estimating population size of large terrestrial carnivores.
This is particularly true in forested areas where sightability is low and when radiocollared
individuals are unavailable in the target population. We used stratified network sampling
to sample wolf (Canis lycaon) tracks in the snow to estimate density in western Algonquin
Park, Ontario in February 2002. We partitioned our 3,425-km2 study area into 137 5 ×
5-km sample units (SU) and stratified SUs as having a high (n=61) or low (n=76) proba-
bility of containing detectable wolf tracks based on the relative amount of watercourses
and conifer cover within each block. We used a Bell 206B helicopter to survey 28 high
(46%) and 17 low (22%) SUs. When fresh tracks were found in a block, we followed the
tracks forward to the wolves themselves and then backward until the tracks were no
longer considered “fresh.” We observed 17 “fresh” track networks within 45 SUs. The
average pack size in the area we surveyed was 4.2±0.4 (SE). These observations result-
ed in an estimate of 87±11.4 (90% CI) wolves in the study area, for a density of 2.5±0.3
wolves/100 km2. We detected no violations of the assumptions of this survey design and
obtained a similar density estimate (2.3 wolves/100 km2) in 2003 using location data
from 24 radiocollared wolves in 10 packs from an area that overlapped our 2002 survey
area. The sampling unit probability estimator (SUPE) provides an objective, accurate, and
repeatable means of estimating wolf density with an associated measure of precision.
However, tracking wolves in forested cover was time-consuming, so costs will be con-
siderably higher per unit area in forested areas relative to the more open cover types
where this technique was originally developed.
Key words aerial survey, Algonquin Park, Canis lupus, density estimation, Ontario, population esti-
mation, probability sampling, radiotelemetry, track surveys, wolves

tral to the conservation of wolves (Canis lupus,
C. lycaon, C. rufus) and other large carnivores.
Becker et al. (1998) presented a novel method of
estimating gray wolf density and statistical confi-
dence intervals over large areas based on stratified
network sampling of wolf tracks in snow.
Population size and statistical confidence intervals
were calculated based on the probability of observ-
ing track networks in snow. This method,called the
sampling unit probability estimator (SUPE), makes
the following assumptions: 1) all animals of interest
move during the study, 2) their tracks are readily
recognizable from survey aircraft, 3) tracks are con-
tinuous, 4) wolf movements are independent of the
sampling process, 5) tracks made within and out-
side the sampling window (pre- and post- snowfall)
can be distinguished, 6) “fresh” tracks in searched
sample units (SU) are not missed, 7) tracks can be
followed forward and backward to determine all
SUs containing those tracks, 8) group size is cor-
rectly enumerated. Because most study areas will
require several days to survey,an additional assump-
tion is that no animals were double-counted by
moving among SUs on subsequent days. Using con-
currently collected radiotelemetry data on 9 wolf
packs in their study area, Becker et al. (1998) did
not detect any violations of these assumptions.
Although promising, there are no published reports
of the application of this method for estimating
density of a large carnivore species in a densely
forested habitat.
At 7,571 km2,Algonquin Provincial Park in south-
central Ontario represents the largest protected
area for the eastern timber wolf (C. lycaon, Wilson
et al. 2000, 2003). Amid concern that wolves may
be declining in Algonquin Park (Theberge 1998,
Vucetich and Paquet 2000), we used the SUPE to
estimate wolf abundance in the park in February
2002. We then compared this estimate with an
independent estimate obtained for the same gener-
al area in winter 2003 using “traditional” methods
based on radiotelemetry (e.g., Fuller and Snow
1988).
Study area
Algonquin Provincial Park (45oN, 78oW) encom-
passed 7,571 km2 on the southern edge of the
Canadian Shield and ranged in elevation from
180–380 m in the east side up to 580 m in the
west (Figure 1). Data were collected primarily in
the western portion of the park. The average
January temperature was –12oC, and temperatures
approaching –40oC were common (Environment
Canada 1993). Mean annual precipitation ranged
from 66–86 cm,with more snowfall in the western
portion of the park (Environment Canada 1993).
Algonquin Park consisted of 2 forests that were
sharply delineated: the eastern third consisted of
white pine (Pinus strobus), red pine (P. resinosa),
and jack pine (P. banksiana) stands on well-
drained sandy outwash and rolling to flat terrain
(Strickland 1993). The park’s west side consisted
primarily of tolerant hardwood forests composed
of sugar maple (Acer saccharum), American
beech (Fagus grandifolia), yellow birch (Betula
alleghaniensis), and eastern hemlock (Tsuga
canadensis), on a glacial till over poorly drained,
rugged terrain. Commercial logging occured in
approximately 75% of the park, and an extensive
network of logging roads covered much of it. No
other large carnivore species were present in the
study area during winter. Although coyotes (C.
latrans) lived immediately south of the park
(Sears 1999), they were rarely found within it. For
example, of the 92 canids (Canis sp.) live-trapped
for research purposes from August 2002 to
February 2004, only one appeared to be a coyote
(B. R. Patterson, Ontario Ministry of Natural
Resources, unpublished data). This animal was
radiocollared but was never relocated in the park.
Medium-sized carnivores that leave tracks in the
area in winter included fishers (Martes pennanti),
red foxes (Vulpes vulpes), and river otters (Lutra
canadensis).
Estimating wolf density in forested areas • Patterson et al. 939
The tracks left by this pair of wolves were followed for about 10
km before the wolves were finally sighted on a lake during the
February 2002 sampling unit probably estimator (SUPE) survey
used to wolf abundance in Algonquin Park, Ontario.