IMPROVING MOOSE POPULATION ESTIMATES IN RUSSIA:
ACCOUNTING FOR DISTANCE BETWEEN RESIDENTIAL AREAS AND
TRACK SIGHTINGS

Vladimir M. Glushkov

Russian Institute of Game Management and Fur Farming, Russian Academy of Agricultural Sciences,
Kirov, Russia.

ABSTRACT: Moose (Alces alces) population density in the Kirov region of Russia is often overesti-
mated when using the relationship between the distance between a residential area and the initial sight-
ing of moose tracks. This paper presents a modified approach to provide better estimates when using
this techinique. Statistically valid density estimation techniques, standardization of estimation points
and routes, landscape characteristics, and time have been addressed in the new approach. Moose den-
sity is estimated once annually based on the distance to the first track, and annual surveys should main-
tain alike protocol. This improved method will provide more accurate population density estimates
critical to prevent regional overharvest of moose.

ALCES VOL. 49: 149–154 (2013)

Key words: Alces alces, density, distribution, moose, population estimate, Russia, track surveys.

INTRODUCTION & BACKGROUND
Spatial distribution of individuals within a

population is generally described as 3 types –
equal, occasional, and grouped (Odum
1986) – that can be affected by regional and
temporal influences (Naumov 1963). In
Estonia, moose (Alces alces) distribution
changes seasonally; moose in summer-
autumn are evenly distributed but in winter
their distribution is sporadic or a “focal
type of distribution” (Ling 1977). Likewise,
moose distribution differs between summer
and winter in the northeast portion of Eur-
opean Russia (i.e., Kirov Region; Glushkov
1982). This seasonal difference is caused
by November migration related to forage
deficiency on summer range (Yazan 1972),
as well as increased moose hunting that
occurs after snow cover (Glushkov
1997, 2001).

The relationship between snow cover
and increased harvest has not been

considered previously relative to population
density estimates (i.e., ecological density;
Bubenik 1965) that are based upon the dis-
tance between a residential area and the
initial sighting of moose tracks. The unique
spatial distribution caused by this relation-
ship is neglected in typical winter route cen-
suses (WRC), creating error in abundance
estimates (Glushkov 2004) and potential
overharvest of moose that threatens popula-
tion stability (Glushkov et al. 2012). This
paper provides the rationale for a modified
approach to account for this relationship
when calculating a population estimate.

Previous studies provide baseline infor-
mation about seasonal moose distribution in
the Kirov region of Russia (Glushkov
1977). Group size is larger in winter (2.8 ±
0.9) than in summer (2.0 ± 0.6), and disper-
sion:density ratios of 2.4 in November ver-
sus 5.1 in March (measured from aerial
surveys within 1 min flight range of 60 ha

Corresponding author: Vladimir M. Glushkov, Russian Institute of Game Management and Fur
Farming, Russian Academy of Agricultural Sciences, Kirov, Russia Email: v.m.glushkov@yandex.ru

149



plots [n = 970]) confirms the more uneven
winter distribution of moose. These early
(November) and late winter (March) data
(1976–1985) were used to construct a graph
of sighting frequency in plots with varied
moose density that described the character
and variance of seasonal moose distribution
in the Kirov region (Fig. 1). There were
fewer unoccupied plots (0 moose) and plots
occupied by ≥4 moose in November than
in March when there were fewer plots with
2–3 moose. These seasonal differences are
statistically different, and specific to both
particular areas (χ2 = 42.7–171.4) and the
Kirov region as a whole (χ2 = 118.1). These
data made it possible to classify summer-
autumn distribution of moose as “occa-
sional” and winter distribution as “grouped
with cluster formation” (Glushkov 2001).

The distance between human settlements
and the initial observation of a moose track
was measured during helicopter surveys;
the area between was assumed absent of
moose. This distance was compared to the
sighting frequency of animals and tracks in
occupied habitat. In the southern area of the
region the correlation was not as strong
(r = − 0.50, tr = 2.15) as in the north
(r = – 0.65, tr = 3.42). Similar tests were con-
ducted with data from 288 terrestrial
straight-line survey routes in 27 regional dis-
tricts (2595 km total length with 288–200 ha
sample plots; November 1996); the distance
to the initial moose track and the population
estimate was inversely related (corr. coeff. =
−0.35; p = 0.002). In 10 of 15 districts sur-
veyed, the average distance to the sighting
of the first track was >7 km, and in 2 districts
it was ∼9 km; tracks were first observed at a
distance of 14, 16, and 25 km on the other 3
routes.

Because physical ability limits the inten-
sity and extent of a terrestrial survey, an
equation was developed (Glushkov 1999)
to calculate the probable distance to the
initial track encountered (L) from the length

of the route travelled where no tracks were
encountered (R0):

L ¼ 0.816R0þ2.98 ð1Þ

The histogram depicting the distribution of
theoretical frequencies of plots with various
moose densities indicated that the proportion
of plots with 2 animals was underestimated
4 times and that of unoccupied plots was
overestimated 2 times. In general, the equa-
tion to estimate population density (P)
from distance (X) had little practical value
(P = 9.17 − 0.54 X). The error in density
estimates was presumably due to insufficient
area in sample plots. A subsequent survey
(1999) was carried out on 14 routes with lar-
ger sample plots (700–2200 ha) at the end of
each route. The following describes this new
survey approach that provides more reliable
population density estimates.

RESULTS AND DISCUSSION
Figure 2 depicts the relationship between

moose population density (D) and the dis-
tance (x) from a residential area to the initial
(recent) track. The predictive equation was
formulated with a logarithmic density

Fig. 1. The relationship between the sight-
ing frequency and the number of observed
moose per plot in early (November) and
late (March) winter in the Kirov region,
Russia.

150

MOOSE POPULATION ESTIMATES IN RUSSIA – GLUSHKOV ALCES VOL. 49, 2013



function that is considered reasonable and
acceptable for sample estimates based on
the inequality criterion of dispersion and
the mean-square deviation of P (Draper and
Smith 1986).

D ¼ � 7.7023 lnðxÞ þ 20.635;
R2¼ 0.8981, P < 0.001 ð2Þ

A verification of this equation was
attempted in 2003–2004 within an experi-
mental hunting farm (63,000 ha) by compar-
ing the “known” moose population estimate
with an estimate derived from survey routes
and sample plots; the estimate was compar-
able and deemed satisfactory. However, this
comparison is general at best because there
was no differentiation between seasonal esti-
mates (early and late winter), and no method
to evaluate extrapolation across a larger area.
In an attempt to verify the method in prac-
tice, and to achieve necessary reduction of
the dispersion value, the number of paired
observations would need to increase to 40
based on equation (2) and the value of coef-
ficient of determination.

In general, the experimental estimates
were not contradictory of the hypothesis

that moose density is directly related to the
distance from a residential area due to their
anthropophobic behavior as a result of
intensive hunting. This new method is more
elementary and easier to implement at the
beginning of winter to estimate moose
density at both the district and regional scale.
Its use is intended for determining abun-
dance trends and setting seasonal harvest
quotas (Glushkov and Buldakov 1997).
Specification of the starting route point,
radial direction, and reference to a sample
plot at the end of the route removed some
associated drawbacks of the traditional
WRC method. The independence of the
“distance” parameter from weather condi-
tions increases not only accuracy but also
comparability of estimates.

A relatively even population distribution
in early winter predetermines reduction of
the estimate error, and defines the “native
population which inhabits a given area dur-
ing summer, autumn, and early winter and
is subject to hunting”, a definition critical
to determine harvest level. The estimate
makes it possible to determine, apart from
ecological density, an area that is actually
used by moose during early winter (extrapo-
lation area), and animal numbers at the dis-
trict and regional levels.

Comparability theory (Yurghenson
1970) can be used as the basis to extrapolate
population density estimates provided that
data are available in a particular region to
estimate density in subsequent years. It is
possible to use equation (2) initially while
simultaneously measuring and calculating
plot estimates to improve the population esti-
mate. If necessary, a locally specific equation
can be developed from a single estimate
from the plots and sample routes; calcula-
tions of the average R value and extrapola-
tion areas are provided in Glushkov (2001).

Application of this new method utilizes
GIS technology and requires preparatory
work to organize permanent estimation

Fig. 2. The relationship between moose
population density and the distance to
the first track sighting in the Kirov region,
Russia.

ALCES VOL. 49, 2013 GLUSHKOV – MOOSE POPULATION ESTIMATES IN RUSSIA

151



points and placement of routes and plots.
The area unused by animals and the extrapo-
lation area are determined with GIS technol-
ogy. An estimation point can be any
“standard” residential area – a village, a
workers’ settlement, or a farm enterprise
with ≥30 people; all are recorded in refer-
ence books, marked on maps, and have a
post office and permanent approach roads.
The principle criteria for selecting these
areas are that they are dead-end locations
on a year-round motorway and representa-
tive of the surveyed lands within the district.

The number of estimation points in a dis-
trict depends on the total area, % forest cover,
land cover diversity, and the number of set-
tlements and their distribution. Ideally, 4
radial routes with plots at the end must cover
the study area completely (see Fig. 3);
the inner circle corresponds to the anthropo-
genic zone with zero moose density. Moose
population density within the ring with
sample plots is equal to the density over
the whole habitation area (outer ring,

Fig. 3). In districts with large forest area
and lack of human settlements, the width
of the ring and the area of land used by ani-
mals (extrapolation area) can be corre-
spondingly large. In districts with small
fragmented forests and densely populated
settlements, the habitation area around
proximate human settlements decreases by
the value of the overlapping anthropogenic
area; i.e., the extrapolation declines.

The area standards for one estimation
point are 100,000 ha for districts with forest
cover >65%, and 60,000 ha otherwise. How-
ever, these standards may require a design
compromise due to conflict with statistical
requirements and the predetermined error
value of the estimation data. For example,
in densely populated districts with little for-
est cover, fragmented forests are often iso-
lated by farming lands, settlements, and
other man-made features. In this case, an
estimation point can be a settlement which
is located near a relatively big forest. The
width of the forest along the line to the near-
est settlement should be ≥2x the average dis-
tance to the first moose track; smaller forests
and forests located in anthropogenic zones
are not subject to estimation. The routes
from such estimation points should be
oriented into the forest not the cardinal direc-
tions (Fig. 4). Reducing the number of routes
and plots to 1–2 per estimation point requires
an adequate increase in the number of esti-
mation points.

Standard route lengths are necessary to
carry out the first estimate that is used for
further corrections (Table 1). The route
length is subsequently corrected from the
distances to the first track in the experimen-
tal estimates. A route is travelled one-fold,
once a year, preferably by vehicle. Choosing
the size and the shape of sample plots is par-
ticular to the size of compartments, config-
uration of forests, and availability of access
routes. It is best to use rectangular plots of

Fig. 3. The principle scheme for establishing 4
radial survey routes with sample plots to
estimate moose density in the Kirov region,
Russia. The center typically represents human
settlement with zero moose density; density is
extrapolated for the area of concentric rings.

152

MOOSE POPULATION ESTIMATES IN RUSSIA – GLUSHKOV ALCES VOL. 49, 2013



2x4 dimension or plots of other shapes in
areas >800 ha (Agafonov et al. 1988).

The new method of field data collection
and subsequent calculation of moose popula-
tion estimations described here will provide
more reliable population estimates than
with previous approaches. This is critical in

the Kirov region of Russia that has experi-
enced population overestimates and subse-
quent overharvest of moose. A coordinated
strategy of using better population estimates,
and measuring calf survival and non-harvest
mortality, including poaching, will benefit
regional moose management in Russia
(Glushkov 2009).

REFERENCES

AGAFONOV, V. A., S. A. KORYTIN, and I. N.
SOLOMIN. 1988. Winter estimate of
game animals on circular routes: rational
methods of game animals study. Pages
17–25 in Russian State Research Institute
of Game Management and Fur Farming,
Kirov, Russia. (In Russian.)

BUBENIK, A. B. 1965. Population Density of
Game Animals, Feed Capacity of Game
Habitat, and Forest Damage Caused by
Animals. Moose Biology and Hunting
2: 265–280. Moscow, Russia. (In
Russian.)

DRAPER, N., and H. SMITH. 1986. Applied
Regression Analysis. Mir, Moscow, Rus-
sia. (Translated from 1981 English
version.)

GLUSHKOV, V. M. 1977. On the Method of
Arial Moose Count. Hunting and Game
Management 12: 14–15. (In Russian.)

———. 1982. The Moose of Vyatka Forests.
Hunting and Game Management 1: 16–
18. (In Russian.)

———. 1997. Extrapolation Area Estimates
for Moose and Wild Boar. Pages 81-83
in Applied Ecology Issues. Proceedings
of Scientific Conference Devoted to 75th

Anniversary of Russian State Research
Institute of Game Management and Fur
Farming. Kirov, Russia. (In Russian.)

———. 1999. Moose. Pages 117–163 in
Management of Game Animals. Proceed-
ings of Russian State Research, Institute
of Game Management and Fur Farming,
Russian Academy of Agricultural
Sciences. Kirov, Russia. (In Russian.)

———. 2001. Moose: Ecology and Manage-
ment of Populations. Russian Academy

Fig. 4. A depiction of how survey routes are
distributed in irregular fragmented forests;
forest width to the nearest settlement should
be >2x the average distance to the first moose
track. Multiple estimation points may be
required to achieve a sufficient number of
survey routes in an area.

Table 1. The stand and length of radial routes
required in varying proportions of forest cover to
estimate moose population density in the Kirov
region, Russia.

% Forest Cover Route Length (km)

80–100 12
70–79 10
55–69 8
40–54 7
25–39 5
<25 4

ALCES VOL. 49, 2013 GLUSHKOV – MOOSE POPULATION ESTIMATES IN RUSSIA

153



of Agricultural Sciences, Kirov, Russia.
(In Russian.)

———. 2004. On route standardization for
moose estimates. Game Management
Bulletin: 195–200. (In Russian.)

———. 2009. Improving population man-
agement and harvest quotas of moose in
Russia. Alces 45: 43–47.

———, and A. L. BULDAKOV. 1997. Princi-
ples of ungulate estimate with the com-
bined method. Pages 83–85 in Issues of
Applied Ecology, Institute of Game
Management and Fur Farming. Proceed-
ings of Scientific Conference Devoted
to 75th Anniversary of Russian State
Research Institute of Game Management
and Fur Farming, Kirov, Russia. (In
Russian.)

———, M. G. DVORNIKOV, V. V. KOLESNI-
KOV, V. G. SAFONOV, A. A. SERGEYEV,
M. S. SHEVNINA, and V. V. SHIRYAEV.
2012. Factors obstructing wild ungulate
management in Russia. Pages 76–83 in
Theoretical and Applied Ecology. (In
Russian.)

LING, H. 1977. Experience of System Analy-
sis of Population Structure and
Dynamics. Pages 7–105 in Proceedings
of Tartu University, Issue 408. University
of Tartu, Tartu, Estonia.

NAUMOV, N. P. 1963. Animal Ecology. Mos-
cow, Russia. (In Russian.).

ODUM, E. 1986. Ecology. Volume 2. Mos-
cow, Russia. (Translated from English.)

SMIRNOV, V. S. 1964. Mammal Estimation
Methods. Proceedings of The Biology
Institute of Ural Academy of Sciences.
(In Russian.)

YAZAN, P. 1972. Game Animals of Pechora
Taiga. Research Institute of Game Man-
agement and Fur Farming. Kirov, Russia.
(In Russian).

YURGHENSON, P. B. 1970. Distribution The-
ory, Spatial Analysis, and Applied Ecol-
ogy of Animals. Bulletin of Moscow
Society of Nature Explorers, Biology
Department, Moscow, Russia. (In
Russian.)

154

MOOSE POPULATION ESTIMATES IN RUSSIA – GLUSHKOV ALCES VOL. 49, 2013


	IMPROVING MOOSE POPULATION ESTIMATES IN RUSSIA: ACCOUNTING FOR DISTANCE BETWEEN RESIDENTIAL AREAS AND TRACK SIGHTINGS
	INTRODUCTION & BACKGROUND
	RESULTS AND DISCUSSION
	REFERENCES