Ecological Modelling 175 (2004) 115–120

Modeling emigration of wolves from a wilderness area into

adjacent agricultural regions
A.L. Jensen∗ , D.H. Miller
School of Natural Resources and Environment, University of Michigan, Ann Arbor, MI 48109-1115, USA
Received 26 November 2002; received in revised form 26 June 2003; accepted 11 July 2003

Abstract
We developed and applied a simple population model to examine the relation between abundance of wolves in a wilderness
area and the numbers that emigrate into adjacent agricultural areas and that may need to be removed on an annual basis. The
model was applied using Minnesota wolf (Canis lupus) data. The gray wolf is emigrating from northern wilderness areas in
the State of Minnesota (USA) into adjacent agricultural and urban areas to the south, and the costs of both wolf control and
compensation to farmers for lost livestock is increasing as the number of wolves increases. Emigration reduces the number of
wolves on a refuge to about 85% of the carrying capacity, and the number of wolves that emigrate into the agricultural area is
6% of the number on the refuge. With control on the refuge the number of wolves killed is about 10% of the carrying capacity or
24% of the wolves on the refuge, but control on the refuge results in less emigration. A conservative control strategy, in which
abundance on the refuge is closer to the carrying capacity and emigration increases slightly, can increase the number of wolves
on the refuge with a small increase in emigration.
© 2003 Elsevier B.V. All rights reserved.
Keywords: Gray wolf; Predator dispersal; Control; Logistic; Livestock depredation

1. Introduction populations resulting from their social structure such

In this study, we applied a simple population model
to examine the relation between abundance of wolves
in a wilderness area and the number that move into ad-
jacent agricultural and suburban areas and that might
need to be removed on an annual basis. We also ex-
amined the possibility of reducing predator emigra-
tion into the agricultural area by controlling the num-
ber in the wilderness area. The model facilitates anal-
ysis of the results of predator control and describes
the outcome of different levels of control but it does
not consider possible effects of removals on predator
∗ Corresponding author. Tel.: +1-734-763-6280.
E-mail address: ajensen@umich.edu (A.L. Jensen).
as those described for wolves by Haber (1996).
The model was applied to the gray wolf (Canis
lupus) population in the northern wilderness area of
the State of Minnesota (USA) which now numbers
between 2500 and 3000. Wolves are emigrating from
the wilderness areas into adjacent Minnesota agri-
cultural areas despite removal of a large number of
wolves each year. About 160 wolves were killed in
1998 and the cost of control and compensation to
farmers was US$ 350,000 (Mech, 1998). The wolf
population in Northern Minnesota has been well stud-
ied (e.g., Mech, 1970; Van Ballenberghe et al., 1975;
Fuller, 1989), and plans have been proposed for con-
trol that consider number of wolves, damage done by
wolves, and cost, and the plans require annual removal

0304-3800/$ – see front matter © 2003 Elsevier B.V. All rights reserved.
doi:10.1016/j.ecolmodel.2003.07.008
116 A.L. Jensen, D.H. Miller / Ecological Modelling 175 (2004) 115–120
of a considerable number of wolves (e.g., Mech,
1998).
2. Development of model
Complex age structured models are available for
wolf populations (e.g., Jensen and Miller, 2001;
Miller et al., 2002), but these models require estima-
tion of many parameters for which there is limited
data. The logistic equation is the simplest population
model that describes the general pattern of animal
population growth (e.g., Volterra, 1926; Krebs, 1985),
and the logistic equation was applied in this study.
In the logistic model population growth is assumed
to be density dependent and the rate of growth is
assumed to be proportional to the level of environ-
mental saturation. The source of density dependence
is not identified and it could be a reproduction func-
tion, food limited growth, predation, disease, or some
other factor. Suppose that a predator population on a
refuge follows the logistic equation
dN rN2
= rN − , (1)
dt K rN − rN2
where N is abundance at time t, K is the environmental
carrying capacity of the refuge, and r is the maximum
intrinsic rate of increase. We assumed that emigration
of predators was density independent and therefore
simply proportional to abundance (Fuller, 1989; Hayes
and Harestad, 2000). We also assumed that predators
that emigrate from the refuge do not return. Adding
emigration to the logistic equation gives the population
growth equation for predators in the refuge as
dN rN − rN2
= , (2)
dt K − hN
where h is an emigration coefficient. Eq. (2) will be
applied to determine the number of predators in the
refuge at equilibrium and the numbers that annually
emigrate to the agricultural area around the refuge.
Solution of Eq. (2) for N when dN/dt = 0, and em-
igration out of the refuge is at equilibrium with growth
of the population in the refuge, gives the number of
predators on the refuge at equilibrium, N∗ , as
(r − h)K
N∗ = . (3)
r
Substitution of Eq. (3) into Eq. (2) gives the annual
emigration out of the refuge at equilibrium E∗ as
h(r − h)K
E∗ = . (4)
r
The number of predators that emigrate out of the
refuge is the number that might need to be removed
annually in the agricultural areas.
It might be more efficient to control emigration by
controlling the number of predators on the refuge than
by dealing with emigrant predators, and the rate of em-
igration will be a minimum when the rate of removal
of predators on the refuge is a maximum. To include
control of predators on the refuge in the model, a term
for predator removal was added to give
dN rN − rN2
= , (5)
dt K − hN − QN
where Q is a variable that measures the removal ef-
fort. Mathematically, Eq. (5) is similar to the surplus
production model applied for fisheries assessments
(e.g., Schaefer, 1954; Jensen, 1976). At equilibrium,
the number of predators removed annually from the
refuge Re is
Re = . (6)
K − hN
As the level of control increases the abundance of
predators on the refuge and emigration both decrease,
but the number of predators removed increases to a
maximum and then decreases. Application of calculus
to Eq. (6) shows that the maximum sustainable rate of
predator removal, MSR, and abundance of the popu-
lation at the MSR, NMSR , are
(r − h)2 K
MSR = , and (7)
4r
(r − h)K
NMSR = . (8)
2r
The emigration rate with the maximum sustainable
rate of predator removal is EMSR = hNMSR . With
removals on the refuge the total number of predators
removed is MSR + EMSR .
In fisheries, where the concept of the maximum
sustainable yield MSY has been widely applied, it has
been found that harvesting at the MSY is not sustain-
able. Harvesting at the MSY tends to be over harvest
fisheries and even small increases in harvest above
 A.L. Jensen, D.H. Miller / Ecological Modelling 175 (2004) 115–120 117

the MSY may result in destruction of a fishery (e.g., effort as



Caddy and Mahon, 1995). Therefore, conservative 1−h K
levels of harvest have been developed to protect fish- Re = K Q− Q2 , (10)
r r
eries (e.g., Caddy and Mahon, 1995). In fisheries sev-
eral different concepts have been applied to develop which is a quadratic equation in Q. From the equations
conservative harvest levels (e.g., Caddy and Mahon, for MSR, Ne , and Re (Eqs. (7), (9) and (10)) it can be
1995). Predator control confronts the two contradic- shown that
tory objectives of maintaining the predator population Re 4Q(r − h − Q)
on the refuge near the carrying capacity for long-term = and (11)
MSR (r − h)2
survival of the predator, and reduction of the preda-
tor population to reduce emigration and the damage Ne (r − h − Q)
= . (12)
caused by emigrants. We developed a target level for K r
removal that maintains population size on the refuge Eqs. (11) and (12) indicate that as Q decreases be-
as close to the carrying capacity as possible while at low (r − h)/2, Ne /K increases along a straight line
the same time reduces emigration as much as possible (Fig. 1) and Re /MSR decreases slowly near the top of
by keeping the numbers removed from the refuge as a parabola (Fig. 1). At first, Ne increases faster than
close to the MSR as possible. Re decreases and when
At equilibrium, the relation between the removal
effort Q, the annual number removed Re , and abun- d(Re /MSR) d(Ne /K)
=− , (13)
dance is dQ dQ
Re
Ne = , (9) Re is decreasing below the MSR at the same rate that
Q Ne is increasing and at this level of effort emigration is
and substitution of Ne for N in Eq. (6) gives the an- minimized by removals while at the same time popu-
nual number removed as a function of the removal lation size is as close to the carrying capacity as possi-

Fig. 1. For the wolf in Northern Minnesota the line is the ratio Ne /K and the parabola is the ratio Re /MSR.
118 A.L. Jensen, D.H. Miller / Ecological Modelling 175 (2004) 115–120
ble. Differentiation of Eqs. (11) and (12) and solution
for Q gives
(r − h)(3r + h)
Q= , (14)
8r
which is the removal effort when population size is as
close to the carrying capacity as possible and at the
same time the emigration rate is as small as possible.
This will be termed the conservative level of removal.
Eq. (14) gives the conservative sustainable removal,
CSR = QNCSR , and abundance at the conservative
sustainable removals, NCSR , (substitution of Eq. (14)
into Eqs. (9) and (10)) as
(r − h)(3r + h)
CSR = NCSR , and (15)
8r
K(r − h)(3r + h) Kh
NCSR = K − − . (16)
8r 2 r Wisconsin and Michigan and were not killed (Mech
The annual emigration rate is ECSR = hNCSR and the
total number removed annually is CSR + ECSR .
3. Results and discussion
The model was applied to the wolf in Northern
Minnesota using data in the literature for estimation
of parameters (e.g., Fuller et al., 1992). The maxi-
mum intrinsic rate of natural increase r is the rate of
increase at small population abundance and r is dif-
ficult to estimate for field populations. Keith (1983)
used data on wolf populations that were increasing in
size to estimate exponential rates of increase of 0.14
and 0.33 per year for Isle Royal, 0.18 per year for Al-
gonquin Park, 0.19 per year in Alberta, 0.27 per year
in Minnesota, 0.29 per year in the Anaktuvak Pass,
Alaska, and 0.38 per year in the Peace River District
of Alberta. These estimates of r based on field studies
almost certainly underestimate the maximum value of
r. Several studies of animals have reported relations
between r and weight (e.g., Fenchnel, 1974; Blueweiss
et al., 1978), and the weight of the wolf is well
known (e.g., Mech, 1970). Fenchnel (1974) showed
that under primarily laboratory conditions r has an
−0.25 scaling with adult body weight, and Blueweiss
et al. (1978) estimated that for primarily laboratory
conditions the relation between body size (kilograms)
and the intrinsic rate of increase was r = 1.6W −0.26
per year. With an average adult weight of 32.5 kg
for male and female wolves (Mech, 1970), this gives
r = 0.64 per year, which is almost certainly an over
estimate of r. Charnov (1993) gives the relation be-
tween adult weight (kilograms) and the intrinsic rate
of increase for mammals under field conditions as r =
0.6W −0.25 , and this gives r = 0.42 per year for the
wolf. We assumed that the estimate with Charnov’s
(1993) equation, r = 0.42 per year, which is between
that of the field and laboratory estimates is the most
reasonable.
The emigration parameter h was estimated with data
reported by Mech (1998). Mech (1998) concluded that
for Northern Minnesota in the year 1998 there were
2520 wolves on the refuge and the number killed in the
agricultural area was 161 and this gives h = E/N =
161/2520 = 0.0639 per year. This is an underestimate
because an unknown number of wolves migrated to
et al., 1995), and some wolves that migrated into areas
in Minnesota may not have killed livestock.
The number of wolves on the refuge and the num-
ber that would emigrate from the refuge into the
agricultural area as the carrying capacity of the wolf
population increased from 3000 to 7000 were deter-
mined with Eqs. (4) and (5) together with the above
estimates of r and h. The number of wolves on the
refuge was always about 85% of the carrying capacity
without removals from the refuge, and the number of
wolves that emigrated into the agricultural area and
that probably would be killed was 6% of the number
on the refuge. Removal of wolves from the refuge
considerably reduced the rate of wolf emigration from
the refuge into the agricultural area. At the maximum
sustainable removal rates abundances on the refuge
were somewhat less than 42% of the carrying capaci-
ties and the maximum sustainable removal rates were
about 21% per year of the numbers on the refuge.
The conservative removal rate increased the number
of wolves on the refuge about 20%, decreased the
number killed on the refuge 4.5%, but increased the
annual emigration about 20%.
If wolves were not removed from the refuge, the
number of wolves on the refuge did not attain the car-
rying capacity because the refuge was open and wolves
emigrated from the refuge as abundance approached
the carrying capacity. The number of wolves on the
refuge was always about 85% of the carrying capac-
ity, and the number of wolves that emigrated into the
 A.L. Jensen, D.H. Miller / Ecological Modelling 175 (2004) 115–120
agricultural area and that probably would be killed
was 6% of the number on the refuge. With a carry-
ing capacity of 7000 wolves, the number of wolves
in the simulations that emigrated was 378 and these
wolves could do considerable damage to livestock.
Removal of wolves from the refuge considerably re-
duced the rate of wolf emigration from the refuge into
the agricultural area, but the total number of wolves
killed was substantially higher when the maximum
sustainable number was removed from the refuge than
when just emigrating wolves were killed. Removal of
wolves from the refuge reduced the rate of wolf emi-
gration from the refuge into the agricultural area but
did not eliminate emigration. Wolves that emigrate
also probably would be killed, so the total number of
wolves killed with control on the refuge was about
10% of the carrying capacity or 24% of the wolves
on the refuge. However, with control of wolves on the
refuge a carrying capacity of 7000 wolves gives an
abundance of 2970 wolves on the refuge, which are
many more wolves than occurs with a carrying ca-
pacity of 3000 and no control. Emigration with con-
trol is only 94 which is much less than 162 that em-
igrate with no control. Therefore, while control on
the refuge results in the killing of a larger number of
wolves, it can also result in a larger population on the
refuge with a smaller emigration and smaller level of
damage.
The maximum sustainable rate of removal is sus-
tainable only in a deterministic environment, and it
is lower in a stochastic environment (Beddington and
May, 1977). If the removal rate were only slightly
higher than the MSR, the population could become
endangered. To protect the wolf population a conser-
vative level of removal could be applied on the refuge.
Several concepts for conservative harvest have been
developed in fisheries (e.g., Caddy and Mahon, 1995);
for control of predators on a refuge it is desirable that
abundance on the refuge be maintained near the car-
rying capacity while at the same time the numbers
emigrating be as small as possible. We developed a
target level for removals that maintains abundance on
the refuge as close to the carrying capacity as pos-
sible while at the same time minimizing the number
emigrating. Conservative removal of wolves from the
refuge increased both the number of wolves on the
refuge and emigration, but it only slightly increased
the number of wolves killed. The total number of
119
wolves killed with a conservative level of removal in-
creases slightly over the number killed at the MSR
because although the number killed on the refuge is
lower, more wolves emigrate.
Although a conservative level of removal on the
refuge will increase the safety of the predator, uncer-
tainty of abundance estimates on the refuge, errors in
parameter estimation, and environmental variation re-
quire that a conservative approach include a plan for
continual update of abundance estimates, estimates of
model parameters, the MSR, and Re .
In the models developed here, the population dy-
namics of the wolf were represented simply in terms
of a carrying capacity and a rate of increase, and the
social organization of the wolf was ignored. More
complex models easily can be constructed (e.g.,
Jensen, 2000; Starfield and Bleloch, 1991), but often
simpler models perform better than more complex
models (e.g., Ludwig and Walters, 1985). Haber
(1996) contends that control of wolf populations may
disrupt social patterns, and that unlike ungulate pop-
ulations wolf populations may not be able to sustain
a high level of removal. Lehman et al. (1992) found
close genetic connections among members of dif-
ferent wolf packs both in Minnesota and in Denali
National Park, Alaska, where turnover of packs was
slow. In the Inuvik Region of the Northwest Territo-
ries, Canada, where rapid turnover occurred due to
hunting of wolves, genetic connections between packs
were lower. Several packs may be removed at once
by hunting, creating an opportunity for invasion by
young dispersing wolves from outside the area. These
young dispersing wolves are likely to have a lower
kinship with individuals from pre-existing packs, and
if experience and long-term residency are important
to the persistence of wolves then newly established
packs may have a lower survivorship and reproduc-
tive success than established packs. Thus, if control is
applied to wolves on the refuge, the population must
be monitored to assure its continued viability and to
continually update parameter estimates.
Haber (1996) has argued that control of wolf popu-
lations may not be necessary at all, and that the wolf
social organization represents an adaptation for popu-
lation self regulation, but wolf populations have a ca-
pacity to increase and to disperse, and have expanded
into agricultural areas of Minnesota even with sub-
stantial control (Mech, 1998). Dispersing wolves are
120 A.L. Jensen, D.H. Miller / Ecological Modelling 175 (2004) 115–120
difficult to census, but they may account for 5–20% of
the wolves in established populations during the win-
ter (Fuller, 1989). The observed spread of wolves from
the wilderness areas of Northern Minnesota into the
adjacent agricultural areas and into the adjacent states
of Wisconsin and Michigan clearly indicates that the
model adequately describes the tendency of wolves to
emigrate into surrounding areas. Mech (1998) has in-
dicated that if the wolf in Minnesota is not controlled
soon control might become difficult.
References
Beddington, J.R., May, R.M., 1977. Harvesting natural populations
in a randomly fluctuating environment. Science 197, 463–
465.
Blueweiss, L., Fox, H., Kudzma, V., Nakashima, D., Peters, R.,
Sams, S., 1978. Relationship between body size and some life
history parameters. Oecologia 37, 257–272.
Caddy, J.F., Mahon, R., 1995. Reference Points for Fisheries
Management. FAO Fisheries Technical Paper 347.
Charnov E.L., 1993. Life History Invariants. Oxford University
Press, New York.
Fenchnel, T., 1974. Intrinsic rate of natural increase: the
relationship with body size. Oecologia 14, 317–326.
Fuller, T.K., 1989. Population dynamics of wolves in north-central
Minnesota. Wildl. Monogr. 105, 1–41.
Fuller, T.K., Berg, W.E., Raddle, G.L., Lenarz, M.S., Joselyn,
G.B., 1992. A history and current estimate of wolf distribution
and numbers in Minnesota. Wildl. Soc. Bull. 20, 12–45.
Haber, G.C., 1996. Biological, conservation, ethical implications
of exploiting and controlling wolves. Conserv. Biol. 10, 1068–
1081.
Hayes, R.D., Harestad, A.S., 2000. Demography of a recovering
wolf population in the Yukon. Can. J. Zool. 78, 36–48.
Jensen, A.L., 1976. Assessment of the United States lake whitefish
of Lake Superior, Lake Michigan, Lake Huron. J. Fish. Res.
Board Can. 33, 747–759.
Jensen, A.L., 2000. Sex and age structured matrix model applied
to harvesting a white tailed deer population. Ecol. Model. 128,
245–249.
Jensen, A.L., Miller, D.H., 2001. Age structured matrix predation
model for the dynamics of wolf and deer populations. Ecol.
Model. 141, 299–305.
Keith, L.B., 1983. Population dynamics of wolves. In: Carbyn,
L.N. (Ed.), Wolves in Canada and Alaska. Report Series 45,
Canadian Wildlife Service, Ottawa.
Krebs, C.J., 1985. Ecology: the Experimental Analysis of
Distribution and Abundance. 3rd edition. Harper Collins, New
York.
Lehman, N., Clarkson, P., Mech, L.D., Meier, T.J., Wayne, R.K.,
1992. The use of DNA fingerprinting and mitochondrial DNA
to study genetic relationships within and among wolf packs.
Behav. Ecol. Sociobiol. 30, 83–94.
Ludwig, D., Walters, C.J., 1985. Are age-structured models
appropriate for catch-effort data? Can. J. Fish. Aquat. Sci. 42,
1066–1072.
Mech, L.D., 1970. The Wolf. University of Minnesota Press,
Minneapolis, MN.
Mech, L.D., 1998. Estimated costs of maintaining a recovered
wolf population in agricultural regions of Minnesota. Wildl.
Soc. Bull. 26, 817–822.
Mech, L.D., Fritts, S.H., Wagner, D., 1995. Minnesota wolf
dispersal to Wisconsin and Michigan. Am. Midland Natural.
105, 408–409.
Miller, D.H., Jensen, A.L., Hammill, J.H., 2002. Density dependent
matrix model for gray wolf population projection. Ecol. Model.
151, 271–278.
Schaefer, M.B., 1954. Some aspects of the dynamics of populations
important to the management of the commercial marine
fisheries. Bull. Inter-Am. Tropic. Tuna Commission 1, 27–56.
Starfield, A.M., Bleloch, A.L., 1991. Building Models for
Conservation and Wildlife Management. Burgess, Edina, MN,
USA.
Van Ballenberghe, V., Erickson, A.W., Byman, D., 1975. Ecology
of the timber wolf in northeastern Minnesota. Wildl. Monogr.
43, 1–43.
Volterra, V., 1926. Fluctuations in the abundance of a species
considered mathematically. Nature 118, 558–560.