Professional Documents
Culture Documents
1986 - Random Walk Models of Movement
1986 - Random Walk Models of Movement
1986 - Random Walk Models of Movement
I. Introduction
Biologists long have sought quantitative models to describe the process of
dispersal : to aid understanding, to guide experimentation, and to facilitate
predict ion. The most common such models are of the random walk type, deriving
from the assumption that individuals move in a series of discrete steps with
probabilities tot ally determined by positional information. Learning is ignored .
The simplest random walk model may be motivated by the following
illustration. Let an organism be located at position 0 at time 0, as in Fig. 1. Assume
further that at discrete times kt, the organism jumps either forward or backward A.
units,
o + A. + 2). + 3A.
and that either event has probability 1/2. All ofthese assumptions may be relaxed,
but will be observed here to make the presentation clearer.
If m and n both are even integers, what is the probability that at time nt the
organism is at position mAo (after its latest jump)? This is just the probability that
(n + m)/2 forward steps have been taken, and (n - m)/2 backward steps, and thus is
given by the appropriate term of the Bernoulli (binomial) distribution :
(1)
Ce -
m2{ 2 n
(2)
where
c=~ .
Biomathemat ics, Vol. 17, Mathem atical Ecology
Edited by T. G . Hallam and S. A. Levin
Sprin ger-Verla g Berlin Heidelber g 1986
150
S. A. Levin
2r)
(3)
C exp ( - 4t . .,1.2
This tends to the limiting form
t)
2
XD
P(x , = C exp ( - 4
t),
where
C= . 1
2VlIDt '
(4)
provided A and r both are allowed to shrink to zero in such a way that the limit
.,1.2
lim -
,t, <-+0
=2D
(5)
exists. Similar forms also can be obtained in higher dimensions (see Okubo, 1980;
Lin and Segel, 1974). D is termed the diffusion coefficient.
(4) represents the approximate limiting distribution to be expected when
individuals spread according to the random walk model proposed above. As
Okubo (1980) observes, the limiting approximation is valid only when "the time of
observation t is much greater than the duration time r of each random step , and
when the scale of observation x is much greater than the length of each random
step." In other words, the use of the diffusion approximation is justified only on
scales that invol ve a great many individual step s. The limiting form (4) can be
derived under more general conditions : it does not depend, for example, on the
assumption that each individual moves at each time step .
Note that the population described by (4) alwa ys is normally distributed (for
t > 0), and has a variance (2Dt) which increases linearly with time. Note further (by
direct differentiation) that at an y point x, the rate of change of population density
is proportional to the second derivative of population density with respect to x;
that is,
(6)
This equation is known as the diffusion equation (or the heat equation), and also
could be derived from first principles. It describes the spre ad of a diffusing
population with any distribution (not just the normal). Moreover, the property
that the variance increases linearly with time at the rate 2D also is true in general.
Let N be the total population size; that is,
00
N= S P(x,t)dx.
(7)
-00
Then
aN = j ap dx= j D a2~dX =D ap I 00 .
- 00 ax
ax - 00
at - 00 at
(8)
lSI
00
m=
J xP(x, t)dxIN.
(9)
- 00
By application of the chain rule, we see that m also is a constant, provided P and
iJPliJx vanish sufficiently rapidly at 00 . Similarly, if the variance V is defined by
V=
00
J (x-m)2p(x, t)dxIN,
(10)
- 00
~ =2D.
(11)
Thus, in general,
(12)
Property (12), that the variance increases linearly with time, is a characteristic
of the simplest (constant coefficient) model of dispersal. Several researchers (e.g.
Kareiva, 1983)have used this property to test the adequacy of the simplest model.
For data on the foraging movements of phytophagous insects, Kareiva estimated
D from the slope of the regression of V on t. He then used this estimate of D to
generate a series of probability distributions for the spread of the insects, and
compared it with what was actually observed. He found that the agreement was
excellent in most cases, but that in some cases a habitat-dependent diffusion model,
iJP
iJt
iJ2(D(x)P)
iJx 2
(13)
iJ2P
at =D iJx 2 +F(P,x, t) ,
(14)
where F(P, x, t) is the local population growth rate. In the next section, I shall
discuss some uses of this equation and its higher dimensional versions in studying
the rates of spread of species entering new habitats. In a later chapter, the
152
S. A. Levin
ap
a2p
at =D ax2 + f(P) ,
(16)
where
f(O)=f(l)=O ,
f>O on (0,1)
(17)
and
1'(0) >
rep) on [0, 1] .
(18)
c>O.
(19)
Kolmogorov et al. showed that there exist monotone wave solutions for all wave
speeds c greater than or equal to the critical value
c* =2VD. 1'(0) ,
(20)
153
and none for c < c*. They showed further that for an initial Heaviside distribution,
I for
Po(x)= { 0 for
x<O
x>O,
(21)
154
References
Aronson, D .G., Weinberger, H.F . (1975). Nonlinear diffusion in population genetics, combustion,
and nerve propagation. pp. 5-49, In : J. Goldstein (ed.), Partial Differential Equations and
Related Topics. Lecture Notes in Mathematics 445. Springer-Verlag, Heidelberg
Bramson, M. (1983). Convergence of solut ions of the Kolmogorov equation to travelling waves.
Mem. Amer. Math. Soc. # 285
Fife, P .C. (1984). Current topics in reaction-diffusion systems . In: M. G . Velarde (ed.),Proceedings
of NATO Conference on Nonequilibrium Phenomena in Physics and Related Fields. Plenum
Fisher, RA. (1937). The wave ofadvance ofadvantageous genes. Ann . Eugen . London 7:355-369
Hadeler, K.P. (1976). Nonlinear diffusion equations in biology . In: W. N . Everett and B. D.
Sleeman (eds.), Ordinary and Partial Differential Equations. Lect. Notes in Mathematics 564,
Springer-Verlag, Heidelberg
Hadeler, K.P. (1984). Spread and age structure in epidemic models in Perspectives in Mathematics . Anniversary ofOberwolfach, 1984. pp . 295-320. Birkhauser-Verlag, Basel
Hadeler, K .P., Rothe, E. (1975). Travelling fronts in nonlinear diffusion equations. J. Math.
BioI. 2:251-263
Hoppensteadt, F . (1975). Mathematical Theories of Populations : Demographics, Genetics and
Epidemics . SIAM Reg. Conf. Series 20, Philadelphia
Kareiva, P. (1983). Local movement in herbivorous insects: applying a passive diffusion model
to mark-recapture field experiments. Oecologia 57:322-327
Kendall, D .G. (1965). Mathematical models of the spread of infection . Mathematics and
Computer Science in Biology and Medicine, London H.M.S.O. pp . 213-225
Kolmogorov, A., Petrovskij, I., Piskunov, N. (1937). Etude de l'equation de la diffusion avec
croissance de la quantite de la matiere et son application a un probleme biologique . Bull.
Univ. Moscou Ser. Intemation., Sec. A, I (6) 1-25
Lin, C.C., Segel, LA. (1974). Mathematics Applied to Deterministic Problems in The Natural
Sciences. MacMillan, New York. 604 + xix pages
Okubo, A. (1980). Diffusion and Ecological Problems : Mathematical Models. Biomathematics 10,
Springer-Verlag, New York
Skellam, J.G. (1951). Random dispersal in theoretical populations. Biometrika 38 : 196-218