Está en la página 1de 10

Ecological Modelling 117 (1999) 305 – 314

Mastering chaos in ecology

Inti Suárez *
Zoologisches Institut der Uni6ersität Basel, Rheinsprung 9, CH-4051 Basel, Switzerland

Received 3 November 1997; accepted 15 December 1998


Recent developments in dynamical system theory consider chaotic fluctuations of a dynamical system as highly
desirable behaviour because fluctuations allow such a system to be easily controlled. This recent development, chaos
control theory, has been tested in several physical situations. The intrinsic instability of the orbits embedded in a
strange attractor make them easy to change to a more stable and predictable behaviour. The existence of chaos in
natural populations, at least in some species of pest insects, makes the topic of importance in pest management. The
techniques proposed to achieve control over a chaotic dynamical system are presented and discussed in an ecological
context. An important point in the application of this technique to real systems is the amount of noise involved in
the estimation of the parameters of the system. It is also important in the determination of the perturbations required
in the control parameters of the system. The theoretical possibility of obtaining these estimations in field population
biology provides opportunities for using these techniques in pest management. The control of chaos theory shows
how chaotic systems can be controlled by stable and linked systems. The possibility that chaotic behaviour has been
seldom found in nature because of the existence of linked phenomena is discussed. © 1999 Elsevier Science B.V. All
rights reserved.

Keywords: Chaos control; Chaos detection; Pest control; Complex population dynamics

1. Introduction previously as bewildering behaviour, can be desir-

able and advantageous since it has been demon-
Amidst the recent advances in the study of strated that a system sensitive to initial conditions
is highly controllable (Ditto and Pecora, 1993).
chaos, the algorithm developed by Ott et al.
This idea has been explored in an evolutionary
(1990a) has become one of the seminal papers in
context by Doebeli (Doebeli, 1993, 1995a,b) and
the area. It has been said that chaos, considered
in medicine by Weis et al. (1993). The object of
this communication is to explore the conse-
* Tel.: +41-61-2673486; fax: + 41-61-2673457. quences of these developments in population ecol-
E-mail address: (I. Suárez) ogy, starting from the uses in other contexts.

0304-3800/99/$ - see front matter © 1999 Elsevier Science B.V. All rights reserved.
PII: S 0 3 0 4 - 3 8 0 0 ( 9 9 ) 0 0 0 0 7 - 1
306 I. Suárez / Ecological Modelling 117 (1999) 305–314

2. Proposed techniques seem to be more robust to noise (Meucci et al.,

1994). They will be discussed later.
The central idea of the method proposed by Ott
et al. (1990a) is to stabilize one of the orbits of the
system’s attractor by making a series of small 3. Chaos control and pest management
perturbations on a control parameter. This is
possible because there is an infinite number of Although the techniques of chaos detection
unstable periodic orbits densely embedded within have been improved (Schaffer, 1984; Sugihara and
the attractor. The proposed algorithm, in general, May, 1990; Ellner and Turchin, 1995), the exis-
is as follows: tence of chaos in natural populations is still con-
1. The system has to be studied through Poincarè troversial, and several reviews about it have been
sections in order to choose a desirable orbit to written (e.g. Hasting et al., 1993). Following
be stabilized. Pimm (1984) the first attempt to detect chaotic
2. With the reconstruction of the attractor in behaviour in nature was important in shifting the
phase space, the direction the system will fol- opinion of ecologists against its existence (Hassell
low in the presence of a determined change in et al., 1976). Nevertheless Hassell et al. (1976) did
the control parameter has to be predicted, find evidence of chaotic behaviour in a pest insect
using the techniques proposed by Lathrop and population, Zeirapthera fagi (Nicholson’s
Kostelich (1989). blowfly). There is some more evidence that it
3. The change in the control parameter that could occur in pest populations (e.g. Phyllaphis
tends to stabilize the desired orbit has to be fagi, Turchin and Taylor, 1992). Then, consider-
made, monitoring the result in phase space, to ing that chaos exists in at least some pest insects,
plan the next change. would it be possible to control these populations
It has been demonstrated than the fast conver- through the control techniques mentioned above?
gence of this algorithm to the desired orbit is a This is the first question that the author intends to
consequence of the instability of the orbits in a answer.
chaotic system (Ott et al., 1990a) and this has Thinking about the applicability of the tech-
been validated in many different experimental sit- niques presented above, one that needs a change
uations (Bayli and Virgin, 1994). in a dynamical variable of the system can be
There are some variations in this method, as compared with a planned harvesting of the popu-
follows: lation (M. Doebeli, personal communication). For
1. Change a dynamic variable from the system, example, to change a dynamical variable in the
not a control parameter from the system. (De- system could be to change the population size as
layed controlling feedback, Pyragas, 1992; dis- convenient (Guemez and Matias, 1993). Because
sipative feedback control, Rulkov et al., 1994). the use of this technique is studied elsewhere
This group of techniques presents less practical (Doebeli and Ruxton, 1997) it will be not dis-
complications than the original one. cussed here. In an ecological context, the main
2. Disturb the control parameter in a periodic advantage of this technique is not needed, because
way, e.g. sinusoidal, (Lima and Pettini, 1990), computer analysis in real time is not a problem,
or with the perturbations in reference to a since the frequency in an ecological system is days
determinate value of the parameter (Hunt, or months, but not nanoseconds. This means that
1991). the human controller can have at least days to
These particular techniques do not need the make the numerical analysis and decide the next
computational resources that the Ott et al. perturbation, whether the system goes on and
(1990a) algorithm does, but their validation is less cycles enough times to allow the next optimal
general. They are more suitable to systems with control. A more relevant point in choosing the
high frequency because they do not need a real appropriate technique is the sensitivity to the
time computer analysis of the system, and they noise present in the determination of the attrac-
I. Suárez / Ecological Modelling 117 (1999) 305–314 307

tor. With regard to this topic Ott et al. (1990b) 1995a). The problem with this kind of manipula-
point out that in the presence of noise, the ampli- tion is in determining exactly what is the magni-
tude of the perturbation that will be made in the tude of the change in the parameter due to the
control parameter must exceed the error length. If controller’s perturbation. If for example, the con-
it is possible to keep the noise bound, it was troller applies some amount of insecticide, he will
shown that the typical outbursts of a chaotic not know immediately the amount of change in
system can be considered as a low probability tail the rate of population growth., but if he main-
case (Ott et al., 1990a), occurring rarely. The tains a monitoring program, he can estimate this
technique proposed by Pyragas (1992) is robust to amount a posteriori, and before making the next
noise. It uses feedback control. If noise is present perturbation.
in the system determination, the feedback in- So far, the feasibility of this kind of control in
creases as the noise increases, achieving control in ecology has been presented. Now, some points
spite of the noise (Pyragas and Tamasevicious, against it will be discussed. Firstly, it must be said
1993). Whatever be the choice between Ott et al.’s that the small perturbations on the system control
(1990a) or Pyragas’s techniques, the possibility of parameters have been permanent in the examples
deciding the amplitudes of the perturbations as a where the statements discussed above were vali-
function of the size of noise needs to be dated. This means that once the parameter is
considered. changed from state p to state p’, the latter will not
To know these amplitudes is, at least theoreti- change to state p’’ or any other until the con-
cally, possible in population field biology. To troller makes the next perturbation. The author
obtain it there are standard techniques to estimate believes that in natural populations it is impossi-
confidence intervals in the determination of popu- ble to affirm the existence of this steadiness.
lation sizes, through capture-recapture methods, Moreover, the opposite cannot be affirmed. The
or another standard technique (Krebs, 1989). Al- resilience of natural populations is still a point of
though a confidence interval, this is not a direct strong debate, with deep implications on manage-
measure of the noise involved in population size ment ecology and conservation (Pimm, 1984).
determination, but it could be used as an indirect Given the disagreements in this debate, the steadi-
measure of it. ness of the control parameter state will not be
Reviewing the proposed techniques, resonant assured until the controlling techniques are ap-
parametric perturbation (Lima and Pettini, 1990) plied to a natural system, and even in that case it
still has to be considered. The control used would will only validate the statements in that particular
have a determined amplitude and frequency, but system and cannot be generalized to other natural
the problem here is that it seems difficult to make systems.
perturbations with well-determined values artifi- Furthermore, the fast convergence of these
cially, in an ecological system. methods has been demonstrated for systems with
Until now there have been presented here some steady changes in the control parameter. Thus,
criteria to choose an adequate technique, and the the convergence speed of the control method re-
possibility of estimating the noise in ecological mains as an open question in ecological systems.
control has been mentioned. The possible exis- Before passing to a numerical example the pro-
tence of a control parameter that could be manip- posed algorithms following from the above argu-
ulated has not been mentioned yet. The logistic ment would be:
model, a system capable of displaying chaotic 1. Obtain historical data on the population to be
behaviour, has only one control parameter: the controlled. With these, build a time series and
population growth rate. It is obvious that this analyze it in phase space, following Schaffer
parameter is susceptible to manipulation, with (1984). It would be useful make additional
biocontrol agents or insecticides. More flexible analyses to find evidence of chaos, following
models, such as the Bellows equation, can also be Ellner and Turchin (1995). Moreover, such an
controlled by perturbing one parameter (Doebeli, analysis permits the derivation of a model of
308 I. Suárez / Ecological Modelling 117 (1999) 305–314

the population dynamics. An estimation of the slope of f at the equilibrium N*, given by f(N*)=
parameters of the model to be used is needed, N*. The slope is given by the first derivative of
to identify stable orbits. f(N) evaluated at N*, N*\0. As this slope
2. Make the perturbation in the control variable, reaches a critical value chaos sets in. The parame-
by, for example, the introduction of an insecti- ter a is not included in the equation for the slope,
cide in the insect’s egg-laying locations. it affects only the equilibrium density, which cor-
3. Follow the amplitude of the change by stan- responds to the carrying capacity of the popula-
dard techniques of estimation of population tion, and will be considered as the control
size. Estimate the size of the noise in this parameter.
determination to plan the size of the next Assuming that parameters l and b have values
perturbation. coding for chaos, it is assumed that it is possible
4. With the new points in the time series, go back to manipulate a in each generation, slightly
to the first step. changing the population ability to cope with the
environment, hence the carrying capacity. In the
case of the control of a pest insect, as discussed
4. A numerical example above, these changes could be reduction of avail-
able egg-laying places, or similar perturbations.
In this section a numerical example of the Monitoring population size, which has a chaotic
control method discussed above is presented. Its dynamic, a generation in which Nt is close to the
efficiency in controlling a biological population equilibrium value, N* will occur. In this genera-
with unstable dynamics will be shown. In the tion, one can approximate the dynamics of the
present case it will be considered that the determi- system linearly:
nation of the population density has a stochastic
component, arising from its estimation in the Nt + 1 − N*= (N ,a ) · (Nt − N*)
wild. The control method is the one proposed by (N t 0
Ott et al. (1990a,b), and it is tried on a simple (f
+ (N*,a0) · (at − a0) (2)
model of population dynamics. The model used (a
was introduced by Maynard-Smith and Slatkin
Introducing the ability to control the change in
(1973), and considered by Bellows (1981) to be the
the parameter a by adjusting it linearly according
most generally applicable, due to its mathematical
to the population density
flexibility. The model and the mathematical con-
siderations for the control of the deterministic al − a0 = c(Nt − N*) (3)
case is briefly described, taken from Doebeli
because c is a constant to be determined, Eq. (3)
could be substituted into Eq. (2), yielding:
Consider a population with discrete generations

whose dynamics are described by: Nt + 1 − N*

Nt + 1 = f(Nt )=Nt
(1) =
(f (f n
(N*,a0)+ c (N*,a0) · (Nt − N*) (4)
1+(aNt )b (N (a
Here Nt + 1 and Nt are population densities in If c is chosen to make the modulus of the
successive generations. l, a and b are demo- expression in brackets less than one, the system
graphic parameters: l \ 1 is the intrinsic growth will then approach equilibrium. Evaluating the

) )
rate, a is a measure of how well the individuals partial derivatives, we get the condition
cope with the environment, and b ] 1 describes
l− 1 b(l−1)(b + 1)/b
the type of competition that leads to density 1− b −c B1 (5)
l a 20l
dependence (b: 1 could be interpreted as contest,
b\ \1 to scramble competition). The dynamic The optimal c value occurs when the above
of the population density is determined by the modulus is zero, finally at arriving the equation
I. Suárez / Ecological Modelling 117 (1999) 305–314 309

l −1 a l 2
copt = 1−b · (6) with two major perturbations, much smaller than
l b(l − 1)(b + 1)/b
the used in Fig. 1a. If it is intended to control a
To specify when control has to be applied, we natural population with this method, an estima-
must provide a criterion to decide when Nt is close tion of b and l from the real data is needed, and
to N*. As can be seen from Eq. (3), the size of the a time series of 30 or fewer data points is enough
perturbation depends on this decision. A practical to provide estimation of these parameters. In the
strategy to decide the value d for which Nt − simulation shown (Fig. 1b) the parameters of Eq.
N* B dcould be to calculate the size of the maxi- (1) can be estimated from the time series before
mal perturbation that can be done in each the control is activated. Nonlinear regression and
particular system, accordingly calculating d. Con- the Levenberg–Marquardt parameter estimation
sidering that an uncontrolled chaotic population method (provided by SPSS statistical software)
dynamic is ergodic, given any d \ 0 eventually a yield estimation of parameters precise to eight
generation will fulfill the condition (Ott et al., decimal places.
1990b). Fig. 1c shows the system with d= 0.06. The
To include stochastic factors in the simulation, condition to apply the control is only fulfilled
Nt from Eq. (1) is calculated by adding a random after 1100 generations. It is still noteworthy the
percentage of itself: size of the perturbation, which is smaller than in
the preceding cases. Only one relevant perturba-
Nt = Nt 9 (7) tion is needed to control the system. In general it
100 is illustrated that the use of smaller values of d
The parameter r is a random number, taken imply a longer time of monitoring the population
from a uniform distribution from zero to a maxi- density to start the control, but once it has been
mal value, fixed at 1, 5, 10 and 20. These extreme made active, the size of the perturbations and its
values are reasonable amounts of noise in the numbers are smaller and the technique works
determination of wild population densities. more efficiently.
(Krebs, 1989). The sign of the added expression Fig. 2 shows the control working with different
was positive with a 0.5 probability each time a levels of noise in the population density. The
population density was calculated, and negative simulations shown are typical of the system, cho-
for the remaining cases. sen after many repetitions.
Fig. 1 shows the effect of choosing different In this case, the technique is robust to signifi-
values of d, namely 10%, 1% and 0.1% of the cant amounts of noise added to the population
maximal population size. In this case no noise in density determination. At increased noise levels,
the population density is simulated. the dynamic is still controlled, but bigger pertur-
Fig. 1a shows the results of the control when bations are needed to keep the system stable.
d = 6 (the maximal population density for the b Even in cases in which the noise is 10% of the
and l values chosen being 60). In the fifth genera- deterministic population density, the method is
tion the control is made active and the equi- still useful for eliminating the fluctuations associ-
librium density is closely attained after three ated with a chaotic regime. These levels of noise
relevant perturbations. The equilibrium orbit is are not enough to force the system out of the
approached asymptotically, so the size of the per- neighborhood of a stable orbit. Only in Fig. 2d,
turbations decreases through time but is not zero. showing a system with 20% noise, is it the case
After the number of perturbations mentioned, the that the stochastic part of a population’s dynamic
changes in them occur at the octave decimal can make the control method no longer useful.
position, and are irrelevant in practice. The reason is that such an amount of noise takes
In Fig. 1b is shown the result of controlling the the system out of the neighborhood of equi-
system when d is set to 0.6. After 30 generations librium. In this case the system returns to chaotic
of uncontrolled chaotic dynamics, the condition is fluctuations. Out of control again, the condition
fulfilled and the population density is controlled to make the perturbation scheme active again
310 I. Suárez / Ecological Modelling 117 (1999) 305–314

Fig. 1. Changing the neighborhood criteria changes the required time for starting the control and the number and size of the
perturbations required. The values of a are shown in the small graphics. The population densities are plotted against generation time
in the main graphics, showing the attainment of the equilibrium density. The parameter values for the figure were as follows:
a0 =0.05, b = 4 and l=5. d= 6 in Fig. 1a, d = 0.6 in Fig. 1b and d= 0.06 in Fig. 1c.
I. Suárez / Ecological Modelling 117 (1999) 305–314 311

Fig. 2. The importance of noise in the control of an unstable dynamic is considered here. The amount of noise increases from Fig.
2a to Fig. 2d. It is noteworthy that in this noisy case, a constant monitoring of the population is needed to keep its dynamics
controlled. As in the previous figure, the a values are shown in the small boxes and the population densities against generation time
in the main graphics. The parameter values for this figure were as follows: a0 =0.05, b =4, l =5 and d=0.6. r =1 in Fig. 2a, r= 5
in Fig. 2b, r= 10 in Fig. 2c and r = 20 in Fig. 2d.
312 I. Suárez / Ecological Modelling 117 (1999) 305–314

takes some time to be fulfilled, but after it hap- ral selection processes). Allen et al. (1993) argue
pens, the system is randomly set out from the that chaotic systems, because of their intrinsic
equilibrium again. noise amplification can amplify the noise of the
This numerical example exemplifies several fea- system, and enhance its natural heterogeneity,
tures of the method mentioned above. It is shown preserving it from extinction.
that knowledge about the levels of noise built into Forgetting the debate about the existence of
the system to be controlled and the maximal size chaos for a moment, let us consider the con-
of the perturbation to be applied are important trolling techniques consisting in the periodic per-
parameters to know before using this control turbation of a main variable of the system (Lima
method properly. It also points to a striking fea- and Pettini, 1990). Now suppose that the periodic
ture of unstable dynamical behavior. To obtain a perturbations occurring in populations, (precisely
stable orbit from a chaotic system, its parameter in those where chaos was expected to be more
does not need to be changed from chaotic values likely, the temperate ones, according to Turchin,
to stable values. Small perturbations at the proper 1993) are the periodic perturbations of a variable
moment are responsible for the change of the of the system. This view can provide a natural
dynamics. For the numerical example shown here chaos control device in an ecological context, but
the changes in the control parameter are displayed there is strong debate about the existence of chaos
in the small windows on the figures. It shows that in the rodent studied (Turchin, 1993, 1995; Falck
the control parameter a is only slightly perturbed. et al., 1995a,b). From this debate, in Turchin’s
In the deterministic case a returns to its original original 1993 work are several reasons to expect
value, but due to the timing of the perturbation chaos in northern populations, but, also in this
the resulting dynamic is stable and not chaotic. northern ecosystem the resources tend to decrease
abruptly during the winter, rising again in the
following seasons, and repeating this pattern in a
5. Chaos control and chaos detection in natural periodic way. Perhaps this seasonal pattern of the
populations: environment can be likened to the parametric
periodic perturbation in the method proposed by
As mentioned at the beginning, the existence of Lima and Pettini. Maybe the non-existence of this
chaos in natural populations is a controversial behaviour in nature is due to periodic, or seasonal
issue. Without dealing with the methodological perturbations, more than the evolutionary argu-
problems for its detection, some arguments ments against chaos, or the non-existence of
against its existence in nature have been pre- proper detection techniques.
sented, by means of natural selection (Berryman This statement, suggested by one of the chaos
and Milstein, 1989). The central argument in this control techniques, agrees with previous works on
work is that if a population behaves chaotically, it metapopulation models. There was controversy
will pass through very small population sizes, on the existence of chaos in connected metapopu-
increasing the probability of extinction, and there- lations (McCallum, 1992; Gonzales-Andujar and
fore cannot be favoured by natural selection. Be- Perry, 1993; Ruxton, 1993, 1996a; Doebeli, 1995a,
sides, it is clearly a group – selectionist argument among others in broader contexts). It hangs on
(and makes invalid the point following Ferriere whether a metapopulation defined as chaotic dis-
and Gatto, 1993, among others), and there are plays chaos when connected (through migration)
models developed that permit us to consider the with another metapopulation whose behaviour
chaotic behaviour of a natural population as an could be, (or not) chaotic. Although there is a
evolutionary stable strategy, the result of natural consensus that it depends on the immigration–
selection acting at the individual level (Ferriere emigration levels, these results are reinforced by
and Fox, 1995, for a review of their views on the the theory developed to explain the synchroniza-
topic; Doebeli, 1993, and Altemberg, 1991 for tion and the control of the chaotic systems (Pec-
alternative views of chaos as an outcome of natu- ora and Carroll, 1990). Furthermore, they are
I. Suárez / Ecological Modelling 117 (1999) 305–314 313

becoming established in a broader theoretical ment and continous discussions about the topic.
frame, the control chaos theory, as pointed out by Two anonymous referees who suggest the addi-
Doebeli (1995b). The fact is that in other disci- tion of a numerical example, improving the read-
plines it has been demonstrated that two poten- ability of the manuscript.
tially chaotic systems, if coupled, can behave in a
periodic fashion, because of their intrinsic chaotic
natures (Pecora and Carroll, 1990). References
As an example of these phenomena in biology,
there are theoretical models of trophic chains that Allen, J.C., Schaffer, W.M., Rosko, D., 1993. Chaos reduces
show chaos (Hastings and Powell, 1991; McCann species extinction by amplifying local population noise.
Nature 364, 229 – 231.
and Yodzys, 1994; Eisenberg and Maszle, 1995; Altemberg, L., 1991. Chaos from linear frequency-dependent
Ruxton, 1996b). A trophic chain implies several selection. Am. Nat. 138, 51 – 68.
populations connected by predation. If we follow Bayli, P.V., Virgin, L.N., 1994. Practical considerations in the
the suggested idea in the case of a metapopulation control of chaos. Phys. Rev. E50, 604 – 607.
connected by migration we can propose the same Bellows, T.S., 1981. The descriptive properties of some models
for density dependence. J. Anim. Ecol. 50, 139 – 156.
idea. In nature it has become difficult to find Berryman, A.A., Milstein, J.A., 1989. Are ecological systems
chaotic behaviour probably because the connec- chaotic — And if not, why not? TREE 4, 26 – 28.
tions in a trophic chain allow the same regulatory Ditto, W.L., Pecora L.M., 1993. Mastering chaos. Scientific
role for populations with strong periodic be- American. August, 78 – 84.
haviour. It is important to say that all the authors Doebeli, M., 1993. The evolutionary advantage of controlled
chaos. Proc. R. Soc. Lond. B254, 281 – 285.
cited here discuss to some extent the importance Doebeli, M., 1995a. Updating Gillespie with controlled chaos.
of coupling as a regulation source, but the point Am. Nat. 146, 479 – 487.
of this paper is to see this regulation by coupling Doebeli, M., 1995b. Dispersal and dynamics. Theor. Popul.
enhanced because of the instability of chaos, fol- Biol. 47, 82 – 106.
lowing Ditto and Pecora (1993). Doebeli, M., Ruxton, G.D., 1997. Controlling spatial chaos in
metapopulations with long-range dispersal. Bull. Math.
Biol. 59, 497 – 515.
Eisenberg, J.N., Maszle, D.R., 1995. The structural stability of
6. Conclusions a three-species food chain model. J. Theor. Biol. 176,
501 – 510.
Exploring the Ott et al. (1990a) method and its Ellner, S., Turchin, P., 1995. Chaos in a noisy world. New
methods and evidence from time-series analysis. Am. Nat.
variants reveals valuable contributions to the the- 145, 343 – 375.
ory of population management. This algorithm is Falck, W., Bjornstad, O.N., Stenseth, N.C., 1995a. Bootstrap
applicable in the control of pests, insects and estimated uncertainty of the dominant Lyapunov exponent
other natural populations, offering a fast converg- for Holartic microtine rodents. Proc. R. Soc. Lond. B261,
159 – 165.
ing solution. The exploration of a simple numeri-
Falck, W., Bjornstad, O.N., Stenseth, N.C., 1995b. Voles and
cal example supports this claim. lemmings. chaos and uncertainty in fluctuating popula-
The existence of linked phenomena in ecology tions. Proc. R. Soc. Lond. B262, 363 – 370.
suggests that the paucity of examples in which Ferriere, R., Fox, G.A., 1995. Chaos and evolution. TREE 10,
chaos is detected in nature could be due to the 480 – 485.
Ferriere, R., Gatto, M., 1993. Chaotic population dynamics
control of chaos by means of periodic fluctuations
can result from natural selection. Proc. R. Soc. Lond.
from the environment of another population. B251, 33 – 38.
Gonzales-Andujar, J.L., Perry, J.N., 1993. The effect of dis-
persal between chaotic and non-chaotic populations within
Acknowledgements a metapopulation. Oikos 66, 555 – 557.
Guemez, J., Matias, M.A., 1993. Control of chaos in unidi-
mensional maps. Phys. Lett. A181, 29 – 32.
I would like to thank Luis Miguel Marquez for Hassell, M.P., Lawton, J.H., May, R.M., 1976. Patterns of
comments in the initial version of this paper. To dynamical behaviour in single species populations. J. An.
Michael Doebeli and Dita Vizoso for encourage- Ecol. 45, 471 – 568.
314 I. Suárez / Ecological Modelling 117 (1999) 305–314

Hastings, A., Powell, T., 1991. Chaos in a three-species food Pimm, S.L., 1984. The balance of nature?: Ecological issues in
chain. Ecology 72, 896–903. the conservation of populations and communities. The
Hasting, A., Hom, C.L.., Ellner, S., Turchin, P., Godfray, University of Chicago Press, Chicago.
H.C.J., 1993. Chaos in ecology: is mother nature a strange Pyragas, K., 1992. Continuos control of chaos by self-con-
attractor? Ann. Rev. Ecol. Syst. 34, 1–33. trolling feedback. Phys. Lett. A170, 421 – 428.
Hunt, E.R., 1991. Stabilizing high-period orbits in a chaotic Pyragas, K., Tamasevicious, A., 1993. Experimental control of
system: The diode resonator. Phys. Rev. Lett. 67, 1953– chaos by delayed self-controlling feedback. Phys. Lett.
A180, 99 – 102.
Rulkov, N.F., Tsimring, L.S., Abarbanel, H.D.I., 1994. Track-
Krebs, C.J. 1989. Ecological Methodology. Harper Collins
ing unstable orbits in chaos using dissipative feedback
Publishers, New York.
control. Phys. Rev. E50, 314 – 324.
Lathrop, D., Kostelich, E.J., 1989. Characterization of an Ruxton, G.D., 1993. Linked populations can still be chaotic.
experimental strange attractor by periodic orbits. Phys. Oikos 68, 347 – 348.
Rev. A40, 4028 – 4031. Ruxton, G.D., 1996a. Temporal scales and the occurrence of
Lima, R., Pettini, M., 1990. Suppresion of chaos by resonant chaos in coupled populations. TREE 10, 141 – 142.
parametric perturbations. Phys. Rev. A41, 726–733. Ruxton, G.D., 1996b. Chaos in a three-species food chain with
Maynard-Smith, J., Slatkin, M., 1973. The stabliity of preda- a lower bound on the bottom population. Ecology 77,
tor – prey systems. Ecology 54, 384–391. 317 – 319.
McCallum, H.I., 1992. Effects of immigration on chaotic Schaffer, W.M., 1984. Stretching and folding in lynx fur
population dynamics 154, 277–284. returns: evidence for an strange attractor in nature? Am.
McCann, K., Yodzys, P., 1994. Biological conditions for chaos Nat. 124, 798 – 820.
in a three species food chain. Ecology 75, 561–564. Sugihara, G., May, R.M., 1990. Nonlinear forecasting as a
Meucci, R., Gadomski, W., Ciofini, M., Arecchi, F.T., 1994. way to distinguish chaos from measurement noise in time
Experimental control of chaos by means of weak paramet- series. Nature 3444, 734 – 831.
ric perturbations. Phys. Rev. E49, R2528–R2531. Turchin, P., 1993. Chaos and stability in rodent populations
dynamics: evidence from nonlinear time series analysis.
Ott, E., Grebogy, C., Yorke, J.A., 1990a. Controlling chaos.
Oikos 68, 167 – 172.
Phys. Rev. Lett. 64, 1196–1199.
Turchin, P., 1995. Chaos in microtine populations. Proc. R.
Ott, E., Grebogy, C., Yorke, J.A., 1990b. Controlling chaos.
Soc. Lond. B262, 357 – 361.
In: Campbell, D.K. (Ed.), Chaos/Xaoc; Soviet American
Turchin, P., Taylor, A.D., 1992. Complex dynamics in ecolog-
perspectives on nonlinear science. American Institute of ical time series. Ecology 73, 289 – 305.
Physics, pp. 153 – 172. Weis, J.N., Garfinkel, A., Spano, M.L., Ditto, W.L., 1993.
Pecora, L.M., Carroll, L.T., 1990. Synchronization in chaotic Chaos and chaos control in biology. J. Clin. Invest. 93,
systems. Phys. Rev. Lett. 64, 821–824. 1355 – 1360.