3275

Ecology, 82(12), 2001, pp. 32753284

2001 by the Ecological Society of America
IS IT TIME TO BURY THE ECOSYSTEM CONCEPT?
(WITH FULL MILITARY HONORS, OF COURSE!)
1
ROBERT V. ONEILL
Environmental Sciences Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831-6036 USA
ROBERT V. ONEILL, MacArthur Award Recipient, 1999
Abstract. The ecosystem concept has become a standard paradigm for studying eco-
logical systems. Underlying the ecosystem concept is a machine analogy derived from
Systems Analysis. This analogy is difcult to reconcile with our current understanding of
ecological systems as metastable adaptive systems that may operate far from equilibrium.
This paper discusses some logical and scientic problems associated with the ecosystem
concept, and suggests a number of modications in the paradigm to address these problems.
Key words: ecosystem; ecosystem stability; ecosystem theory; ecotone; Homo sapiens; natural
selection; system dynamics; Systems Analysis.
INTRODUCTION
The term ecosystem was coined by Tansley in 1935.
But as Botkin (1990) points out, the underlying concept
goes back at least to Marsh (1864). Nature was viewed
as relatively constant in the face of change and repaired
itself when disrupted, returning to its previous balanced
state. Clements (1905, 1916) and Elton (1930) offered
plant and animal succession as basic processes that
permitted relative constancy by repairing damage.
Forbes (1925) described the northern lake as a micro-
Manuscript received 22 June 2000; revised 5 March 2001;
accepted 6 March 2001.
1
Presented 9 August 1999 in Spokane, Washington, USA.
cosm, a relatively closed, self-regulating system, an
archetypic ecosystem.
Science emerged from the Second World War with
a new paradigm, Systems Analysis (e.g., Bode 1945),
which seemed uniquely suited for this balance of na-
ture concept, and t well with earlier work on the
stability of interacting populations (Nicholson and Bai-
ley 1935). Systems Analysis dealt with complex sys-
tems as interconnected components with feedback
loops (Hutchinson 1948) that stabilized the system at
a relatively constant equilibrium point. Systems Anal-
ysis can be seen underlying E. P. Odums (1953) def-
inition of the ecosystem as a . . . natural unit that
includes living and nonliving parts interacting to pro-
3276 ROBERT V. ONEILL Ecology, Vol. 82, No. 12
P
e
r
s
p
e
c
t
i
v
e
s
duce a stable system in which the exchange of materials
between the living and nonliving parts follows circular
paths . . . .
The machine analogy, inherent in Systems Analysis,
became a central paradigm for many ecologists (Odum
1971, Holling 1973, Waide and Webster 1976). The
paradigm offered a practical approach to the enormous
complexity of natural systems (Teal 1962, Van Dyne
1969). The paradigm helped harness the power of the
computer in ecosystem models (Olson 1963). The par-
adigm permitted a holistic view of system properties
such as nutrient cycling (Webster et al. 1974). The
familiarity of the machine analogy facilitated the com-
munication of ecological concepts to the public.
If the ecosystem concept has held such a central
place in ecology and been so productive of new ideas,
why call it into question? The simple fact is that the
ecosystem is not an a posteriori, empirical observation
about nature. The ecosystem concept is a paradigm
(sensu Kuhn 1962), an a priori intellectual structure, a
specic way of looking at nature. The paradigm em-
phasizes and focuses on some properties of nature,
while ignoring and de-emphasizing others. After a half
century of application, the paradigm is showing some
rust. Limitations in the concept are becoming more
apparent and leading to a vigorous backlash toward
ecosystem concepts in particular, and ecology in gen-
eral.
BACKLASH AGAINST THE ECOSYSTEM CONCEPT
Part of the backlash results from the apocalyptic fer-
vor of the environmental movement over past decades.
Ecology oversold its ability to predict doom, and is
now seen as unnecessarily constraining human freedom
and economic growth. Inuential opponents simply
dismiss the prophecy (Simon 1980), and offer opti-
mistic counterclaims (Naveson 1993). Human inge-
nuity is seen as sufcient to feed, clothe, and supply
energy to an ever-growing population for the next seven
billion years (Myers and Simon 1994)!
Clearly, there are ideological underpinnings to this
backlash. Anything that limits human development is
immediately suspect. Nevertheless, the critique high-
lights important limitations in ecological theory
(Shrader-Frechette and McCoy 1993, Shrader-Frech-
ette 1995, Callicott 1996). The critics claim that the
ecosystem concept isnt a scientic theory at all, simply
a statement about physical constraints on living things
(Sagoff 1997). Concepts like stability and ecosystem
are ambiguous and dened in contradictory ways. In
fact there is no such thing as an integrated, equilibrial,
homeostatic ecosystem: It is a myth (Soule and Lease
1995)!
If there is no stable equilibrium, why bother to con-
serve? Protecting and restoring endangered species is
unnecessary, species go extinct all the time. How do
you restore ecosystems when you dont know what to
restore them to? Fragmentation actually increases spe-
cies diversity by adding edge species. Eutrophication
leads to lakes with greater productivity. There isnt any
environmental crisis at all!
It is important to recognize that not all of the back-
lash has an ideological bias. The ecosystem concept is
also broadly criticized within ecology. One of the clear-
est statements is given by Pickett et al. (1992): The
classical paradigm in ecology, with its emphasis on the
stable state, its suggestion of natural systems as closed
and self-regulating, and its resonance with the nonsci-
entic idea of balance of nature, can no longer serve
as an adequate foundation for conservation.
There is the temptation, of course, to respond de-
fensively to the backlash. After all, the ecosystem the-
ory being attacked is a sort of historical artifact, a
straw man. Ecosystems are now seen as disequili-
brial, open, hierarchical, spatially patterned, and scaled
(ONeill et al. 1982, Pickett et al. 1992, Levin 1999).
Many of the criticisms have been addressed as the the-
ory matured. But rather than provoking a defensive
reaction, perhaps the backlash should motivate a care-
ful re-examination of the ecosystem concept.
SCIENTIFIC PROBLEMS
The ecosystem concept is a paradigm, i.e., a con-
venient approach to organizing thought. Like any par-
adigm, it is a product of the human minds limited
ability to understand the complexity of the real world.
In the case of ecological systems, we are faced with
hundreds to thousands of interacting populations. The
systems vary through time in complex ways, and they
are spatially heterogeneous at every scale.
The ecosystem concept takes these impossibly com-
plex phenomena and focuses on a small subset: the
average or integrated properties of all the populations
within a specied spatial area. This approach has the
advantage of identifying emergent properties such
as energy ow and nutrient cycling, and permits study
of the relative stability of this abstract structure and its
function.
But in order to gain these advantages, the concept
accepts a set of assumptions that limits our thinking
and determines the questions we ask. Therefore, we
must continuously examine the assumptions, and con-
sider the questions they might keep us from asking.
Most importantly, we must examine whether the as-
sumptions limit our ability to answer the very ques-
tions, such as relative stability, that the concept was
designed to address.
The proposed exploration is rendered difcult by the
ambiguity of terms like complexity, ecosystem, and
stability. Pimm (1984) pointed out that a change in
these denitions can lead to signicantly different con-
clusions about stability. At present, the terms eco-
system and ecosystem theory are used in many
different ways. At one extreme, ecosystem is a con-
venient term, relatively free of any assumptions, that
indicates the interacting organisms and abiotic factors
December 2001 3277 MACARTHUR AWARD LECTURE
P
e
r
s
p
e
c
t
i
v
e
s
in an area. At the other extreme, ecosystem is a pre-
cisely dened object of a predictive model or theory.
As a result, any limiting assumption that is offered here
may be disavowed by some subset of ecologists. Nev-
ertheless, the critical examination is important, even if
no single ecologist would admit allegiance to the total
set of assumptions. With this difculty in mind, let us
consider some implicit assumptions and their impact
on addressing stability.
Spatial closure
The ecosystem concept considers a specic spatial
unit: classically, a small watershed for terrestrial sys-
tems and a lake for aquatic systems. The concept looks
within these boundaries to locate the signicant dy-
namics that require explanation and the signicant pro-
cesses that will explain the dynamics. The boundaries
may be open to the exchange of organisms, energy, and
matter. Nevertheless, the ecosystem concept assumes
that the interactions and feedback loops necessary and
sufcient to explain dynamics occur within the bound-
aries.
The problem with this assumption is that the spatial
distributions of the component populations may be
much larger than the ecosystem boundaries. Indeed,
even the home ranges of individuals may be larger than
the ecosystem, particularly for predators. This leads to
anomalies, such as food web calculations determining
that an ecosystem can support one half of a top car-
nivore, without specifying which end. But importantly,
dispersal from outside system boundaries is a critical
mechanism for system stability. Thus, an internal pro-
cess, recovery, is not explicable by feedbacks occurring
within the system specications. The critical obser-
vation is that a forest plot surrounded by continuous
forest behaves differently from that same forest plot in
isolation. At the minimum, the spatial context of the
system and all its component populations must be in-
cluded in the specications of the ecosystem. The sta-
bility properties of an ecological system cannot be ex-
plained by a paradigm that only considers dynamics
occurring within the ecosystem boundaries.
Spatial homogeneity
A second assumption of the concept is spatial ho-
mogeneity. Spatial heterogeneity within the spatial unit
is averaged in order to focus on integrated or emergent
properties. And yet it is the internal heterogeneity, or
the heterogeneity of the larger spatial context, that
maintains the full range of populations needed to main-
tain stability. Without the heterogeneity, for example,
pioneer species are not maintained, and recovery be-
comes impossible or follows an unpredictable course.
A homogeneous ecosystem, like an overspecialized
species, cannot respond to change and is inherently
unstable. The critical observation is that two forest
plots may have identical average properties, but dif-
ferent relative stability if they differ radically in spatial
heterogeneity. The stability of an ecological system
cannot be predicted by a theory that ignores hetero-
geneity.
Latin binomials are substitutable
Most ecologists consider the species list critical to
the denition of an ecosystem, often designating eco-
system types by their dominant species. Nevertheless,
some degree of substitutability is implicit in the eco-
system concept. For example, an impacted ecosystem
would be considered recovered if succession re-
placed the dominant species, and brought the system
back to the same physiognomy and functional attri-
butes. But the species list of soil organisms, for ex-
ample, need not be identical. The functional properties
of the system are restored, but with an altered species
list. In fact, in much of ecosystem theory, the stated
variables are functional groups, such as trophic levels,
and which of several species perform the function is
not considered.
The ambiguity introduced by species substitutability
is often unrecognized. But the inconsistencies are
brought out by the seemingly inane question: Do eco-
systems die? Consider, for example, a northern lake
that has undergone eutrophication. If the ecosystem is
dened as a functional system at a spatial location, then
the lake is the same ecosystem, albeit altered by
changed conditions. On the other hand, if the ecosystem
is dened by the species list, then the oligotrophic eco-
system has been killed and replaced by a eutrophic
ecosystem. The ecosystem dened by a species list is
almost always unstable because it rarely, if ever, re-
covers to the identical list of species.
If the ecosystem is dened by the species list, the
only stable systems are found in extreme conditions
with impoverished species lists, such as the Arctic tun-
dra. In such conditions, recovery occurs to the same
species list because only a few species can survive. But
this leads to the anomalous conclusion that stability is
inversely related to biodiversity. Although the biodi-
versitystability relationship is questionable (see e.g.,
Huston 1994), the relationship suddenly reverses and
becomes crystal clear. The probability of a system re-
covering to an identical species list is dependent on
how long the list is, making tundra ecosystems more
stable than tropical forests!
The dilemma would seem to be solved by dening
the ecosystem by the rate processes plus dominant spe-
cies. But problems still arise in dealing with stability.
Consider a marine fouling community (Sutherland
1974) that may recover to the same rate processes, but
any of several different species lists. In SerengetiMara
woodlands, elephants and re interact to produce a va-
riety of stable states (Dublin et al. 1990, Dublin 1995).
Are these stable ecosystems with several nal states?
Or are these unstable ecosystems?
The problem is compounded by evidence from the
pollen record. Over glacial cycles, the record shows
3278 ROBERT V. ONEILL Ecology, Vol. 82, No. 12
P
e
r
s
p
e
c
t
i
v
e
s
that species respond individually to changing condi-
tions (Davis 1976). Different species, even dominants,
move in or out of an area based on their individual
responses to temperature and other changes (Delcourt
and Delcourt 1987). As a result, intact communities of
organisms do not move as a unit, and the collection of
interacting species at a location is continuously chang-
ing. The conservative conclusion is that the entity
ecosystem is unstable. The more radical conclusion
is that ecosystems, dened by their species list, dont
exist and never have.
The solution to the problem would seem to involve
ignoring the species list altogether and assuming per-
fect substitutability. The stable entity is the collection
of functional groups that recover to the same rate pro-
cesses, feedbacks, and complex organization. Indeed,
this solution is the one implicitly adopted in many ap-
plications of the ecosystem concept. But this solution
also leads to anomalous conclusions when applied to
ecological phenomena other than stability.
Consider, for example, the phenomenon known as
the ecotone. The ecotone is a tension zone where one
vegetation type changes suddenly into another, e.g.,
grassland into scrub (Hobbs 1986) or forest (Hansen
et al. 1994). These transitions have long attracted the
attention of ecologists (e.g., Clements 1897, 1905, Liv-
ingstone 1903, Griggs 1914). But the ecotone is dened
by a change in the species list, particularly in the dom-
inants. Since environmental constraints are very similar
on either side of the ecotone, rate processes would also
be expected to be very similar. If the ecosystem is
dened strictly in terms of function, there is not a dif-
ferent ecosystem on either side of the ecotone. In fact,
the ecotone doesnt exist!
Nevertheless, if we limit consideration to stability
phenomena, a functional denition with an assumption
of species substitutability seems the logical choice. The
abstraction and the associated assumptions limit the
utility of the theory for explaining other phenomena,
such as ecotones. Of course, this means that the eco-
system concept is not adequate as a general theory of
ecological phenomena, but at least it permits expla-
nation of stability. The functional theory denes an
entity that maintains a trophic structure, recycles nu-
trients, and recovers from disturbance. Minimizing the
role of latin binomials seems to permit a consistent
framework for dealing with stability.
Natural selection minimized or ignored
The assumption of species substitutability minimizes
the role of natural selection. This may be the concepts
most serious limitation in dealing with stability. Nat-
ural selection is relegated to a background role causing
component populations to optimize or maximize their
share of resources. As a result, functional groups can
be assumed to be operating at the rates set by physical
and chemical constraints. Ecosystem dynamics can
then be predicted from these constraints. Stated another
way, natural selection is assumed to operate slowly.
Therefore, its dynamics can be assumed to be constant
over the time scales relevant to ecosystem behavior.
But the advantage gained may not outweigh the losses.
Natural selection is the most powerful predictive theory
available to ecology.
In the extreme, ecosystem specication may simply
ignore the identity of component populations. But this
extreme admits of a reductio ad absurdum. The forest
ecosystems of eastern North America have lost the cou-
gar, woodland bison, wolf, and bear. The presettlement
Passenger Pigeon population is estimated at 3 to 5 bil-
lion (Schorger 1955). In the 1870s, Audubon estimated
a single ock at 136 million. It moved . . . like a tor-
nado through the forest . . . breaking off trees up to
two feet in diameter when it roosted (Schorger 1955).
It is absurd to maintain that rate processes and stability
were not altered by replacing these woodland species
with Homo sapiens. In fact, many invasive species have
this same disruptive property. Examples include cattle,
kudzu, coconut, and zebra mussel.
The simple empirical fact is that ecosystems are col-
lections of interacting populations. The component
populations have been shaped by natural selection. The
resulting biotic potential determines ecosystem dynam-
ics just as much as chemical and physical constraints.
Critically, natural selection is one of the processes that
determines system stability.
Stability is a scaled concept
A different ambiguity arises because the concept of
stability cannot be dened independently of the scale
of observation. The dependence on scale is revealed
when one considers the full spatiotemporal spectrum
of disturbances. Disturbances that are frequent and
smaller in spatial scale than the dened ecosystem
boundaries can be counteracted by internal mecha-
nisms. The ecosystem would be considered stable to
these disturbances. Up to some point, the ecosystem
also recovers from larger scale disturbance, i.e., dis-
turbances that have greater spatial extent and occur less
frequently. Recovery now involves mechanisms, such
as dispersal, which are not ordinarily considered as
internal ecosystem processes. Nevertheless, we could
still consider the ecosystem as responding stably.
Ultimately, of course, the ecosystem is unstable. It
is only a matter of time until a disturbance of sufcient
intensity and spatial extent overwhelms the ecosys-
tems ability to respond. Examples include broad-scale
desertication and rare asteroid collisions. Over a suf-
ciently long period of time, the cumulative probability
of a catastrophic event approaches 1.0.
The problem would seem to be solved by specifying
the time period of concern, e.g., the ecosystem is stable
over millennia. But ambiguities remain. If stability de-
pends on the spatial extent of disturbances, then the
size of the ecosystem matters. The relative stability of
two systems that differ radically in size cannot be ex-
December 2001 3279 MACARTHUR AWARD LECTURE
P
e
r
s
p
e
c
t
i
v
e
s
plained by internal mechanisms. If the disturbance re-
gime is the same, the smaller ecosystem is logically
less stable because it can sustain itself over a shorter
period of time. If the disturbance regime is radically
different, the relative stability of similar-sized ecosys-
tems cannot be explained by internal mechanisms. The
ecosystem in the harsh environment is logically less
stable because it can sustain itself over a shorter period
of time. To deal with stability resulting from internal
mechanisms requires that the denition of the ecosys-
tem itself specify (1) the disturbance regime, and (2)
the size of the specic system under consideration.
Another way to view the scale problem is to consider
how we use the term disturbance. Once the bound-
aries of the ecosystem are delimited, the continuous
distribution of environmental variability is divided into
internal conditions and external disturbances. Small-
scale variability, such as the daily light cycle, seasonal
temperature changes, or the occasional death of a dom-
inant individual are considered to be part of ecosystem
dynamics. Larger scale variability, such as wildre, is
not considered to be part of ecosystem dynamics.
By focusing on the ecosystem as a self-regulating,
integrated system, we are led to focus on that systems
resistance to disturbances, minimizing the impacts, and
resilience or recovery from disturbance back to some
optimal state or states. Disturbances are viewed as a
damaging external interference. Yet it is well docu-
mented that preventing damage can be catastrophic.
The obvious examples are management practices that
prevent small wildres. The spatiotemporal spectrum
of environmental variability determines ecosystem sta-
bility just as surely as internal feedback mechanisms.
Therefore, ecosystem stability cannot be explained un-
less the disturbance regime is part of the specications
of the ecosystem.
Homo sapiens is not a component of the ecosystem
The ecosystem concept typically considers human
activities as external disturbances to the ecosystem.
Other invasive pests, such as kudzu and brown rats,
are considered as ecosystem components, and their im-
pact on structure and function considered explicitly.
Homo sapiens is the only important species that is con-
sidered external from its ecosystem, deriving goods and
services rather than participating in ecosystem dynam-
ics.
If there was ever a species that qualied as an in-
vasive pest, it is Homo sapiens The litany is familiar
but merits repetition. Since 1850, the human population
has quintupled, and per capita energy use has quadru-
pled (Holden 1991). Human inputs of nitrogen now
exceed natural rates of nitrogen xation (Vitousek
1994), and nitrogen saturation leads to the loss of other
important nutrients, such as calcium (Vitousek et al.
1997). Inputs of nitrogen to the North Atlantic have
increased by a factor of two to twenty compared to pre-
industrial inputs (Howarth et al. 1996). Average tem-
perature is increasing faster than it has in the last 10 000
yr (Arrhenius and Waltz 1990). The human economy
uses 40% of net primary production (Vitousek et al.
1986). Soil erosion is nearly universal, with soil losses
exceeding soil formation rates by at least 10-fold (Pi-
mentel 1993).
It is clear that Homo sapiens has altered the physical
environment of the ecological system. We have
changed process rates ranging from productivity to dis-
persal. We have changed ecological structure by elim-
inating our competitors, e.g., timber wolves, and even
food species, e.g., Passenger Pigeon. It becomes fa-
cetious to talk about sustainability and continued ex-
traction of goods and services when we cannot specify
with any scientic rigor how Homo sapiens has already
altered the stability properties of the system.
Summary
The importance of the denitional ambiguities be-
comes apparent if we consider the reasons for coming
up with the ecosystem concept in the rst place. The
theory provides an explanatory framework for ecolog-
ical phenomena. Without reverting to a naive concept
of the balance of nature, the relative stability of eco-
logical systems represents a fundamental phenomenon
to be explained. Therefore, an adequate theory must
be able to deal with stability. At present, the ambiguity
involved in the denition of an ecosystem leads to basic
problems in explaining stability. Therefore, the ambi-
guities undercut the very purpose for which the concept
was devised.
The focus on internal dynamics and stability creates
a mindset that excludes relevant phenomena. In fact,
the most fundamental observation is sustainability un-
der conditions of constant change. The stasis implied
by the ecosystem concept is self-limiting. The critical
property is the ability to change state in response to a
continuous spectrum of change and variability. Sus-
tainability of ecological systems involves two anti-
thetical elements: (1) local and short-term stability in
the sense of recovery from disturbance, and (2) exi-
bility in the sense of maintaining variability of structure
in space and time because conditions will change.
ELEMENTS OF A NEW PARADIGM
None of the criticisms offered in the previous section
are new. Many would seem to be addressed by so-
phisticated developments in ecosystem theory involv-
ing, for example, nonlinear dynamics and fuzzy set
theory. Unfortunately, the developments make the eco-
system theory more intriguing for mathematicians, but
less useful and intuitive for biologists. I would offer
for consideration that none of the developments address
all of the ambiguities, and that we are putting splints
and patches on an old horse.
What is proposed here is not a complete theory. What
is offered is a set of principles that might lay the foun-
dation for such a theory. In essence, these principles
3280 ROBERT V. ONEILL Ecology, Vol. 82, No. 12
P
e
r
s
p
e
c
t
i
v
e
s
constrain the theory to (1) be explicitly scaled, (2) in-
clude variability, (3) consider long-term sustainability
in addition to local stability, and (4) include population
processes as explicit system dynamics. The rst step
is to include the spatial context in the system denition.
1) An ecological system is composed of a range of
spatial scales, from the local system to the potential
dispersal range of all of the species within the local
system.
The recovery of the local ecological system depends
critically on the ability of individuals and populations
to disperse into an impacted area. Dispersal is a critical
ingredient in explaining the stability of ecological sys-
tems (Huffaker 1958). Therefore, the minimal area re-
quired to explain recovery is not the boundary of the
local ecosystem, but the dispersal range of its com-
ponent biota.
2) The potential dispersal range is set by (a) the
environmental constraints (biotic and abiotic) for each
species, by (b) dispersal barriers, and by (c) species
dispersal mechanisms.
The critical area is not the total distribution of the
species, but that portion of the range that is accessible
to the local ecological system of interest. A dispersal
barrier, e.g., human land use, may mean that the po-
tential dispersal range is much smaller than the total
distribution of the species. On the other hand, human
activities may expand the potential range by providing
new dispersal routes that permit the invasion of non-
indigenous species.
Thus, human use of space may have critical impacts
on stability by creating dispersal barriers. In constrict-
ing the potential dispersal range, society limits the total
range of environmental variability to which the local
ecosystem can respond stably. For example, a small
dam makes the upstream ecosystem unstable to any
uctuation that kills sh and other organisms that must
migrate back into the area to permit recovery. Here the
scaled impact is insidious because immediately follow-
ing construction of the dam, the upstream ecosystem
and its internal feedback mechanisms appear intact.
Thus, Homo sapiens is a keystone species, like the
beaver (Naiman et al. 1986) or starsh (Paine and Levin
1981), that alters the structure of its ecological system.
By introducing dispersal barriers, Homo sapiens ex-
cludes populations even though the site is within their
potential dispersal range. And by constructing invasion
pathways, Homo sapiens introduces populations that
would not otherwise occur. As a result, land use change
becomes a component of stability dynamics. Homo sa-
piens, as a dynamic component of the ecological sys-
tem, may have more far-reaching impacts on stability
than society viewed as extracting goods and services
from the local system.
3) The potential dispersal range is not constant or
uniform.
Over time, this potential dispersal range can change,
for example, with changes in climate. In addition, geo-
logic events may create or remove dispersal barriers.
These changes may happen slowly and monotonically,
permitting adaptive responses by component popula-
tions, or they may occur rapidly and produce cata-
clysmic changes. In addition, conditions within the po-
tential dispersal range are not uniform. They vary on
a large scale along environmental gradients, and they
vary locally due to soils, topography, aspect, etc. Thus,
the spatiotemporal variability that can be ignored with-
in the boundaries of the local system may be critical,
and must be explicitly considered before the stability
of the local system can be predicted.
4) Within the potential dispersal range is an effective
dispersal range that is time-scaled to the problem at
hand.
Typically, an individual study focuses on a limited
range of scales. For simplicity, consider a regularly
repeating impact on a local ecological system. The in-
terval between impacts determines an effective dis-
persal range, i.e., the distance over which the popu-
lations needed for recovery can move and become es-
tablished before the next impact. Over geologic time,
the effective dispersal range might equal the complete
potential dispersal range. For smaller scales or more
frequent impacts, the effective dispersal range would
be much smaller.
5) Within the local system, populations interact to
maximize biotic potential.
A local ecological system, such as a watershed, forest
stand, or lake, has three factors that determine dynam-
ics. First, the physical conditions at the site form a
constraint set that determines the dynamic potential.
Second, the potential biota are constrained to the pop-
ulations whose distributions overlap at this site. Within
these physical and biotic constraints, the populations
then interact to form complex networks and feedback
loops.
Competition and dominance determine what subset
of the potential biota actually function on the site and
may permit a variety of different states for the site. As
a result, the list of species currently resident on a site
is not necessarily sufcient to explain the range of
potential system responses to impact. Indeed, exibility
in the species list is critical to the sustainability of
ecological systems across the geologic scale of vari-
ability.
Competition operates within the constraints of phys-
ical laws, such as thermodynamics, and the interacting
populations move toward maximizing the biotic po-
tential of the site. The biotic potential forms an at-
tractor, and interactions, such as nutrient cycling, move
the set of populations toward this potential. But nutrient
cycling is an emergent property that cannot be ex-
plained by competition alone. Energy ow and nutrient
cycling operate on the same spatiotemporal scale as
competition, and it is the combination of the two dy-
namics that results in ecosystem function.
6) As the local ecological system approaches the
December 2001 3281 MACARTHUR AWARD LECTURE
P
e
r
s
p
e
c
t
i
v
e
s
biotic potential of a site, it increases local stability,
but decreases global sustainability.
As the system develops feedback loops and complex
networks of interactions, it becomes better adapted to
local conditions and more stable to local impacts. But,
over longer time scales, conditions change continu-
ously. So as the system approaches the present attractor
by maximizing biotic potential, it becomes suboptimal
to the changed conditions. A process analogous to spe-
cies specialization occurs.
The existence of local attractors becomes conspic-
uous at the ecotone. A small change in the environment
changes the competitive relationship between domi-
nants, and as disturbances destroy the existing vege-
tation and open the opportunity for new vegetation to
take over the site, a different system of interacting pop-
ulations gains a competitive advantage. The critical
observation is that a small change in conditions has
shifted the local attractor. The system that converged
to one attractor becomes suboptimal when the attractor
changes. It follows that the tendency to converge to a
local attractor may also make the system vulnerable to
changes in the attractor.
What maintains the sustainability of ecological sys-
tems over long periods of time is heterogeneity in time
and space. Heterogeneity is not an annoyance that com-
plicates experimental designs, it is a critical ingredient
in explaining the stability of ecological systems (Roff
1974a, b). Consider, as an extreme case, that a uniform
plain with no variation in time could result in the loss
of pioneer and successional species required for re-
covery when the inevitable impact does occur.
The consideration of heterogeneity recommends an
important change in ecosystem theory. The stability of
the system depends on two complementary and scaled
processes. Stability to smaller scale impacts depends
on the systems ability to resist change and recover
with resilience. But long-term stability or sustainability
depends on a exibility of response that can only be
maintained in an environment that varies in time and
space.
7) Stability of the local ecological system depends
on the time scale of observation and the critical balance
between (a) rates of change in environmental condi-
tion, and (b) rates of change in the biota.
Over short intervals of observation, environmental
variability is likely to be small and the local system
will appear stable. Over geologic time, the probability
of a catastrophic event approaches 1.0 (Crowley and
North 1988), and the local system is unstable. Even in
the absence of catastrophic events, gradual change in
the environment may eventually move the local system
across a critical threshold, change the local attractor,
and make the local system unstable. Thus, the relative
stability of two systems is not simply measured by the
rate of recovery from a disturbance, but also by the
expected length of time until the next catastrophic
event.
If rates of change in the conditions are slow, or the
recurrence interval of disturbances is long relative to
the rates of response by the local system, then the sys-
tem is stable. If rates of change are rapid or the re-
currence interval is short, relative to rates of response,
the ecological system is unstable. If rates of change are
nearly equal to rates of response, the system will appear
to be highly variable or even chaotic (Phillips 1996).
For this reason, human manipulations of the distur-
bance regime are just as important as disrupting the
structural and functional integrity of the local system
itself.
8) Stability of the local ecological system depends
on the spatial scale of observation and the critical
balance between (a) the size distribution of distur-
bances, and (b) the effective dispersal ranges of the
biota.
If the spatial extent of the ecological system under
consideration is large, for example, approaching the
size of an ecoregion, then only extremely large and
rare disturbances can overcome its ability to respond
stably. If a local system of small extent is being con-
sidered, for example, an isolated forest plot, then the
probability of a destabilizing event increases propor-
tionately.
If the spatial extent of disturbances is small relative
to the size of the ecological system, then spatial het-
erogeneity will be maintained within the effective dis-
persal ranges of biota, and the system will be stable.
If the disturbances are large and approach the effective
dispersal range of pioneer and successional species, the
system will be unstable. If the disturbance size and
effective dispersal ranges are nearly equal, the local
system will appear highly variable or chaotic.
Clearly, time and space scales are related (ONeill
1988). Stability depends on disturbance intervals rel-
ative to recovery rates, and the spatial extent of dis-
turbances relative to the spatial extent of the effective
dispersal range (Turner et al. 1993). The critical ob-
servation is that the internal interactions and feedback
mechanisms within the local system are only one of a
number of processes determining stability. The life his-
tory and dispersal ability of component populations,
and the heterogeneity of the landscape must also be
considered.
9) Homo sapiens is a keystone species that changes
system stability by altering environmental constraints,
rate processes, and biotic structures.
The impact of Homo sapiens is not limited to the
quantity of goods and services extracted from the eco-
system. The long-term impact of this keystone species
will likely be determined by the way it alters the sta-
bility properties of ecological systems.
Homo sapiens changes the frequency distribution of
disturbances. For example, the suppression of small
wildres changes the competitive advantage of re-
resistant dominants and alters biotic structure (Botkin
1990, Buell et al. 1954). Fire suppression also decreas-
3282 ROBERT V. ONEILL Ecology, Vol. 82, No. 12
P
e
r
s
p
e
c
t
i
v
e
s
es the small-scale heterogeneity of the landscape and
increases the probability of larger res. All of these
impacts change the stability properties of the system.
By fragmenting the spatial structure of the landscape,
Homo sapiens alters habitat connectivity and dispersal
rates (Gardner et al. 1993). By creating dispersal bar-
riers, this species decreases the potential dispersal
range of endemic species. By creating dispersal path-
ways, this species greatly increases the effective dis-
persal range of exotics (DAntonio and Vitousek 1992).
Changes in landscape structure, such as habitat loss,
can be particularly insidious if an extinction debt is
incurred, and observable species loss does not occur
until much later (Tilman et al. 1994, Loehle and Li
1996). Once again, these changes impact the stability
properties of the system even though they are not direct
effects on the interactions and feedback mechanisms
within the local ecological system.
Homo sapiens is changing the physical and chemical
constraint space within which ecological systems op-
erate. The average changes, e.g., in mean temperature,
may be small and result in minor changes in average
rate processes. But even small changes can switch the
local attractor and move the system to a new state.
Goldemberg et al. (1996) point out that even modest
increases in temperature can exclude crops that are in-
tolerant to the additional few extreme days. Similarly,
the temporal extremes of temperature at the spatial ex-
tremes of system distribution are likely to move eco-
tones by switching the local attractor.
The assertion that Homo sapiens is impacting the
stability properties of ecological systems is not extreme
or apocalyptic. Sackcloth and ashes are not required,
just a review of documented examples that have already
occurred (e.g., Loehle 1989): Extensive coral reefs
have been destabilized (Hughes 1994), and grazing has
destabilized semi-arid savannas (Walker et al. 1969,
Loehle 1985). In the extreme, overgrazing turns the
savanna to barren desert (Hills 1966). Nutrient addi-
tions have destabilized oligotrophic lake ecosystems
(Recknagel 1985, Carpenter et al. 1998). Harvesting
has destabilized competitive interactions leading to the
extinction of many sh populations (Watt 1968, Jones
and Walters 1976). Extensive agriculture destabilized
large areas of the United States to drought, resulting
in massive erosion and the Dust Bowl (Kahn 1995).
Theoretical studies provide the explanation for such
destabilization in complex ecological systems in gen-
eral (May 1977), and ecosystems in particular (ONeill
et al. 1982, 1989). Tainter (1988) points to the possi-
bility of similar mechanisms in the collapse of human
societies.
The fundamental problem is that Homo sapiens is
moving ecological systems outside the envelope of
conditions that have existed over evolutionary history.
This is terra incognita and the assumption that ecolog-
ical systems will respond stably is unjustied.
CONCLUSIONS
Is it time to bury the ecosystem concept? Probably
not. But there is certainly need for improvement before
ecology loses any more credibility. This paper suggests
some of the key problems. Spatial pattern, extent, and
heterogeneity are critical to stability. You cannot get a
predictive theory if you assume them away. Temporal
variability and scale are critical to stability. You cannot
get a predictive theory if you assume them away either.
It is the interplay of natural selection and internal feed-
back mechanisms that determines dynamics. Again,
you cannot get a predictive theory if you assume either
away. Basically, all the processes and constraints need-
ed to explain stability are not encompassed within the
boundaries of the local ecological system.
An improved paradigm would have many implica-
tions for ecological applications, such as conservation.
Increasing the size of an isolated preserve only in-
creases the length of time until the cumulative prob-
ability of a disruption approaches 1.0. Maintaining dis-
persal pathways might better conserve sustainability
by keeping the potential dispersal range near its orig-
inal, undisturbed scale.
There are also important implications for monitoring.
Current theory leads us to focus on average rates and
standing crops at a location. Yet scale and variability
in space and time may be more important in determin-
ing sustainability. Mean values at two locations may
indicate that no signicant change has occurred, but if
dispersal pathways between the sites have been dis-
rupted, one has reduced by orders of magnitude the
scale of a catastrophic disturbance.
Perhaps the most important implication involves our
view of human society. Homo sapiens is not an external
disturbance, it is a keystone species within the system.
In the long term, it may not be the magnitude of ex-
tracted goods and services that will determine sustain-
ability. It may well be our disruption of ecological
recovery and stability mechanisms that determines sys-
tem collapse.
Certainly, we dont want to dismiss the current the-
ory prematurely. But we must understand that the ma-
chine analogy is critically limited. In so far as the local
system maximizes environmental potential, it neces-
sarily sacrices stability when that potential changed.
The challenge to the ecological system is optimization
to a moving target. Optimize too rapidly and the system
is trapped in a local attractor and, like an overspecial-
ized species, cannot adapt when conditions change. So,
it would not be wise to send the old dobbin to the glue
factory before we determine how well the new one
takes the bit. But it certainly seems to be time to start
shopping for a new colt.
ACKNOWLEDGMENTS
Oak Ridge National Laboratory is operated for the U.S.
Department of Energy by UT-Battelle under contract number
December 2001 3283 MACARTHUR AWARD LECTURE
P
e
r
s
p
e
c
t
i
v
e
s
DE-AC05-00OR22725. This is Environmental Sciences Di-
vision publication no. 5070.
LITERATURE CITED
Arrhenius, E., and T. W. Waltz. 1990. The greenhouse effect:
implications for economic development. Discussion Paper
number 78. World Bank, Washington, D.C., USA.
Bode, H. W. 1945. Network analysis and feedback amplier
design. Van Nostrand Reinhold, Princeton, New Jersey,
USA.
Botkin, D. 1990. Discordant harmonies. Oxford University
Press, New York, New York, USA.
Buell, M. F., H. F. Buell, and J. A. Small. 1954. Fire in the
history of Mettlers woods. Torreya 81:253255.
Callicott, J. B. 1996. Do deconstructive ecology and socio-
biology undermine Leopolds land ethic? Environmental
Ethics 18:369372.
Carpenter, S., N. F. Caraco, D. L. Correll, R. W. Howarth, A.
N. Sharpley, and V. H. Smith. 1998. Nonpoint pollution
of surface waters with phosphorus and nitrogen. Issues in
Ecology Number 3. Ecological Society of America, Wash-
ington, D.C., USA.
Clements, F. E. 1897. Peculiar zonal formations of the Great
Plains. American Naturalist 31:968970.
Clements, F. E. 1905. Research methods in ecology. Uni-
versity Printing, Lincoln, Nebraska, USA.
Clements, F. E. 1916. Plant succession: an analysis of the
development of vegetation. Carnegie Institute of Washing-
ton Publication 242:1512.
Crowley, T. J., and G. R. North. 1988. Abrupt climate change
and extinction events in earth history. Science 240:996
1002.
DAntonio, C. M., and P. M. Vitousek. 1992. Biological in-
vasions by exotic grasses, the grass/re cycle and global
change. Annual Review of Ecology and Systematics 23:
6387.
Davis, M. B. 1976. Pleistocene biogeography of temperate
deciduous forests. Geoscience and Man 13:1326.
Delcourt, P. A., and H. R. Delcourt. 1987. Long-term forest
dynamics of the temperate zone. Springer-Verlag, New
York, New York, USA.
Dublin, H. T. 1995. Vegetation dynamics of the Serengeti-
Mara ecosystem: the role of elephants, re and other fac-
tors. Pages 7190 in A. R. E. Sinclair and P. Arcese, editors.
Serengeti II: dynamics, management, and conservation of
an ecosystem. University of Chicago Press, Chicago, Illi-
nois, USA.
Dublin, H. T., A. R. E. Sinclair, and J. McGlade. 1990. El-
ephants and re as causes of multiple stable states in the
Serengeti-Mara woodlands. Journal of Animal Ecology 59:
11471164.
Elton, C. 1930. Animal ecology and evolution. Clarendon
Press, Oxford, UK.
Forbes, S. A. 1925. The lake as a microcosm. Illinois Natural
History Survey Bulletin 15.
Gardner, R. H., R. V. ONeill, and M. G. Turner. 1993. Eco-
logical implications of landscape fragmentation. Pages
208226 in M. J. McDonnell and S. T. A. Pickett, editors.
Humans as components of ecosystems. Springer-Verlag,
New York, New York, USA.
Goldemberg, J., R. Squitieri, J. Stiglitz, A. Amano, X. Shaox-
iong, and R. Saha. 1996. Introdution: scope of the assess-
ment. Pages 2151 in J. P. Bruce, H. Lee, and E. F. Haites,
editors. Climate change 1995. Volume 3: Economic and
social dimensions of climate change. Cambridge University
Press, Cambridge, UK.
Griggs, R. F. 1914. A botanical survey of the Sugar Grove
region. Ohio Biological Survey Bulletin 1:248340.
Hansen, A. J., F. di Castri, and R. J. Naiman. 1994. A new
look at ecotones. Biology International 17:946.
Hills, E. S., editor. 1966. Arid lands: a geographical ap-
praisal. Methuen, London, UK.
Hobbs, E. R. 1986. Characterizing the boundary between
California annual grassland and coastal sage scrub with
differential proles. Vegetatio 65:115126.
Holden, J. P. 1991. Population and the energy problem. Pop-
ulation and Environment 12:231255.
Holling, C. S. 1973. Resilience and stability of ecological
systems. Annual Review of Ecology and Systematics 4:1
24.
Howarth, R. W., G. Billen, D. Swaney, A. Townsend, N.
Jaworski, K. Lajtha, J. A. Downing, R. Elmgren, N. Caraco,
T. Jordan, F. Berendse, J. Freney, V. Kudeyarov, P. Mur-
doch, and Z. Zhao-liang. 1996. Regional nitrogen budgets
and riverine N & P uxes for the drainages to the North
Atlantic Ocean: natural and human inuences. Biogeo-
chemistry 35:75139.
Huffaker, C. B. 1958. Experimental studies on predation:
dispersion factors and predator-prey oscillations. Hilgardia
27:343383.
Hughes, T. P. 1994. Catastrophes, phase shifts, and large-
scale degradation of a Caribbean coral reef. Science 265:
15471551.
Huston, M. A. 1994. Biological diversity: the coexistence of
species on changing landscapes. Cambridge University
Press, New York, New York, USA.
Hutchinson, G. E. 1948. Circular causal systems in ecology.
Annals of the New York Academy of Science 50:221246.
Jones, D. D., and C. J. Walters. 1976. Catastrophe theory
and sheries regulation. Journal of the Fisheries Research
Board of Canada 33:28292833.
Kahn, J. R. 1995. The economic approach to environmental
and natural resources. Dryden Press, Orlando, Florida,
USA.
Kuhn, T. S. 1962. The structure of scientic revolutions.
University of Chicago Press, Chicago, Illinois, USA.
Levin, S. A. 1999. Fragile dominion. Perseus Books, Read-
ing, Massachusetts, USA.
Livingstone, B. E. 1903. The distribution of the upland plant
societies of Kent County, Michigan. Botanical Gazette 35:
3655.
Loehle, C. 1985. Optimal stocking levels for semi-desert
range: a catastrophe theory model. Ecological Modelling
27:285297.
Loehle, C. 1989. Catastrophe theory in ecology: a critical
review and an example of the buttery catastrophe. Eco-
logical Modelling 49:125152.
Loehle, C., and B. Li. 1996. Habitat destruction and the ex-
tinction debt revisited. Ecological Applications 6:784789.
Marsh, G. P. 1864. Man and nature: physical geography as
modied by human action. Belknap Press, Cambridge,
Massachusetts, USA.
May, R. M. 1977. Thresholds and breakpoints in ecosystems
with a multiplicity of stable states. Nature 269:471477.
Myers, N., and J. Simon. 1994. Scarcity or abundance. Nor-
ton, New York, New York, USA.
Naiman, R. J., J. M. Melillo, and J. E. Hobbie. 1986. Eco-
system alteration of boreal forest streams by beaver. Ecol-
ogy 67:12541269.
Naveson, J. 1993. Humanism for humans. Free Inquiry
Spring 1993:24.
Nicholson, A. J., and V. A. Bailey. 1935. The balance of
animal populations. Proceedings of the Zoological Society
of London 105:551598.
Odum, E. P. 1953. Fundamentals of ecology. W. B. Saunders,
Philadelphia, Pennsylvania, USA.
Odum, H. T. 1971. Environment, power and society. Wiley-
Interscience, New York, New York, USA.
Olson, J. S. 1963. Analog computer models for movement
of nuclides through ecosystems. Pages 121125 in V.
3284 ROBERT V. ONEILL Ecology, Vol. 82, No. 12
P
e
r
s
p
e
c
t
i
v
e
s
Schultz and A. W. Klements, editors. Radioecology. Van
Nostrand Reinhold, Cincinnati, Ohio, USA.
ONeill, R. V. 1988. Hierarchy theory and global change.
Pages 2945 in T. Rosswall, R. G. Woodmansee, and P. G.
Risser, editors. Spatial and temporal variability in bio-
spheric and geospheric processes. John Wiley & Sons, New
York, New York, USA.
ONeill, R. V., R. H. Gardner, and D. E. Weller. 1982. Chaotic
models as representations of ecological systems. American
Naturalist 120:259263.
ONeill, R. V., A. R. Johnson, and A. W. King. 1989. A
hierarchical framework for the analysis of scale. Landscape
Ecology 3:193205.
Paine, R. T., and S. A. Levin. 1981. Intertidal landscapes:
disturbance and the dynamics of pattern. Ecological Mono-
graphs 51:145178.
Phillips, J. D. 1996. Time lags and emergent stability in
morphogenic/pedogenic system models. Ecological Mod-
elling 78:267276.
Pickett, S. T. A., V. T. Parker, and P. L. Fiedler. 1992. The
new paradigm in ecology: implications for conservation
above the species level. Pages 6588 in P. L. Fiedler and
Jain, editors. Conservation biology: the theory and practice
of nature conservation, preservation and management.
Chapman & Hall, New York, New York, USA.
Pimentel, D., editor. 1993. World soil erosion and conser-
vation. Cambridge University Press, Cambridge, UK.
Pimm, S. L. 1984. The complexity and stability of ecosys-
tems. Nature 307:321326.
Recknagel, F. 1985. Analysis of structural stability of aquatic
ecosystems as an aid for ecosystem control. Ecological
Modelling 27:221234.
Roff, D. A. 1974a. Spatial heterogeneity and the persistence
of populations. Oecologia 15:245258.
Roff, D. A. 1974b. The analysis of a population model dem-
onstrating the importance of dispersal in a heterogeneous
environment. Oecologia 15:259275.
Sagoff, M. 1997. Muddle or muddle through? College of
William and Mary Law Reviews 38:825993.
Schorger, A. W. 1955. The Passenger Pigeon. University of
Wisconsin Press, Madison, Wisconsin, USA.
Shrader-Frechette, K. S. 1995. Practical ecology and foun-
dations for environmental ethics. Journal of Philosophy 92:
621.
Shrader-Frechette, K. S., and E. D. McCoy. 1993. Method
in ecology. Cambridge University Press, London, UK.
Simon, J. 1980. Resources, population, environment: and
oversupply of false bad news. Science 208:14351436.
Soule, M., and G. Lease. 1995. Reinventing nature? Island
Press, Washington, D.C., USA.
Sutherland, J. P. 1974. Multiple stable points in natural com-
munities. American Naturalist 108:859873.
Tainter, J. A. 1988. The collapse of complex societies. Cam-
bridge University Press, New York, New York, USA.
Tansley, A. G. 1935. The use and abuse of vegetational con-
cepts and terms. Ecology 16:284307.
Teal, J. M. 1962. Energy ow in the salt marsh ecosystem
of Georgia. Ecology 43:614624.
Tilman, D., R. M. May, C. L. Lehman, and M. A. Nowak.
1994. Habitat destruction and the extinction debt. Nature
371:6566.
Turner, M. G., W. H. Romme, R. H. Gardener, R. V. ONeill,
and T. K. Kratz. 1993. A revised concept of landscape
equilibrium: disturbance and stability on scaled landscapes.
Landscape Ecology 8:213227.
Van Dyne, G. M., editor. 1969. The ecosystem concept in
natural resource management. Academic Press, New York,
New York, USA.
Vitousek, P. M. 1994. Beyond global warming: ecology and
global change. Ecology 75:18611876.
Vitousek, P. M., J. Aber, R. W. Howarth, G. E. Likns, P. A.
Matson, D. W. Schindler, W. H. Schlesinger, and G. D.
Tilman. 1997. Human alteration of the global nitrogen cy-
cle: causes and consequences. Ecological Applications 7:
737750.
Vitousek, P. M., P. R. Ehrlich, A. H. Erlich, and P. A. Matson.
1986. Human appropriation of the products of photosyn-
thesis. BioScience 36:368373.
Waide, J. B., and J. R. Webster. 1976. Engineering systems
analysis: applicability to ecosystems. Pages 329371 in B.
C. Patten, editor. Systems analysis and simulation in ecol-
ogy. Academic Press, New York, New York, USA.
Walker, B. H., D. Ludwig, C. S. Holling, and R. M. Peterman.
1969. Stability of semi-arid savanna grazing systems. Ecol-
ogy 69:473498.
Watt, K. E. F. 1968. Ecology and resource, Management.
McGraw-Hill, New York, New York, USA.
Webster, J. R., J. B. Waide, and B. C. Patten. 1974. Nutrient
cycling and the stability of ecosystems. Pages 127 in F.
G. Howell, J. B. Gentry, and M. H. Smith, editors. Mineral
cycling in Southeastern ecosystems. ERDA SymposiumSe-
ries, CONF-740513.