Content uploaded by Emily Nicholson
Author content
All content in this area was uploaded by Emily Nicholson on Jan 22, 2018
Content may be subject to copyright.
1
© The Ecological Society of America www.frontiersinecology.org
REVIEWS REVIEWS REVIEWS
Developing a standardized definition of
ecosystem collapse for risk assessment
Lucie M Bland1,2*, Jessica A Rowland1, Tracey J Regan2,3, David A Keith4,5,6, Nicholas J Murray4,
Rebecca E Lester7, Matt Linn1, Jon Paul Rodríguez8,9,10, and Emily Nicholson1
The International Union for Conservation of Nature (IUCN) Red List of Ecosystems is a powerful tool for
classifying threatened ecosystems, informing ecosystem management, and assessing the risk of ecosystem
collapse (that is, the endpoint of ecosystem degradation). These risk assessments require explicit definitions of
ecosystem collapse, which are currently challenging to implement. To bridge the gap between theory and
practice, we systematically review evidence for ecosystem collapses reported in two contrasting biomes –
marine pelagic ecosystems and terrestrial forests. Most studies define states of ecosystem collapse quantitatively,
but few studies adequately describe initial ecosystem states or ecological transitions leading to collapse. On the
basis of our review, we offer four recommendations for defining ecosystem collapse in risk assessments: (1)
qualitatively defining initial and collapsed states, (2) describing collapse and recovery transitions, (3) identify-
ing and selecting indicators of collapse, and (4) setting quantitative collapse thresholds.
Front Ecol Environ 2018; doi: 10.1002/fee.1747
Ecosystems are dynamic by nature, but concern arises
when they undergo substantial loss of biodiversity and
re- organization of ecological processes (Scheffer et al.
2001). Such large detrimental changes, collectively
termed “ecosystem collapse” (see Panel 1 for a glossary of
terms), have important implications for conserving biodi-
versity and maintaining ecosystem services, and are funda-
mental to assessing risks to ecosystems (Keith et al. 2013).
Understanding the risks of ecosystem collapse is critical
to ecosystem management, and requires consideration of
an ecosystem’s exposure and vulnerability to various haz-
ards (Burgman 2005). With respect to biodiversity con-
servation, two tools are commonly used to assess risks to
ecosystems and species: Red Lists (decision- rule- based
protocols) and probabilistic models. Red Lists assign eco-
systems to ranked categories of risk (eg Vulnerable,
Endangered, or Critically Endangered) based on decision
rules that incorporate multiple symptoms of threat expo-
sure and vulnerability, such as rates of decline in spatial
and functional indicators (Nicholson et al. 2009). Red
Lists for both ecosystems and species have strong theoret-
ical foundations (Burgman 2005; Mace et al. 2008; Keith
et al. 2013), and ecosystem Red Lists have been imple-
mented in many countries, including Finland, South
Africa, and Australia (Nicholson et al. 2009). In contrast,
probabilistic models quantitatively estimate the risk of
ecosystem collapse based on a mathematical representa-
tion of ecosystem dynamics, threats, and social–ecologi-
cal relationships (Bland et al. 2017). The International
Union for Conservation of Nature (IUCN) Red List of
Ecosystems was endorsed by IUCN in 2014 and is the
only global protocol for assessing risks to ecosystems. The
protocol is based on five rule- based criteria, one of which
pertains to estimating the probability of collapse through
models (Keith et al. 2013).
Risk is defined as the probability of an adverse outcome
within a specified time frame (Burgman 2005). Whether
using decision rules or probabilistic models, defining the
characteristics of a collapsed ecosystem is essential to esti-
mating risk. In the absence of a clear theoretical frame-
work and practical recommendations, defining collapse is
often perceived as judgement- laden and impractical
(Boitani et al. 2015; Cumming and Peterson 2017). Early
1School of Life and Environmental Sciences, Deakin University,
Burwood, Australia *(l.bland@deakin.edu.au); 2School of
BioSciences, The University of Melbourne, Parkville,
Australia; 3The Arthur Rylah Institute for Environmental
Research, Department of Environment, Land, Water and Planning,
Heidelberg, Australia; 4Centre for Ecosystem Science, School of
Biological, Earth and Environmental Science, University of New
South Wales, Kensington, Australia; 5New South Wales Office of
Environment and Heritage, Hurstville, Australia; 6Long Term
Ecological Research Network, Terrestrial Ecosystem Research
Network, Australian National University, Canberra,
Australia; 7Centre for Rural and Regional Futures, Deakin
University, Waurn Ponds, Australia; continued on last page
In a nutshell:
• The difficulty of defining ecosystem collapse has challenged
the classification of threatened ecosystems
• We reviewed 85 studies of collapse in two biomes to
inform the design of a robust framework to better define
ecosystem collapse
• Most studies defined collapsed ecosystem states quantita-
tively, but many lacked a description of ecosystem processes
leading to collapse
• Our recommended framework can be applied to define
ecosystem collapse in IUCN Red List of Ecosystems
assessments and national ecosystem risk assessments
2
www.frontiersinecology.org © The Ecological Society of America
LM Bland et al.How to define ecosystem collapse
ecosystem assessment protocols defined collapsed ecosys-
tem states poorly (Nicholson et al. 2015), severely limit-
ing the consistency and robustness of risk assessments.
Defining collapsed states can be difficult because ecosys-
tem collapse is expressed through symptoms that may
vary across ecosystems and scales of investigation, and
may be characterized by subtle rather than clear- cut
changes. For instance, collapse of mountain ash
(Eucalyptus regnans) forests in southeast Australia is char-
acterized by the loss of large cavity- bearing trees, and not
just through reductions in the forests’ distributional
extent (Burns et al. 2015). Decisions on whether to clas-
sify an ecosystem as collapsed depend on the objectives of
the risk assessment and on the needs of the decision mak-
ers. Accordingly, lists of threatened ecosystems are
focused on averting the loss of characteristic biota and
ecological function (eg Keith et al. 2013).
The first step in ecosystem risk assessment is to describe
initial or baseline states that reflect the natural range of
spatial and temporal variability in ecosystems (Panel 1).
In the IUCN Red List of Ecosystems, these baselines are
defined for three different time frames (“current”,
“future”, and “historic”; Keith et al. 2013) and provide
important contextual information for understanding how
the defining features of an ecosystem change during a
transition to collapse (Sato and Lindenmayer 2017). The
second step is to identify potential pathways of collapse
and symptoms of degradation (Scheffer et al. 2001). This
step can be informed by ecological models (textual,
diagrammatic, or mathematical), and is key to selecting
indicators (Panel 1) that reflect changes in ecosystem
states (Rumpff et al. 2011) and that can be used as proxies
for risk in Red Lists or probabilistic models. In the third
step, collapsed states should be defined with quantitative
decision thresholds (Panel 1) in key indicators, which
can be informed by observation, experimentation, mode-
ling, or expert elicitation. Uncertainty in the resulting
thresholds may be substantial and is quantifiable with
upper and lower bounds (Keith et al. 2013).
Despite challenges in defining ecosystem collapse, a
large amount of evidence exists for collapses in a variety
of ecosystems (Washington 2013), but much of this
evidence – which could help inform ecosystem risk
assessment – has yet to be synthesized. For example, over-
harvesting and burning of forest ecosystems on Easter
Island in the 16th century CE led to the extinction of the
foundation (ie habitat- forming) species of palm and to
the transformation of forests to grasslands, with extreme
consequences for the human population (Diamond
2007). Similarly, water extraction from the Aral Sea
caused a 92% reduction in water volume between 1960
and 2010, leading to the extirpation of most fish and
invertebrate species, the disappearance of reed beds and
associated waterbirds, and a transformation to saline lakes
and desert plains (Micklin 2010). Such extreme transfor-
mations, and losses of defining biological and environ-
mental features, are characteristic of ecosystem collapse
as it is conceptualized in ecosystem risk assessments.
Panel 1. Glossary
Bounded threshold: Represents uncertainty in the occurrence of the collapsed state based on two or more values of an eco-
system indicator (Bland et al. 2016).
Decision threshold: Value of an indicator above or below which a decision differs (Martin et al. 2009), such as a decision to list
an ecosystem as collapsed.
Ecological model: Written, pictorial, or mathematical representation of key ecosystem components and processes, which
effectively summarizes ecosystem dynamics to a broad audience (Suter 1999).
Ecological threshold: Value of an indicator above or below which non- linear or specific changes in ecosystem dynamics occur.
For example, small changes in an environmental indicator can produce disproportionately large responses in biotic indicators (Mac
Nally et al. 2014).
Ecosystem: Although definitions vary, the IUCN Red List of Ecosystems defines ecosystems as: a biotic complex or assemblage of
species, an associated abiotic environment or complex, the interactions within and between those complexes, and a physical space
in which they operate (Bland et al. 2016). These four elements assist in identifying and classifying ecosystems and understanding their
susceptibility to threats.
Ecosystem collapse: Indicates a transition beyond a bounded threshold in one or more indicators that define the identity and
natural variability of the ecosystem (Bland et al. 2016). Collapse involves a transformation of identity, loss of defining features, and/
or replacement by a novel ecosystem. It occurs when all ecosystem occurrences (ie patches) lose defining biotic or abiotic features,
and characteristic native biota are no longer sustained.
Indicator: Metrics that quantify complex changes in ecosystem structure, composition, and function (Niemeijer and de Groot
2008). In different risk assessment protocols, spatial, biotic, or abiotic indicators may quantify threats to ecosystems and/or ecosys-
tem responses to threats. Thresholds in indicators indicate ecosystem collapse.
Natural variability: The ecological conditions, and the spatial and temporal variation in these conditions, within a period of time
and geographical area appropriate for the study objectives (Keane et al. 2009).
State: A numerical description of multiple biotic components of an ecosystem, typically including values of species abundances or
biomasses and of ecosystem processes, such as primary production and respiration (Bestelmeyer et al. 2003).
Transition: Can describe both a change in ecosystem state, and the value of a driver at which the change in ecosystem state occurs
(Bestelmeyer et al. 2003).
3
© The Ecological Society of America www.frontiersinecology.org
LM Bland et al. How to define ecosystem collapse
Here, we bridge the gap between theory and practice by
critically examining how ecologists have defined ecosys-
tem collapse in two globally important biomes – marine
pelagic ecosystems and temperate forests – in part by
exploring the ecosystem features that were studied and
the methods used. We then present a standardized frame-
work for defining ecosystem collapse, based on the out-
comes from the review and the needs of ecosystem risk
assessments such as the IUCN Red List of Ecosystems
(Bland et al. 2016). Importantly, our review neither
defines ecosystem collapse (but see Panel 1 and Bland
et al. 2016 for a definition) nor directly addresses the
decision rules or models used in ecosystem risk
assessments; rather, it focuses explicitly on developing a
systematic method for defining collapse so that ecosystem
risk assessments may be applied more effectively.
JLiterature review
We systematically reviewed the scientific literature on
ecosystem collapse using a standard method (Pickering
and Byrne 2014) that complies with the guidelines of
the Preferred Reporting Items for Systematic reviews and
Meta- Analyses (PRISMA) statement (Moher et al. 2009).
We reviewed publications on marine pelagic and tem-
perate forest ecosystems that reported ecosystem collapses,
regime shifts, or trophic cascades with a strong focus on
loss of biodiversity and ecosystem function (rather than
economic or societal losses; Cumming and Peterson 2017),
to conform to the objectives of most ecosystem risk as-
sessment protocols (Nicholson et al. 2009).
We searched Web of Science and Science Direct on 9
Sep 2016 with standardized search terms (WebPanel 1).
We screened papers based on abstracts and then full text
according to set criteria (WebPanel 1), recording the num-
ber of papers retained at each screening stage according to
the PRISMA statement (WebTable 1). Our final selection
included 35 publications reporting collapses in 37 marine
pelagic ecosystems (hereafter referred to as “studies”) and
48 publications reporting collapses in 48 temperate forest
ecosystems (Figures 1 and 2). In those publications, we
reviewed: (1) what research methods were employed to
identify collapse; (2) whether initial and collapsed states
were described, and what ecosystem features were exam-
ined; (3) whether studies used ecological models to
describe pathways to collapse; (4) what mechanisms were
involved in the transition to collapse; (5) what variables
were identified as useful indicators of collapse; (6) whether
studies defined quantitative thresholds of collapse; and (7)
whether studies accounted for uncertainty in their defini-
tions of collapse.
We found significant differences between biomes in
research methods used to define ecosystem collapse
(WebTable 2). Spatial comparisons between initial and
collapsed states were applied more often in temperate for-
est studies (21%) than in marine pelagic studies (0%),
where temporal comparisons were always used. In addition,
21% and 49% of temperate forest and marine pelagic stud-
ies, respectively, described all four features of initial eco-
system states required in IUCN Red List of Ecosystems
assessments (biota, abiotic environment, processes, and
spatial distribution; Bland et al. 2016). Conversely, col-
lapsed states were quantitatively described more often in
temperate forests (79%) than in marine pelagic ecosys-
tems (65%). Approximately one- half and three- quarters
of temperate forest studies (54%) and marine pelagic stud-
ies (78%), respectively, relied on ecological models of
transitions to collapse. In both biomes, text descriptions of
ecological processes were more often used than concep-
tual diagrams (WebTable 2). Trophic restructuring (43%)
and climatic shifts (40%) were the most common transi-
tions to collapse in marine pelagic ecosystems, whereas
distribution shifts (52%) and climatic shifts (23%) were
more common transitions to collapse in temperate forests.
The application of indicators and quantitative thresh-
olds to measure transitions to collapse differed
significantly between temperate forests and pelagic
ecosystems. Spatial indicators (eg distribution size, distri-
bution limits) were used in 60% of temperate forest
studies, whereas biotic or abiotic indicators were used in
100% of marine pelagic studies (WebTable 2). Spatial
thresholds of collapse were frequently quantified in
temperate forests (90%), but not in marine pelagic
ecosystems. Collapse thresholds in biotic and abiotic
indicators were quantified in most studies, except for
biotic indicators in temperate forest studies (31%). All
marine pelagic studies (100%) and almost all temperate
forest studies (94%) accounted for uncertainty by meas-
uring collapse with multiple indicators. One study
accounted for uncertainty in ecological models and none
Figure 1. Cumulative publication numbers through time, for
marine pelagic (n = 35; blue line) and temperate forest (n = 48;
green line) publications selected for the systematic literature review.
No papers published prior to 1995 met the selection criteria.
4
www.frontiersinecology.org © The Ecological Society of America
LM Bland et al.How to define ecosystem collapse
accounted for uncertainty in quantitative collapse
thresholds.
JDiscussion
Consistent methods for defining ecosystem collapse are
needed to meet increasing demand for tools to monitor
the status of biodiversity from local to global scales
(Keith et al. 2013; CBD 2014; Nicholson et al. 2015).
Of the marine pelagic and temperate forest studies
reviewed here, most defined quantitative collapse thresh-
olds with empirical data or predictive models, but often
failed to adequately define features of the initial eco-
system state and ecological processes leading to collapse,
especially for temperate forests. The use of conceptual
models was more common in marine pelagic ecosystems,
illustrating a stronger focus on ecosystem functioning
compared to forests. Indicator selection protocols were
also applied more often in marine ecosystems, where
clear guidelines have been developed to inform ecosystem-
based management of fisheries (Rice and Rochet 2005).
Marine pelagic studies quantified collapse with biotic
and abiotic indicators, while temperate forest studies
typically quantified collapse with spatial and biotic in-
dicators, reflecting a focus on multiple trophic levels
in marine pelagic ecosystems (eg plankton, planktivorous
fish, and piscivorous fish) and on the distribution of
foundation tree species in temperate forests.
On the basis of our literature review, we propose a com-
mon, systematic framework for defining ecosystem collapse
in four key steps: (1) qualitatively defining initial and col-
lapsed states, (2) describing collapse and recovery transi-
tions, (3) identifying and selecting indicators of collapse,
and (4) setting quantitative collapse thresholds. Such a
framework can be applied to representative marine pelagic
and temperate forest ecosystems
(WebPanel 2) as well as to other
types of ecosystems not reviewed here
(Figure 3). Our recommendations are
particularly relevant to the IUCN
Red List of Ecosystems (Keith et al.
2013), but can inform other risk
assessment protocols and manage-
ment tools for ecosystems.
(1) Qualitatively defining initial
and collapsed states
We found that few studies described
initial ecosystem states. This fun-
damental omission makes it harder
to identify characteristic ecosystem
features, infer patterns of natural
variability as being distinct from
directional change, and select ade-
quate indicators of ecosystem
change. Most ecosystem assessment
protocols specify baselines to quantify initial ecosystem
states (Nicholson et al. 2009). Indeed, the IUCN pro-
tocol specifies three temporal baselines: historic (since
1750 CE), current (within the past 50 years), and future
(50 years from the present day or “any 50- year period
including the present and future”) (Keith et al. 2013).
An ecosystem’s initial state may be characterized by a
large degree of uncertainty (especially with respect to
historical baselines), which may affect the definition of
key ecological features and processes lost during a tran-
sition to collapse (Figure 2 in Keith et al. 2015). It is
therefore important to examine the sensitivity of as-
sessment outcomes to the plausible range of initial states.
The majority of studies defined ecosystem collapse
quantitatively, suggesting that there may be greater
consensus on collapsed states than on initial ecosystem
states. For example, “limit reference points” are often
defined in marine ecosystem management (Cury et al.
2005), while “thresholds of probable concern” are de-
fined in river management (Rogers and Biggs 1999).
These quantitative definitions reflect the large amounts
of empirical data on collapsed ecosystems in specific
biomes (as opposed to a perceived lack of data in more
general reviews; Sato and Lindenmayer 2017), which
can inform the implementation of ecosystem risk as-
sessment worldwide. Identifying intermediate or transi-
tion states toward collapse can be informative in some
ecosystems (eg woodlands; Rumpff et al. 2011), but no
studies identified intermediate states in our review.
Direct evidence of historical and geographic variation
can help to establish bounds of natural variability in eco-
system features (Keane et al. 2009). Most studies relied on
temporal rather than spatial comparisons to define ecosys-
tem states, suggesting that time- series analysis is a preferred
method to detect deviations from natural variability
Figure 2. Spatial distribution of marine pelagic studies (n = 37; blue circles) and temperate
forest studies (n = 41; green circles), with circle size indicating the number of studies. One
marine pelagic publication included studies of three different ecosystems. Seven temperate
forest studies of global extent were not mapped. Studies focusing on all ecosystems from a
certain biome within a country were mapped as the centroid of the country.
5
© The Ecological Society of America www.frontiersinecology.org
LM Bland et al. How to define ecosystem collapse
(eg with sequential t-test algorithms;
Litzow and Mueter 2014). Only two
temperate forest studies used infor-
mation on locally collapsed patches,
although these comparisons can
inform clear definitions of collapsed
states. For instance, invasion by arc-
tic foxes (Alopex lagopus) reduced
populations of seabirds and nutrient
transport on some Aleutian Islands
and led to major vegetation transfor-
mations from grasslands to tundra
(Croll et al. 2005), providing a com-
parative framework between fox- free
and fox- infested islands to define
ecosystem collapse. Analogous but
collapsed ecosystems can provide a
basis for delineating collapsed states
in a focal ecosystem. For instance,
collapse thresholds for the extant
southern Benguela upwelling ecosys-
tem can be inferred from the col-
lapsed northern Benguela, which
underwent a regime shift in the 1970s
due to overfishing and environmen-
tal pressures (Roux et al. 2013). For
ecosystems dominated by foundation
species, environmental tolerances of
those species can be used as proxies to
define the abiotic components of ini-
tial states (eg bioclimatic correlates of
the distribution of temperate forests;
WebPanel 2; Peñuelas and Boada
2003). Few studies relied on experi-
ments to define collapsed states, most
likely due to the difficulty of manipu-
lating large interconnected ecosys-
tems as compared to relatively smaller or isolated estuaries
and lakes (Scheffer et al. 2001; Mac Nally et al. 2014). No
studies involved expert elicitation to define ecosystem
states, although expert- derived data can help identify
transitions to collapse (Rumpff et al. 2011).
(2) Describing collapse and recovery transitions
A qualitative understanding of ecosystem processes is es-
sential for defining transitions to collapse and selecting
indicators of ecosystem change (Bland et al. 2016), yet
many studies did not employ ecological models to describe
transitions. For example, most vegetation distribution
models predict the presence or absence of suitable envi-
ronmental conditions (eg temperature, precipitation), but
do not explicitly describe ecosystem processes leading to
collapse (Feng et al. 2013). Understanding the pathways
to collapse helps to identify intermediate ecosystem states,
indicators of risk, and corresponding collapse thresholds,
especially for ecosystems threatened by declines in eco-
logical function such as changes in fire or hydro-
logical regimes, or species invasions (Nicholson et al.
2015).
Representations of ecosystem dynamics such as concep-
tual diagrams are particularly useful for risk assessment,
given that these can effectively summarize initial and
collapsed states, clearly depict assumptions and uncer-
tainties about ecosystem processes, and are less prone to
semantic uncertainties than written descriptions (Suter
1999). Two types of conceptual diagrams are commonly
used in risk assessment: state- and- transition models
(which explicitly depict transitions between ecosystem
states based on various drivers) and cause–effect models
(which depict interactions and dependencies among eco-
system components and processes; Bland et al. 2016).
Uncertainty in ecological models potentially affects all
subsequent components of risk assessment but was
addressed in only one study (forests of southwest
Tasmania; Wood and Bowman 2012). If plausible
alternative models of ecosystem dynamics exist, multiple
Figure 3. Recommendations for defining ecosystem collapse in practice, using coral reefs
as an example. (a) Collapse of the reef is qualitatively defined as an algae- dominated
state, with very low coral cover. The initial state is defined as a coral- dominated state.
(b) Drivers of ecosystem transitions to collapse include warming, exploitation, and
acidification (shown as red boxes) in a cause–effect conceptual model. (c) Indicators of
ecosystem collapse are identified based on the ecological model. (d) Bounded collapse
thresholds are defined based on ecological evidence: here, a coral cover <1% indicates an
algae- dominated state. (e) Uncertainty affects each step in the framework. This
framework was used to support the application of the IUCN Red List of Ecosystems
criteria to the Meso- American Reef (Bland et al. 2017).
R Ferrari
6
www.frontiersinecology.org © The Ecological Society of America
LM Bland et al.How to define ecosystem collapse
ecological models should be applied in replicated risk
assessments (Burgman 2015).
Ecological models help to organize evidence on ecosys-
tem degradation and recovery, thereby providing
management- related insights. Four studies – two from
marine pelagic and two from forest ecosystems – demon-
strated potential recovery from a collapsed state, so recov-
ery may be possible for some collapsed ecosystems following
effective restoration actions (Keith et al. 2013). Clearly,
ecosystems that have experienced global extinctions of key
species are unlikely to recover (eg foundation palm species
on Easter Island; Diamond 2007). Reversal of collapse may
be possible in other ecosystems if all ecological components
are available and re- assembled through active management
(as was the case in both marine examples), or if granted
sufficient time for recovery (in both temperate forest exam-
ples). The interpretation of novel ecosystems as collapsed
states of antecedent ecosystems may be relevant for ecosys-
tems subject to a variety of threats, including climate
change. Antecedent ecosystem states may or may not be
recoverable (Keith et al. 2015). For example, the current
dominance of haddock (Melanogrammus aeglefinus) over
cod (Gadus morhua) in the eastern Scotian Shelf suggests
that the species composition of the ecosystem “recovering”
from overfishing may be different from that of the pre-
collapse ecosystem (Frank et al. 2011). The likelihood of
ecosystem recovery can be estimated based on the charac-
teristics of initial and collapsed states, transitions to
collapse, and possible hysteretic (ie path- dependent) mech-
anisms maintaining collapsed states (Frank et al. 2011). Yet
in the absence of an ecological model, determining the
likelihood of ecosystem recovery is difficult.
(3) Identifying and selecting indicators of collapse
Informative and sensitive indicators of ecosystem collapse
act as proxies for niche diversity, habitat availability,
and stabilizing biotic interactions that are key to the
persistence of ecosystem biodiversity. To accommodate
different mechanisms of collapse, the IUCN Red List of
Ecosystems requires assessors to define ecosystem- specific
biotic and abiotic indicators (Bland et al. 2016). Effective
comparison and selection of indicators relies on rigorous
protocols (Niemeijer and de Groot 2008), but few studies
applied explicit protocols to select indicators of ecosystem
collapse. No studies used quantitative criteria to score
and compare indicators, limiting the exploration of trade-
offs among indicators, which can be important in eco-
system risk assessment (Bland et al. 2017).
Because ecosystems can collapse through multiple
pathways and exhibit different symptoms of degradation,
definitions of collapse based solely on one type of indica-
tor may underestimate risks to ecosystems (Bland et al.
2017). Although the majority of studies accounted for
uncertainty in indicator selection by applying multiple
indicators within each type (spatial, biotic, or abiotic),
few studies included different types of indicators (eg
spatial indicators and biotic indicators). We found
important differences in indicator use between biomes,
with biotic and abiotic indicators commonly used in
marine pelagic ecosystems, and spatial and biotic indica-
tors commonly used in temperate forests. Although this
may reflect genuine differences in mechanisms of col-
lapse, definitions of collapse in marine pelagic ecosystems
may ignore spatially explicit transitions to collapse (such
as species distribution shifts; Coetzee et al. 2008), while
definitions of collapse in temperate forests may ignore
changes in the physical environment. Ecological models
can help diagnose multiple symptoms of collapse
(Figure 3), inform indicator selection according to
explicit criteria, and identify trade- offs among indicators.
(4) Setting quantitative collapse thresholds
The outcomes of ecosystem risk assessment hinge on
the definition of discrete endpoints to ecosystem deg-
radation (ie transitions beyond quantitative collapse
thresholds in one or more indicators). Despite applying
quantitative indicators of ecosystem change, many studies
did not use quantitative thresholds to define collapsed
states (eg 38% of marine pelagic studies applying abiotic
indicators), falling one step short of requirements for
ecosystem risk assessment.
A distinction exists between decision thresholds, which
specify the conditions required to invoke a decision (eg
classification as collapsed), and ecological thresholds,
which describe non- linear changes in ecosystem dynamics
(Panel 1). Collapse thresholds qualify as decision thresh-
olds because they inform the assessment decision to assign
an ecosystem to a particular category of risk. Most studies
we reviewed did not provide clear ecological rationales for
setting collapse thresholds. An implicit collapse threshold
of 0 km2 when measuring percent declines in spatial distri-
bution, for example, may be inappropriate if ecosystems
lose the ability to sustain their native biota below a cer-
tain distribution size (eg fish reproductive volume;
Möllmann et al. 2008). Many marine pelagic studies relied
on regime shift- detection algorithms to pinpoint the tim-
ing of ecosystem collapse and thresholds in indicators,
with little consideration for the ecological importance (if
any) of these thresholds (eg regime shifts in the North
Pacific; Hare and Mantua 2000). Unless derived explicitly
and based on ecological evidence (Cumming and Peterson
2017), collapse thresholds may provide misleading esti-
mates of risk and little insight into possible management
actions to revert ecosystem degradation. When applied
collectively, spatial, biotic, and abiotic indicators can pro-
vide a comprehensive description of collapsed states, but
very few studies assigned quantitative collapse thresholds
to multiple indicator types (WebTable 2). Careful com-
parisons are needed to derive consistent collapse thresh-
olds among multiple indicators, given that inconsistent
thresholds can severely affect estimates of degradation and
outcomes of risk assessments (Payet et al. 2013).
7
© The Ecological Society of America www.frontiersinecology.org
LM Bland et al. How to define ecosystem collapse
Collapse thresholds should be bounded to represent
inevitable uncertainty in collapsed states (Panel 1; Bland
et al. 2016), but no studies characterized uncertainty in
collapse thresholds. In both species and ecosystem risk
assessment, bounded thresholds accommodate a suite of
uncertainties related to the timing, likelihood, and effects
of management actions on extinction or collapse (Regan
et al. 2009). Explicit consideration of these uncertainties
can help risk assessors identify appropriate collapse
thresholds for each indicator, and sensitivity analyses can
help identify which collapse thresholds trigger changes in
risk assessment outcomes. For example, in simulations of
the mountain ash forest in southeast Australia (Burns
et al. 2015), the collapse threshold would need to decrease
by 30% from an average of 1.0 to 0.7 cavity- bearing trees
per hectare to modify the risk assessment outcome from
Critically Endangered to Endangered.
JConclusions
Previous studies have highlighted the difficulty of de-
fining ecosystem collapse (Boitani et al. 2015) and the
perceived scarcity of descriptions of collapsed ecosystems
(Sato and Lindenmayer 2017). Here, we revealed a
wide array of studies of ecosystem collapse that could
support ecosystem risk assessments and found that these
studies often overlooked the ecological processes leading
to collapse, despite the importance of such processes
for inclusion within risk assessments. A consistent frame-
work to define ecosystem collapse (Figure 3 and WebPanel
2) promotes practical comparisons between ecosystems,
which can be applied to ecosystem risk assessment pro-
tocols from local to global scales. To improve definitions
of ecosystem collapse for biodiversity risk assessment,
we recommend: (1) qualitative description of initial
and collapsed states (based on defining biotic and abiotic
components, ecological processes and distributions) to
provide a robust assessment of characteristic features;
(2) use of ecological models (in particular conceptual
diagrams) to diagnose mechanisms and pathways of
ecosystem change and thus inform indicator selection;
(3) application of spatial, biotic, and abiotic indicators
to capture multiple symptoms of collapse, with careful
consideration of indicator selection and consistency
among collapse thresholds; and (4) explicit definition
of quantitative collapse thresholds based on ecological
evidence, quantifying uncertainty with bounded thresh-
olds and sensitivity analyses. Our recommendations are
particularly relevant to scientists and managers applying
IUCN Red List of Ecosystems assessments (Keith et al.
2013) and other ecosystem assessment tools.
JAcknowledgements
This work was supported by Australian Research
Council Linkage Project (LP130100435). EN and LMB
acknowledge the support of Veski and the Victorian
Government through an Inspiring Women Fellowship.
JAR was supported by a Deakin University Postgraduate
Research Scholarship. We thank R Ferrari for coral
reef photos.
JReferences
Bestelmeyer BT, Brown JR, Havstad KM, et al. 2003. Development
and use of state- and- transition models for rangelands. J Range
Manage 56: 114–26.
Bland LM, Keith DA, Miller RM, et al. 2016. Guidelines for the
application of IUCN Red List of Ecosystems categories and
criteria, Version 1.0. Gland, Switzerland: IUCN.
Bland LM, Regan TJ, Dinh MN, et al. 2017. Using multiple lines of
evidence to assess the risk of ecosystem collapse. P Roy Soc B
Biol Sci 284: 20170660.
Boitani L, Mace GM, and Rondinini C. 2015. Challenging the
scientific foundations for an IUCN Red List of Ecosystems.
Conserv Lett 8: 125–31.
Burgman MA. 2005. Risks and decisions for conservation and
environmental management. Cambridge, UK: Cambridge
University Press.
Burgman MA. 2015. Trusting judgements: how to get the best out
of experts. Cambridge, UK: Cambridge University Press.
Burns EL, Lindenmayer DB, Stein J, et al. 2015. Ecosystem assess-
ment of mountain ash forest in the Central Highlands of
Victoria, south- eastern Australia. Austral Ecol 40: 386–99.
Coetzee JC, Van der Lingen CD, Hutchings L, and Fairweather
TP. 2008. Has the fishery contributed to a major shift in the
distribution of South African sardine? ICES J Mar Sci 65:
1676–88.
CBD (Convention on Biological Diversity). 2014. Aichi
Biodiversity Targets. www.cbd.int/sp/targets/default.shtml.
Viewed 7 Nov 2017.
Croll DA, Maron JL, Estes JA, et al. 2005. Introduced predators
transform subarctic islands from grassland to tundra. Science
307: 1959–61.
Cumming GS and Peterson GD. 2017. Unifying research on
social–ecological resilience and collapse. Trends Ecol Evol 32:
695–713.
Cury P, Mullon C, Garcia S, and Shannon L. 2005. Viability the-
ory for an ecosystem approach to fisheries. ICES J Mar Sci 62:
577–84.
Diamond J. 2007. Easter Island revisited. Science 317: 1692–94.
Feng Q, Liang T, Huang X, et al. 2013. Characteristics of global
potential natural vegetation distribution from 1911 to 2000
based on comprehensive sequential classification system
approach. Grassland Sci 59: 87–99.
Frank KT, Petrie B, Fisher JA, and Leggett WC. 2011. Transient
dynamics of an altered large marine ecosystem. Nature 477:
86–89.
Hare SR and Mantua NJ. 2000. Empirical evidence for North Pacific
regime shifts in 1977 and 1989. Prog Oceanogr 47: 103–45.
Keane RE, Hessburg PF, Landres PB, and Swanson FJ. 2009. The
use of historical range and variability (HRV) in landscape
management. Forest Ecol Manag 258: 1025–37.
Keith DA, Rodríguez JP, Brooks TM, et al. 2015. The IUCN Red
List of Ecosystems: motivations, challenges and applications.
Conserv Lett 8: 214–26.
Keith DA, Rodríguez JP, Rodríguez-Clark KM, et al. 2013.
Scientific foundations for an IUCN Red List of Ecosystems.
PLoS ONE 8: e62111.
Litzow MA and Mueter FJ. 2014. Assessing the ecological impor-
tance of climate regime shifts: an approach from the North
Pacific Ocean. Prog Oceanogr 120: 110–19.
Mac Nally R, Albano C, and Fleishman E. 2014. A scrutiny of the
evidence for pressure- induced state shifts in estuarine and near-
shore ecosystems. Austral Ecol 39: 898–906.
8
www.frontiersinecology.org © The Ecological Society of America
LM Bland et al.How to define ecosystem collapse
Mace GM, Collar NJ, Gaston KJ, et al. 2008. Quantification of
extinction risk: IUCN’s system for classifying threatened spe-
cies. Conserv Biol 22: 1424–42.
Martin J, Runge MC, Nichols JD, et al. 2009. Structured decision
making as a conceptual framework to identify thresholds for
conservation and management. Ecol Appl 19: 1079–90.
Micklin P. 2010. The past, present, and future Aral Sea. Lakes
Reservoirs Res Manage 15: 193–213.
Moher D, Liberati A, Tetzlaff J, and Altman DG. 2009. Preferred
Reporting Items for Systematic reviews and Meta- Analyses:
the PRISMA statement. PLoS Med 6: e1000097.
Möllmann C, Müller-Karulis B, Kornilovs G, and St John MA.
2008. Effects of climate and overfishing on zooplankton
dynamics and ecosystem structure: regime shifts, trophic cas-
cade, and feedback loops in a simple ecosystem. ICES J Mar Sci
65: 302–10.
Nicholson E, Keith DA, and Wilcove DS. 2009. Assessing
the threat status of ecological communities. Conserv Biol 23:
259–74.
Nicholson E, Regan TJ, Auld TD, et al. 2015. Towards consistency,
rigour and compatibility of risk assessments for ecosystems and
ecological communities. Austral Ecol 40: 347–63.
Niemeijer D and de Groot RS. 2008. A conceptual framework for
selecting environmental indicator sets. Ecol Indic 8: 14–25.
Payet K, Rouget M, Esler KJ, et al. 2013. Effect of land cover and
ecosystem mapping on ecosystem- risk assessment in the Little
Karoo, South Africa. Conserv Biol 27: 531–41.
Peñuelas J and Boada M. 2003. A global change- induced biome
shift in the Montseny Mountains (NE Spain). Glob Change Biol
9: 131–40.
Pickering C and Byrne J. 2014. The benefits of publishing system-
atic quantitative literature reviews for PhD candidates and
other early- career researchers. High Educ Res Dev 33: 534–48.
Regan T, Taylor B, Thompson G, et al. 2009. Developing a struc-
ture for quantitative listing criteria for the US Endangered
Species Act using performance testing, phase 1 report. NOAA
technical memorandum. Rockville, MD: NOAA. NOAA-
TM-NMFS-SWFSC-437.
Rice JC and Rochet MJ. 2005. A framework for selecting a suite
of indicators for fisheries management. ICES J Mar Sci 62:
516–27.
Rogers K and Biggs H. 1999. Integrating indicators, endpoints and
value systems in strategic management of the rivers of the
Kruger National Park. Freshwater Biol 41: 439–51.
Roux JP, van der Lingen CD, Gibbons MJ, et al. 2013. Jellyfication
of marine ecosystems as a likely consequence of overfishing
small pelagic fishes: lessons from the Benguela. B Mar Sci 89:
249–84.
Rumpff L, Duncan DH, Vesk PA, et al. 2011. State- and- transition
modelling for adaptive management of native woodlands. Biol
Conserv 144: 1224–36.
Sato CF and Lindenmayer DB. 2017. Meeting the global ecosys-
tem collapse challenge. Conserv Lett; doi.org/10.1111/
conl.12348.
Scheffer M, Carpenter S, Foley JA, et al. 2001. Catastrophic shifts
in ecosystems. Nature 413: 591–96.
Suter GW. 1999. Developing conceptual models for complex
ecological risk assessments. Hum Ecol Risk Assess 5: 375–96.
Washington H. 2013. Human dependence on nature: how to help
solve the environmental crisis. New York, NY: Routledge.
Wood SW and Bowman DM. 2012. Alternative stable states and
the role of fire–vegetation–soil feedbacks in the temperate
wilderness of southwest Tasmania. Landscape Ecol 27: 13–28.
JSupporting Information
Additional, web-only material may be found in the
online version of this article at http://onlinelibrary.
wiley.com/doi/10.1002/fee.1747/suppinfo
8Centro de Ecología, Instituto Venezolano de Investigaciones
Científicas, Caracas, Venezuela; 9Provita, Caracas, Venezuela;
10IUCN Commission on Ecosystem Management and IUCN
Species Survival Commission, Gland, Switzerland
