You are on page 1of 15

Total Environment Research Themes 3–4 (2022) 100011

Contents lists available at ScienceDirect

Total Environment Research Themes


journal homepage: www.elsevier.com/locate/totert

Ecological dynamics, resilience and sustainability


William Grace
Australian Urban Design Research Centre, 1002 Hay Street, Perth, Western Australia 6000, Australia

A R T I C L E I N F O A B S T R A C T

Keywords: This article describes a simple generic system dynamics model of resource dependency in ecosystems to explore
Model the potential range of behaviours that can arise purely due to system structure. The model produces a small
Equilibrium number of behaviours that replicate the observed real‐world behaviour of ecosystems but does not produce
Ecosystem the less plausible outcomes produced in simpler predator–prey models (e.g. Lotka Volterra) such as exponential
Collapse
growth or exploding oscillations. The more limited range of behaviours is attributed to the inclusion in the
Predator
Prey
model of species dependence on abiotic resource flows. The simulations explore the response of systems in
Resilience equilibrium to various types of ‘shock’ and the outcomes are used to reflect on the concepts of resilience
Sustainability and sustainability. The results challenge the normative understanding of these terms and their relationship
to each other. The results also reinforce the notion that shocks to the abiotic system flows at the foundation
of all ecosystems present the greatest threat to ‘resilience’ as they affect all trophic levels.

1. Introduction may be value in informing research, the vagueness of the terms is


unhelpful in promoting a general understanding of what is sustainable
The concepts of sustainability and resilience share a common focus and / or resilient and what is not. In order to answer this question, it is
‐ they both involve study of the dependency of agents on resources of necessary to have a model that reflects the observed dynamic beha-
one type or another within a system, including in two of the well‐ viour of ecosystems in nature.
known systems archetypes: limits to growth and tragedy of the com- The biological resource dependency literature most often cited
mons (Braun, 2002). Resource constraints are the reason ‘nothing relates to the interactions between predators and prey. A Google Scho-
grows forever’. All animals, including humans, are dependent on a lar search for ‘Predator prey models’ returns 900,000 results. Most of
wide range of biotic and abiotic resources. Predators depend on prey, these are mathematical treatments of the species interactions using dif-
plants require air, water, minerals, and sunlight to grow. ferential equations, the genesis of which is the widely cited Lotka‐
The term ‘sustainability’ has in some cases been used interchange- Volterra equations which are based on work originally published in
ably with ‘resilience’, a term credited to C.S. Holling from his seminal 1910 (Lotka, 1910).
article Resilience and Stability of Ecological Systems (C. S. Holling, 1973). Swart (1990) reports on the insight gained from using system
Although Holling’s early work was associated with understanding the dynamics modelling to explore system behaviour arising from the
persistence of species in nature, the term ‘resilience’ has been Lotka Volterra equations. Swart’s model is configured as shown in
expanded by him and others since to include concepts such as ‘trans Fig. 1.
formability’ and ‘adaptability’ in socio‐ecological systems (C. Models of this type produce a range of behaviours. Terminology
Holling, Walker, Carpenter, & Kinzig, 2004; Carl et al., 2010). Several varies, but in this article the various terms have the following
articles have sought to draw connections between the terms ‘sustain- meaning.
ability’ and ‘resilience’ (Berkes & Folke, 1998; Fiksel, 2006; Ludwig,
Walker, & Holling, 1997; Redman, 2014), but none offer a satisfactory Equilibrium a condition in which stock (e.g. population) levels
explanation of the meaning of each term or their relationship to each are constant over time.
other in terms of system behaviour. Oscillation periodic variation in stock levels which may or may
The lack of clarity about what is meant by these terms has led to not vary in amplitude and / or frequency.
them being described as ‘boundary objects’, defined by some authors Periodic a form of oscillation in which the amplitude and
as ‘a theoretical perspective explaining the role of objects in inter and trans-
disciplinary research’ (Lundgren, 2020). While it is accepted that there (continued on next page)

https://doi.org/10.1016/j.totert.2022.100011
Received 20 May 2022; Revised 30 August 2022; Accepted 30 August 2022

2772-8099/© 2022 The Author(s). Published by Elsevier B.V.


This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
W. Grace Total Environment Research Themes 3–4 (2022) 100011

orbit period is regular and repetitive. N and P are the populations of prey and predator respectively
Limit cycle a form of periodic orbit that is closed (i.e. “that a r0 is the prey growth rate,
point moving along the cycle will return to its K is the prey carrying capacity,
starting position at fixed time intervals and thus c is the predator consumption rate
execute periodic motion” Edelstein‐Keshet (1988) d is a constant regulating equilibrium (satiation) levels
δ is the predator death rate
χ is the predator–prey conversion rate

Each of these conditions may be stable (i.e. the system has a tendency This configuration incorporates the so‐called “Holling Type‐II Func-
to that condition independent of initial conditions and perturbations), tional Response” which characterises how the predation rate responds
or unstable. to prey density.
As reported by Swart, the Lotka Volterra structure can produce a This model eliminates most of the obvious deficiencies in the pure
wide range of behaviours depending on its parameterisation. These Lotka‐Volterra model and can produce a range of plausible behaviours
include exponential growth of both species, extinction of the preda- including: stable equilibrium populations of predators and prey,
tors, damped oscillations toward stable equilibrium, explosive oscilla- damped oscillations leading to equilibrium, predator extinction and
tions, and non‐isolated closed orbits. Although the Lotka‐Volterra limit cycles. In fact, the model produces two forms of stable limit cycle,
equations are known for their ability to produce closed orbits, these one in which populations oscillate in an elliptical manner similar to
occur only with a specific set of parameters. As Swart notes: ‘In the neu- the unstable Lotka‐Volterra orbits (Fig. 2a) and one in which limit
trally stable Lotka‐Volterra model, both populations undergo constant oscil- cycles ‘emerge’ with population levels that fall to almost zero (see
lations whose amplitudes bear no relation to the biology of the species but Fig. 2b). This behaviour occurs at elevated values of K – the prey car-
only to the initial values.’ As identified by many other authors (Toro & rying capacity, and is known as the ‘paradox of enrichment’
Aracil, 1988), the equations are insufficient to characterise any real (Rosenzweig, 1971).
world examples of limit cycle behaviour, as the cycles created by the Apparently stable, limit cycle behaviour is observed in predator –
equations are inherently unstable. prey interactions in nature. Probably the most famous example of this
The existence of exponential growth of predators and prey is clearly derives from fur records of lynx (predator) and snow hares (rey) in
indicative of missing structure in the Lotka Volterra model. As nothing Canada analysed by Elton and Nicholson(1942) based on data com-
can grow forever, the agents in question would be limited by balancing piled by MacLulich (1937). The data is illustrated in Fig. 3 and illus-
feedbacks which are simply not represented in the model structure. trates approximately ten year cycles.
Similarly, exploding oscillations are rarely, if ever observed, again Other examples are reported in the literature including (Gilg,
probably because there are balancing feedbacks missing from the Hanski, & Sittler, 2003) in respect of the collared lemming in the Arc-
model that would damp such behaviour. tic tundra, and several others by Begon (2009) p.298. Replicating this
Many others have produced mathematical models that modify and behaviour in the laboratory has been largely unsuccessful but recent
expand the Lotka‐Volterra equations. The most well known variation work by Blasius et al (2020) has produced limit cycles from interac-
on Lotka Volterra is the Rosenzweig‐Macarthur model (Rosenzweig tions between the aquatic invertebrate Brachionus calyciflorus (preda-
& MacArthur, 1963). A version of that model after Turchin (1997) is tor) and its prey, the green algal species Monoraphidium minutum
set out below in equation form. (prey). The evidence indicates that although stable limit cycles are
  observed in nature, they cannot be reproduced by the simple Lotka‐
dN N cNP
¼ r0N 1   ð1aÞ Volterra two‐species predator–prey model. As noted above, stable limit
dt K dþN
cycles can be produced from the Rosenzweig‐Macarthur type models.
Charles Krebs from the Department of Zoology at the University of
dP cNP
¼χ  δP ð1bÞ British Columbia has published several papers on the lynx‐hare inter-
dt dþN
actions in conjunction with other authors. In (Stenseth, Falck,
where:

Fig. 1. Predator – prey model structure (Swart 1990) This simplified stock and flow diagram illustrates populations as ‘stocks’ and inflows / outflows from the
stocks as ‘rates of change’ in the stocks.

2
W. Grace Total Environment Research Themes 3–4 (2022) 100011

Fig. 2.

Fig. 3. Hare-lynx limit cycle behaviour.

Bjørnstad, & Krebs, 1997), the authors report that a simple preda- flows in an ecosystem model. In the Rosenzweig‐Macarthur model,
tor–prey interaction is insufficient to explain the observed behaviour. prey are regulated from below by the K value (see Equation 1) which
represents all lower trophic levels under the term ‘carrying capacity’.
“We have furthermore found an asymmetry in the way the lynx and the
Models commonly include the dynamic impact of such effects as “re-
hare are positioned within the ecosystem: The snowshoe hare appears to
source subsidies” (Leroux & Loreau, 2008; Nevai & Van Gorder,
be regulated from below and above (by a variety of Predators including
2012) which may include allochthonous resources (e.g. carcasses that
the lynx). The lynx, in contrast, seems to be regulated only from below,
seasonally float across the system boundary) or autochthonous
and primarily by the hare. Thus, from the hare’s point of view, the food
resources (e.g. dissolved organic matter in lakes (Lau, Sundh, Vrede,
chain is a vegetation–hare–Predator chain, whereas from the lynx point
Pickova, & Goedkoop, 2014)).
of view, the hare–lynx interaction dominates.”
The term ‘resilience’, at least in respect of ecological systems, was
This suggests that any model that contains only predator–prey coined by C.S. (Buzz) Holling in his seminal (1973) article ‘Resilience
interactions without prey‐food interactions may be inadequate to and Stability of Ecological Systems’. Holling was interested in the dif-
reproduce all facets of species behaviour. Snow hares’ food resources ferent forms of dynamic behaviour observable in the interactions
include a large variety of plant types which are also biotic resources between species and their environment. In his terms, ‘stability’ refers
which in turn rely on photosynthesis, a process fuelled by abiotic to ecological systems that are in a relatively stable dynamic equilib-
resources: air, water, sunlight, and nutrients (minerals). The availabil- rium, meaning a return to equilibrium ‘after a temporary disturbance’.
ity of air and sunlight can be considered as variable flows (at least at By ‘resilience’ he meant ‘the persistence of systems and of their ability
temporal scales relevant to humans). Water and minerals cycle contin- to absorb change and disturbance and still maintain the same relationships
uously and so abiotic resources can be conceptually represented as between populations or state variables’.

3
W. Grace Total Environment Research Themes 3–4 (2022) 100011

Fig. 4. Simplified stock and flow diagram of 2 level system.

The emergence of resilience theory has led to a significant research The initial stage model simulates plant dependency on soil mois-
community associated with the Resilience Alliance,1 spawning studies ture, and parameters were varied (including with extreme values) to
spanning ecology, social and economic systems; and a dedicated journal test system behaviour, namely: the relative growth rates of plants
‐ Ecology & Society. Various modelling studies (Kar, Pal, & Ghosh, 2019; and soil moisture, the unit consumption rate of moisture by the plants
Pal, Kar, Yamauchi, & Ghosh, 2020; Tromeur & Loeuille, 2017) have and the graphical functions that regulate the inflows and outflows to
considered the effect on resilience from human harvesting of resources, the stocks. The parameters were then finally fixed at the parameter
most notably in marine ecosystems, where conventional management set that produced a smooth dynamic transition to equilibrium. The
policies pursue Maximum Sustainable Yield (MSY) of the target species model was then extended to add herbivores at a time after equilibrium
rather than resilience of the multiple species making up that ecosystem. had been established between plants and soil moisture. Parameters
The objective of the work described here is twofold, firstly to create relating to the behaviour of herbivores were then varied to test system
a simple yet plausible generic model of ecosystems that explains behaviour patterns, as in the previous step. Once again, parameters
observed behaviour; and secondly to use this model to unpack the were then finally fixed at the values that produced a smooth transition
meanings of, and relationship between, sustainability and resilience. to equilibrium across the expanded system. Thereafter, predators were
added and the process repeated. Specific details of each step are set out
in the following sections, and the equilibrium parameter values are
2. Methods
included in Appendix B.
Here I describe a simple hierarchical system dynamics model
2.1.1. Plant dependency on Soil Moisture
(Randers, 1997) of a food chain built sequentially in several phases to
Consider a barren field that is colonised by a single plant species (a
explore the essential dynamics of biotic dependence on resources,
biotic resource) that is dependent (inter alia) on Soil Moisture (an abi-
including the fundamental dependency of all life on abiotic resources.
otic resource).
The model is deliberately simplified in order to provide insights into
Fig. 4 depicts the dynamics of this two‐level system, which can be
the generic behaviour of ecosystems based on the inherent structure of
read as follows:
dependency. The model has not been parameterised to reflect real world
behaviour of any specific species interactions, rather to explore the
range of possible system behaviours arising from that generic structure. Loop B1 The stock of Soil Moisture (SM) is (on average)
The model seeks to reflect the endogenous behaviour of such systems but constant over time prior to plant colonisation, i.e. there
doesn’t include allochthonous resources or attempt to incorporate the is zero inflow. As SM is consumed by plants it is
various autochthonous resources that exist in real systems. replenished by inflows regulated by function fSR (SM)
The model was constructed using the Vensim Professional software up to a maximum rate (RSM).
(Version 9.2.2 using a very small time step (0.03125) and R4K Auto Loop B2 SM is depleted by moisture transpired by the Plant
integration type to ensure model stability over a wide range of param- Mass (PM) at a rate controlled by its maximum
eterisation. The model structure is set out below with further explana- transpiration rate (CSM), regulated by function fSC(SM).
tion of the parameterisation of the model, in particular the stock Loop B3 The Plant Mass (PM) regeneration rate is a function
control functions, contained in Appendix A. The equations for the com- fPR(SM) of Soil Moisture up to a maximum regeneration
plete model are included in Appendix B. fraction (RPM) and regulated by Loop R1.
Loop R1 The product of the regeneration rate and the stock
value of PM.
2.1. Biotic hierarchy modelling

2.1.1. Model construction


In order to approximately replicate the notional sequential intro- dSM
¼ RSM  f SR ðSM Þ  PM  CSM  f SC ðSM Þ ð2Þ
duction of biotic resources in a landscape (i.e. plants followed by her- dt
bivores followed by predators), the model was constructed in discrete
steps. dPM
¼ PM  RPM  f PR ðSM Þ ð3Þ
dt
1
https://www.resalliance.org/.

4
W. Grace Total Environment Research Themes 3–4 (2022) 100011

where. In this configuration the equation for Plant mass is modified as


shown in Equation (6) and the equation for Herbivores is added (Equa-
RSM, CSM, RPM are non‐zero constants tion (7)).
fSR is a decreasing function with domain {0 ≤ SM/SMi ≤ 1} and
range {0 ≤ fSR ≤ 1} dPM
¼ PM  RPM  f PR ðSM Þ  H:CPM :f PC ðPM Þ ð6Þ
fSC is an increasing linear function with domain {0 ≤ SM/SMi ≤ 1} dt
and range {0 ≤ fSC ≤ 1}
fPR is an increasing linear function with domain {0 ≤ SM/SMi ≤ 1} dH
and range {‐1 ≤ fPR ≤ 1} ¼ H:RH :f HR ðPM Þ ð7Þ
dt

Images of the functions are included in Appendix A, and explained At equilibrium.


further in the following.
RPM :f PR ðSM Þ
fSR regulates the inflow to Soil Moisture in response to its stock H ¼ PM: ð8Þ
level. At t = 0 soil is saturated, and therefore fSR = 0, meaning there CPM :f PC ðSM Þ
is no inflow, thus the system starts in equilibrium. As Soil Moisture is The additional functions fPC (that regulates Plant Mass outflow in
depleted by plants through evapotranspiration, Soil Moisture is replen- response to consumption by Herbivores) and fHR (that regulates Herbi-
ished and fSR > 0, meaning there is a non‐zero inflow whenever Soil vore inflow in response to Plant Mass) are illustrated graphically in
Moisture falls below its initial value. Appendix A. fPC is identical to fSC and fHR is structured similarly to
fSC regulates the outflow from Soil Moisture. Initially there is no fPR. CPM and RH are constants.
restriction to outflow (i.e. fSC = 1) but as Soil Moisture is depleted,
the outflow from the stock diminishes. Should Soil Moisture be com-
pletely depleted, fSC = 0, the outflow is zero (which ensures that Soil 2.1.3. Predator dependency on Herbivores
Moisture stock cannot fall below zero). Consider now a further resource dependent agent, a Predator,
fPR regulates the inflow to the Plant Mass stock in response to the entering the system (Fig. 6) after equilibrium has been reached
Soil Moisture level. This inflow represents the net regeneration rate between Herbivores, Plants and Soil.
of Plant Mass and therefore must have a positive and negative range The equations for this system follow.
(set here to −1 to + 1) in order for the stock of Plant Mass to both
increase and decrease. The inflow must be zero at equilibrium (i.e. dH
¼ H  RH  f HR ðPM Þ  Pr:CPr :f HC ðH Þ ð9Þ
when fPR = 0). As configured here, this occurs when Soil Moisture dt
reaches a quarter of its initial value.
At equilibrium: dPr
¼ Pr:RPR :f PR ðH Þ ð10Þ
dt
RSM  f SR ðSM Þ
PM ðt Þ ¼ ð4Þ
CSM  f SC ðSM Þ And at equilibrium.

RH :f HR ðPM Þ
f PR ðSM Þ ¼ 0 ð5Þ PR ¼ H: ð11Þ
CPR :f HC ðH Þ
The additional functions fHC (that regulates Herbivore outflow in
2.1.2. Herbivore dependency on plants response to predation) and fPr (that regulates Predator inflow in
Consider now the situation where plants have been established and response to Herbivore stocks) are illustrated graphically in Appendix
the system is in equilibrium, when a small number of Herbivores enter A. fHC is identical to fSC and fPC, and fPr is structured similarly to fPR
the area. This expands the model structure as illustrated in Fig. 5. and fHR. CPr and RPR are constants.

Fig. 5. Simplified stock and flow diagram of 3 level system.

5
W. Grace Total Environment Research Themes 3–4 (2022) 100011

Fig. 6. Simplified stock and flow diagram of 4 level system.

2.2. Resilience testing - post-equilibrium disturbances At very high ratios of RPM to RSM, the amplitude of the oscillations
becomes so extreme that Plant Mass is diminished almost to zero, i.e.
To explore the concept of resilience requires modelling of the sys- overshoot and collapse behaviour is apparent (Fig. 8). However, as
tem response to disturbances or shock. This was tested by introducing noted above, Soil Moisture cannot be indefinitely depleted, and so this
changes after equilibrium has been achieved in the hierarchy depicted apparent behaviour is misleading. If the timeframe of the model is
in Fig. 6. extended, it can be seen that what appears to be overshoot and col-
lapse in Fig. 8 is actually very long period damped oscillations leading
3. Results eventually to equilibrium at the same stock levels of Plant Mass.2
Examination of Equation (3) illustrates that an equilibrium level of Plant
3.1. Plant and Soil Moisture interaction Mass (PM) cannot fall to zero indefinitely. As Plant Mass grows, Soil
Moisture is depleted but is continually replenished, even if the stock
The two stock model described in Fig. 4 starts in equilibrium with level reduces to zero. This is the outcome that results from the inflow
zero net inflow and outflow of soil moisture (i.e. the soil is saturated). to Soil Moisture being a ‘rate’ rather than a fraction of the stock level.
Plant mass is introduced at time zero with an initial small stock. This In summary, the Plant Mass ‐ Soil Moisture system inevitably trends
commences the dynamic responses of Plant mass and Soil moisture. to a unique equilibrium, albeit with damped oscillatory behaviour evi-
Sensitivity testing identifies that the behaviour of this system configu- dent with some combinations of variables. Extreme value testing of the
ration is limited to a small number of outcomes. model parameters confirms that other modes of behaviour such as
When unconstrained by limitations in Soil Moisture the Plant Mass exponential growth, limit cycles or exploding oscillations cannot be
expands at its maximum growth rate (RPM). Eventually the growth rate produced with this model structure.
of the Plant Mass becomes constrained by limitations in Soil Moisture, This behaviour is reflective of the dependent resource being an abi-
i.e. the Plant Mass regeneration rate drops to zero. At this stage Plant otic rather than biotic resource. The inflow (growth rate) of a biotic
Mass and Soil Moisture are in equilibrium, meaning the rates of inflow resource is some fraction of the stock level (i.e. the higher the stock
and outflow of Soil Moisture are equal. All of this assumes the only level the faster the growth rate and vice versa) which creates reinforc-
impact on the stock of Plant Mass is Soil Moisture. ing feedback. If the Plant Mass was to fall to zero (e.g. through a catas-
The system’s eventual equilibrium is reached when fPR = 0 (Equa- trophic fire) the Plant Mass would become locally extinct. In contrast,
tion (4)). As parameterised, at that point fSC = 0.25 and fSR = 1 which the inflow to an abiotic resource reduces with the stock level (i.e. the
means that Soil Moisture (SM) is a quarter of its initial value and higher the stock of Soil Moisture the slower the rate of inflow) creating
inflow and outflow are identical at the rate of RSM. Accordingly, the balancing feedback. If the Soil Moisture were to fall to zero (e.g. years
product of Plant Mass (PM), fSC and CSM is equal to RSM. Increasing of drought) it would recover as soon as rainfall returned. This simple
or decreasing RSM changes the equilibrium value of the Plant Mass model exemplifies the dependence of all photosynthetic organisms
(PM) but otherwise has no effect. on the abiotic resources of sunlight, nutrients and water.
The dynamic behaviour of the system is controlled by the relative
magnitudes of RPM (the growth fraction of Plant Mass) and RSM (Soil
Moisture inflow rate) ‐see Fig. 7. Below a certain ratio of those param-
eters (RPM / RSM) the system exhibits a smooth transition to equilib-
rium, irrespective of other system parameters. However, above that
threshold the system exhibits damped oscillations prior to equilibrium 2
Of course in the real world this level of depletion may lead to extinction of Plant mass,
becoming established at the same stock values. the atto‐Fox problem.

6
W. Grace Total Environment Research Themes 3–4 (2022) 100011

Fig. 9. 3 level system – smooth transtion to equilibrium.


Fig. 7. 2 level system - dynamic behaviour arising from variation of RPM to
RSM.

Fig. 8. 2 level system - overshoot and collapse arising from extreme ratio of Fig. 10. 3 level system – oscillations arising from increasing ratio of RH to RPM.
RPM to RSM.

viour as the plant resource is exhausted (Fig. 11). These outcomes


3.2. Herbivore, Plant and Soil Moisture interaction result from the parameterisation of the graphical function fHR, more
specifically the Plant Mass level that triggers a negative Herbivore
Herbivores are introduced to the model at time 50 (the three stock growth rate (i.e. population decline). The difference in Herbivore
system described in Fig. 5) with the Plant mass and Soil moisture response to Plant mass scarcity between these three situations, i.e.
parameters set at the values that produced a smooth transition to equi- immediate decline, equilibrium or overshoot and collapse, is depicted
librium in the previous stage (see Fig. 7). in the function fHR for each case (Fig. 12).
The behaviour of is dominated by the response of Herbivores to the If a (relatively) fast Herbivore growth rate is combined with a
availability of Plant Mass. This is regulated by the Herbivore regener- ‘slow’ response to resource scarcity then overshoot and collapse beha-
ation fraction (RH. fHR), or more specifically the level of Plant Mass that viour transitions to very long period and deep stable limit cycles
can sustain a positive Herbivore growth rate. If the Plant Mass is insuf- (Figs. 13a and b). These limit cycles are of the form evident in the
ficient to create a positive growth rate (births minus deaths) the Her- phase plot shown in Fig. 2b).
bivores immediately decline to extinction (zero population). Above This behaviour (Fig. 13) occurs at the higher (relative) Herbivore
that threshold value, the system comes to equilibrium smoothly growth rates because the collapse happens so quickly that the Herbi-
(Fig. 9) or via damped oscillations, depending on the relative values vore stock declines sufficiently to be negligible before the Plant Mass
of RH and RPM (Fig. 10). The higher this ratio, the more the system is completely exhausted, leading to a slow recovery of Plant Mass
overshoots as it seeks equilibrium. Nevertheless, the oscillations are and concomitant (eventual) recovery of Herbivores. In the overshoot
damped. and collapse case (Fig. 11), the Plant Mass is exhausted first, condemn-
However, if the level of Plant Mass that can sustain a positive Her- ing Herbivores to extinction.
bivore growth rate is lower than the threshold level, the system Extreme value testing of the model parameters illustrates that this
reaches a bifurcation point and exhibits overshoot and collapse beha- structure does not produce exponential growth or exploding oscilla-

7
W. Grace Total Environment Research Themes 3–4 (2022) 100011

Fig. 11. 3 level system – overshoot and collapse arising from extreme ratios of Fig. 13b. 3 level system – phase plot associated with Fig. 13a).
RH to RPM.

tions, nor stable limit cycles of the Rosenzweig‐Macarthur model type


illustrated in Fig. 2a).

3.2. Predator, Herbivore, plant and Soil Moisture interaction

Predators are introduced to the model (the four stock system


depicted in Fig. 6) at time 200 with the Plant mass, Soil moisture
and Herbivore parameters set at the values that produced a smooth
transition to equilibrium in the previous stage (see Fig. 9).
The behaviour mirrors the earlier introduction of Herbivores, i.e. it
is dominated by the response of Predators to scarcity of Herbivores.
Above the threshold that leads to equilibrium, the system either does
so smoothly (Fig. 14) or with damped oscillations (Fig. 15). As for the
previous Herbivore / Plant mass case, below that threshold, the system
exhibits either: overshoot and collapse (Fig. 16) of both Predators and
Herbivores; or very long period deep limit cycles (Fig. 17). Again,
extreme value testing confirms that neither exponential growth or
exploding oscillations are evident with this parameterisation of the
model, nor stable elliptical limit cycles. As for the previous case only
Fig. 12. 3 level system – graphical function fHR(PM). very long period and deep stable limit cycles are produced by the
model.

Fig. 13a. 3 level system – limit cycles arising from extreme ratios of RH to RPM
combined with fHR(PM) = 0 at lower levels of PM. Fig. 14. 4 level system – smooth transition to equilibrium.

8
W. Grace Total Environment Research Themes 3–4 (2022) 100011

3.3. Resilience testing

Persistence of the system in stable equilibrium (as shown in Fig. 14)


was tested by introducing a ‘spike’ in each stock of around 20 % of its
equilibrium level at time 600. This level of disturbance has no long‐
term impact on the system, whether or not the spike is an increase
or decrease. In each case stock levels return to their initial equilibrium
levels with minor damped oscillation. Fig. 18 represents a sudden
increase in Herbivore numbers, which initially triggers increases in
Predators and decrease of Plant Mass. This confirms that the equilib-
rium that is established is stable.
The next test of stability is the response of the system to slower
changes in abiotic flows. This case was tested by slowly increasing
and decreasing (through ramping) the flows that regulate Soil Mois-
ture, following the establishment of equilibrium. Decreasing the inflow
of Soil Moisture has the predictable effect of reducing Plant Mass and
therefore both Predator and Herbivore numbers, with Predators at the
top of the trophic scale being most impacted (Fig. 19). At larger reduc-
tions of Soil Moisture flows, Predators become extinct (Fig. 20) and at
Fig. 15. 4 level system – oscillations arising from increasing ratio of RPr to RH. even larger reductions so do Herbivores. This begs the question about
whether resilience is a property of the system or the species. The land-
scape has returned to its original form, but animals have vanished.

4. Discussion

The simulations demonstrate that system behaviour is determined


by:

• The relative maximum (i.e. unconstrained) population growth rates


of the dependent agent and its resource, in combination with:
• the dependent agent’s response to extreme resource scarcity, i.e.
the resource level at which the agent’s population growth rate
becomes negative through increased mortality and / or reduced fer-
tility – referred to here as ‘growth rate zero’.

There are distinct categories of agent behaviour observed in the


simulations (see Fig. 12 and Table 1).
The deep limit cycles evident in the experiments where agent and
resource populations decline to near zero, occur near the ‘growth rate
zero’ boundary between a transition to equilibrium and overshoot and
collapse, and are influenced by the unconstrained growth rate of the
Fig. 16. 4 level system – overshoot and collapse arising from extreme ratios of dependent agent.
RPr to RH. For systems exhibiting immediate decline or a transition to equilib-
rium, the unconstrained growth rates of the dependent agent affect the

Fig. 17. 4 level system – limit cycles arising from extreme ratios of RPr to RH
combined with fPr(H) = 0 at lower levels of H. Fig. 18. Post equilibrium Herbivore injection.

9
W. Grace Total Environment Research Themes 3–4 (2022) 100011

Fig. 19. Gradual small reduction in Soil Moisture.

Fig. 21. Phase plot of stable elliptical cycles (modelled).

due to cyclical variation in resource availability. In these deep cycles,


agents and resource stocks decline to near zero (Fig. 22).
Continuous models (such as that described here) that can produce
very small population levels are sometimes referred to as having the
“atto‐fox problem”, a term that derives from criticism by Mollison
(1991) of a model by Murray and Seward (1992) of fox rabies in which
foxes decline to 10−18 (atto‐) foxes per square kilometre. In the real
world such deep cycles will in many cases likely lead to extinction.
Nevertheless, these deep limit cycles seem to be consistent with the
cycles identified in the Hare‐lynx data (see Fig. 23), the Blasius
(2020) data (see Fig. 24) and those described by Holling in the spruce
budworm example.
These occur when the decline of the dependent agent in response to
critical resource scarcity is so fast that population levels approach zero
before the resource stock is completely depleted (compare Figs. 11 and
13a). Accordingly, both stocks eventually recover and the cycle is
repeated. These cycles are also referred to as ‘relaxation oscillations’
Fig. 20. Gradual larger reduction in Soil Moisture. that occur when the growth rates of the dependent agent and resource
vary significantly in so called slow‐fast systems (Saha, Pal, & Banerjee,
2021). In contrast, when the response to scarcity is ‘slower’, the
rates of population growth/decline and oscillatory behaviour (which resource stocks decline to zero first, leading to overshoot and collapse,
occurs at higher relative growth rates) but not the equilibrium magni- ensuring the extinction of both species (see Figs. 11 and 16).
tude of the stocks. Rosenzweig‐Macarthur type models can produce a similar type of limit
I have referred to the observed limit cycles as ‘deep’ as they are not cycle but in that case it is associated with the resources available to
(stable) Rosenzweig‐Macarthur type elliptical cycles in which popula- prey at high levels of carrying capacity (“nutrient enrichments”) rather
tion levels remain well above zero (Fig. 21), or oscillations that arise than their response to predation.

Table 1
Summary of simulation results.

Growth rate zero Relative unconstrained System behaviour Descriptor


population growth rates

If ‘growth rate zero’ occurs above a resource level which is insufficient to Influences rate of decline Exponential decay of initial dependent agent Immediate
support a positive growth rate from the outset of the agent’s emergence only population to zero decline
If ‘growth rate zero’ occurs below a resource level at which continuing agent Low Dependent agent and resource populations Overshoot
population growth leads to severe depletion of the resource stock. decline to zero and collapse
High Dependent agent and resource decline to near- Deep limit
zero and slowly recover cycles
If ‘growth rate zero’ occurs at a resource level between the upper threshold and Low Smooth transition to a stable equilibrium for Transition to
a lower threshold. dependent agent and resource populations Equilibrium
High Damped oscillations to a stable equilibrium for
dependent agent and resources populations

10
W. Grace Total Environment Research Themes 3–4 (2022) 100011

Fig. 22. Phase plot of deep limit cycles (modelled).

Fig. 24. Phase plot of Blasius data cycles Data from the C2 experiments.

oscillations or unconstrained exponential growth which are found in


the original Lotka Voltera model are neither evident in nature or in
these experiments. Other more sophisticated models derived from
Lotka Volterra eliminate those problems but retain some behaviours
that are not reproduced in the complete hierarchical model presented
here (e.g. stable elliptical limit cycles).
It is posited here that the reason for plausible behaviour is the
inclusion of abiotic flows (in this case water) at the base of all ecosys-
tems. This means the regeneration of this abiotic resource is regulated
by a balancing feedback loop, whereas in most predator–prey models,
regeneration of the prey is the product of the stock itself and its growth
fraction, which is a reinforcing feedback loop. This suggests that the
modelling of biotic systems without their underlying dependency on
abiotic resources such as air, water, nutrients (minerals) and solar radi-
ation may be insufficient to explain observable behaviour in the field.
This is not to say that other models could not be configured in a way
that gives rise to similar outcomes.
Studies on resilience such as those conducted in fisheries manage-
ment (Pal et al., 2020) might benefit from a more explicit inclusion of
Fig. 23. Phase plot of Hare Lynx data cycles.
species hierarchy such as in the model presented here.

5. Model validation
As noted above, the empirical data seems to support the existence
of deep limit cycles (relaxation oscillations) but it is interesting to The process of model validation is to build confidence in its ability
question whether there is empirical evidence of stable elliptical limit to achieve its particular purpose (Schwaninger and Groesser, 2020;
cycles such as those produced by Rosenzweig‐Macarthur type models. Senge & Forrester, 1980). The model presented here can be thought
The model presented here does not produce oscillations of that kind, of as in ‘insight’ model, or a ‘causal‐descriptive’ model (Barlas,
nor have they been identified in the literature reviewed for this study. 1996). Unlike many system dynamics models it is not aimed at identi-
All real‐world systems mimic the essential hierarchical resource fying the “effectiveness of alternative policies”; its purpose is to
dependency described in these simple experiments, although with explore the set of behaviours that emerge from a plausible generic
much more complexity, including competition between species at each model structure. It does not seek to replicate the exact dynamic beha-
level, omnivores operating at both the Herbivore and Predator level, viour of a specific hierarchical food chain, rather to depict the condi-
and with fungi, bacteria, insects and birds affecting all levels together tions that give rise to the persistence or otherwise of each level in the
with irregular allochthonous or autochthonous resource subsidies. chain. Accordingly, it has not been calibrated against empirical data
Despite the simplifications, the experiments replicate the real‐world for any particular set of species, nor contains additional structure to
behaviour observed in nature including: smooth transitions to equilib- represent the myriad of interdependencies that exist in real ecological
rium, damped oscillations, overshoot and collapse, and deep limit systems, most obviously competition at each trophic level. However,
cycles and (Begon, 2009; C. S. Holling, 1973; Klein, 1968). Exploding the general behaviour produced by the model can be tested against

11
W. Grace Total Environment Research Themes 3–4 (2022) 100011

the empirical evidence, including in respect of predator‐ prey dynam- Holling defined resilience as ‘..the persistence of systems and of their
ics and the existence of limit cycles. Establishing its consistency with ability to absorb change and disturbance and still maintain the same rela-
that evidence is the prime purpose of the study, and citations to the rel- tionships between populations or state variables’.
evant literature are included in the text of the article to support its This definition implies a system in dynamic equilibrium prior to a
achievement of that aim. Because the model does not seek to simulate disturbance; in the case of the budworm ‘the chance consequence of cli-
specific real‐world species interactions, questions of boundary ade- matic conditions’. However, as these experiments show, the ability of
quacy (i.e. does the model contain sufficient structure to satisfy its pur- a species within an ecosystem to ‘persist’ is merely a function of the
pose) are primarily satisfied by producing plausible behaviours, i.e. scale and nature of the disturbance. If the disturbance is to the fun-
those observable in nature. damental abiotic flows or lower trophic levels, the impact on the
The key to the model behaviour is the parameterisation of the higher levels is larger. Such disturbances may lead to the extinction
inflows and outflows to the stocks, structured as constants modified of one or more species in an ecosystem while leaving others in place,
dynamically through graphical functions. With the sequential addition albeit in a reconfigured equilibrium. Such a change would not meet
of each level of the hierarchy, the parameters and graphical functions Holling’s resilience test as the ‘same relationships between populations’
were iteratively varied to evaluate the “window” within which the would no longer exist. However, later papers by resilience theorists
dependent agent could establish, and extreme value testing used to have proposed a difference between ‘specific resilience’ and ‘general
explore the full range of system behaviours following establishment resilience’. In somewhat of a summary of resilience and related con-
– effectively representing behaviour sensitivity testing at each stage cepts Folke et al (2010) describe the former as ‘resilience of some par-
of model development. The model does not purport to replicate the ticular part of a system, related to a particular control variable, to one or
actual growth rates of the individual stocks, rather it establishes the more identified kinds of shocks’ and the latter as ‘the resilience of any
relative rates between each agent that govern persistence or otherwise. and all parts of a system to all kinds of shocks, including novel ones.’ This
Thus, the development of the model incorporated both parameter ver- represents somewhat of a qualification of the original Holling
ification and extreme value testing, while being dimensionally definition.
consistent. The results presented here also question whether resilience is a
property of a ‘system’ or a specific species within a system. It could
be said that (for example) plants are resilient to virtually any shock
(prior to the appearance of Herbivores) if they are underpinned by a
6. Conclusions constant flow of Soil Moisture. However, the resilience of the animals
at a higher trophic level is more limited.
A simple model of a four‐level ecosystem provides insights into The focus of the resilience community is on human intervention
plausible system behaviour (i.e. that which is observable in the real intended to retain the persistence of a given system (including so
world). It is argued here that this is due to the founding of the biotic called socio‐ecological systems3). Based on the findings set out here,
resources on an abiotic flow, in this case water, which creates a balanc- intervening in systems subject to some specific shock (i.e. specific resili-
ing feedback loop at the lowest system level and limits the behaviour ence) would seem to be the more fruitful endeavour. As noted elsewhere
of higher trophic levels to a small number of outcomes, namely equi- (Willliam Grace & Pope, 2015) no system is likely to withstand a direct
librium achieved in a smooth transition or with tightly damped oscil- asteroid hit or catastrophic climate change. Therefore, what is referred
lations, overshoot behaviour leading to collapse, or deep limit cycles. to as general resilience is really just specific resilience to a wide range
As all life on earth is founded on such flows, this suggests that any of plausible disturbances.
ecosystem models that neglect such flows may in some cases produce It is over thirty years since the Brundtland report “Our Common
unrealistic system behaviour. Future” (World Commission on Environment and Development,
Second, that stable limit cycle behaviour can be produced by sys- 1987) produced the oft‐quoted explanation of sustainable develop-
tem structure alone, in the circumstances where overshoot does not ment: 'development that meets the needs of the present without compro-
quite create collapse, but rather leads to very small stock levels mising the ability of future generations to meet their own needs'. If our
and slow recovery. These cycles may initially expand in amplitude prime interest is the ‘persistence’ of humanity with an aspiration to
but inevitably settle into stable limit cycles. Holling’s original 1973 meet the Brundtland objective (i.e. the achievement of a decent stan-
paper suggested that the budworm / balsam fir interactions were dard of living for all), then the global socio‐ecological system must
an example of ‘…a distinct domain of attraction determined by the inter- first achieve a stable dynamic equilibrium. Exponential increases in
action between budworm and its associated natural enemies, which is peri- population and the increasingly unsustainable depletion of renewable
odically exceeded through the chance consequence of climatic conditions’. and non‐renewable resources indicates that humanity is nowhere
While that interpretation is not disproven by these experiments, they near such an equilibrium, as illustrated by the Limits to Growth stud-
do illustrate that this behaviour can be created by the same system ies first published in 1972 (Meadows, Meadows, Randers, & Behrens,
structure that produces other observable behaviour patterns without 1972) and is now in a state of overshoot as more recent work
random variations. explains (William Grace, 2015; Meadows, Meadows, & Randers,
Dynamic equilibrium conditions are observable in nature, and so 2004). Sustainability can be thought of as a quest for such an equi-
are the damped oscillations that occur in response to natural varia- librium. Resilience of the human (or any) species can be thought
tion and disturbances. Are such systems therefore ‘sustainable’? It of as the ability to recover from shocks and disturbances once equi-
is more helpful to consider whether the outflows from dependent librium has been achieved.
stocks are sustainable, i.e. whether net consumption of the dependent Shocks and disturbances to the lower trophic levels will have more
agent’s resource can be sustained over time. This depends both on impact on humans (e.g. the loss of pollination services (Kevan & Viana,
the net depletion rate of the resource (i.e. population levels and con- 2003)), than those to higher trophic levels. However, shocks to the
sumption rates of dependent agents), and the stability of the inflows abiotic resources that underpin life on earth will have even more
(regeneration rates) to all the resources at lower trophic levels, impact, as they threaten the entire dependent biotic system. Changing
including the abiotic resources that support all life on earth. Accord-
ingly, sustainability is a time limited concept. Nothing stays the same
forever, climate conditions change, systems re‐configure on the
appearance of new agents, species become extinct through disease, 3
This descriptor is open to the accusation of tautology as humans are obviously part of
and so on. ecological systems.

12
W. Grace Total Environment Research Themes 3–4 (2022) 100011

the radiation balance in the earth’s atmosphere is just such a distur- Appendix B. Model equations
bance, making climate change our most urgent challenge, albeit not
the only one. The model behaviour is dependent on the graphical functions and
constants, which were set to ensure the biotic agents (Plant Mass, Her-
Declaration of Competing Interest bivores and Predators) could reach a stable equilibrium after their
sequential appearance in the model.
The authors declare that they have no known competing financial The key variables that affect behaviour are the constants that con-
interests or personal relationships that could have appeared to influ- trol stock inflow (regeneration) rates, i.e. RSM, RPM, RH, and RPR. A mul-
ence the work reported in this paper. tiplier for each of these constants was included in the model to
facilitate sensitivity and extreme value testing of the model. The
Appendix A. - model parameterisation results of these tests are described in the article text.
The following sets out the equations for the full 4 level model. This
Graphical functions. should be read in conjunction of the figures and equations in the text.
Equations are set out in the order that the model was constructed, i.e.
fSR = GF1 (SM/SMi)
With consistent rainfall and without plants, Soil Moisture will • Plant mass and soil moisture interactions
vary through the year but reach an equilibrium state on an • Additional interactions following introduction of Herbivores
average annual basis. It is assumed in this model for simplicity • Additional interactions following introduction of Predators
that this is saturation (fSR = 0) so the model commences in
equilibrium. As Soil Moisture declines below a threshold, inflow The full Vensim model is available on request.
increases to a maximum value of RSM. Modelled as a linear
function but testing with non‐linearity show minor impact on Model settings
behaviour
fSC = GF2 (SM / SMi)
Conventional stock control. Soil moisture outflow is linearly
constrained as stock level falls. Non‐linearity affects the INITIAL TIME = 0
equilibrium level of Plant Mass. This is because concave GF2 Units: Year
constrains Soil Moisture loss – keeping Soil Moisture stock The initial time for the simulation.
higher, and hence plant regen higher for longer. Convex GF2 FINAL TIME = 200
causes the opposite. Essential model behaviour is though Units: Year
unchanged. The final time for the simulation.
fPR = GF3 (SM / SMi) TIME STEP = 0.03125
Plant mass regeneration rate is constrained once SM / SMi falls Units: Year [0,?]
below 0.5. Equilibrium obtained when fSR = 0 at SM / The time step for the simulation.
SMi = 0.25. Regeneration rate becomes negative at lower levels
of SM / SMi. Effect of non‐linearity: convex GF3 slows the
transition to equilibrium as restriction on PM inflow occurs
earlier. Concave is the opposite. Essential model behaviour is Soil moisture equations
though unchanged.
fPC = GF4 (PM / PMi)
As for GF2
fHR = GF5 (PM / PMi)The domain determines the envelope of Soil moisture = INTEG (SM inflow‐SM outflow, Initial soil
persistence of Herbivores. If there is insufficient Plant Mass to moisture)
support a positive Herbivore generation rate then the Herbivore Units: SM perc
influx immediately declines to zero. At the other extreme, if the Initial soil moisture = 200000
transition from positive to negative regeneration rates occurs at Units: SM perc
a lower bound threshold, overshoot and collapse SM inflow= “Max SM inflow (Rsm)”*”fSR(SM)”*SM Multiplier
occurs.As for GF3, non‐linearity has no significant impact on Units: SM perc/Year
essential behaviour (assuming zero point is maintained) “Max SM inflow (Rsm)”= 50,000
. Units: SM perc/Year
fHC = GF6 (H / Hi) SM Multiplier = 1
As for GF2 and GF4 Units: Dmnl
FPr = GF7 (H / Hi)As for GF5. If the transition from positive to [Note this variable was used to vary Rsm for sensitivity testing
negative regeneration rates occurs at a lower bound threshold, of model]
overshoot and collapse “fSR(SM)”=
occurs.As for GF3 and 5, non‐linearity has no significant impact “GF1 SM‐inflow”(Soil moisture/Initial soil moisture)
on essential behaviour (assuming zero point is maintained) Units: Dmnl
. “GF1 SM‐inflow”([(0,0)‐(1,1)],(0,1),(0.5,1),(1,0))
Units: Dmnl
SM outflow = max(Plant mass,0)*”P cons (Csm)”*”fSC(SM)”
Units: SM perc/Year

(continued on next page)

13
W. Grace Total Environment Research Themes 3–4 (2022) 100011

Soil moisture equations (continued) Herbivore equations (continued)

Soil moisture = INTEG (SM inflow‐SM outflow, Initial soil Herbivores = INTEG (H inflow + Initial H A H outflow, 0)
moisture) Units: Herbivores
Units: SM perc Multiplier
“P cons (Csm)”= 10 Units: Herbivores/Year
Units: SM perc/(Year*PMtonnes) “H‐regen‐fr (Rh)”= 0.075
“fSC(SM)”= “GF2 SM‐outflow”(Soil moisture/Initial soil Units: 1/Year
moisture) H Multiplier = 1
Units: Dmnl Units: 1
“GF2 SM‐outflow”([(0,0)‐(1,1)],(0,0),(1,1)) [Note this variable was used to vary Rh for sensitivity testing
Units: Dmnl of model]
“fHR(PM)”= “GF5 H‐inflow ‐ smooth”(zidz(Plant mass, Initial
plant mass))
Units: Dmnl
Plant mass equations “GF5 H‐inflow ‐ O&C”(
[(0,‐2)‐(2000,1)],(0,‐2),(100,‐0.85),(300,0),(630,0.55),(870,0.
8),(1600,1))
Plant mass = INTEG (PM inflow‐PM outflow, Initial plant mass) Units: Dmnl
Units: PMtonnes [Note this version of the function produces the herbivore
Initial plant mass = 25 overshoot and collapse mode of the model]
Units: PMtonnes “GF5 H‐inflow ‐ smooth”([(0,‐2)‐(3200,1)],(0,‐2),(500,0),(750,1
PM inflow = Plant mass*“PM‐regen‐fr (Rpm)”*“fPR(SM)”*PM ),(1600,1))
Multiplier Units: Dmnl
Units: PMtonnes/Year [Note this version of the function produces the herbivore
“PM‐regen‐fr (Rpm)”= 0.2 smooth or damped oscillatory mode of the model]
Units: 1/Year H outflow= “Pr cons (Ch)”*“fHC(H)”*Predators
PM Multiplier = 1 Units: Herbivores/Year
Units: Dmnl “Pr cons (Ch)”= 0.25
[Note this variable was used to vary Rpm for sensitivity testing Units: Herbivores/Predators/Year
of model] “fHC(H)”= GF6 H outflow (zidz(Herbivores,“H‐injection”))
“fPR(SM)”=”GF3 PM‐inflow”(Soil moisture/Initial soil Units: Dmnl
moisture) GF6 H outflow([(0,0)‐(1,1)],(0,0),(1,1),(100,1))
Units: Dmnl Units: Dmnl
“GF3 PM‐inflow”([(0,‐1)‐(1,1)],(0,‐1),(0.5,1),(1,1))
Units: Dmnl
PM outflow = Herbivores*“H cons (Cpm)”*“fPC(PM)”
Units: PMtonnes/Year Predator equations
“H cons (Cpm)”= 4.5
Units: PMtonnes/Herbivores/Year
“fPC(PM)”= “GF4 PM‐outflow”(zidz(Plant mass, Initial plant
Predators = INTEG (Initial Pr + Pr inflow ‐ Pr outflow, 0)
mass))
Units: Predators
Units: Dmnl
Initial Pr = 0 + STEP(“Pr‐injection”/“1 year”,200)‐STEP(“Pr‐
“GF4 PM‐outflow”([(0,0)‐(1,1)],(0,0),(1,1),(1000,1))
injection”/“1 year”,201)
Units: Dmnl
Units: Predators/Year
[Note this equation was used to introduce Predators to the
model at t200]
“1 year”= 1
Herbivore equations
Units: Year
“Pr‐injection”= 10
Units: Predators
Herbivores = INTEG (H inflow + Initial H A H outflow, 0) Pr inflow = Predators*“Pr‐regen‐fr (Rpr)”*“fPrR(H)”*Pr
Units: Herbivores multiplier
Initial H = 0 + STEP(“H‐injection”/“1 year”,50)‐STEP(“H‐inje Units: Predators/Year
ction”/“1 year”,51) + STEP(“H‐spike” /“1 year”,300)‐STEP(“ “Pr‐regen‐fr (Rpr)”= 0.03
H‐spike” /“1 year”,301) Units: 1/Year
Units: Herbivores/Year Pr multiplier = 1{6}
[Note this equation was used to introduce Herbivores to the Units: Dmnl
model and t50 and then to introduce a spike for resilience [Note this variable was used to vary Rpr for sensitivity testing
testing at t300] of model]
“H‐injection”= 10 “fPrR(H)”= “GF7 Pr‐inflow‐smooth”(zidz(Herbivores,“H‐injec
Units: Herbivores tion”))
“H‐spike”= 10 Units: Dmnl
Units: Herbivores “GF7 Pr‐inflow O&C”([(0,‐2)‐(100,1)],(0,‐2),(15,0),(40,1),(100,
H inflow = max(Herbivores,0)*“H‐regen‐fr (Rh)”*“fHR(PM)”*H

14
W. Grace Total Environment Research Themes 3–4 (2022) 100011

Predator equations (continued) Klein, D.R., 1968. The Introduction, Increase, and Crash of Reindeer on St. Matthew
Island. J. Wildl. Manag. 32 (2), 350–367. https://doi.org/10.2307/3798981.
Predators = INTEG (Initial Pr + Pr inflow ‐ Pr outflow, 0) Lau, D.C.P., Sundh, I., Vrede, T., Pickova, J., Goedkoop, W., 2014. Autochthonous
resources are the main driver of consumer production in dystrophic boreal lakes.
Units: Predators Ecology (Durham) 95 (6), 1506–1519. https://doi.org/10.1890/13-1141.1.
1)) Leroux, S.J., Loreau, M., 2008. Subsidy hypothesis and strength of trophic cascades
Units: Dmnl across ecosystems. Ecol. Lett. 11 (11), 1147–1156. https://doi.org/10.1111/j.1461-
0248.2008.01235.x.
[Note this version of the function produces the predator
Lotka, A.J., 1910. Contribution to the theory of periodic reactions. J. Phys. Chem. 14
overshoot and collapse mode of the model] (3), 271–274.
“GF7 Pr‐inflow‐smooth”([(0,‐2)‐(100,1)],(0,‐2),(25,0),(40,1),(10 Ludwig, D., Walker, B., & Holling, C. S. (1997). Sustainability, Stability, and Resilience.
Ecol. Soc., 1(1 C7 - 7). 10.5751/es-00012-010107..
0,1))
Lundgren, J., 2020. The Grand Concepts of Environmental Studies Boundary objects
Units: Dmnl between disciplines and policymakers. J. Environ. Stud. Sci.. https://doi.org/
[Note this version of the function produces the predator 10.1007/s13412-020-00585-x.
smooth or damped oscillatory mode of the model] MacLulich, D.A., 1937. Fluctuations in the Number of the Varying Hare (Lepus
americanus). University of Toronto Press.
Pr outflow = 0 Meadows, D.H., Meadows, D.L., Randers, J., Behrens, W.W.I., 1972. The Limits to
Units: Predators/Year Growth. Universe Books, New York.
Meadows, D. H., Meadows, D. L., & Randers, J. (2004). Limits to growth the 30 year
update. White River Junction, VT: Chelsea Green Publishing Company..
Mollison, D., 1991. Dependence of epidemic and population velocities on basic
parameters. Math. Biosci. 107 (2), 255–287. https://doi.org/10.1016/0025-5564
(91)90009-8.
Murray, J.D., Seward, W.L., 1992. On the spatial spread of rabies among foxes with
References immunity. J. Theor. Biol. 156 (3), 327–348. https://doi.org/10.1016/S0022-5193
(05)80679-4.
Barlas, Y., 1996. Formal aspects of model validity and validation in system dynamics. Nevai, A.L., Van Gorder, R.A., 2012. Effect of resource subsidies on predator-prey
System Dynamics Review. J. Syst. Dynam. Soc. 12 (3), 183–210. population dynamics: a mathematical model. J. Biol. Dyn. 6 (2), 891–922. https://
Begon, M., 2009. Ecology From Individuals to Ecosystems. Wiley. doi.org/10.1080/17513758.2012.677485.
Berkes, F., & Folke, C. (1998). Linking social and ecological systems for resilience and Pal, D., Kar, T. K., Yamauchi, A., & Ghosh, B. (2020). Balancing maximum sustainable
sustainability. Linking social and ecological systems: management practices and social yield and ecological resilience in an exploited two-predator one-prey system.
mechanisms for building resilience, 1(4).. BioSystems, 187, 104064-104064. 10.1016/j.biosystems.2019.104064..
Blasius, B., Rudolf, L., Weithoff, G., Gaedke, U., Fussmann, G.F., 2020. Long-term cyclic Randers, J., 1997. Elements of the System Dynamics Method. J. Operat. Res. Soc. 48
persistence in an experimental predator–prey system. Nature (London) 577 (7789), (11), 1144–1145. https://doi.org/10.1057/palgrave.jors.2600456.
226–230. https://doi.org/10.1038/s41586-019-1857-0. Redman, C., 2014. Should sustainability and resilience be combined or remain distinct
Braun, W. (2002). The system archetypes. The Systems Modeling Workbook (Vol. 2002, pursuits?. Ecol. Soc. 19 (2), 1. https://doi.org/10.5751/ES-06390-190237.
pp. 27).. Rosenzweig, M.L., 1971. Paradox of Enrichment: Destabilization of Exploitation
Carl, F., Stephen, R.C., Brian, W., Marten, S., Terry, C., Johan, R., 2010. Resilience Ecosystems in Ecological Time. Sci. (Am. Associat. Adv. Sci.) 171 (3969),
Thinking: Integrating Resilience, Adaptability and Transformability. Ecol. Soc. 15 385–387. https://doi.org/10.1126/science.171.3969.385.
(4), 20. https://doi.org/10.5751/es-03610-150420. Rosenzweig, M.L., MacArthur, R.H., 1963. Graphical Representation and Stability
Edelstein-Keshet, L., 1988. Mathematical models in biology. Random House, New York. Conditions of Predator-Prey Interactions. Am. Nat. 97 (895), 209–223. https://doi.
Elton, C., Nicholson, M., 1942. The ten-year cycle in numbers of the lynx in Canada. J. org/10.1086/282272.
Animal Ecol. 11 (2), 215. Saha, T., Pal, P.J., Banerjee, M., 2021. Relaxation oscillation and canard explosion in a
Fiksel, J., 2006. Sustainability and resilience: toward a systems approach. Sustainab.: slow–fast predator–prey model with Beddington–DeAngelis functional response.
Sci., Pract. Policy 2 (2). Nonlinear Dyn. 103 (1), 1195–1217. https://doi.org/10.1007/s11071-020-06140-
Gilg, O., Hanski, I., Sittler, B., 2003. Cyclic dynamics in a simple vertebrate predator- 1.
prey community. Science 302 (5646), 866–868. Schwaninger, M., & Grösser, S. (2020). System dynamics modeling: validation for
Grace, W., & Pope, J. (2015). A systems approach to sustainability assessment. In quality assurance. System Dynamics: Theory and Applications, 119-138..
Morrison-Saunders A, Pope J, & B. A (Eds.), Handbook of Sustainability Assessment: Senge, P. M., & Forrester, J. W. (1980). Tests for building confidence in system dynamics
Edward Elgar.. models. System dynamics, TIMS studies in management sciences, 14, 209-228..
Grace, W. (2015). Simulating sustainability: a resources perspective. J. Nat. Resour. Pol. Stenseth, N.C., Falck, W., Bjørnstad, O.N., Krebs, C.J., 1997. Population regulation in
Res., 191-220. 10.1080/19390459.2015.1050202. snowshoe hare and Canadian lynx: asymmetric food web configurations between
Holling, C.S., 1973. Resilience and Stability of Ecological Systems. Annu. Rev. Ecol. Syst. hare and lynx. Proc. Natl. Acad. Sci. 94 (10), 5147–5152.
4 (1), 1–23. Swart, J., 1990. A system dynamics approach to predator-prey modeling. System
Holling, C., Walker, B., Carpenter, S., Kinzig, A., 2004. Resilience, adaptability and Dynamics Rev. 6 (1), 94–99.
transformability in social-ecological systems. Ecol. Soc. 9 (2), n/a. https://doi.org/ Toro, M., Aracil, J., 1988. Qualitative analysis of system dynamics ecological models.
10.5751/es-00650-090205. System Dynamics Rev. 4 (1–2), 56–80.
Kar, T.K., Pal, D., Ghosh, B., 2019. Managing yield and resilience in a harvested tri- Tromeur, E., Loeuille, N., 2017. Balancing yield with resilience and conservation
trophic food chain model. J. Theor. Biol. 469, 35–46. https://doi.org/10.1016/j. objectives in harvested predator–prey communities. Oikos 126 (12), 1780–1789.
jtbi.2019.02.006. https://doi.org/10.1111/oik.03985.
Kevan, P.G., Viana, B.F., 2003. The global decline of pollination services. Biodiversity World Commission on Environment and Development. (1987). Our Common Future.
(Nepean) 4 (4), 3–8. https://doi.org/10.1080/14888386.2003.9712703. Retrieved from Oxford University Press..

15

You might also like