Ecosystems (2017) 20: 229–236

DOI: 10.1007/s10021-016-0071-2
Ó 2016 Springer Science+Business Media New York

20th Anniversary Paper

Next-Generation Individual-Based
Models Integrate Biodiversity and
Ecosystems: Yes We Can, and Yes We
Must
Volker Grimm,1,2* Daniel Ayllón,1 and Steven F. Railsback3

1
Department of Ecological Modelling, Helmholtz Centre for Environmental Research-UFZ, Permoserstr. 15, 04318 Leipzig, Germany;
2
German Centre for Integrative Biodiversity Research (iDiv) Halle-Jena-Leipzig, Deutscher Platz 5e, 04103 Leipzig, Germany; 3Lang
Railsback & Associates, 250 California Avenue, Arcata, California 95521, USA

ABSTRACT
Ecosystem and community ecology have evolved individual-based models will increasingly use
along different pathways, with little overlap. standardized and re-usable submodels to represent
However, to meet societal demands for predicting behaviors and mechanisms such as growth, uptake
changes in ecosystem services, the functional and of nutrients, foraging, and home range behavior.
structural view dominating these two branches of The strategy of pattern-oriented modeling then
ecology, respectively, must be integrated. Biodi- helps make such ecosystem models structurally
versity–ecosystem function research has addressed realistic by developing theory for individual
this integration for two decades, but full integration behaviors just detailed enough to reproduce and
that makes predictions relevant to practical prob- explain patterns observed at the system level. Next-
lems is still lacking. We argue that full integration generation ecosystem scientists should include the
requires going, in both branches, deeper by taking individual-based approach in their toolkit and focus
into account individual organisms and the evolu- on addressing real systems because theory devel-
tionary and physico-chemical principles that drive opment and solving applied problems go hand-in-
their behavior. Individual-based models are a ma- hand in individual-based ecology.
jor tool for this integration. They have matured by
using individual-level mechanism to replace the Key words: biodiversity research; emergence;
demographic thinking which dominates classical first principles; individual-based ecology; individ-
theoretical ecology. Existing individual-based ual-based modeling; pattern-oriented modeling;
ecosystem models already have proven useful both predictions; structural realism; theory develop-
for theory and application. Still, next-generation ment.

INTRODUCTION
In Richard Wagner’s opera ‘‘Siegfried,’’ the dwarf
Received 7 May 2016; accepted 17 July 2016; Mime tries to weld together the pieces of the magic
published online 10 November 2016
but broken sword Nothung. After welding fails,
Author contributions VG, SR, and DA wrote the paper; DA did the
literature review of Table 1. Siegfried, a headstrong and uncouth young man,
*Corresponding author; e-mail: [email protected]

229
230 V. Grimm and others

tries an entirely new approach: he reduces No- How could this integration be achieved? Loreau
thung to filings, melts them down, and succeeds in (2010) proposes ‘‘community evolution models,’’ in
re-forging the sword entirely anew. The broken which trait compositions evolve in response to
sword reminds us of the separate disciplines of environment and species interactions. This approach
ecosystem and community ecology, not yet inte- has led to interesting results, but the models are
grated via traditional approaches such as aggre- usually aggregated at the population or species level
gated system-level modeling. We certainly do not and again often only weakly linked to ecosystem
suggest reducing ecosystem and community ecol- ecology and evolutionary principles. A fundamen-
ogy to filings, but the metaphor lies in our con- tally different approach was suggested almost
viction that we need to revert to the building blocks 30 years ago: individual-based modeling. In their
of communities and ecosystems, individual organ- still intriguing vision, Huston and others (1988, p.
isms, before a true integration can be achieved. 690) wrote: ‘‘Individual-based models link all of
The separation of ecology into largely disjunct these separate levels in the ecological hierarchy. The
fields has been bemoaned for decades, for example responses of individuals to their local environment
from the perspective of modeling: ‘‘Each level of are based on physiological and behavioral responses.
organization has traditionally been a separate disci- The aggregation of all individuals of a species pro-
pline (for example, physiological ecology, behav- duces the population dynamics of that species.
ioral ecology, population ecology, and ecosystem The aggregation of all individuals of many species
ecology). Each field has its own set of phenomena to interacting with each other and with their environ-
explain, and most have their own distinctive types of ment produces community dynamics. Ecosystem
models.’’ (Huston and others 1988, p. 690). dynamics result from the aggregation of individual-
Ecosystem ecology, for example, focuses on biogeo- environment interactions into large-scale material
chemistry and the stocks and flows of energy and lim- and energy fluxes’’.
iting nutrients, whereas community ecology focuses on Huston and others (1988) believed that individ-
organisms and structure in terms of species richness and ual-based models will unify ecological theory and
composition. Organisms and species were ignored in predicted their rapid development. Progress,
ecosystem ecology, while energy and nutrients were though, has been slow (Grimm 1999; Grimm and
ignored in community ecology. Over the last two dec- Railsback 2005; DeAngelis and Grimm 2014). Re-
ades, though, a merger has started—referred to as bio- cent attempts to outline how structure and function
diversity–ecosystem function (BEF) research (Loreau could be integrated in large-scale ecosystem models
2010; Cardinale and others 2012)—which aims at do not even mention the individual-based approach
integrating structural and functional perspectives. This (Loreau 2010; Mokany and others 2016). IBMs are
merger was driven by the societal demand to under- an established tool for population ecology, but many
stand and predict how ecosystem function and services ecologists consider IBMs unsuitable for communities
respond to changes in land use and climate. To meet this and ecosystems. Although we acknowledge why this
demand, neither energy and nutrient constraints nor notion exists, we consider it fundamentally wrong.
organisms and how they interact with each other and Here, we discuss why progress integrating com-
their environment can be ignored (Evans and others munity and ecosystem ecology has been slow and
2013a; Mokany and others 2016). how, nevertheless, major obstacles have been
Still, there is no true integration of ecosystem overcome recently. Our central tenet is that, iron-
and community ecology yet. BEF research often ically, ecosystem and community ecology can only
relates biodiversity patterns to ecosystem functions be fully and mechanistically integrated when we
such as productivity and stability properties, but revert to their building blocks and acknowledge
mostly via small-scale experiments or meta-analy- individual organisms and their traits and adaptive
ses, which have limited ability to elucidate the behaviors. We believe, as Huston and others
mechanisms underlying observed relationships. (1988), that a unified ecology must start from
These experiments have also focused on grasslands, individual-based ecology (Grimm and Railsback
so research over more systems and larger scales is 2005). This belief does not imply that all ecological
still needed. Moreover, BEF research often does not models should be individual-based but that indi-
explicitly refer to biogeochemistry and energy vidual-based models must play a more central role
considerations. Loreau (2010) and others before by complementing more aggregated approaches.
him (Schulze and Mooney 1993; Tilman 1999) We discuss what this perspective means for com-
have therefore called for a more mechanistic inte- bining empirical study with modeling and theory in
gration of biodiversity research, which is largely ecosystem ecology and for the training of ecosys-
community ecology, with ecosystem ecology. tem ecologists for the 21st century.
 Next-Generation Individual-Based Models Integrate Biodiversity and Ecosystems 231

FROM DEMOGRAPHIC THINKING TO FIRST 2013; Sibly and others 2013) uses the same model for
PRINCIPLES all heterotrophic species, with species differing only in
parameterization. Likewise, ecophysiological forest
Classical theoretical ecology focuses on demographic gap models like FORMIND are based on general
rates of populations. Even if these rates depend on submodels of photosynthesis and competition for
size, stage, or age in structured population models, light (Köhler and Huth 1998; Fischer and others
they remain empirical (Figure 1). They reflect the 2016). Moreover, unlike system-level models, IBMs
observed net outcome of survival and reproduction are naturally suited for adaptive mechanisms from life
in a certain group of organisms over a certain period history selection (for example, Railsback and Harvey
of time. These rates can be used to extrapolate pop- 2013) to microevolution (for example, Ayllón and
ulation growth rates, but only for those environ- others 2016).
mental conditions and densities at which the rates The research fields dealing with these first princi-
were observed. This dependence on empirical rates ples usually focus on the individual organism and
limits the usefulness of classical theory for predicting ignore the population context. Next-generation
responses to novel conditions and for linking eco- individual-based models (Grimm and Berger 2016)
logical dynamics to first principles like fitness seek- link these principles to higher levels: populations
ing and energy budgets. and communities. Because model individuals can
Although most early IBMs also used demographic respond to and process nutrients and energy
rates as input, newer IBMs increasingly have demo- explicitly, IBMs of this type link community and
graphics that emerge from mechanisms such as fit- ecosystem ecology. Pioneering prototypes of such
ness-seeking behavior (Railsback and Harvey 2002; models were already developed more than 20 years
Stillman and Goss-Custard 2010), energy budgets ago, about moose foraging affecting plant commu-
(Martin and others 2012, 2013; Sibly and others 2013; nities and then nutrient cycling dynamics (Pastor
van der Vaart and others 2016), stoichiometry (Kaiser and Naiman 1992; Pastor and others 1998). IBMs
and others 2014; Smith and others 2014), or photo- also link community and ecosystem ecology by
synthesis (Fischer and others 2016). These mecha- allowing adaptation and evolution as driving pro-
nisms make IBMs more complex but also more cesses (Loreau 2010).
coherent because they are based on first princi- The integration of community and ecosystem
ples—processes we can observe in the laboratory in ecology must start from the community side, as in
real organisms or believe from the fundamental BEF research, as this is the less aggregated ap-
concepts of biochemistry, evolution, and so on. The proach. Integration has thus to add function to
same mechanisms can be used for different species structure, or biogeochemistry to demographics.
and contexts. For example, the energy budget theory However, classical models of community ecology,
of Kooijman (2010; see also Martin and others 2012, still mostly Lotka–Volterra type, do not provide
mechanistic interfaces to nutrients and energy. This
missing link can be made by going down two levels
of organization: to individuals and the principles
that drive their behavior.

FROM POPULATIONS TO COMMUNITIES AND

ECOSYSTEMS: COPING WITH COMPLEXITY
IBMs of populations can be complex if they are
based on first principles and designed to predict the
response of populations to heterogeneous and
changing environments. Developing them can be
‘‘big science,’’ requiring years for development and
testing (Stillman and others 2015).
Extrapolating from this experience with popula-
Figure 1. The structure and dynamics of ecological sys- tion IBMs makes IBMs of species-rich communities
tems emerge from how individual organisms perform and ecosystems seem impossible. However,
and how they interact with their biotic and abiotic
ecosystem IBMs have existed for decades, in par-
environment. Vital and growth are used in models as a
ticular for forests (Botkin and others 1972; Shugart
summary representation of processes at the individuals’
level, but they are linked to specific environmental 1984; Bugmann 2001) and, to a lesser extent,
conditions. marine fish communities (Shin and Cury 2001;
232 V. Grimm and others

Giacomini and others 2009, 2013; Rose and others mechanisms controlling them, biodiversity, and
2010). The trick to keep complexity at bay for ecosystem services). However, the design of these
multiple species is to use the same submodels for all models is still very heterogeneous, reflecting dif-
species within a functional group, for example, ferent backgrounds, modeling traditions, and re-
trees, and the same submodel for interactions search questions of their developers. We expect
among individuals of all species. Forest gap models that in the future individual-based modeling of
(Botkin and others 1972), for example, use the ecosystem will be more coherent by using generic
same growth equation for all trees and represent representation of individual organisms and their
species differences via parameters. Competition interactions.
among trees for light depends only on the vertical
leaf area distribution. Other plant models represent
FROM OBSERVED PATTERNS TO STRUCTURAL
competition via overlapping zones of influence
(Weiner and others 2001; May and others 2009; REALISM, PREDICTIONS, AND THEORY
Lin and others 2012), which can also be used for As key features of next-generation ecological
animals (Booth 1997; Piou and others 2007). In models, Grimm and Berger (2016) list emergence,
marine fish community models, trophic interac- structural realism, and prediction. Emergence from
tions are based on size differences among individ- first principles is needed to predict responses to
uals, independent of species (Shin and Cury 2001; novel conditions, but how can we achieve struc-
Giacomini and others 2009, 2013). Grimm and tural realism? Striving for structural realism means
Berger (2016) describe elements of ‘‘next-genera- trying to make models reproduce observed patterns
tion ecological modeling’’ that help cope with for the right reasons instead of forcing the right
complexity while aiming for structural realism and output via calibration: being able to parameterize a
predictions. Lotka–Volterra community model to reproduce, for
example, realistic cycles does not mean the model
YES, WE CAN: MODELS LINKING captures the mechanisms actually causing the cy-
cles.
BIODIVERSITY AND ECOSYSTEMS EXIST Pattern-oriented modeling (POM; Grimm and
Quite a few models linking communities and others 1996; Wiegand and others 2003; Grimm and
ecosystems already use the individual-based ap- others 2005; Grimm and Railsback 2012) is a gen-
proach. However, these models are, perhaps be- eral strategy to achieve structural realism. Multiple
cause of their complexity, not yet generally patterns, observed at different scales and levels of
perceived as good starting points for integrating observations, are used in model design as filters for
biodiversity and ecosystems. In biodiversity re- (1) selecting variables and processes, (2) designing
search, they seem to largely be ignored. In Table 1, submodels for mechanisms including individual
which was inspired by a similar table in Mokany behavior, and (3) parameterizing full models. POM
and others (2016), we compiled an overview of is not unique to ecological IBMs: using observed
spatially explicit, dynamic IBMs that link ecosys- patterns (referred to as ‘‘stylized facts’’ in eco-
tem structure and function with varying levels of nomics) to select variables and processes, and to
integration. These models use a wide range of ap- parameterize models (known also as ‘‘inverse
proaches and scopes and address the full spectrum modeling’’), is common in many kinds of model-
of scales of application and resolution. They de- ing, often unconsciously. Moreover, the pattern-
scribe composition and diversity either by treating oriented development of submodels corresponds to
each of multiple species differently or by applying the rationale of ‘‘strong inference’’ introduced by
trait-based functional groupings, and thus are Platt (1964). What is new is using patterns to find
capable of simulating driving ecological processes at IBM submodels that link individual-level mecha-
both species and ecosystem level. nisms to higher levels of organization. In particular,
These IBMs incorporate, to different degrees, realistic mechanisms can be designed via a
many of the essential modeling strategies and fea- hypothesis-testing approach: posing alternative
tures of next-generation ecological models and submodels for individual behavior and falsifying
therefore have the potential to provide not only them if they do not cause the IBM to reproduce
predictions of ecosystem dynamics in a changing patterns observed at higher levels.
world but also insights into fundamental theoreti- There are several examples of this POM strategy
cal and applied issues in ecosystem ecology (for for developing models of individual traits that ex-
example, stability properties and resilience and the plain population or higher-level ecological
Table 1. Main Features Defining the Structural Realism of a Number of Existing Spatially Explicit, Dynamic Individual-Based Models Linking
Ecosystem Structure and Function
Element Features/Models1 aDGVM2 Madingley OSMOSE Giacomini iLand FORMIND WIST WEAVER DOVE

Scales2 Spatial scale of application Global Global Regional Regional Landscape Stand Local Micro-site Virtual
Fine spatial resolution - - - Pot. + + + + Na
Fine temporal resolution + + + + + + + + Na
Ecosystem Multitrophic - + + + – – + + +
structure Multispecies/functional groups + + + + + + + + +
and function Parameterized at species level + - + + + + + + +
Non-trophic interactions + - - - + + + + -
Mass/energy fluxes across trophic - + + + - - + + +
levels
Biogeochemistry + - + - + + + - -
Model struc- First principles + + + + + + + + +
tural realism Spatio-temporal heterogeneity + + + + + + - + -
Trait-based approach + + - + - + + + +
Microevolution + - – + - - - + +
Resilience3 + (a,b) + (a) + (a,c) + (d) + (e) + (e) + (e) + (e) Nd
Model allows for simulation of + Pot. + Pot. + + Pot. + Pot.
multiple-stressors
Link species distribution to mecha- + + - + + + Pot. + Pot.
nisms
Modeling Pattern-oriented theory develop- + Nd Nd Nd Nd Nd Nd Nd Nd
strategies ment
Use of standardized generic sub- + + + + + + + + +
models
Use of standard models of interac- + + + + + + + + +
tions
Pattern-oriented calibration/testing + + + - + + - - -
Robustness analysis4 Nd Nd Nd Nd + Nd Nd Nd Nd
Human Can explicitly incorporate human Pot. Pot. + Pot. + + Pot. Pot. Pot.
dimension management actions

‘‘+’’ means ‘‘yes,’’ ‘‘-‘‘ means ‘‘no’’.

1
Model scope and reference: aDGVM2 (dynamic global vegetation; Scheiter and others 2013), Madingley (global ecosystem; Harfoot and others 2014), OSMOSE coupled with ROMS-N2P2D2Z2 (end-to-end marine ecosystem; Shin and Cury
2001; Travers-Trolet and others 2014), Giacomini (marine fish community; Giacomini and others 2009, 2013), iLand (forest landscape; Seidl and others 2012), FORMIND (forest gap; Köhler and Huth 1998; Fischer and others 2016),
WIST (terrestrial food web; Parrott and Kok 2002), WEAVER (soil food web; Moya-Laraño and others 2014), DOVE (terrestrial food web; Peacor and others 2007).
2
The spatial scale shown refers to the typical scale of application the model was designed for, but most models can be applied at different scales. Fine spatio-temporal resolution refers, for example, to <1 km2 and <1 month, respectively.
3
Resilience theories addressed, examples: (a) system stability and persistence, (b) shifts in trait diversity after fire, (c) regime shifts, (d) stability and persistence under environmental gradients, (e) recovery and persistence after natural/
anthropogenic disturbance events.
4
Next-Generation Individual-Based Models Integrate Biodiversity and Ecosystems

Robustness analysis refers to the systematic deconstruction of a model by forcefully changing the model’s parameters, structure, and representation of processes to detect the model mechanisms that explain a certain phenomenon break
down (Grimm and Berger 2016).
5
Nd Not documented, Pot. Potentially, Na not apply.
233
234 V. Grimm and others

dynamics (for example, Railsback and Harvey so on) in the laboratory, and the ecological theory
2002; Amano and others 2006; Cury and others we have most confidence in is that individuals
2008; Railsback and Johnson 2011). In a study behave to increase their future fitness. Therefore,
designed in part to illustrate POM, Railsback and as Huston and others (1988) pointed out, letting
Johnson (2011) developed a model of bird foraging system dynamics emerge from what individuals do
behavior for an IBM of how pest-control services is a productive way to integrate ecological levels. If
by songbirds depend on land use on coffee farms. we focus on individual-based approaches to
They identified nine patterns observed at the indi- ecosystems, how does our research and education
vidual bird, bird population, and pest population change?
levels. Then they posed four alternative submodels The biggest overall change we need is learning to
of foraging behavior as hypothesized traits. These think and model across ecological levels: instead of
traits included a ‘‘null’’ model (random foraging), a building models and testing them with data just on
simple model that assumed birds can sense local ecosystems, we need to think about and model
prey availability, a simple model that assumed birds how ecosystem characteristics emerge from char-
can sense prey over large distances, and a model acteristics and behaviors of individuals. But when
based on classical optimal foraging theory. They modeling and measuring individuals, we need to
implemented each of the four submodels in a sep- search for simple representations of individuals that
arate version of their IBM and conducted simula- are useful for explaining higher-level phenomena.
tion experiments to see which versions reproduced We do not want, for example, everything we know
which observed patterns. They found that only the about individual energetics in an IBM but only just
simple model with local sensing of prey availability enough to model energy flow through an ecosys-
could reproduce all the observed patterns. The high tem of unique and adaptive individuals. Seeing not
level of structural realism achieved with this ap- only the forest for the trees, but also the trees for
proach made it possible to predict bird populations, the forest is an apt metaphor for this kind of inte-
their pest-control services, and coffee production grative individual-based ecology of communities
under alternative scenarios of land sharing versus and ecosystems.
land sparing (Railsback and Johnson 2014). A second change is focusing more on real sys-
POM is the link we need between empirical sci- tems and applied problems, and making predictions
ence and theory: we use empirically observed pat- useful to managers. Classical theoretical ecology
terns to develop theory (submodels; Grimm and has sacrificed prediction for generality, which has
Railsback 2005) for individual-level mechanisms delayed the development of predictive theories for
that explains, via the IBM, observed patterns at the decades (Evans and others 2013b). Ecosystem
population, community, and ecosystem levels. ecology traditionally has a strong applied focus
Theory development via POM is a cycle: after one (Mokany and others 2016), but biodiversity re-
round of testing hypothesized submodels against search is still largely focused on theoretical issues
observed patterns, we can identify additional like stability–diversity relationships. A stronger fo-
observations that would help further refine the cus on applied questions and realistic settings is
theory. Therefore, POM is a productive way to important not just to make ecology more relevant,
identify the right mix of empirical science and but also because it produces better models and
theory in ecosystem ecology: we should collect science (Stillman and others 2015). Addressing
empirical observations that help us develop and applied problems not only forces us to tackle hard
refine theory via the POM cycle. and, therefore, unexplored questions, but also
makes productive strategies such as POM possible.
POM greatly facilitates model and theory develop-
NEXT-GENERATION ECOSYSTEM ment but only works in real systems with real
ECOLOGISTS: SEEING THE TREES FOR THE observations.
Using POM also implies a closer integration of
FOREST, AND THE FOREST FOR THE TREES
theoretical and empirical research. Individual-
We claim here that ecosystem ecology can be based ecology and POM provide a productive
integrated with community ecology and biodiver- framework for designing empirical research that
sity research by going all the way down to the directly supports theory and model development.
individual level. This idea may seem irrational until This approach is a way to design laboratory and
we remember that individuals are where theory is field studies at the individual level, where research
most firmly tied to reality: we can easily measure is much less expensive and uncertain, that still
and model individual mechanisms (physiology, and contribute directly to system-level understanding.
 Next-Generation Individual-Based Models Integrate Biodiversity and Ecosystems 235

We also continue to need system-level observa- tion: test of individual-based model in geese. Ecol Monogr
tions, but can focus on (often, qualitative and 76:601–16.
short-term) patterns useful for developing and Ayllón D, Railsback SF, Vincenzi S, Groeneveld J, Almodóvar A,
Grimm V. 2016. InSTREAM-Gen: modelling eco-evolutionary
testing IBMs instead of more extensive and quan- dynamics of trout populations under anthropogenic environ-
titative data. In the current fixation on ‘‘big data’’ mental change. Ecol Model 326:36–53.
and huge and ever-growing databases of, for Botkin DB, Janak JF, Wallis JR. 1972. Some ecological conse-
example, plant traits it is important to remember quences of a computer model of forest growth. J Ecol 60:849–72.
that data by themselves are useless; it is multiple Booth G. 1997. Gecko: a continuous 2-D world for ecological
patterns that we need to identify and decode. As modeling. Artif Life J 3:147–63.
Platt (1964) put it: ‘‘Many—perhaps most—of the Bugmann H. 2001. A review of forest gap models. Clim Change
great issues of science are qualitative, not quanti- 51:259–305.
tative, even in physics and chemistry. Equa- Cardinale BJ, Duffy JE, Gonzalez A, Hooper DU, Perrings C,
Venail P, Narwani A, Mace GM, Tilman D, Wardle DA, Kinzig
tions and measurements are useful when and only AP, Daily GC, Loreau M, Grace JB, Larigauderie A, Srivastava
when they are related to proof: but proof or dis- DS, Naeem S. 2012. Biodiversity loss and its impact on
proof comes first and is in fact strongest when it is humanity. Nature 486:59–67.
absolutely convincing without any quantitative Cury PM, Shin YJ, Planque B, Durant JM, Fromentin JM, Kra-
measurement.’’ (p. 352). mer-Schadt S, Stenseth NC, Travers M, Grimm V. 2008.
A final change is a culture change in theoretical Ecosystem oceanography for global change in fisheries. Trends
Ecol Evol 23:338–46.
ecology, so ‘‘theoretical’’ no longer means ‘‘hypo-
DeAngelis DL, Grimm V. 2014. Individual-based models after
thetical’’ but instead refers to the kinds of theory four decades. F1000Prime Reports 6(39):6.
used in hard sciences: models shown via testing to Evans MR, Bithell M, Cornell S, Dall SRX, Diaz S, Emmott S,
be useful and general for solving real problems. The Ernande B, Grimm V, Hodgson DJ, Lewis SL, Mace GM,
POM theory development cycle is a hypothesis- Morecroft M, Moustakas A, Murphy E, Newbold T, Petchey O,
testing approach for developing and documenting Smith M, Travis JMJ, Benton TG. 2013a. Predictive systems
the ability of IBMs to explain and predict ecosystem ecology. Proc R Soc B 280:20131452.
phenomena. Evans MR, Grimm V, Johst K, Knuuttila T, de Langhe R, Lessells
CM, Merz M, O’Malley MA, Orzack SH, Weisberg M, Wilkinson
Returning to the broken sword metaphor from DJ, Wolkenhauer O, Benton TG. 2013b. Do simple models lead
the Introduction, we are convinced that the bot- to generality in ecology? Trends Ecol Evol 28:578–83.
tom-up approach we suggest is needed to make Fischer R, Bohn F, Dantas de Paula M, Dislich C, Groeneveld J,
progress in ecosystems ecology and is feasible. Gutiérrez AG, Kazmierczak M, Knapp N, Lehmann S, Paulick
Simple and elegant models that can be formulated S, Pütz S, Roedig E, Taubert F, Köhler P, Huth A. 2016. Les-
and analyzed within a few weeks are still important sons learned from applying a gap model to complex forests
and their carbon dynamics. Ecol Model 326:124–33.
as heuristic tools, but predictive modeling of
Giacomini HC, De Marco P, Petrere M. 2009. Exploring com-
ecosystems, to tackle both applied and basic ques- munity assembly through an individual-based model for
tions, will to a large degree be ‘‘big science,’’ trophic interactions. Ecol Model 220:23–39.
requiring long-term collaboration of interdisci- Giacomini HC, DeAngelis DL, Trexler JC, Petrere M. 2013. Trait
plinary teams. contributions to fish community assembly emerge from
After re-forging the magic sword Nothung, trophic interactions in an individual-based model. Ecol Model
Siegfried demonstrates its strength by cutting, with 251:32–43.
one strike, the anvil in two pieces. This is where the Grimm V, Frank K, Jeltsch F, Brandl R, Uchmanski J, Wissel C.
1996. Pattern-oriented modelling in population ecology. Sci
metaphor ends, but it confirms our vision of a Total Environ 183:151–66.
powerful predictive and integrated ecosystem and Grimm V. 1999. Ten years of individual-based modelling in
biodiversity research, where structure and function ecology: what have we learned and what could we learn in
are linked via the individual-based perspective. the future? Ecol Model 115:129–48.
Grimm V, Railsback SF. 2005. Individual-based modeling and
ACKNOWLEDGEMENTS ecology. Princeton: Princeton University Press.
Grimm V, Revilla E, Berger U, Jeltsch F, Mooij WM, Railsback
We thank Monica Turner and Steve Carpenter for SF, Thulke HH, Weiner J, Wiegand T, DeAngelis DL. 2005.
their invitation to contribute to this special feature Pattern-oriented modeling of agent-based complex systems:
of Ecosystems and for their comments on an earlier lessons from ecology. Science 310:987–91.
draft. Grimm V, Railsback SF. 2012. Pattern-oriented modelling: a
‘multi-scope’ for predictive systems ecology. Philos Trans R
Soc B 367:298–310.
REFERENCES
Grimm V, Berger U. 2016. Structural realism, emergence, and
Amano T, Ushiyama K, Moriguchi S, Fujita G, Higuchi H. 2006. predictions in next-generation ecological modelling: synthesis
Decision-making in group foragers with incomplete informa- from a special issue. Ecol Model 326:177–87.
236 V. Grimm and others

Harfoot MBJ, Newbold T, Tittensor DP, Emmott S, Hutton J, Railsback SF, Johnson MD. 2011. Pattern-oriented modeling of
Lyutsarev V, Smith MJ, Scharlemann JP, Purves DW. 2014. bird foraging and pest control in coffee farms. Ecol Model
Emergent global patterns of ecosystem structure and function 222:3305–19.
from a mechanistic general ecosystem model. PLoS Biol Railsback SF, Johnson MD. 2014. Effects of land use on bird
12:e1001841. populations and pest control services on coffee farms. Proc
Huston M, DeAngelis D, Post W. 1988. New computer models Natl Acad Sci 111:6109–14.
unify ecological theory. Bioscience 38:682–91. Railsback SF, Harvey BC. 2013. Trait-mediated trophic interac-
Kaiser C, Franklin O, Dieckmann U, Richter A. 2014. Microbial tions: is foraging theory keeping up? Trends Ecol Evol 28:119–
community dynamics alleviate stoichiometric constraints 25.
during litter decay. Ecol Lett 17:680–90. Rose KA, Allen JI, Artioli Y, Barange M, Blackford J, Carlotti F,
Köhler P, Huth A. 1998. The effects of tree species grouping in Cropp R, Daewel U, Edwards K, Flynn K, Hill S, Hille Ris
tropical rainforest modelling: simulations with the individual- Lambers R, Huse G, Mackinson S, Megrey BA, Moll A, Rivkin
based model FORMIND. Ecol Model 109:301–21. R, Salihoglu B, Schrum C, Shannon L, Shin Y, Smith SL,
Kooijman SALM. 2010. dynamic energy budget theory for me- Smith C, Solidoro C, St John M, Zhou M. 2010. End-to-end
tabolic organisation. Cambridge: Cambridge University Press. models for the analysis of marine ecosystems: challenges, is-
Lin Y, Berger U, Grimm V, Ji Q-R. 2012. Differences between sues, and next steps. Mar Coast Fish 2:115–30.
symmetric and asymmetric facilitation matter: exploring the Scheiter S, Langan L, Higgins SI. 2013. Next-generation dynamic
interplay between modes of positive and negative plant global vegetation models: learning from community ecology.
interactions. J Ecol 100:1482–91. New Phytol 198:957–69.
Loreau M. 2010. Linking biodiversity and ecosystems: towards a Schulze ED, Mooney HA, Eds. 1993. Biodiversity and ecosystem
unifying ecological theory. Philos Trans R Soc B 365:49–60. function Berlin. Berlin: Springer.
Martin BT, Zimmer EI, Grimm V, Jager T. 2012. Dynamic energy Seidl R, Rammer W, Scheller RM, Spies TA. 2012. An individual-
budget theory meets individual-based modelling: a generic based process model to simulate landscape-scale forest
and accessible implementation. Methods Ecol Evol 3:445–9. ecosystem dynamics. Ecol Model 231:87–100.
Martin BT, Jager T, Nisbet RM, Preuss TG, Grimm V. 2013. Shin YJ, Cury P. 2001. Exploring fish community dynamics
Predicting population dynamics from the properties of indi- through size-dependent trophic interactions using a spatial-
viduals: a cross-level test of dynamic energy budget theory. ized individual-based model. Aquat Living Res 14:65–80.
Am Nat 181:506–19. Shugart HH. 1984. A theory of forest dynamics: the ecological
May F, Grimm V, Jeltsch F. 2009. Reversed effects of grazing on implications of forest succession models. New York: Springer.
plant diversity: the role of below-ground competition and size Sibly RM, Grimm V, Martin BT, Johnston ASA, Kułakowska K,
symmetry. Oikos 118:1830–43. Topping CJ, Calow P, Nabe-Nielsen J, Thorbek P, DeAngelis
Mokany K, Ferrier S, Connolly SR, Dunstan PK, Fulton EA, DL. 2013. Representing the acquisition and use of energy by
Harfoot MB, Harwood TD, Richardson AJ, Roxburgh SH, individuals in agent-based models of animal populations.
Scharlemann JPW, Tittensor DP, Westcott DA, Wintle BA. Methods Ecol Evol 4:151–61.
2016. Integrating modelling of biodiversity composition and Smith B, Wårlind D, Arneth A, Hickler T, Leadley P, Siltberg J,
ecosystem function. Oikos 125:10–19. Zaehle S. 2014. Implications of incorporating N cycling and N
Moya-Laraño J, Bilbao-Castro JR, Barrionuevo G, Ruiz-Lupión limitations on primary production in an individual-based dy-
D, Casado LG, Montserrat M, Melián CJ, Magalhães S. 2014. namic vegetation model. Biogeosciences 11:2027–54.
Eco-evolutionary spatial dynamics: rapid evolution and iso- Stillman RA, Goss-Custard JD. 2010. Individual-based ecology of
lation explain food web persistence. In: Moya-Laraño J, coastal birds. Biol Rev 85:413–34.
Rowntree J, Woodward G, Eds. Advances in ecological re- Stillman RA, Railsback SF, Giske J, Berger U, Grimm V. 2015.
search, Vol. 50. Oxford: Academic Press. p 75–143. Making predictions in a changing world: the benefits of
Parrott L, Kok R. 2002. A generic, individual-based approach to individual-based ecology. Bioscience 65:140–50.
modelling higher trophic levels in simulation for terrestrial Tilman D. 1999. The ecological consequences of changes in
ecosystems. Ecol Model 154:151–78. biodiversity: a search for general principles. Ecology 80:1455–
Pastor J, Naiman RJ. 1992. Selective foraging and ecosystem 74.
processes in boreal forests. Am Nat 139:690–705. Travers-Trolet M, Shin YJ, Field JG. 2014. An end-to-end cou-
Pastor J, Dewey B, Moen R, Mladenoff DJ, White M, Cohen Y. pled model ROMS-N2P2Z2D2-OSMOSE of the southern Ben-
1998. Spatial patterns in the moose–forest–soil ecosystem on guela foodweb: parameterisation, calibration and pattern-
Isle Royale, Michigan, USA. Ecol Appl 8:411–24. oriented validation. Afr J Mar Sci 36:11–29.
Peacor SD, Allesina S, Riolo RL, Hunter TS. 2007. A new com- van der Vaart E, Johnston AS, Sibly RM. 2016. Predicting how
putational system, DOVE (Digital Organisms in a Virtual many animals will be where: how to build, calibrate and
Ecosystem), to study phenotypic plasticity and its effects in evaluate individual-based models. Ecol Model 326:113–23.
food webs. Ecol Model 205:13–28. Weiner J, Stoll P, Müller-Landau H, Jasentuliyana A. 2001. The
Platt JR. 1964. Strong inference. Science 146:347–53. effects of density, spatial pattern, and competitive symmetry
Piou P, Berger U, Hildenbrandt H, Grimm V, Diele K, D’Lima C. on size variation in simulated plant populations. Am Nat
2007. Simulating cryptic movements of a mangrove crab: 158:438–50.
recovery phenomena after small scale fishery. Ecol Model Wiegand T, Jeltsch F, Hanski I, Grimm V. 2003. Using pattern-
205:110–22. oriented modeling for revealing hidden information: a key for
Railsback S, Harvey B. 2002. Analysis of habitat selection rules reconciling ecological theory and conservation practice. Oikos
using an individual-based model. Ecology 83:1817–30. 100:209–22.