Simulation Modeling — Florian Hartig

Opening paragraph

With the rise of computers, simulation models have emerged beside the more traditional statistical and mathematical models as a third pillar for ecological analysis. Broadly speaking, a simulation model is an algorithm, typically implemented as a computer program, which propagates the states of a system forward. Unlike in a mathematical model, however, this propagation does not employ the methods of calculus but rather a set of rules or formulae that directly prescribe the next state. Such an algorithmic model specification is particularly suited for describing systems that are difficult to capture or analyze with differential equations such as: (a) systems that are highly nonlinear or chaotic; (b) discrete systems, for example networks or groups of distinct individuals; (c) systems that are stochastic; and (d) systems that are too complex to be successfully treated with classical calculus. As these situations are frequently encountered in ecology, simulation models are now widely applied across the discipline. They have been instrumental in developing new insights into classical questions of species’ coexistence, community assembly, population dynamics, biogeography, and many more. The methods for this relatively young field are still being actively developed, and practical work with simulation models requires ecologists to learn new skills such as coding, sensitivity analysis, calibration, validation, and forecasting uncertainties. Moreover, scientific inquiry with complex systems has led to subtle changes to the philosophical and epistemological views regarding simplicity, reductionism, and the relationship between prediction and understanding.

What the paper shows and why it matters

(AI-generated)

Ecologists work with statistical and mathematical models, but many ecological systems are hard to capture with differential equations: they are nonlinear or chaotic, made of discrete units such as individual organisms, driven by chance, or too complex for calculus. Simulation models take another route. They are computer programs that move a system forward step by step by explicit rules, and they are used across the discipline. Working with them requires skills such as programming, sensitivity analysis (testing how outputs respond to changed inputs), calibration (fitting parameters to data), validation and forecasting under uncertainty. Scientific inquiry with complex systems has also led to subtle changes in philosophical views on simplicity, reductionism and the relation between prediction and understanding. The article responds to the need for simulation models of such systems and for the concepts and methods to build and test them.

This article, an annotated bibliography in the Oxford Bibliographies in Ecology, presents simulation modelling (the use of computer programs that step a system forward by explicit rules) as a third pillar of ecological analysis beside statistical and mathematical models, and guides readers through the literature on it. It begins with overviews, textbooks and journals, then covers core concepts: models of discrete entities (cellular automata, networks, individual-based models, which follow single organisms), emergence, stochasticity and chaos, and what it means for a model to be valid. A sample of applications follows, from population and community models to movement and evolution, explicitly a cross-section rather than a complete survey. The route ends with the working steps: programming languages, checking that the code does what is intended, sensitivity analysis, calibration, and forecasting with uncertainty analysis. On the philosophy of models, the article suggests that modellers have become bolder: simulations had come to be used frequently to test or compare hypotheses, which implicitly assumes that they are structurally realistic, going beyond the view that a model only has to be valid for its purpose.

The article is a reference map rather than a new finding. For ecologists who want to work with simulation modelling, it connects what these models are good for with the practical steps of building, fitting and forecasting with them, and it notes that the methods of this relatively young field were still under development.