Fig. 2: Comparison between simple and hyper-real environments
in ABMs. (A) The Simple Environment represents a grid-based, homogeneous
landscape with abstract interactions. (B) The Hyper-Real Environment
shows the Bohemian Forest ecosystem, incorporating spatially explicit
data layers such as administrative boundaries, habitats, elevation, and
species territories for more realistic simulations.
Implementing these parameters into the model requires careful
sensitivity analysis. For instance, directly applying mortality rates
derived from monitoring data may lead to unrealistic outcomes, such as
population crashes in simulations. By iteratively testing parameter
values, sensitivity analysis ensures that the model aligns with
ecological expectations, such as population stability under known
conditions. For example, while monitoring data may suggest a 1% annual
mortality rate, sensitivity testing might reveal that a slight
adjustment (e.g., 0.8%) ensures population stability in the model,
reflecting real-world dynamics. This process is critical to ensure that
parameters remain ecologically sound and do not introduce unintended
biases into the simulation.
By leveraging these tools and processes, ABMs achieve a balance between
data-driven realism and adaptive flexibility. The capacity to
parameterise agents and environments using empirical data enhances model
accuracy and relevance while maintaining the unpredictability essential
for simulating emergent ecological trends. This, combined with rigorous
sensitivity testing and collaborative data preparation, provides a
robust foundation for answering complex ecological and management
questions.
Parameterisation
Agent decision-making should be informed by all available relevant
input data and parameter databases (Fig. 3), encompassing both static
input data and continuous access to external parameter databases. For
instance, the movement procedure illustrated in Fig. 4 highlights how
agents can use dynamic data access to make habitat-informed decisions.
In this example, an agent (e.g., a badger) selects a movement step
length and turning angle from a telemetry-derived parameter database.
Once a step is initiated, the agent evaluates the habitats within its
radius, accessing an external resource selection function (RSF) database
to determine the relative suitability of each habitat type. If multiple
habitats are available, the agent moves toward the one with the highest
RSF value. Conversely, if only one habitat is present, the agent moves
stochastically while adhering to data-driven parameters such as step
length and turning angle. This dynamic integration ensures that movement
is both realistic and adaptive, replicating observed patterns while
allowing emergent behaviours.