Integrating agent-based models into the ensemble ecosystem modelling framework: a rewilding case study at the Knepp Estate, UK.
Abstract
Ensemble ecosystem models (EEMs) are powerful tools for assessing the impacts of ecological management decisions made under uncertainty. However, their simplicity means they can omit finer-scale spatial interactions and individual-level behaviours, potentially limiting the "what if" scenario experiments available to managers and the ability of EEMs to replicate observed dynamics in relatively data-rich case studies. In this article, we introduce a spatial agent-based model (ABM) to explore the benefits of integrating a more complex model into the EEM framework. The ABM-EEM is used to assess the impacts of rewilding at the Knepp Estate, UK, and predict future ecosystem states under various management intervention scenarios. A wide range of questions is explored, for example, how have species reintroductions impacted the ecosystem, and how might they continue to in the future? What is the chance of wood-pasture habitats being maintained in the long term? Can the addition of new species or changes to stocking densities affect this probability? The ABM-EEM approach had several advantages. Compared with a previous EEM of Knepp, calibrating the ABM-EEM was more efficient and identified a wider range of acceptable outputs. Incorporating space and individual-level dynamics enabled a more nuanced evaluation of the possible impacts of hypothetical European bison or elk reintroductions, and it revealed an emergent behaviour: woodland regenerated more frequently in thorny scrubland when large herbivores were present. Practical implication. While the ABM-EEM approach comes with disadvantages such as higher computational demands and reduced transparency, it shows promise for conservation planning and decision-making, particularly for relatively data-rich case studies that may require a more complex model.