Developing multidisciplinary mechanistic models: challenges and approaches

Current biodiversity models often struggle to represent the complexity of global crises, as the affected ecosystems are shaped by many different ecological, physical, and social processes. To understand these dynamics better, we will need to build larger and more complex ecological models, and couple ecological models to models produced by other disciplines, such as climate science, economics, or sociology. However, constructing such integrated models is a significant technical undertaking, which has received little attention by ecological modellers so far. We review literature from computer science and several other environmental modelling disciplines to identify common challenges and possible strategies when creating large integrated models. We show that there is a software-architectural trade-off between modularity and integration, where the former is required to keep the technical complexity of a model manageable, and the latter is desirable to represent the scientific complexity of a studied system. We then present and compare five different software engineering techniques for navigating this trade-off. Which technique is most suitable for a given model depends on the model’s aims and the available development resources. The larger a model becomes, the more important it is to use more advanced techniques, such as integrating models from different domains using a model coupling framework. Our review shows that ecological modellers can learn from other modelling disciplines, but also need to invest in increased software engineering expertise, if they want to build models that can represent the numerous processes affecting ecosystems and biodiversity loss. IN PRESS. ARTICLE ACCEPTED FOR PUBLICATION 20/8/2024. FULL ARTICLE WILL BE AVAILABLE ONLINE SOON.