Fusing and disaggregating models, data and analysis tools for a dynamic science–society interface

Abstract

AbstractSociety requires rapid, most-probable predictions for specific and/or multifaceted questions related to environmental and geological science. In principle, models that encapsulate disciplinary knowledge are useful tools for making predictions and testing theory, but academic rewards favour disciplinary specialism and a proliferation of often insufficiently tested models. Decision makers have to assess the quality and robustness of predictions for complex environmental issues, and may prefer a model that performs accurately in a case study to a more parsimonious and generalizable model. Predictive ecosystem models tend to grow, as more processes are considered, even when a simpler model may be more appropriate and give results that are easier to interpret within a policy-relevant timeframe. Model fusion provides a practical way to combine knowledge from different disciplines, but can accelerate model growth. How then can we facilitate the evolution of useful predictive models? Coherent design is essential. When combining models it is often necessary to resolve overlapping scope, so tools need to allow for the disaggregation of model implementations as well as their fusion. Modelling software and integration frameworks can help resolve technical constraints, but to make models useful and used it is essential to involve stakeholders in their design and interpretation.