Uncertainty pervades all aspects of modelling in marine ecology. It undermines our ability to forecast future dynamics, and it leads us to make incorrect conservation management decisions. It demands a fundamental break with classical approaches to modelling, and a new, orthogonal approach to making management decisions.
I will start by demonstrating that our problem is not risk (i.e., uncertainty about the future that can be quantified probabilistically), but Knightian uncertainty – an inability to represent novel, sparse-data, non-stationary systems. This form of uncertainty doesn't just affect our ability to make accurate forecasts, it renders us unable to make qualitative predictions. Is option A better than option B? Should I act at all?
Using three different separate examples - whole-ecosystem models, biophysical dispersal models, and marine spatial planning tools - I show that these problems are fundamental to the ecological questions we are asking, and to the conservation decisions that we need to make. I will argue that our standard toolbox is counter-productive when faced with these problems, and outline solutions that have been proposed in other contexts that may prove more effective.