Projecting the future state of marine ecosystems, "la grande illusion"?

Using numerical models to project the state of marine ecosystems several decades into the future is commonly advocated, in particular for investigating the possible effects of climate change. Numerical models are useful to explore how ocean climate and other drivers may regulate the dynamics of marine ecosystems and constitute indispensable tools to test our conceptual representations of how marine systems function. However, I argue here that these models might be of limited use to project the future state of marine ecosystems decades into the future because several factors limit predictability. These include stochasticity, deterministic chaos, enablement vs. entailment, non-ergodicity, ecological surprises, irreducibility, and limits to upscaling. Many simulations of ecosystem states in the distant future may be no more than a "grande illusion" until explicit evaluations of how uncertainties increase with the time horizon of projection are performed.