Group Leader UiB, Professor, PhD

Dag L. Aksnes

Mechanistic modeling in biological oceanography

A major endeavor in ecology is to gain understanding of how biological processes, organisms, and ecosystems respond to environmental changes. Such responses are mapped in experimental and field studies and frequently quantified by means of equations that are fitted to the observations. The choice of equation is often ad hoc with little or no claim of particular mechanistic content. Ecological models, and in particular ecosystem models, rests on a number of such empirical equations that reduces the explanatory, but also the predictive, power of such models. Thus it is a challenge to derive mechanistic models founded on hypotheses, to test, to modify, to test again, and so on (i.e. strong interference). In this way, general explanatory (and predictive) models emerge in a stepwise manner.

An example: Michaelis-Menten type saturation curves are widely used for observed phenomena such as growth rate, food intake, microbial nutrient uptake versus resource concentration, and to predict outcome of competition. This modeling paradigm involves observed half saturation constants commonly assumed species specific. The assumption of constancy, however, appears unfounded in most modeling contexts. Thus a scientific challenge is to find an alternative formulation – that proves better.

Why mechanistic modeling: Equations that are derived mechanistically, i.e. based on explicit hypotheses concerning causality for the observed phenomenon, have three main advantages over ad hoc equations. First, they include parameters that are physically and biologically interpretable. Second, they provide expectations on how observed quantities respond to e.g. environmental changes. Third, the equation and its underlying hypotheses can be tested and modified through strong interference with experiments.