Towards a robust concept for modelling zooplankton migration

We develop a model framework for handling some aspects of zooplankton behaviour in biophysical simulation models. The vertical behaviour of zooplankton (their diel vertical migration pattern, or DVM) is modelled as a stimuli-response process where inputs from the environment (i.e. presence of light, food, and predators) are used as decision parameters. By use of simple artificial neural networks to control behaviour and optimising these by genetic algorithms we obtain close to optimal sets of DVM strategies. We show how adding increasing levels of environmental complexity (number of stimuli, stochasticity, and seasonality) to the model affects predicted behaviour. Within 50 generations of adaptation the model strategies perform reasonable well under the levels of environmental complexity used in simulations. The performance of strategies tended to increase with the complexity level of inputs used to control behaviour. By adapting strategies to environments with a stochastic variability we show that the strategies become more plastic to unpredictable changes in environmental forcing. Also, we discuss how plastic strategies with high degree of robustness to environmental fluctuations of the magnitude encountered in nature can be derived from this method. Results are discussed in relation to the behavioural patterns observed in the field.