Inferred support for disturbance-recovery hypothesis of North Atlantic phytoplankton blooms

Analyses of satellite-derived chlorophyll data indicate that the phase of rapid phytoplankton population growth in the North Atlantic (the ‘spring bloom') is actually initiated in the winter rather than the spring, contradicting Sverdrup's Critical Depth Hypothesis. An alternative disturbance-recovery hypothesis (DRH) has been proposed to explain this discrepancy, in which the rapid deepening of the mixed layer reduces zooplankton grazing rates sufficiently to initiate the bloom. We use Bayesian parameter inference on a simple Nutrient-Phytoplankton-Zooplankton (NPZ) to investigate the DRH and also investigate how well the model can capture the multiyear and spatial dynamics of phytoplankton concentrations and population growth rates. Every parameter in our NPZ model was inferred as a probability distribution given empirical constraints, this provides a more objective method to identify a model parameterisation given available empirical evidence, rather than fixing or tuning individual parameter values. Our model explains around 75% of variation in the seasonal dynamics of phytoplankton concentrations, 30% of variation in their population rates of change, and correctly predicts the phases of population growth and decline. Our parameter-inferred model supports DRH, revealing the sustained reduction of grazing due to mixed layer deepening as the driving mechanism behind bloom initiation, with the relaxation of nutrient limitation being another contributory mechanism. Our results also show that the continuation of the bloom is caused in part by the maintenance of phytoplankton concentrations below a level that can support positive zooplankton population growth. Our approach could be employed to formally assess alternative hypotheses for bloom formation