MEPS 640:63-77 (2020)  -  DOI: https://doi.org/10.3354/meps13237

ABSTRACT: Supported by chl a satellite data in the North Atlantic (and phytoplankton division rates computed from that data), the disturbance-recovery hypothesis for the initiation of phytoplankton blooms posits that the change in chl a concentration is proportional to the relative change in the phytoplankton division rate. We used this hypothesis, introduced by Behrenfeld, as a principal model assumption and constructed a non-autonomous ordinary differential equation model for seasonally varying chl a concentrations. Our quantitative comparison between model simulations and in situ measurements of chl a and primary production collected from a Swedish fjord was 2-fold: first, using approximate Bayesian computations, we found distributions of values for the 3 model parameters that best described the chl a data. Then, we validated our model by comparing the simulated (not fitted) division rate to the division rate determined from the data. Our minimalistic model was able to capture (1) the yearly trend in the chl a concentration, (2) the pattern of growth and decline in the phytoplankton division rate, and (3) the decreasing trend in the relative change of the division rate exhibited in the data for several individual years. Moreover, the modeling efficiency was positive (between 0.3 and 0.9 with an average of 0.63) for all 11 yr included in this study. We conclude that the change in chl a concentration being proportional to the relative change in the division rate is a possible explanation for the bloom dynamics in the Gullmar fjord. In addition, our work provides a simple and empirically based differential equation for representing yearly dynamics of primary production, e.g. for generating ecological hypotheses using models of other trophic levels.

KEY WORDS: Non-autonomous ordinary differential equations · Primary production · Chlorophyll a · Division rate · Gullmar fjord · Approximate Bayesian computation inference