Threshold quantification and short-term forecasting of Anabaena, Aphanizomenon and Microcystis in the polymictic eutrophic Lake Müggelsee (Germany) by inferential modelling using the hybrid evolutionary algorithm HEA

Abstract

Forecasting models for Anabaena, Aphanizomenon and Microcystis have been developed for the hypertrophic phase from 1979 to 1990 and the eutrophic phase from 1997 to 2012 of the polymictic Lake Muggelsee by means of the hybrid evolutionary algorithm HEA. Comparisons of limnological parameters of the two phases revealed not only a distinct seasonal extension of N-limitation but also higher water temperatures that rose earlier and lasted longer between spring and autumn from 1997 to 2012. These differences were reflected by threshold conditions and sensitivity functions of the cyanobacteria-specific models evolved by HEA for the two phases. Seven-day-ahead forecasts matched well timings of peaking biomass observed for the three cyanobacteria but partially failed to predict accurate magnitudes, whereby coefficients of determination r 2 ranged between 0.48 and 0.76 for models in Phase I and between 0.42 and 0.69 in Phase II. The threshold conditions of the models quantified ranges of key predictor variables such as water temperature and transparency, concentrations of NO3-N and PO4-P that were symptomatic for sudden outbreaks of high biomass of the three cyanobacteria. Sensitivity functions extracted from 20 best performing models for each of the three cyanobacteria in both phases indicated different abundances between N-fixing Anabaena and Aphanizomenon compared to non-N-fixing Microcystis in response to strengthened N-limitation in Phase II.