Ecological relationships, thresholds and time-lags determining phytoplankton community dynamics of Lake Kinneret, Israel elucidated by evolutionary computation and wavelets

Abstract

The hybrid evolutionary algorithm (HEA) was implemented to model and analyze population dynamics of the different phytoplankton phyla (chlorophyta, bacillariophyta, cyanophyta and dinophyta) in relation to physical, chemical, and biological determinants and their combinations in a large lake. Biweekly measurements over a 12-year period were used as input. The validation of models obtained with HEA showed the best results for bacillariophyta and dinophyta resulting in coefficients of determination (r2) between the modeled and measured data of 0.54–0.79 and 0.29–0.76 for these phyla, respectively, suggesting good predictability of their dynamics. The lowest adequacy of HEA models was found for cyanophyta (r2 of 0.28–0.46). Models that combined physical, chemical and biological inputs scored highest, whilst zooplankton-based models scored lowest in all experiments and indicated that top-down control of algal biomass could have only secondary effect. The input sensitivity analysis was used for testing the best phytoplankton models with threshold values determining high or low algal biomass and inhibitory–excitatory effects of specific parameters. Wavelets were tested to analyze two extreme cases of dinophyta dynamics in years of its exceptionally high and low developments to gain insights into lag times between the exert of key factor and algae response. Lag times extracted from daily interpolated data of highly correlated inputs of dinophyta in 1998 varied between 2 and 4 days.