Modelling and forecasting the heterogeneous distribution of picocyanobacteria in the tropical Lajes Reservoir (Brazil) by evolutionary computation

Abstract

Five years of water quality data from six stations across the mesotrophic and oligomictic Lajes Reservoir (Brazil) were utilized to develop 7-day ahead forecasting models for the picocyanobacteria Cyanodictyon imperfectum, Cyanogranis ferruginea and Synechococcus sp. by means of the hybrid evolutionary algorithm HEA. The data included physical and chemical water quality parameters as well as abundance data of the three picocyanobacteria. Models based on site-specific data of six monitoring stations forecasted population dynamics of Synechococcus with coefficients of determination (r 2) between 0.58 for and 0.88, of Cyanodictyon with r 2 between 0.5 and 0.89 and of Cyanogranis with r 2 between 0.53 and 0.77. Despite phosphorus limiting conditions the sensitivity analysis revealed that the three picocyanobacteria responded much stronger to nitrate rather than to phosphate concentrations throughout the Lajes Reservoir suggesting that cyanobacteria may have adopted the sulphur-for-phosphorus strategy by utilizing sulfolipids instead. Cyanogranis displayed a negative relationship with increasing water temperature indicating its higher competitiveness at internal nutrient supply and low light levels during winter turnover. The resulting models will inform operational intervention and prevention of fast growth and dispersal of picocyanobacteria in Lajes Reservoir, and reveal environmental thresholds for outbreaks of such events.