Response of Microcystis and Stephanodiscus to Alternative Flow Regimes of the Regulated River Nakdong (South Korea) Quantified By Model Ensembles Based on the Hybrid Evolutionary Algorithm (HEA)

Abstract

This study demonstrates the use of inferential models for scenario analyses by simulating direct and indirect effects of predictor variables on state variables through model ensembles. Two model ensembles have been designed to predict the response of the cyanobacterium Microcystis aeruginosa and the diatom Stephanodiscus hantzschii to modified flow regimes of the River Nakdong (Korea) by a scenario analysis. Whilst flow‐independent predictor variables of growth of Microcystis and Stephanodiscus such as water temperature and pH remain unchanged during the scenario analysis, flow‐dependent predictor variables such as turbidity, electrical conductivity, phosphate, nitrate, silica and chlorophyll a are forecasted by inferential models. In the course of scenario analysis, flow‐independent and flow‐dependent predictor variables feed into the Microcystis and Stephanodiscus models to make sure that both direct and indirect effects of altered flow regimes are taken into account. The eight inferential models that were incorporated into the model ensembles have been developed by the hybrid evolutionary algorithm based on 19 years of time‐series monitored in the River Nakdong between 1993 and 2012. The models achieved good accuracy in terms of timing and magnitudes reflected by coefficients of determination r2 = 0.94 for Microcystis and r2 = 0.83 for Stephanodiscus. The scenario analysis revealed that extreme summer blooms of Microcystis as observed between 1994 and 1997, and winter blooms of Stephanodiscus as observed between 1994 and 1997 and in 2004 can be prevented in the River Nakdong by adaptive management of seasonal water release from adjacent dams. Copyright © 2017 John Wiley & Sons, Ltd.