Development of an algal bloom prediction algorithm by integrating CE-QUAL-W2 and optimization techniques

This study developed an algorithm to predict algal blooms in river systems by integrating the two-dimensional CE-QUAL-W2 water quality model with optimization techniques. The study area was the Nakdong River system in South Korea, and the model was constructed for eight multipurpose weir sections. Sensitivity analysis was used to identify key parameters influencing algal concentrations, and an optimization algorithm was developed using the bagging ensemble method. The algorithm aimed to minimize the relative error (<i>%Difference</i>) between simulated and observed chlorophyll-a concentrations, which served as the target variable. The model achieved a "Very Good" rating in the overall efficiency assessment across all target weir sections. Furthermore, a separate evaluation of temporal algal trends was conducted, which showed that while some sections received a 'Poor' rating during low-concentration periods, most sections achieved a 'Good' or higher rating overall. In addition, the model effectively captured temporal patterns and spatial heterogeneity, demonstrating its adaptability to complex hydrodynamic conditions. By integrating machine learning techniques into the physically-based modeling framework, the proposed algorithm is expected to enhance model reliability and improve optimization efficiency. These improvements are applicable to water quality prediction and management systems.