Water eutrophication assessment relied on various machine learning techniques: A case study in the Englishmen Lake (Northern Spain)

Algal atypical productivity, also called eutrophication, is a process where the phosphorus content in the water, together with aquatic flora, increases, causing high Chlorophyll levels and affecting the water quality and its possible applications. Therefore, it is important to be able to anticipate such circumstance to avoid subsequent hazards. In this paper, a model that estimates the conditions where an abnormal growth of algae in reservoirs and lakes takes place is built. This method combines artificial bee colony and support vector machines algorithms to predict the eutrophication taking into account physical-chemical and biological data sampled in the Englishmen Lake and posterior analysis in a laboratory. The support vector machines parameters are tuned by means of the artificial bee colony algorithm, improving the accuracy of the procedure. For comparison sake, two other methods have been used to construct additional models, the M5 model tree and multilayer perceptron network. Two objectives are covered by this study: the forecasting of the algal proliferation by means of the model and, the ranking of the relative importance of the independent variables. Indeed, coefficients of determination of 0.92 for the Chlorophyll and 0.90 for the Total phosphorus concentrations were obtained using this hybrid method that optimizes the regression parameters. Furthermore, the results obtained with M5 model tree and multilayer perceptron network techniques were clearly worse. Finally, conclusions of this work are drawn in the final section.