Evaluation of models generated via hybrid evolutionary algorithms for the prediction of Microcystis concentrations in the Vaal Dam, South Africa

Abstract

Cyanobacteria are responsible for many problems in drinking water treatment works (DWTW) because of their ability to produce cyanotoxins that potentially can have an adverse effect on consumer health. Therefore, the monitoring of cyanobacteria in source waters entering DWTW has become an essential part of drinking water treatment management. Managers of DWTW rely heavily on results from physical, chemical and biological water quality analyses, from grab samples, for their management decisions. However, results of water quality analyses may be delayed from 3 h to 14 days depending on a magnitude of factors such as sampling, distance and accessibility to laboratory, laboratory sample turn-around times, specific methods used in analyses, etc. Therefore, the benefit to managers and production chemists to be able to forecast future events of high cyanobacterial cell concentrations in the source water is evident. During this study, physical, chemical and biological water quality data from samples taken from 2000 to 2009 in the Vaal Dam, supplying South Africa’s largest bulk drinking water treatment facility, were used to develop models for the prediction of the cyanobacterium Microcystis sp. in the source water (real-time prediction together with 7, 14 and 21 days in advance). Water quality data from the Vaal Dam from 2010–2012 were used to test these models. The model showing the most promising results for incorporation into a ‘Cyanobacterial Incident Management Protocol’ is the one predicting Microcystis sp. 7 days in advance. This model showed a square correlation coefficient (R²) of 0.90 when tested with the testing dataset (chosen by bootstrapping from the 2000–2009 input dataset) and a R² of 0.53 when tested with the 3-year ‘unseen’ dataset from 2010–2012.