Forecasting cyanobacterial concentrations using B-spline networks

Artificial neural networks have been used successfully in a number of areas of civil engineering, including hydrology and water resources engineering. In the vast majority of cases, multilayer perceptrons that are trained wit h the back-propagation algorithm are used. One of the major shortcomings of this approach is that it is difficult to elicit the knowledge about the in put/output mapping that is stored in the trained networks. One way to overc ome this problem is to use B-spline associative memory networks (AMNs), bec ause their connection weights may be interpreted as a set of fuzzy membersh ip functions and hence the relationship between the model inputs and output s may be written as a set of fuzzy rules. In this paper, multilayer percept rons and AMN models are compared, and their main advantages and disadvantag es are discussed. The performance of both model types is compared in terms of prediction accuracy and model transparency for a particular water qualit y case study, the forecasting (4 weeks in advance) of concentrations of the cyanobacterium Anabaena spp. in the River Murray at Morgan, South Australi a. The forecasts obtained using both model types are good. Neither model cl early outperforms the other, although the forecasts obtained when the B-spl ine AMN model is used may be considered slightly better overall. In additio n, the B-spline AMN model provides more explicit information about the rela tionship between the model inputs and outputs. The fuzzy rules extracted fr om the B-spline AMN model indicate that incidences of Anabaena spp, are lik ely to occur after the passing of a flood hydrograph and when water tempera tures are high.