Prediction of riverine suspended sediment discharge using fuzzy logic algorithms, and some implications for estuarine settings

The ability of fuzzy logic algorithms to model relationships between stream flow and suspended sediment discharge was investigated using daily measurements of stream flow and suspended sediment discharge for the Escanaba River mouth station, situated on the shore of Lake Michigan and operated by the US Geological Survey. Three different configurations of inputs were applied, whereby the inputs were fuzzified into fuzzy subsets of variables by means of triangular membership functions. The relationships between inputs and suspended sediment discharge (output) were represented by a set of fuzzy rule expressed in IF–THEN format. The weighted average method served for defuzzification. The commonly used sediment rating curve was also applied to the data, and its performance compared with that of the three models by means of statistical analyses. For all three models, suspended sediment discharge predicted by the fuzzy logic algorithm was in satisfactory agreement with observations. Furthermore, the fuzzy logic algorithms performed better than the sediment rating curve, particularly at higher rates of suspended sediment discharge (in this study, more than 50  × 106 g/day). Considered collectively, the use of fuzzy logic algorithms is suggested as a simple and effective approach for better prediction of suspended sediment discharge, also for estuaries.