Abstract
It is important to evaluate the inflowing pollutant loads for the conservation of water quality of closed water bodies. In this study, system correlations between pollutant concentrations and water characteristics of river flow (such as signals of optical sensors, water temperature, etc.) were generated by ANN models. Continuous time series of COD, T-N and T-P were estimated well by the trained ANN models with the continuous measurements of water characteristics. The estimation accuracy of pollutant loads during single storm events were improved by employing the estimated time series of water qualities. Four rivers from two different watersheds (Lake Ogarawa and Lake Kasumigaura) were studied. The results of regression analysis showed that high correlations existed between specific accumulative loads and specific accumulative river flow rate. The regression coefficients were in the same order for the four rivers and most of them had high correlations with the area ratio of paddy field.
