Abstract
Chlorophyll-a is a major water quality indicator for an algal bloom in streams and lakes. The purpose of the study is to estimate chlorophyll-a concentration in tributaries of the Seonakdonggang by an artificial neural network (ANN). As the tributaries are ungauged streams, a watershed runoff and quality model was used to simulate water quality parameters. The tributary watersheds include urban area and thus Storm Water Management Model (SWMM) was used to simulate TN, TP, BOD, COD, and SS. SWMM, however, can not simulate chlorophyll-a. The chlorophyll-a series data from the tributaries were estimated by the ANN and the simulation results of water quality parameters using SWMM. An assumption used is as follows: the relation between water quality parameters and chlorophyll-a in the tributaries of the Seonakdonggang would be similar to that in the mainstream of the Seonakdonggang. On the assumption, the measurement data of water quality and chlorophyll-a in the mainstream of the Seonakdonggang were used as the learning data of the ANN. Through the sensitivity analysis, the learning data combination of water quality parameters was determined. Finally, chlorophyll-a series were estimated for tributaries of the Seonakdonggang by the ANN and TN, TP, BOD, COD, and temperature data from those streams. The relative errors between the estimated and measured chlorophyll-a were approximately 40 ~ 50%. Though the errors are somewhat large, the estimation process for chlorophyll-a may be useful in ungauged streams.