Eutrophication forecasting and management by artificial neural network: a case study at Yuqiao Reservoir in North China

Abstract

Yuqiao Reservoir is the potable water supply source for a city with a population of more than 14 million. Eutrophication has threatened the reliability of drinking water supplies and, therefore, the forecasting systems for eutrophication and sound management become urgent needs. Water temperature and total phosphorus have long been considered as the major influencing factors to eutrophication. This study used the artificial neural network (ANN) model to forecast three water quality variables including water temperature, total phosphorus, and chlorophyll-a in Yuqiao Reservoir. Two weeks in advance for forecasting was chosen to ensure a sufficient preparation response time for algae outbreak. The Nash–Sutcliffe coefficient of efficiency (R2) was between 0.84 and 0.99 for the training and over-fitting test data sets, while it was between 0.59 and 0.99 for the validation data set. To better respond to the algae outbreak, a number of management scenarios formed by orthogonal experimental design were modeled to assess the responses of chlorophyll-a and an optimal management scenario was identified, which can reduce chlorophyll-a by 23.8%. This study demonstrates that ANN model is potentially useful for forecasting eutrophication up to 2 weeks in advance. It also provides valuable information for the sound management of nutrient loads to reservoirs.