A cumulative-risk assessment method based on an artificial neural network model for the water environment

To analyze the cumulative risks to the water environment, the backpropagation artificial neural network (BP-ANN), a self-adapting algorithm, was proposed in this study. A new comprehensive indicator of cumulative risks was formed by combining the water risk assessment tool proposed by the World Wide Fund for Nature or World Wildlife Fund (WWF), Deutsche Investitions und Entwicklungsgesellschaft mbH (DEG), and the cumulative environmental risk assessment system proposed by the US Environmental Protection Agency (USEPA). Eleven training algorithms were selected and optimized based on the mean square error (MSE) of prediction results. Data concerning evaluating indicators and cumulative risk indexes of the Liao River collected from 2005 to 2017 in the cities of Tieling, Shenyang, and Panjin, China, were used as input and output data to train, validate, and test the BP-ANN. Levenberg Marquardt backpropagation was the most accurate algorithm, with an MSE of 3.33 x 10(-6). After optimization, there were six hidden layers in the model. The correlation coefficient of the BP-ANN with LM exceeded 80%. These findings suggest that the BP-ANN model is applicable to prediction of cumulative risks to the water environment. The model was sensitive to the number of wastewater treatment facilities and the wastewater treatment rate along the river. Based on the sensitivity analysis, the contributing factors can be controlled to reduce the cumulative risk.

Automated summary

In this study, a self-adapting algorithm called BP-ANN was proposed to analyze cumulative risks to the water environment by combining tools from WWF, DEG, and USEPA. After optimization, the model had six hidden layers and showed a correlation coefficient with LM exceeding 80%. The findings suggest that the BP-ANN model is applicable for predicting cumulative risks and sensitive to factors such as the number of wastewater treatment facilities and treatment rate along the river.