Estimation of fish assessment index based on ensemble artificial neural network for aquatic ecosystem in South Korea. (March 2022)

• Record Type: Journal Article
• Authors: Kang, Hyeongsik; Jin Jeon, Dong; Kim, Soohong; Jung, Kichul
• Is Part Of: Ecological indicators. Volume 136(2022)
• Publication Date: 2022-03
• Subjects: Fish Assessment Index -- Ecological Health -- Ensemble Artificial Neural Network -- Support Vector Machine -- South Korea; Environmental monitoring -- Periodicals; Environmental management -- Periodicals; Environmental impact analysis -- Periodicals; Environmental risk assessment -- Periodicals; Sustainable development -- Periodicals; 333.71405
• DOI: 10.1016/j.ecolind.2022.108708
• Languages: English
• ISSNs: 1470-160X

Abstract

Highlights: Fish assessment index estimation model was developed for river health assessment. Input parameters for the model were selected based on correlation analyses. 5-fold and 10-fold cross validation techniques were used for model validation. Ensemble artificial neural network model performed better than support vector model. Abstract: Fish assessment index (FAI) is commonly used to perform ecological health assessments of fish communities in rivers. However, the estimation of this index is difficult, considering the complexity of relationships among water quality, water quantity, and hydrologic properties. In this study, we propose an ensemble artificial neural network (EANN) based on various hydrological and environmental variables, and their correlation, to improve the FAI estimation for the ecological health assessment of streams and water resource management. The optimal input variables of the model were determined through correlation analyses. These values were used to train the ANN model using the Levenberg-Marquardt algorithm. The EANN model was structured with varying number of neurons and ensemble sizes. This model provides better generalization and model performance than the single ANN model for estimation. The proposed model was applied to 143 monitoring sites across Han, Nakdong, Geum, Yeongsan, and Seomjin rivers in South Korea. The 5- and 10-fold cross-validation approaches were used to evaluate the results of the model based on the Nash criterion, relative root mean squared error, and relative mean bias. For performance comparison, support vector machine (SVM) was used as the other feasible alternative because this method shows good performance in regression analysis and solves problems mathematically, unlike other machine learning algorithms. The results indicate that EANN is suitable for solving large complex problems and provides better FAI estimates than the SVM, which can be used to manage ecological conditions.