A comprehensive methodology for assessing river ecological health based on subject matter knowledge and an artificial neural network

• Published in: Ecological informatics Vol. 77; p. 102199
• Main Authors: Liu, Chao, Pang, Zonglin, Ni, Guoqing, Mu, Ruolan, Shen, Xiang, Gao, Weijun, Miao, Sheng
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.11.2023
• Subjects: Artificial neural network; case studies; China; environmental health; expert opinion; hydrology; Knowledge elicitation; neural networks; regression analysis; River health assessment; rivers; Text mining; China; Text mining; MAE; ANN; Knowledge elicitation; CBOW; River health assessment; Artificial neural network; AHC; BP
• ISSN: 1574-9541
• DOI: 10.1016/j.ecoinf.2023.102199

Assessing river health poses a significant challenge due to its dependency on multiple hydrological, chemical, and biological factors. While several methods have been presented, a comprehensive and systematic assessment of a river's ecological health necessitates the integration of various factors. In this paper, we propose a novel methodology to establish river health assessing system that leverages expert knowledge and objective indices. With the employment of text mining techniques, we identify five key groups of evaluation indices related to river health. Subsequently, 17 indices for river evaluation are determined by combining current regional metrics and relevant findings from recent related studies. Considering the interactive evaluation and nonlinear relationship between data and expert knowledge, an Artificial Neural Network (ANN) is employed to develop a river health evaluation model. To validate the utility and effectiveness of the proposed method, we applied it with a case study in China. The proposed method yielded a Mean Absolute Error (MAE) of 4.78, while a multiple linear regression baseline model yielded an MAE of 8.653. This finding indicates that the assessment system can achieve significantly better comprehensiveness, robustness, accuracy, and applicability. •Present a more robust and accurate river health assessment method.•Combine objective data with the subjective expert knowledge in the system.•Utilize machine learning approaches in the proposed method.