Integrating principal component analysis, fuzzy inference systems, and advanced neural networks for enhanced estuarine water quality assessment

Abstract

This study focuses on the estuarine region of Ilaje in the Niger Delta, Nigeria. The research develops a comprehensive framework for assessing estuarine water quality by integrating Principal Component Analysis (PCA), Fuzzy Inference Systems (FIS), and advanced neural network models, specifically Long Short-Term Memory (LSTM) and a hybrid LSTM-Convolutional Neural Network (CNN). The study employs SHAP (SHapley Additive exPlanations) analysis to interpret the contributions of individual water quality parameters to model predictions, addressing the challenge of handling large and complex datasets from water quality monitoring programs and aiming to provide robust predictions and insights into water quality dynamics. The hybrid LSTM-CNN model demonstrated superior predictive performance, achieving RMSE values lower than 10 % and R² values exceeding 0.90 across various predictive tasks, indicating high accuracy in forecasting water quality parameters. This capability is crucial for the Ilaje region, which is experiencing rapid industrialization and urban expansion. The predictive insights gained can significantly aid in water management and pollution control, helping to address the dearth of such frameworks in the area. This study highlights the importance of integrating advanced neural network architectures in environmental monitoring, offering a reliable tool for managing estuarine water quality under the pressures of development and environmental change. • Developed PCA, FIS, LSTM, and LSTM-CNN framework for water quality assessment. • Hybrid LSTM-CNN showed superior predictive accuracy over standalone LSTM. • HCA and ANOVA revealed site-specific water quality patterns for interventions. • SHAP analysis highlighted significant features influencing model predictions. • WQI and FWQI effectively captured the complexity of estuarine water quality.

Topics & Concepts