Water Quality Assessment using Machine Learning Algorithms from Remote Sensed Data under Non-Stationary Climate Conditions

Accurate estimation of water quality is critical to assess the adverse effects of environmental pollutants on the natural processes of the river systems with changing climate. While there is a growing interest in the community for Machine Learning (ML) models, in many ways, there is still a non-evidence based preference for physical process based models as there is a lack of interpretability and expansibility in terms of applying ML models in real-world application. In this regard, this study aims to develop a ML based methodology that thoroughly addresses the underlying physical science for water quality modeling using Long Short Term Memory (LSTM) approach to predict the changes in the water quality drivers under non-stationary climate conditions. For pilot demonstrating purposes, the methodology will be applied to the Ohio River. The results will show confidence in the application of ML based approaches as well as improvements in understanding of the associated climate uncertainty for assessing the water quality in general.