The investigation of high frequency water quality monitoring and prediction

Abstract

Water quality monitoring and prediction are essential to understanding the complex dynamics of water ecosystems, the impact of human infrastructure on them and to ensure good management of water resources for drinking, recreation and transport in future scenarios. With new technology developed in the areas of smart cities and wireless sensors networks, water quality monitoring and prediction have the potential of becoming more and more a data-rich science. Despite the advancement in technologies and high temporal resolution datasets for water quality in surface water becoming more widely available, there is still insufficient knowledge on how to determine the appropriate sampling frequency for water quality parameters and which frequency is needed to build effective prediction.
In particular, this work presents: 1) A methodology for the identification of the most suitable noise removal technique for high frequency water quality datasets able to retain a high signal information content while minimising the contaminating noise content. 2) A statistical approach based on frequency analysis for determining the appropriate sampling frequency for water quality parameters providing also a practical tool to understand how different sampling frequencies are representative of the water quality changes in the monitored water body. 3) The investigation of a machine learning based approach aimed at identifying the suitable modelling frequency needed for building water quality machine learning prediction models.
This thesis focuses on surface water quality and uses the data recorded in Bristol’s Floating Harbour for water temperature, dissolved oxygen (DO), fluorescent dissolved organic matter (fDOM), turbidity and conductivity collected with a five minute temporal resolution as part of the local UKRIC Urban Observatory activities.
This work is the first of its kind to use a high temporal resolution water quality dataset to explore benefits and limitations of collecting water quality data with a high frequency sampling timestep.

Date of Award	18 Jun 2024
Original language	English
Awarding Institution	University of Bristol
Supervisor	Claire Gronow (Supervisor), Dawei Han (Supervisor) & Theo Tryfonas (Supervisor)