Oral Presentation Society of Environmental Toxicology and Chemistry Australasia 2023

Hyperspectral analysis “Water Fingerprinting” improving the in-situ measurements of water quality that discharge to the Great Barrier Reef lagoon (#143)

Cheuk Yat Ng 1 , Michael Warne 1 2 , Ryan Turner 1 2 , Alan Huang 1 , Daniel Livsey 3
  1. Reef Catchment Science Partnership, School of Earth and Environmental Sciences, University of Queensland, Brisbane, Queensland, Australia
  2. Water Quality and Investigations, Queensland Department of Environement and Science, Brisbane, Queensland, Australia
  3. Queensland University of Technology, Brisbane, Queensland

Both nitrogen and phosphorous are essential nutrients for species in the aquatic environment, and sometimes can be found at elevated concentrations due to different land management practices. These elevated concentrations can lead to eutrophication and potentially affect species abundance (e.g., increased crown of thorns starfish), furthermore, pesticides at elevated concentrations can also be deleterious to the environment. Therefore, it is important to monitor waterways to understand processes occurring in catchments (e.g., land management practices) to derive data and knowledge to help decrease nutrient and pesticide concentrations. Currently, traditional monitoring techniques typically take at least three months between sampling and reporting to stakeholders. A potential solution to this would be using real-time monitoring probes, allowing stakeholders to have quick access to data and information to help adjust their land management practices. TriOS OPUS probes – a high-resolution real-time hyperspectral (200 nm to 360 nm resolution of 0.8 nm) probe – are currently being used in water monitoring stations to detect nitrate (NO3 - N) and total suspended solids. We will determine if the TriOS hyperspectral data can be used to detect and quantify phosphates and pesticides in waterways discharging to the Great Barrier Reef. The data used in this study includes three years’ (2019 – 2022) worth of hyperspectral data with paired (date and time) discreet laboratory-based measurements from Sandy Creek and the Johnstone and Pioneer rivers. This matched data will be analysed by chemometrics and machine learning techniques, specifically Classification and Regression Trees will be used to identify wavelengths that correspond to phosphate and pesticides. The spectral data at these wavelengths will then be used to predict pesticide and phosphate concentrations at sites with TriOS OPUS probes at 60-minute intervals. If successful this approach make phosphate and pesticide monitoring more affordable and provide far greater temporal resolution of their concentrations.