Integrated physical system and AI/ML modelling of water quality and quantity

Project leaders: Klaus Joehnk and Peter Toscas

Project Description: Development of a novel, hybrid AI/ML early warning and forecasting system using causal relationships for modelling water quality and quantity.  Decision and policy makers can utilise the tool for short-, seasonal and long-term decision planning for better water security outcomes.

Project Goals: To improving Australia’s water security, particularly under changing climatic conditions that could result in more frequent and more severe natural events such as: drought, flood, and fire.

We are developing methodology, tools and services that allow for:

  • Online access to curated historical water quality data, remote sensing, land use, and catchment process information
  • Faster and smarter ways to model physical systems underpinning water security that capture causal relationships
  • Forecasting tools for short-term, seasonal, and long-term forecasts under changing climatic conditions
  • Short-term, seasonal and long-term decision-making tools
  • Web accessible (dashboard and web services), set of specific simplified pretrained hydrologic and hydrodynamic models and architectures (ML, hybrid) for use-cases.

Project outputs will include:

  • Data access dashboard for curated datasets
  • Re-useable Python code for pre-trained hydrologic and hydrodynamic models and architectures
  • Trained hydrologic and hydrodynamic models for water quality parameters
  • Tool for AI/ML modelling of physical systems underlining water quality and quantity
  • Methodology and software for forecasting water quality in river systems across selected river reaches of Australia
  • Stakeholder workshop to co-design tools and use-cases
  • Uses-cases for selected river reaches and standing water bodies:
    • To predict persistent temperature stratification and low dissolved oxygen events
    • To predict algal bloom outbreaks
    • To predict bushfire generated sediment loads.
  • Technical architecture for hybrid modelling of water quality across scales

For more information, please contact either Klaus Joehnk (t: +61 2 6246 5636, e: Klaus.Joehnk@csiro.au) or Peter Toscas (t: +61 3 9545 8054, e: Peter.Toscas@csiro.au).

Fish kill in the Darling River at Menindee 2023. Copyright: Graeme McCrabb