Earth observation data-driven coastal water quality forecasting
This project aims to develop state-of-the-art coastal water quality forecasting model using machine learning methods, multi-sensor satellite earth observations, in-situ data streams, ecosystem modelling outputs and cloud computing infrastructure.
Water quality challenges in coastal ecosystems
Water quality degradation in coastal aquatic ecosystems is a global problem. In Australia, reduced water quality due to climate driven changes along with anthropogenic activity impacts coastal ecosystems (total annual asset value >$900 billion) and related marine industries (which contribute >$50 billon/year to Australian economy). Degradation of water quality changes the structure, composition and extent of aquatic habitats and has flow-on effects such as fish populations and species diversity in some cases driving species to extinction (Figure 1). Degrading marine environments eventually leads to disruption of marine ecosystem functioning, resulting in socio-economic loss to human populations that depend on aquatic resources. To support better decision-making for coastal aquatic ecosystem preservation and restoration, it is necessary to have knowledge of water quality degradation (as scenario modelling projections and forecasts). Forecasting capability achieved in this project will help the AquaWatch mission deliver solutions (hotspot detection, targeted monitoring actions and mitigation strategies) to support better ecosystem management and achieve sustainable development.
Limitations with current practices
Current in-situ sampling programs and ecosystem models have many limitations in offering observations and forecasts that are relevant for management and decision making. Monitoring programs involving seasonal in-situ sampling and fixed location measurements only offer limited spatial and temporal knowledge of water quality. Such methods are also limited in providing
information and knowledge about extreme events due to physical constraints and hazardous conditions.
Coastal ecosystem model-based forecasts are very expensive to establish, time consuming to develop and are available only in limited locations. Thus, due to the limitations in existing monitoring programs, and lack of region-specific land-to-ocean ecosystem models, it is difficult to offer coastal water quality forecasts (across Australia) with appropriate lead times that are best aligned with management and industry requirements.
New approach: Earth observation data-driven coastal water quality forecasting.
To significantly advance our capability to forecast coastal water quality, the proposed research aims to apply novel approaches to synergistically use: 1) large volumes of satellite observations of multiple biogeophysical parameters (describing variability in climate, land, and ocean); 2) in-situ sensor network; and 3) three-dimensional ecosystem model outputs in a hierarchical
spatiotemporal modelling framework (Figure 3). This project will generate new knowledge on the integration of multi-sensor earth observations, the biogeophysical connections between climate, land, and ocean systems, as well as new modelling approaches for rapid generation of coastal water quality forecasts. Such a research approach (integrating climate, land and ocean) has not been attempted in Australia and is in initial stages of development internationally.
Data-driven forecasting approaches have not been tested rigorously until now due to the limitations in the availability of observations and computing infrastructure. However, the current availability of large volumes of analysis-ready satellite-based observations, in-situ sensor networks and cloud-computing infrastructure (e.g. Earth Analytics Science and Innovation platform (EASI)) enables the development and implementation of data-driven forecasting models for complex coastal ecosystems. CSIRO also has access to some of the complex coastal ecosystem models, that can provide training datasets. Recent advances in machine learning models, and in particular physics-informed models, also provide strong confidence in the success of the earth observation (EO) data-driven coastal forecasting models proposed in this project.
Benefits to society, economy and environment
The Earth observation data-driven forecasting models developed and implemented in this project will revolutionise the coastal water quality monitoring practices and management approaches applied in Australia.
Impact and outcomes of this project will be seen in the following areas (Figure 4):
– Environment: Outputs delivered by this project will help regional ecosystem managers identify region-specific drivers and pressures. Such knowledge will help managers develop and implement suitable management plans and aid the recovery of coastal ecosystems.
– Economy: Australian blue economy is valued at $68 billion per annum and industry participants depend on the preservation and improvement of water quality. New forecasting and scenario modelling products from this research project will support the sustainable development of these industries (such as aquaculture and coastal tourism).
– Community and wellbeing: Traditional owners and the wider Australian population have long highlighted the benefits of healthy coastal ecosystems. The forecasts and modelling outputs from this project will help achieve healthy coastal ecosystems and thus directly help human wellbeing in these coastal communities.
In addition, the proposed research provides an opportunity to leverage and advance the application of cloud computing, data cube technology, satellite earth observation and spatiotemporal modelling to rapidly forecast coastal water quality changes. The outcomes of the project will be of great interest to the wider coastal ecosystem research community, scientists (ecologists and oceanographers), ecosystem managers and policy managers.
The proposed project is structured into four work packages (Figure 5). Each work package is designed to have its own deliverables and uptake pathway. The work packages in the project are:
- WP1: Climate, Land, Ocean and Underwater Datacube (CLOUD) development
- WP2: Continental-scale Ocean water quality forecasts
- WP3: Forecasting Land-to-Coast (L2C) discharges
- WP4: EO data-driven high-resolution coastal water quality forecasting model
Leads: Rob Woodcock (Minerals), Foivos Diakogiannis (D61), Nathan Drayson (Environment), Eric Lehmann (D61) and Nagur Cherukuru (AquaWatch Mission and Environment)
Experts: Simon Collings (D61), Yanan Fan (D61), Ash Shokri (Environment), Durga Lal Shrestha (Environment), Lachlan Phillips (Environment), Karen Wild-Allen (Environment), Kesav Unnithan (Environment), Yiqing Guo (D61)
Collaborators: AquaWatch Mission, Prof. Wei Xang (La Trobe University), Amazon web services, CSIRO EASI