Computer model sees algal problems before they bloom
PhD student Gavin Bowden
A new computer model, developed at Adelaide University, is showing a high rate of success in predicting blue-green algal blooms in the River Murray up to four weeks before they occur.
Dr Holger Maier, whose PhD work formed the basis of this study, is now supervising another PhD student - Mr Gavin Bowden - who is developing the model within the Department of Civil & Environmental Engineering. The modelling process mimics the human brain by using a neural network program that 'learns' the key factors contributing to an algal outbreak.
Outbreaks of toxic blue-green algae pose a major threat to the quality of water.
"If the model forecasts a bloom four weeks in advance, industry can get the appropriate treatment processes ready and tackle the problem straight away, saving time, money, and maintaining the quality of the water." said Bowden.
To test the model, Mr Bowden has data from water samples collected every week over the last 20 years at Morgan in the Riverland.
Early results are extremely promising. When the computer's forecasts are measured against actual events, the predictions come close to the mark. Although not 100% accurate, they clearly show the peaks and troughs in algal bloom development.
A range of environmental factors in the water is taken into account, including nutrients such as nitrogen and phosphorous, the flow and temperature of the water, turbidity and colour.
The computer model can also advance understanding of the factors involved in toxic algal outbreaks by revealing the principal factors causing algal blooms.
Bowden is now looking to adapt the computer model for other uses such as salinity.
"If you knew a few weeks in advance that a large volume of highly saline water was moving down the Murray, water could be pumped to the Adelaide reservoirs before or after that to avoid pumping the saline water." said Bowden
But the applications of the computer model are even more far reaching, according to Bowden, who believes the model could be used throughout the world to predict problems of water quality.
"All you need is the right data for a particular water supply and you can apply the same modelling processes" he said.
Bowden will deliver a paper on his findings at the Hydro 2000 conference in Perth in November this year, and at the 2nd International Conference for Applications of Machine Learning to Ecological Modelling in Adelaide in November/December.
