29 June 2010 Academic keynote speaker at Paris UNESCO meeting

Solving problems by computer in a way that mimics the working of the human brain has given scientists an improved method for the management of rivers where flow rates and other data is not recorded by gauges.

“Accurate river flow data is essential for effective water resource management,” says Associate Professor Sandhya Samarasinghe of Lincoln University's Department of Environmental Management.

“However, when you have rivers without gauges in place to provide flow and other data, and this relates particularly to rivers in inaccessible areas, then water management becomes a challenging problem.

“One standard solution has been to make predictive calculations based on a concept known as 'Hydrological Regionalisation' - essentially exploiting the geomorphological and hydrological similarities of nearby catchments in a region as the basis for estimating unknown data.
 
“What we have done at Lincoln University to solve the problem of estimating water flows in ungauged rivers is to apply an artificial neural networks approach to Hydrological Regionalisation. In particular, we have developed and applied an improved neural network that far exceeds the efficiency of currently available neural networks and other related approaches.”

Lincoln University, New Zealand's specialist land-based university,  has been employing the new science of “neural networks" for a number of years, with its Centre for Advanced Computational Solutions (C-fACS) playing a prominent role. Associate Professor Samarasinghe has used its theoretical foundations as the basis for innovative research in a range of industrial and biological applications. The river management work is an example of these applications.

“Today at Lincoln University neural networks are a serious and powerful research method with theoretical advancements and high-level practical applications,” she says.

In simple terms “neural networks” are computational methods for solving various complex problems modelled on the information processing systems in the human brain where the “neuron” or “nerve cell” is the basic computational element.

By putting many of these elements together, a rich array of neural networks can be developed on a computer to address various types of prediction, classification and forecasting problems. 

In Computer Science, the term “neural networks” refers to a whole lot of artificial neurons all connected in a manner emulating the information processing system in the brain.

While emulating information processing in the brain, neural computing has strong links with statistics and Associate Professor Samarasinghe has been trying to merge statistics and neural networks to establish a platform with other methods such as “soft computing”. The aim has been to enhance the capability of neural networks in modelling non-linear systems with many variables that interact with one another. Soft Computing methods are a set of computational methods that encapsulate human reasoning embedded in linguistic expressions to solve problems that are harder to solve using traditional mathematics.

Associate Professor Samarasinghe's work on the application of neural networks to ungauged river situations has led to an invitation to be a keynote speaker at a prestigious academic colloquium at UNESCO's headquarters in Paris early in July. The 10th Kovacs Colloquium, hosted by UNESCO and the International Association of Hydrological Sciences has the theme “Hydrocomplexity: novel approaches to solving 'wicked' water resource problems”. 

At the colloquium, to an audience invited by UNESCO's Sustainable Water Resources Development and Management Section, Associate Professor Samarasinghe will explain how her novel neural networks approach provides improved predictions for ungauged river flows, using a seven-basin case study of the Canterbury region. Her paper is entitled Advanced neural networks for water systems analysis.

“Interest in the neural networks research we have been doing at Lincoln University is strong and encouraging,” she says.

Professor Samarasinghe's work, presented at international conferences and in publications, has been followed over the years by the Chief of UNESCO's Sustainable Water Resources Development and Management Section, Professor Shahbaz Khan, who invited her to speak at the colloquium for the novel contribution her work makes to the field. She has also been invited to attend an international panel meeting on global teaching and research needs in water resources in Paris the day before the colloquium.

“I am confident that the advanced neural networks approach we have developed at Lincoln University offers new directions for addressing complex issues of water resources management. Also, the new neural networks paradigm that we are developing based on neural network and soft computing will, I believe, lift water resources management to a new level of efficacy and sustainability.”

UNESCO is following Associate Professor Samarasinghe's work with interest for the on-going contribution it can make to resolving some of the world's critical water management issues.