A farmer’s daughter who studied peanuts in India for her PhD is heading up a new division for the CSIRO, and she’s hoping data-driven research will help to facilitate real world decision-making for industry and the government.
Bronwyn Harch, who will lead CSIRO’s Computational Informatics (CCI) division, joined the organisation in 1995 as a statistician.
She originally studied environmental science at Griffith University in Queensland and later completed a PhD in plant breeding, working with the Queensland government and Indian government examining peanut food security.
“In India and a lot of other countries in the Asian continent peanuts are a staple food and they were concerned about … the diversity of peanuts and [that] they weren’t capturing plants that they could keep for prosperity for breeding,” Harch told CIO.
As a result of working with data around peanut plant bleeding, she said eventually programs were changed based on the information she was collecting.
Harch is hoping to have a similar impact on decision-making in Australia through CSIRO’s new division.
“Often people are only optimising one bit within a value chain – they’re just doing [sensing] or they’re just doing analytics. It’s not a feedback kind of loop,” she said.
Eventually, she will know the division has been successful when collaborators with the division say “‘if it wasn’t for this division we wouldn’t have been able to invest in this as a company – we would not have been able to think about this information in the way that we do now’.”
Harch’s interest in data analytics came about when she was growing up on a horticultural farm in the Lockyer Valley in Queensland and watching her father make decisions about the farm. While her father spoke the “language” of data, he wasn’t collecting any of it.
But it was only when she carried out her PhD at the University of Queensland in statistical science that her fascination with data was really piqued.
“I found that the important thing was you can make claims about different things happening, but if you didn’t have the evidence for it, you didn’t have a [leg] to stand on,” Harch said.
While it might appear a big jump from studying peanuts to heading a research division with 440 staff, Harch said there has been a common thread.
“The common thing is making sure when we collect information, it’s robust and that people can make decisions [about] it,” she said. “We can do a lot of research, but if it’s not being used by someone, that’s not very good.”
Transferring research into real world applications will be a particular focus at CCI. The division was created due to the data deluge which is now hitting different industries and sectors, including government.
The aim of CCI, according to Harch, will be to translate different data sets into useful information that is relayed at the right time and to the right people.
It will focus on four key areas – autonomous robotics, complex systems modelling and design, decision-making under certainty, and big data analytics.
Research around autonomous robotics could include helping users navigate complex environments, such as caves or inside buildings where it can be hard to get spatial signals. Meanwhile, research around modelling systems could help manufacturing companies to speed up the design process by allowing them to carry out models of their design instead of having to go through lengthy trials.
Other focus points will include researching how humans interact with computing equipment and technology and how to turn huge amount of datasets into useful information.
Harch conceded that sometimes the amount of data that is available can be overwhelming, but said scientists live for having large datasets “coming at them”.
She said one of the keys of success to processing large amounts of datasets into something worthwhile is to make sure the context of the analysis is well understood. For example, research might be carried out around vegetation in an area.
“If trees only grow on an annual basis, there’s not much point in getting minute-by-minute information. You need to understand the context in space and time to understand whether data is relevant or not to the question,” she said.
“The challenge is you can have a lot of data but not much information.”