Predicting the Burn: Data61’s Global Bushfire expertise

By May 25th, 2018

Bushfire is a profound, serious and frequent concern for much of the Australian population, with the total annual cost of natural disasters in Australia expected to increase from $9 billion to $33 billion by 2050. The CSIRO has a long history across a range of areas of researching the many facets of this threat. In the 21st century, new digital tools can catalyse efforts to understand bushfire, and position the CSIRO’s Data61 unit as a world-leader in modelling and predicting the spread of bushfire.

Lighting the Spark

Spark is a generic platform. It lets people simulate bushfires, fast”. James Hilton, Senior Research Scientist at Data61, works on a piece of software developed by CSIRO’s Data61 that draws in a range of data, including weather, geography and environmental information and uses state-of-the-art simulation science to predict and visualise the spread of a bushfire, and spoke to us about his research.

“Spark can be used by risk planners and decision makers when planning infrastructure, it can be used by people working in operations to inform real-time resource deployment, or it can be used by researchers to develop new fire models”, says James. “The fact we’ve developed a common platform means there’s a narrower gap between research and operational practice, with both areas feeding into one another”.

The technology, which is currently under discussion for use as an operational tool and is used by a range of planners in emergency services across the country, provides more than predictions of bushfire spread – it creates a closer connection between research, product and deployment.

Spark is also being used to model the impact of electrical transmission lines on the probability of bushfire spread – this has been a key issue in Australia since the Black Saturday [pdf · 849kb] bushfires of 2009 which has been predicted to cost $4 billion in losses and claimed 173 lives. “For the team’s power lines project, we take every powerline in Victoria, assume a fire starts at every single power pole, and work out the worst-case footprint of that fire using our supercomputer resources”, James tells us. “We can do a couple of million simulations, composite them, and display them on a map”.

Spark technology has also been used by the NRMA Fireblanket technology , a prototype wireless sensor network that smells bushfire, and tracks and predicts where the fire could go.

During the trial of the technology, multiple sensor arrays were spread throughout the burn location in different configurations (close, medium distance and widespread) as determined by the CSIRO Spark Modelling/Scenario generation. This helped to detect and map bushfires by acquiring the data about fire behaviour. Data61 also provided advice on fire behaviour and delivered an options analysis of the features and technology requirements needed in comparison with existing products available in the market.

Our global fire expertise

The 2017 California wildfire season was the most destructive on record, with five of the twenty most destructive fires in the state’s history resulting in many deaths and significant loss of property estimated to cost individuals and businesses $9.4 billion. The fires continued through to winter, with significant fires impacting the state in December 2017.

A maze animation that fills with colour according to a fire prediction algorithm

Solving a maze using Spark (maze credit:  © Spark Research Team

“Spark’s flexibility means it can be used anywhere in the world, and we’ve already had interest from other countries in adopting this tool for managing bushfire”, says Dr Simon Dunstall, Research Director in the Decision Sciences Porogram . “We’ve spoken with organisations responsible for fire management in California and Chile who are very interested in our world-leading research and in particular, the difference that Spark has made through the insights and analysis it delivers and in November we’ll be attending the world’s largest wildfire conference – the VIII International Conference on Forest Fire Research

In Chile, Data61 has worked with Arauco , one of the largest suppliers of forest and related products in the Southern Hemisphere. We provided them with integrated fire and data science tools (in collaboration with CSIRO’s Land and Water business unit) that they are using to understand and effectively respond to wildfire threats. This work was commissioned by Arauco after a quarter of their forest holdings were destroyed by a devastating bushfire in January 2017. After the success of this work, Arauco is seeking to engage with Data61 to develop a cloud based implementation of Spark to calculate their fire danger risk on an operational basis.

Spark serves as an excellent example of the leadership opportunities that emerge through Data61’s work in digital innovation and the impact that data science can have in terms of reducing damage to property and ultimately saving lives.