NEWS
How can crowdsourcing help us respond better to disasters?
Natural disasters trigger bursts of crowdsourced data that often emerge faster than official statements, but sifting facts from the flood of social noise has been a challenge for authorities.
Now researchers have developed a reliable new way to track disasters by crowdsourcing data from people directly affected by an emergency.
The research team led by RMIT University automatically tracked active fire front locations in real time using data from calls to emergency services.
The research may be used in future to help authorities determine the extent and direction of an area affected by an emergency, to support the response to the disaster and help save lives and property.
Led by RMIT geographic information scientist Professor Matt Duckham, the team used data from Triple Zero (000) calls during Victoria’s Black Saturday bushfires to track the growth and movement of the blazes that claimed 173 lives.
Duckham, Associate Dean of Geospatial Science, said new public sources of emergency information made available since the 2009 bushfire disaster could add to the arsenal of data sources for authorities dealing with such threats.
“Timely and accurate information about natural disasters helps emergency responders and the public minimise the risks, but in such an emergency, real-time data may be in scarce supply,” he said.
“Authorities already rely on ground-based observations, high-resolution satellites and airborne infrared scanners, all of which play an important role in bushfire emergency planning and response.
“But in the case of bushfires, there are no authoritative information sources that can always and reliably generate up-to-date and accurate information about bushfire perimeters.
“The precise boundaries of a bushfire are highly dynamic and emergency management processes have to capture the progression as accurately and quickly as possible.”
Duckham and his team used data from Black Saturday and more recent bushfires to show how bushfire perimeters can be tracked in real time based on crowdsourced data about emergency incidents.
While less accurate than airborne infrared sensors, the new technique allows information to be continually and automatically updated in real time, day and night.
“Normally, user-generated social media data lacks the necessary structure and trustworthiness for reliable automated processing,” Duckham said.
“Our research centred around developing an automated technique for real-time tracking of bushfire perimeters based on publicly available ‘curated’ crowdsourced data, from telephone calls to the emergency services.”
The team used data from 000 calls during Victoria’s Black Saturday bushfires to track the growth and movement of the fires.
He said authoritative information sources may suffer from delays and bottlenecks, while predictive bushfire behaviour models are inevitably limited in accuracy by the quality of input data.
In contrast, an emergency event such as a natural disaster often triggers a burst of crowdsourced data more rapidly than authoritative, official information.
“That is why crowdsourcing holds such potential for real-time detection and monitoring of natural disasters including wildfires, floods and earthquakes.”
Duckham said telephone calls to emergency services were less noisy and more reliable than many other sources of crowdsourced data, such as social media.
“We analysed real-time data concerning the location and subject of emergency calls to identify spatial and temporal patterns of calls that could be used to estimate the progression of active bushfires.”
Duckham said the calls still required a trained human operator to estimate incident locations from calls which, in the case of fires, could be visible for many kilometres.
“But our research relied on constructing an evolving ‘footprint’ for the bushfire perimeter based on the incident locations and timing estimated from emergency calls and it can also integrate relevant and authoritative demographic and environmental information, such as population density and dynamic wind fields in order to improve estimation accuracy.
“Our analysis operates in real-time for near instantaneous perimeters based on the latest RSS feed data.”
Written by Duckham with University of Melbourne researchers Xu Zhong, Derek Chong and Kevin Tolhurst., the research paper, Real-time estimation of bushfire perimeters from curated crowdsourcing, was published in Scientific Reports.
The research was funded by the Australian Research Council, with funding from Emergency Management Victoria and IBM Research Australia as industry partners.
Story: Gosia Kaszubska
Video: Sarah Adams