Preventing the Burn – Predicting to Protect Against Wildfires
The sight of New Yorkers donning masks this summer, with its resonance of the pandemic, signalled a crisis.
On 7 June 2023, air pollution in New York was the worst of any city in the world, due to 400 wildfires blazing across Canada that had already burnt 8.15m acres of land, and displaced 120,000 people.
Unfortunately this is a situation that is only going to get worse. The United Nations Environment Programme (UNEP) predicts that extreme wildfires will increase by 30% by 2030, and that wildfires will become more frequent and intense.
Addressing this, and mitigating the associated risks, requires a recalibration of approach. Traditionally the focus, and critically the funding – a good indicator of priorities – has been on the response to wildfires.
A good example of this can be found in Canada. In 2021 British Columbia experienced a particularly hot summer, during which they British Columbia spent $600 million USD on fighting forest fires. The budget for wildfire prevention in 2023 is $24 million!
The UNEP are rightly trying to promote the prevention agenda primarily through their ‘Fire Ready Formula’, which calls on governments to rebalance funding so that two-thirds are devoted to prevention and recovery, leaving one third for response.
Any move to a more preventative approach will only work if we strengthen our predictive capabilities to anticipate and prevent wildfires before they escalate. It is this that underpins any decisions around prevention activity, the deployment of firefighting resources, and firefighting tactics to limit the spread of wildfires.
The concept of the Early Warning System is well established in Disaster Risk Reduction. The Sendai Framework includes an ambition for everyone to be protected by a Multi Hazard Early Warning System by 2030. These systems typically include devastating climatic events such as Tsunamis, Storms and Floods.
The last update in 2022 reported that only half of the countries across the world were protected and not many have included wildfires in the list of anticipated events. To address this gap, it is critical that governments invest in the development of predictive capabilities, to include encompassing wildfire events in their Early Warning Systems.
There is reason for optimism about this being achievable. We have a much better understanding of the risk factors for wildfires than ever before, and applying Artificial Intelligence techniques to data around weather including humidity, vegetation, population and infrastructure brings the potential to deliver high quality predictions.
There are a number of promising initiatives in this space. The World Economic Forum’s Climate Technology Team is working with the Turkish Ministry of Forestry on developing a dynamic wildfire risk map; and academics at Stanford University recently developed a wildfire prediction system based on a machine learning framework which achieved a 86% success rate in predicting wildfires.
However, challenges remain in integrating AI for wildfire predictions. Securing real-time, high quality data from a mix of open source and proprietary datasets is an essential prerequisite. As is the need to ensure comparability across data classifications and complex algorithms that attach ‘weights’ to individual risk factors to provide an accurate prediction.
This requires government action. This should include convening partnerships with the NGO and Private sectors, make public sphere data available in usable formats, and fund the technological innovations required to make this work.
There is a cost to doing nothing. Simply responding to wildfires does little to negate the environmental and societal impacts. We must reimagine our approach and use data and technology to mobilise prevention with the same urgency with which we despatch the fire truck when we receive the emergency call.
