The role of machine learning and the Fire Weather Index for wildfire danger prediction

Wildfires are increasingly causing significant economic and environmental damage, becoming more frequent and intense due to climate change. This article explores how our team of climate intelligence experts are using state-of-the-art software to produce wildfire hazard and risk scoring data products. These innovations are helping businesses across various sectors, communities and governments to better understand wildfire danger predictions. 

What are wildfires and how is the risk measured?

Wildfires are unforeseen fires that burn natural areas such as forests, grasslands or prairies.  The Fire Weather Index (FWI) is a globally recognised meteorological index used to estimate fire danger.  It comprises different components that consider the effects of fuel moisture and wind on fire behaviour and spread. The higher the FWI, the more favourable the meteorological conditions are for triggering a wildfire.

Our global present-day and future climate change FWI data is available now for multiple climate scenarios and time points. The present-day data has been built using high-resolution climate reanalysis data, while the future climate scenario data has been built using high-resolution downscaled and bias-corrected data from CMIP6 Global Climate Models.

How are insurers and mortgage lenders using wildfire hazard and risk scoring data?

Insurers and mortgage lenders utilise wildfire hazard and risk scoring data to assess and manage the risk associated with properties in wildfire-prone areas. This data is critical for:

• Underwriting and pricing
• Risk assessment
• Regulatory compliance
• Claims management
• Mitigation incentives

Benefits of data and technology for communities, governments, insurers, and lenders

By leveraging wildfire hazard and risk scoring data, communities and governments can enhance resilience, protect lives and properties, and ensure sustainable development in wildfire-prone areas. Insurers and lenders can better assess and mitigate risks, leading to more informed decision-making.

Limitations of the Fire Weather Index

While the FWI is a powerful tool, it does have limitations. It does not account for factors such as vegetation, topography, and human activity, which are significant to wildfires. Additionally, most wildfires globally are caused by arson, which can be challenging to predict.

How is Royal HaskoningDHV using machine learning to improve wildfire predictions?

At Royal HaskoningDHV, we are leveraging machine learning (ML) and remote sensing to enhance wildfire danger predictions globally. Machine learning automatically learns relationships between variables influencing wildfire danger and improving prediction accuracy. Our methodology, validated on a European scale, predicted >90% of wildfire danger accurately compared to burned area observations. This methodology is also shown to be more accurate than using the Fire Weather Index alone when predicting fire danger.

We are currently developing machine learning-driven Combined Wildfire Hazard and Risk data that incorporates additional variables, such as topography, population and land use/cover.

Ultimately, our Wildfire insights will be available for 8 combinations, as shown in the table:

Product	Hazard	Risk
Fire Weather Index (FWI)	Present Day	Present Day
	Climate Change	Climate Change
Combined Wildfire	Present Day	Present Day
	Climate Change	Climate Change

Wildfire is just one of the many hazards in our suite of data products and software. We currently offer data on over 19 hazards, with plans to include more. We provide risk scores for various emissions scenarios and future time periods. Organisations worldwide, from financial services and industries to utilities and critical infrastructure, rely on our climate intelligence data to manage risks and boost resilience. Our data is easy to use through our app or by integrating it into your systems via API.

To learn more about how our data can assist you in becoming more climate resilient, please contact us.