Datawatch: Fighting fires with big data analytics

Following record-breaking heatwaves in the UK, our latest Datawatch blog explores how firefighters are using analytics to predict and prevent fire outbreaks.

The UK has recently recorded its hottest days ever, with 33 locations breaking the previous record of 38.7 degrees, and one location in Lincolnshire even reaching 40.3 degrees.

In some parts of the UK, the heat forced rail services to close. And across London, fires quickly began to spread, destroying around 15-20 homes.

The most concerning part is that events like these are just the start of what’s to come. As global temperatures continue to rise as part of a worldwide climate crisis, countries like the UK will experience more and more high-risk fire outbreaks. 

The problem is, unlike locations in the US such as California, Texas, and North Carolina that are familiar with fire outbreaks, the UK doesn’t have the appropriate infrastructure and investment in fire services to respond quickly and effectively enough. And this problem will only be amplified as the scale of these outbreaks continues to grow.

That means fire services will need to find new ways to use their resources to their full potential. And many are already discovering a solution in big data analytics.

Using data in the UK to fight fires


With limited resources, firefighters need the knowledge and foresight to know where they should be and where they should focus their efforts — ideally before a fire even starts. 

The Met Office’s Fire Severity Index has been keeping the UK’s fire services informed for years. The index identifies how severe a fire could become if one were to start, using data on wind speed, temperature, time of year, and rainfall to produce an accurate assessment. 

It’s an approach based on Canada’s Fire Danger Rating System, and it’s an extremely valuable resource for helping fire departments prepare for the worst-case scenarios in specific locations. 

In recent years, some fire departments in the UK have extended their use of analytics even further. London Fire Brigade has been working with predictive analytics company Geoplace to build an address catalogue with complex details of every building in London, which can be used to help predict the spread and impact of fire outbreaks in the city.

Geoplace uses unique property reference numbers (UPRNs) to gather diverse data sets from a wide range of sources, including ordinance survey data, LIDAR data, ONS data, energy performance certificates, and buildings’ own fire histories. Paired with machine learning models, this data helps teams like the London Fire Brigade predict where fires could break in the city, see where they’ll spread to, and assess where they need to deploy their teams. 

As fire outbreaks become more commonplace in the UK, methods like these will become crucial to keeping citizens safe. However, these approaches are just the beginning. Fire departments in the US have been using analytics to fight fires for years — and their innovative approaches have already helped save thousands of lives. 

New ways to fight fires in the US

Many fire services in the US have much more experience tackling fast-growing fires than we do in the UK. And over time, they’ve come up with new tools to manage and mitigate them more effectively.

Los Angeles, for example, has experienced multiple wildfires over the past few years due to the high air pressure that builds over the western US. But its fire department is quickly adapting. 

When The Thomas Fire struck in December 2017, it spread over nearly 282,000 acres, destroying more than 1,000 homes. The Los Angeles Fire Department was able to respond rapidly, and monitor the fire’s growth as it spread, using an innovative tool named WIFIRE. 

WIFIRE is a web-based platform that merges satellite imagery and real-time data from cameras and sensors to create an accurate image of the fire, the conditions surrounding it, and its trajectory. It takes into consideration crucial elements such as wind conditions, compares them with historical fires, and uses AI modelling to accurately predict what the fire is going to do next. 

In the case of The Thomas Fire, the Los Angeles Fire Department used WIFIRE to share essential data across its teams on which roads and topography would be affected. In some cases, this data was even shared with the public to help citizens spot fires before they reached their homes. 

Insights like these have already helped save hundreds of lives and homes throughout the US — and some research bodies are taking firefighting innovation even further.  

Granular insights for precise firefighting

One project at Stanford University is taking firefighting intelligence to a new level, looking at the granular data points available in leaves to gather insights about a fire’s potential growth. 

Earth System Science PHD student Krishna Rao has created a deep learning algorithm that estimates the ‘fuel moisture’ — a value that describes the moisture content in living vegetation — in individual leaves. The fuel moisture value shows how much ‘fuel’ is available to burn, and how much would be consumed during a wildfire. 

Using values like these with the right AI tools, Krishna believes firefighters will be able to accurately predict when fires will start, how big they’ll get, where they might spread, and how they can best keep them under control. It’s an approach that’s still in development, but it’s been tested in 12 states prone to wildfire, and on more than 100 species of trees so far. 

If methods like these are combined with other data gathering techniques currently in use today — such as satellite and drone imagery — firefighters will have a wealth of intelligence on hand to fight fires more efficiently than ever before. And for places like the UK, where resources are limited, analytics projects like these could be lifesaving. 

For more fascinating insights into the role analytics plays in managing real world problems, read more of our Datawatch and Inside Analytics blogs here.