While using predictive software to help law enforcement is a new trend, modeling and forecasting wildfires has been around for decades. This year’s forest fire season is off to a violent start, and we already see many of these analytical tools in play in New Mexico, which has its largest forest fire in history burn through over 390 square miles in the Gila National Forest since May 16 and is still only 37% contained. An article in NextGov notes the various software suites running on nearly 100 computers at the incident command post to model and predict the growth of the fire.
WindNinja, for example, is a computer program that computes spatially varying wind fields for combating wildfires. To do so, it uses elevation, wind speed and direction, and information on the vegetation in the area. Another tool being used to understand weather conditions for the New Mexico fire is FX-Net, which provides the full Advanced Weather Interactive System visualization and data analysis capability from the National Weather Service as a service over the Internet so that it can be used on a laptop rather than at a command station.
To model the fire itself, firefighters are using the BehavePlus fire modelling system and FARSIGHT. BehavePlus runs on a PC and uses a collection of models to produce tables and graphs. BehavePlus is an update on the Behave system from the 80s, and came out in 1999. BehavePlus has many modelling capabilities such as surface fire intensity and spread, crown scorch height, transition from surface to crown fire, and safety zone size. FARSIGHT is a fire behavior and growth simulator that bases predictions on spatial and temporal data about topography, fuels, and weather. FARSIGHT is widely by the U.S. Forest Service and National Park Service but is too complex for anyone without fire behavior training and experience to use.
While these software suites have served firefighters well, there is Big Data analytics potential still to be tapped with more modern software. Buxton, which typically helps businesses better understand consumers, recently published a use case where they applied analytics to Philadelphia households, creating a Fire Vulnerability Index for the Philadelphia Fire Department. As with most police departments, fire departments are being asked to improve service with a stagnant or declining budget, requiring a more efficient use of resources. And, again as with policing, one way to do this is by being preventative, but the PFD could not work with every household in the city. With Buxton’s help, they used individual lifestyles, life stages, behaviors, attitudes, and finances that correlated with fire incidents to choose high risk households. As a result, they narrowed the list of vulnerable households from 48,000 to 2,000.
Just as with predictive policing, predictive firefighting is still expanding and has tremendous promise. For example, a solution like the one used in Philadelphia could be applied on a larger scale to forests, incorporating even more warning factors to narrow down the likely sites of potential forest fires and stop them before they spread through dozens or hundreds of miles. And, just as with business intelligence, firefighting analytics will need to be easy to use beyond specially trained analysts as well as fast and portable enough to be used in the field.
