In a landmark peer-reviewed validation study, Pyrologix predicts where most wildfires will occur with proven, actionable accuracy – areas that actually burned were predicted to be 3.5X more likely to do so. 

Through a newly published paper, Pyrologix has recently demonstrated the best predictive performance of burn probability modeling published to date. Across insurance, utilities, land management, emergency management, community planning, and beyond, managers assessing wildfire risk can rely on our well-calibrated burn probability maps to support their ongoing wildfire risk assessments – throughout California and other fire prone regions.  

Proof Points: 

1. Tested over a 4 year period in California, 57 to 80% of the burned area occurred in the top 20% of predicted locations across 5 equal-area burn probability classes. Furthermore, 74% to 96% of the burned area occurred within the top 50% of predicted locations.
2. Each year we updated vegetation and fuel data to better reflect current conditions for simulations, which improved accuracy by an average of 42%.
3. We published the paper as Open Access in the journal Scientific Reports and hosted our data on the Open Science Framework.

What Risk Managers Should Know: 

1. Burn probability is an essential ingredient in understanding wildfire risk across a landscape. 
2. To build targeted – and defensible – risk mitigation strategies, managers would benefit from validated, proven burn probability methods. 
3. Pyrologix wildfire simulation data can be confidently used to identify at-risk communities and landscapes, inform actuarial analyses, and prioritize effective mitigations.

 

Supporting Our Nation’s Rising Stakes

At Pyrologix we have long stressed the importance of rigorous data curation, calibration, and validation. We take great pains to ensure our risk modeling results are trusted and used. We frequently work with fire managers and scientists for their expert review, submit our work for peer-review, and strive for continuous improvement. We’re also committed to openly sharing this data to help communities across the country understand their risk. 

For example, Pyrologix helped design and develop the data for the Wildfire Risk to Communities  project, which identified many at-risk communities that subsequently experienced devastating fires. In the figure below, we illustrate how communities that experienced wildfire had disproportionately higher burn probability values than those that didn’t. We specifically identify a set of severely impacted communities whose burn probability values were among the most extreme nationwide.

 

 

In a previous blog post, we provided an overview of burn probability modeling and how it provides essential information for risk management, along with some salient examples demonstrating how our modeling identified high-risk areas including those burned in the Smokehouse Creek and Lahaina fires. But, especially in the wake of the 2025 Los Angeles fires, the stakes have never been higher. Wildfire risks are increasing, not just for wildfire managers, but also for power and insurance providers. 

Pressures are growing across these sectors for managers to be more efficient, targeted, and outcome-focused in their risk mitigation strategies, and they expect more scrutiny and a need for greater transparency and defensibility. To support this growing industry need and ensure we’re supporting communities with the best available, most accurate data, Pyrologix pressure tested our burn probability methods in a more expansive validation effort. To do so we confronted 4 years of annual burn probability maps against 4 subsequent years of fire activity in California (2020–2023).

 

Validation Analysis Provides Confidence in Results

Our primary focus in the validation study was evaluating how well our wildfire simulations could spatially differentiate areas of higher and lower fire likelihood, which is essential information for targeted, localized mitigation. We were also interested in how annual updates to fuel conditions captured the impacts of landscape changes – the largest driver being prior wildfire – on fire spread potential and spatial patterns of burn probability. To perform the simulations we used the FSim system, which was developed by the USDA Forest Service and has been widely used for strategic assessment and planning efforts including the National Cohesive Wildland Fire Management Strategy, the Wildfire Crisis Strategy, and the FEMA National Risk Index. 

 

 

Results showed strong predictive performance of our annual burn probability maps. We demonstrated greater proportions of burned areas in higher burn probability classes and higher average burn probabilities in burned than unburned areas. We also demonstrated statistically significant differences between burn probability distributions in burned and unburned areas, and showed that annual updates to the landscape fuel layers improved accuracy. Areas that burned had burn probability values that were up to 3.5 times higher on average compared to areas that didn’t burn. In general, we established a performance benchmark up to 80% of burned areas occurred in the top ~20% of mapped burn probability areas. We believe these results indicate the strongest burn probability modeling performance published to date. For a deeper analysis, review the paper here. 

 

How Pyrologix Can Help You Understand Risk From All Angles 

The results of our peer-reviewed burn probability work are a testament to our continued commitment to developing high quality and transparent wildfire risk data, and we will continue to share our findings, such as this work, broadly with our community. 

There are dozens of additional applications through which this kind of  work can be utilized, helping communities, insurers, utilities, and land managers understand their risk from a variety of angles. Here are just a few ways we can support:

• Understand changing landscape conditions: We can provide annual updates of burn probability maps that reflect changed landscape conditions and fire spread potential, particularly important in areas recently affected by wildfire or fuels reduction.
• Model catastrophic events: We can generate event sets with millions of simulated wildfires under variable conditions to quantify right tail risk and identify potential for extreme events. 
• Delineate dangerous fire spread corridors: With our custom-built Pyrosity™  we can identify destructive fire corridors spreading into communities. 

Our team of fire scientists, engineers, and analysts are dedicated to continue surfacing current, accurate, and actionable data that supports efforts to tackle wildfire risk as quickly as possible. To learn about using this data for your wildfire risk project, reach out to us at contact@pyrologix.com.