Compare mitigation.
Evaluate scenarios, quantify uncertainty and compare changes in exposure to help choose where to act.
Platforms / SWIMWildfire uncertainty.
Stochastic Wildfire Impact Model (SWIM)
Wildfire uncertainty.
Informed decisions.
SWIM is a cutting-edge wildfire modeling tool that uses epidemic-inspired
stochastic simulations to predict how wildfires are likely to spread across
the landscape, based on real-world environmental conditions.
Recorded SWIM simulation · Berkeley, 21 October 2026.
ATS’ SWIM is rigorously tested for accuracy against active wildfires in the United States
Updated
SWIM Total WildFire Prediction Accuracy
——Correctly predicted
—Wildfires evaluated
How accuracy is measured
SWIM success requires predictions that satisfy two conditions:
- SWIM’s 90th percentile predicted wildfire perimeter encompasses a wildfire’s actual final perimeter.
- SWIM's predicted wildfire perimeter exceeds a shape geometry similarity threshold with respect to the actual final perimeter.
Condition (1) mandates that SWIM predict the correct burn areas, while condition (2) serves as a control mechanism to prevent arbitrarily large estimations.
Together, SWIM is judged on the accuracy and usefulness of its predictions.
SWIM accuracy is measured as the proportion of times both conditions are satisfied.
2D prediction & consequence modeling
The forecast.
The consequences.
See more than a fire perimeter. Examine possible outcomes alongside the buildings, communities and infrastructure they could affect.
P75 selected. 901.4 modeled acres.
Compare recorded outcomes
One scenario.
A range of possibilities.
Compare saved P50, P75 and P90 outputs. All three predicted perimeters stay visible as you compare the heat maps and their modeled consequences.
901.4acres
modeled extent
modeled extent
Saved scenario · 21 October 2026
Explore live predictionsInvestigate an ignition.
Forensic ignition analysis works backward from a burned perimeter to investigate candidate ignition locations.
Keep risk in view.
Dynamic risk maps connect geospatial context to the infrastructure and communities that need attention.
Explore wildfire risk intelligenceWhat Sets SWIM Apart?
Fine-grained inputs.
A range of possible outcomes.
Fine-Scale Resolution
Most wildfire models operate on large grid cells. SWIM models use 0.2-acre polygons – allowing fine-grained accuracy that matters for risk modeling.
Data-Rich Simulations
SWIM integrates wind speed and direction, temperature, humidity, elevation, fuel composition, moisture, and more. It’s powered by millions of wildfire records and terrain data.
Probabilistic Risk Outputs
Instead of a single deterministic forecast, SWIM outputs a distribution of outcomes – critical for utilities, insurers, and emergency responders.
Nationwide Coverage
SWIM is built to scale. From California to Florida, SWIM can simulate fire behavior anywhere in the U.S. given an ignition point.
Applications of SWIM
What will you
decide next?
The same wildfire scenario can inform different decisions. Start with the question your organization needs to answer.
Prioritize mitigation investments.
Compare where wildfire could spread with the infrastructure and communities in its reach. Evaluate mitigation options and focus investment where it can make the greatest difference.
- Identify and mitigate wildfire risk from overhead assets
- Evaluate tradeoffs in mitigation investments
Prepare before the fire.
Test ignition scenarios, evaluate wildfire exposure and inform preparedness strategies. Use probabilistic outcomes to plan for a range of possible conditions.
- Prioritize fuel reduction and firebreak efforts
- Develop zoning or evacuation plans informed by probabilistic wildfire pathways
Understand portfolio exposure.
Examine how potential fire spread intersects a portfolio. Support underwriting decisions, geographic accumulation management and comparison of mitigation strategies.
- Incorporate fine-grained wildfire spread probabilities into actuarial models
- Quantify monetized risk under different mitigation strategies
Protect agricultural operations.
Use wildfire spread scenarios to compare protective measures for crops, livestock and working land.
- Optimize crop and livestock protection strategies based on wildfire spread forecasts
- Assess tradeoffs between fire-resilient practices and operational costs
Assess a property in context.
Consider wildfire exposure alongside development, acquisition and risk-management decisions.
- Evaluate wildfire risk for prospective developments and acquisitions
- Incorporate probabilistic fire spread into property valuation and insurance negotiations
Case studies across four continents
Wildfire intelligence.
Across four continents.
Explore ATS wildfire risk work in Australia, Canada, Italy and Chile.
Predicting Wildfire Like an Epidemic
Inspired by Epidemiology, Engineered for Wildfire
We use the same mathematics that predict disease outbreaks to forecast the path of wildfire.
Epidemic Model Insight
- The single greatest predictor for whether you will be infected by a disease is whether someone near you is infected.
- Epidemiologists have refined tools over decades, rooted in probabilistic spread models, to assess and mitigate disease risk.
- These tools use stochastic simulations to represent uncertainties in disease transmission outcomes.
Stochastic epidemic spread model – widely
used by epidemiologists to predict disease
outcomes.
The SWIM Breakthrough
- At Aerospace Technical Services (ATS), we have taken this proven modeling approach and applied it to wildfire.
- Our Stochastic Wildfire Impact Model (SWIM) predicts how a wildfire is likely to spread given an ignition point and real-world variables.
- SWIM is a U.S.-wide, high-resolution model that can be applied anywhere in the country.
SWIM’s wildfire spread model – built using
similar stochastic dynamics and informed by
geophysical and weather data.
Why Stochastic Modeling?
Because Wildfire Is Uncertain – Just Like Disease Spread
| Epidemic Models | SWIM Wildfire Models |
|---|---|
|
Use random sampling of disease
transmission likelihoods | Use random sampling of wildfire spread likelihoods |
| Factor in individual-level data (age, health, exposure) | Factor in geospatial data (fuel moisture, elevation, wind, slope, etc.) |
| Perform millions of simulations to build infection risk maps | Perform millions of simulations to build wildfire impact probability maps |