Better Data Is Transforming Wildfire Prediction
Why the future of wildfire forecasting belongs to those who listen to the land — and the data.
Wildfires are no longer seasonal anomalies. They are recurring disasters, reshaping landscapes and lives from the Amazon to Alberta, from Greece to California. As the stakes rise, so does the urgency to improve how we predict them — not just where the weather looks dangerous, but where fires will actually start.
A recent study published in Nature marks a turning point. Led by Di Giuseppe, Global Data-driven Prediction of Fire Activity, explores a global, machine learning–based approach to fire activity prediction.
The punchline: traditional fire weather indices are outdated. They over-predict danger in deserts and under-predict actual ignition in places where human behavior, fuel buildup, and climate whiplash are the real drivers. These results are akin in the US Forest Services’ Wildfire Hazard Potential maps which overpredict wildfire risk in some locations, yet 25%+ of all wildfires occur in places the WHP shows as low risk.
The researchers at ECMWF (European Centre for Medium-Range Weather Forecasts) show that a properly trained machine learning (ML) model — particularly one using the mid-complexity algorithm XGBoost — can forecast where fires will ignite with surprising accuracy. The secret isn’t the model itself. It’s the data.
Why Forecasts Fail: A Lesson from Los Angeles
In January 2025, wildfires tore through Los Angeles fueled by dry Santa Ana winds and an unusual buildup of vegetation — remnants of two unusually wet springs. The Canadian Fire Weather Index (FWI), widely used in fire management, lit up nearly all of Southern California as “extreme.” Helpful? Barely. It flagged everything and nothing.
Meanwhile, the ECMWF’s data-driven system — trained on observed fire ignitions, fuel loads, and ignition triggers — accurately pinpointed the Palisades and Eaton fire areas days in advance. This wasn’t magic. It was the result of feeding the model with the full fire triangle: weather, fuel, and ignition.
Voice of the Acre®: Athena Intelligence’s Answer
This principle — that data quality matters more than model complexity — is one that Athena Intelligence knows well. At Athena, we call our approach Voice of the Acre®. It’s a geospatial intelligence platform that listens to the land itself: integrating fuel dynamics, ignition risks, moisture, land use, and even insurance and infrastructure overlays. It’s designed not just to warn that fire might happen, but to reveal where if an ignition occurs, the wildfire will be large an catastrophic.
More importantly, Voice of the Acre® focuses on what matters. Showing customers (utilities, communities and financial organizations) where these fires are likely to occur 12+ months in advance, giving them time to proactively reduce the risk.
Where ECMWF’s model proves this is possible globally and with a few days of notification, Voice of the Acre brings that same predictive power to decision-makers on the ground — especially electric utilities and insurers.
With 30 sq meter pixel resolution, integrating private and proprietary data, and aligning forecasts with operational needs, Athena delivers what we call actionable pre-wildfire geospatial intelligence.
We don’t just model potential; we model probabilities and operational priorities.
Fuel Is the Foundation
In Global Data-driven Prediction of Fire Activity, Di Giuseppe (et al.) found that fuel availability and moisture are the most powerful predictors of fire activity — more so than temperature or even ignition sources in many regions. Their analysis showed that models using weather alone suffer a 30% drop in predictive skill. With only ignition data, it’s similar. But with fuel? The drop is just 15%, and when all three inputs are combined, forecast reliability peaks.
At Athena, we’ve taken this a step further. Voice of the Acre integrates terrain and vegetation data with human ignition proxies like roads and community density, for miles around each location. Our utility customers integrate this with their asset management. This allows our partners to know not just that fire risk is high, but why, where, and what they can do about it.
From Hazard to Risk to Readiness
There’s a fundamental difference between forecasting and operationalizing data. Fire weather indices like FWI or WHP are valuable — but they’re abstractions.
Athena’s platform closes the loop: from abstract indices to precise probabilities. We help utilities prioritize their limited dollars to the areas with the largest impact. We also provide data for Public Safety Power Shutoffs and Risk Spend Efficiency and Risk to Financial Impact reporting. At the same time, these probabilities which were developed for insurance actuaries help underwriters refine wildfire risk pricing and reinsurance strategies. We can help banks managed their credit risk and mortgage portfolios.
Because while danger is generic, wildfire has terroir by bioregion and humans have multiple types of risk exposure.
Global Vision, Local Action
The ECMWF’s global system is a proof of concept — and an impressive one. It correctly predicted the chaotic Canadian wildfire season of 2023 and the Los Angeles fires of 2025. Its reliance on MODIS satellite data, weather reanalysis, and physically derived fuel models shows that even in the absence of perfect observations, smart proxies and smart training can yield powerful results.
Voice of the Acre® has a longer forecast and offers localized intelligence. We fill in the missing pieces with private data, infrastructure overlays, and operational relevance. We’re have not just built a better wildfire forecast, our customers are using an intelligence layer for proactive wildfire prevention.
Listen to the Land
In the end, the most advanced machine learning system won’t help if it doesn’t listen — to data, to fuel, to history, to human patterns. That’s the philosophy behind both the research out of ECMWF and Athena’s geospatial intelligence.
The future of wildfire prediction isn’t just global. It’s hyperlocal, data-rich, and purpose-driven. It doesn’t issue alerts. It answers questions: Where will it burn? When? And what should I do about it?
Welcome to AI at its best, the Voice of the Acre®.
Athena Intelligence is a data vendor with a geospatial, conditional, profiling tool that pulls together vast amounts of disaggregated wildfire and environmental data to generate spatial intelligence, resulting in a digital fingerprint of wildfire risk.
Clients include electric utilities, communities and financial services companies, where Athena’s geospatial intelligence incorporated into multiple products that can be accessed through an online portal. Athena’s data is currently used in wildfire mitigation plans (WMP) and public safety power shutoffs (PSPS), Community Wildfire Protection Plans (CWPP), property insurance underwriting and portfolio risk optimization.
You can learn more on our YouTube Channel @Athena_Intelligence Athena Intelligence (AthenaIntel.io) — YouTube Reach out to me at Elizabeth@AthenaIntel.io and follow us on LinkedIn
