Get the most reliable climate data
Oops! Something went wrong while submitting the form.
At Ambee, we’ve always believed in using data to make better decisions. Over the past two years, we’ve been working on something groundbreaking: a predictive model capable of forecasting wildfire risks weeks in advance. And the results are already proving to be a game-changer.
While still in development, on December 16, 2024, the model identified several regions in Southern California (highlighted in yellow) as high-risk areas for wildfires over the coming four weeks.
On January 7, 2025, a series of catastrophic wildfires erupted across Los Angeles and San Diego County, forcing more than 200,000 people to evacuate and reducing over 57,000 acres to ash. Over the next three weeks, the fires became some of the deadliest in California’s history, claiming at least 29 lives and destroying 18,000 homes and structures.
In the aftermath, a closer look at the data revealed a strong correlation with Ambee’s predictions. The model had identified high-risk zones with 96% accuracy, meaning that 96% of all the active fires (highlighted in red) occurred within the areas it had flagged a month in advance.
This level of foresight isn’t just a massive technical achievement —it’s an opportunity to transform the way authorities, communities, and businesses prepare for wildfires. With the right data, we can anticipate, prepare, and, most importantly, save lives.
The January California wildfires not only cost billions of dollars in losses, they also took 29 lives. Effectively predicting wildfires can ensure that not only money is saved, but lives are saved too.
- Jaideep Singh Bachher, CEO and Co-founder of Ambee
But making these predictions is far from simple.
Unlike hurricanes or heatwaves, which follow somewhat predictable meteorological patterns, wildfires are chaotic, driven by a tangled web of factors. Some are ignited naturally by lightning, while others stem from human activity—whether intentional or accidental. Terrain, fuel availability, wind speed, humidity, and broader climate trends all influence fire behavior, making it extremely difficult to anticipate when and where a wildfire will occur.
Existing operational fire prediction models struggle with these complexities. They often assume that fire behavior can be predicted solely based on local conditions—such as temperature, wind speed, and vegetation density—without accounting for the broader environment.
At Ambee, we’ve taken a radically different approach. We've built a predictive model that captures local and nonlocal influences on fire risk.
.svg){kind=icon}
.svg){kind=icon}
{
"message": "success",
"data": [
{
"week": 9,
"fwi_today": 0,
"ffmc_today": 0,
"ffmc_temp_precip": -0.007227499999998876,
"weekly_seasonality": 0,
"temperature": 27.1,
"precipitation": 0,
"days_since_last_fire": 32,
"predicted_risk": 0,
"predicted_risk_category": "low",
"created_time": "2025-02-28T01:08:55.126Z",
"timeLocal": "2025-02-27T19:08:55-06:00",
"timezone": "America/Mexico_City"
},
{
"week": 10,
"fwi_today": 0,
"ffmc_today": 0,
"ffmc_temp_precip": -0.05158999999999879,
"weekly_seasonality": 0,
"temperature": 19.9,
"precipitation": 0.1,
"days_since_last_fire": 39,
"predicted_risk": 0,
"predicted_risk_category": "low",
"created_time": "2025-02-28T01:08:55.126Z",
"timeLocal": "2025-02-27T19:08:55-06:00",
"timezone": "America/Mexico_City"
},
{
"week": 11,
"fwi_today": 0,
"ffmc_today": 0,
"ffmc_temp_precip": -0.008574999999998667,
"weekly_seasonality": 0,
"temperature": 27.2,
"precipitation": 0,
"days_since_last_fire": 46,
"predicted_risk": 0,
"predicted_risk_category": "low",
"created_time": "2025-02-28T01:08:55.126Z",
"timeLocal": "2025-02-27T19:08:55-06:00",
"timezone": "America/Mexico_City"
},
{
"week": 12,
"fwi_today": 0,
"ffmc_today": 0,
"ffmc_temp_precip": 0,
"weekly_seasonality": 0,
"temperature": 26.5,
"precipitation": 0,
"days_since_last_fire": 53,
"predicted_risk": 0,
"predicted_risk_category": "low",
"created_time": "2025-02-28T01:08:55.126Z",
"timeLocal": "2025-02-27T19:08:55-06:00",
"timezone": "America/Mexico_City"
}
]
}Sample response from Ambee’s wildfire forecast API. Read the wildfire forecast documentation.
We began by dividing North America into uniform grids, within which we incorporated a range of variables—temperature, vegetation density, precipitation, the Fire Weather Index (FWI), the Fine Fuel Moisture Code (FFMC), slope, elevation, and other meteorological factors. But predicting wildfire risk goes beyond current conditions; it’s about understanding long-term patterns.
To do this, we analyzed over 30 years of historical fire data to capture seasonality for each grid, helping us pinpoint the times when regions are most vulnerable to fires. We also considered the number of days since the last fire, a key indicator of fuel accumulation and fire ignition likelihood. Additionally, we engineered more variables to fine-tune our risk assessments.
The model processes this data to assign a high, moderate, or low wildfire risk level to each location. Unlike traditional models, it also accounts for nonlocal influences, identifying risks in nearby grids to deliver a more accurate and comprehensive prediction.
Data and coverage details.
It must be noted that there is no universal standard for wildfire forecasting and risk levels at this scale–because no one has been able to do it until now.
Our forest fire forecast system is a cutting-edge AI-driven solution that analyzes over 15 years of historical data to provide highly accurate risk assessments across the U.S. By integrating geospatial intelligence, fire weather indices, satellite imagery, and real-time weather patterns, it predicts high-risk areas with precision weekly.
– Chandrasekar D, Director of Engineering
This breakthrough technology is just the beginning. We’re continuing to refine the model to expand its accuracy, coverage, and reliability. As we advance, we aim to bring this technology to more regions and provide even more detailed forecasts, giving authorities and communities the tools they need to prepare for the worst—and, ultimately, save lives.
Ambee’s wildfire forecast API is changing the landscape of wildfire preparedness. It’s not just a prediction; it’s the beginning of a new era in disaster response. We’re ready to make this vision a reality, and we invite you to be part of it.
Imagine a future where:
To learn more about Ambee’s wildfire risk forecast and see how it can help you prepare for the future, explore our API and documentation, or reach out to our team to get started. The future of wildfire preparedness is here, and it’s powered by data.
