Revolutionizing Wildfire Prediction and Management with WindTL

In a world where wildfires pose an escalating threat, traditional methods of forecasting and managing these disasters fall short. Enter WindTL by Improving Aviation, a pioneering wildfire and ember spotting model that is setting new standards in the field by combining in-situ data from drones, remote sensors, and state-of-the-art science.

As the threat of wildfires continues to escalate globally, traditional methods of predicting and managing these disasters are proving insufficient. In response to this pressing issue, Improving Aviation introduces WindTL, a cutting-edge wildfire and ember spotting model using in-situ data. Led by atmospheric scientist Daniel Sunvold , the development of WindTL has undergone numerous iterations, testing against historical data, and implementation of emerging data analysis techniques. In Daniel’s words, “WindTL stands out as a versatile tool, deployable in the field or accessible remotely. It synergizes physics, machine learning, and high-scale resolution in-situ data to revolutionize how we predict and manage wildfires”.

According to the Insurance Institute for Business & Home Safety (IBHS), embers cause up to 90% of home and business ignitions during wildfire events. An ember is a small, burning debris shed from burning material lifted within the fire plume and transported laterally by the wind. Embers landing on a combustible material ignite and initiate a local fire far from the original location, causing a spotting event as much as 6 miles ahead of the main fire front.

Embers are highly unpredictable, and as of today, remain among the most challenging problems regarding wildfire prediction. A major reason for this is the lack of high-resolution in-situ meteorological observations. During wildfire incidents, on-call meteorologists perform only limited vertical sampling of the atmospheric parameters at a single location. These observations are used together with numerical estimates by incident managers to assess the atmospheric processes in the fire plume and guide firefighting operations. However, the prediction is based on sparse measurements from weather sensors in the area and not in the vicinity of the wildfire front. In addition, wildfires create their own atmospheric conditions and are often unpredictable. Therefore, current numerical weather prediction models struggle to simulate these conditions accurately.

The unpredictable nature of ember spread, coupled with the lack of high-resolution in-situ meteorological observations, has historically left a significant gap in wildfire management. Improving Aviation’s WindTL model fills this gap by combining physics, machine learning, and in-situ data, predicting the progression of the fire front and accurately assessing areas where embers are likely to ignite and create secondary fires.

WindTL utilizes ground weather station data, aircraft equipped with weather sensors, and autonomous aerial vehicles, commonly known as drones, to generate a comprehensive understanding of atmospheric processes. To capture atmospheric information from drones, Improving Aviation has developed a ruggedized atmospheric sensor suite that can be deployed on commercial off-the-shelf drones. The information captured with the sensor suite communicates in real-time with a back-end station in the field. In addition, drones are equipped with cameras, so fire signatures are located using machine learning methods. This data provides the initial conditions utilized by the WindTL model.

By combining in-situ data with various layers of third-party data, including satellite fire signatures, topography, fuel load, and drought conditions, WindTL generates high-resolution maps of fire risk and ember spread locations. As opposed to current alternatives, this model operates at a high resolution of 30 meters. Additionally, the model integrates already-developed programs including FlamMap and Farsite to simulate rate of spread, flame length, and projected fire front perimeters to utilize within the model itself. This negates the need to “reinvent the wheel”, allowing for an umbrella application to be developed that houses various applications as well as adding our own developments.

What sets WindTL apart is its real-time application by field operators. Through our field-deployable hardware, firefighters and decision-makers are equipped with the latest, most accurate information for resource deployment, fire management, and evacuations. They can generate wildfire risk maps for their areas of interest in real-time. In addition, WindTL is easily accessible for remote decision-makers through the Improving Aviation web-based platform, skytlcloud.com.  Additionally, the model can be seamlessly integrated into other platforms, enhancing decision support systems.

Currently, the model is in the testing phase with projected commercial deployment during Q3 of 2024. In the upcoming months, Daniel Sunvold and the team will be going to several prescribed fires and wildfires to obtain more empirical data for model testing, tuning, and refinement.

In essence, WindTL is reshaping the landscape of wildfire prediction and management. By harnessing in-situ and remotely sensed data and blending it with cutting-edge science and technology, WindTL provides a critically needed accurate forecast of fire and ember spread risk. With ongoing testing and plans for full deployment in late 2024, WindTL holds the promise of making wildfire incidents more manageable and safer, marking a significant leap forward in our fight against these natural disasters.

Written by Rocio Frej Vitalle March 05 2024

Previous
Previous

Google invests in Latino-led AI startup in Tampa

Next
Next

Navigating Wildfires: Improving Aviation's Cutting-Edge Risk Model for Precision Emergency Planning Amid Texas Fire Risk