top of page
Writer's pictureOverwatch Imaging

Beyond Firefighting: Leveraging multiband imaging for post-fire analysis and simulation

A detailed case study by Overwatch Smart Sensor user, Aria Firefighting

France-based Aria Firefighting offers an innovative and pragmatic approach to firefighting. Its know-how extends from the operation of water bomber aircraft to the provision of tactical intelligence directly inspired by military techniques of coordinated air/ground combat. The company is also dedicated to post-fire prevention and analysis as a crucial aspect of wildland fire fighting.


Aria created following case study after an early 2024 wildfire that broke out in Corsica, leveraging multispectral imagery collected from their Overwatch Imaging TK Series Smart Sensor


On January 3rd, 2024, the particularly unfavorable weather conditions associated with a significant water deficit in Corsica (a situation of recurrent winter drought for 3 years) were particularly conducive to the outbreak and spread of a fire.



At 11:30 a.m., a fire broke out in the town of Barbaggio.

Thanks to the combined action of firefighters, foresters, civil security personnel and soldiers, as well as two Canadairs from the Nîmes security base and a water bomber helicopter from South Corsica, the fire was brought under control a little more than 24 hours after it broke out.


Our goal is to provide an in-depth debriefing and a detailed understanding of fires.

In this article, we want to highlight the innovative tools and methodologies we are deploying in this critical phase.

 

1. Mapping the site of the fire and creating the outline of the fire


Our first tool is an advanced (Overwatch Imaging) sensor that allows us to create an accurate mapping of the disaster site in multiple spectral bands. This technology is essential for understanding the extent and progression of the fire. Thanks to this mapping, our teams can analyze the impact of the fire on the ground and extract the final outline of the fire.


A planned flight a few days after the fire was brought under control allowed us to make the following renderings and quickly obtain a definitive calculation of the burned area.


RGB ortho mosaic and final outline of the disaster (in red)

The high resolution of the data collection makes it possible to analyse the passage of the fire and its possible impact in detail.


A closer look at the burn area with the sensor's full resolution imagery

2. Multiband Orthophotography (RGB,NIR,LWIR) and NDVI Analysis


The ability of our sensor to perform orthophotography in multiple spectral bands is a key asset in our arsenal. It allows us to observe the affected area in multiple spectral bands outside the visible spectrum (Near InfraRed - NIR and Long Wave InfraRed - LWIR), thus providing a detailed view of the fire damage that would not be attainable in the visible range only.


NIR (Near Infrared) Burned Zone

In addition, thanks to the NDVI (Normalized Difference Vegetation Index) analysis, we can assess the condition of burned vegetation. This analysis is crucial to understand the ecological impact of the fire and plan for vegetation regeneration.


NDVI mapping of the fire area (in shades of pink, areas of dead vegetation)

3. Terrain modelling and estimation of the volume of biomass destroyed.


In our post-fire analysis approach, we have integrated an innovative methodology combining several state-of-the-art technologies to accurately assess the impact on vegetation and estimate the volume of biomass destroyed.


  • Creating a Digital Height Model (DTM) from Digital Terrain Models (DTM) and Area Models (DSM):


We started by relying on digital elevation models (DTMs) and surface models (DSMs) made using high-definition LIDAR data provided by the IGN - French Mapping Agency (National Geographic Institute). These models allowed us to understand in detail the topography and surface structure of the affected area.


From the DEMs and DSMs, we developed a Digital Height Model (DSM). This model is essential for distinguishing tree and vegetation elevations from the ground, providing a clear perspective on the three-dimensional structure of vegetation prior to the fire.


Difference between the ground (in Blue) and the tops of the vegetation (in black)

  • Data Reliability with 3D Photogrammetry

To enhance the reliability of our analyses, we also performed a 3D reconstruction of the area using photogrammetry. This approach validated and refined the data obtained by LIDAR, providing a detailed and up-to-date visualization of the post-fire area.


Photogrammetric reconstruction of the disaster area
  • Estimated Volume of Biomass Destroyed

Then, we used our NDVI mapping to delineate the areas of vegetation that were actually destroyed. By cross-referencing these data with the MNH, we were able to accurately determine the extent and degree of impact on vegetation.

Through the combination of these technologies, we estimated the volume of biomass destroyed in the affected area. Although approximate, this estimate is crucial for assessing the ecological impact of the fire and planning reforestation and ecological restoration efforts.

 

4. Fire Spread Simulation and Impact Analysis


The second major tool at our disposal is Firecaster, a fire propagation simulation software developed by Jean-Baptiste Filippi within the CNRS laboratory  of the University of Corsica Pascal Paoli and for which we have acquired the sub-licenses through the SATT Sud-Est.


This software is essential for analysing the potential behaviour of the fire and the impact of the firefighters' action.


For the following simulations, the global weather data comes from the European Forecasting Centre, modified locally with the station data. The calculation of the 30 hours of propagation took 30 seconds on 1 processor without atmospheric coupling.


  • Operational Use


From an operational point of view, it can be used by firefighters to support their decision by illustrating critical points. On the image below it is possible to easily identify the important "rendezvous" at the level of the RD 81 to avoid an overflow into the town of Furiani.


Firecaster API simulation based on expected weather conditions (adverse scenario). In blue the critical area to be defended at the level of the RD 81, in white the final outline of the fire.

  • Analysis and debriefing


Secondly, the simulations help us to understand how the fire could have spread without intervention and to evaluate the effectiveness of the fighting strategies deployed.


This analysis can be carried out by replaying the fire according to various weather scenarios but also by analyzing the predicted firepower.


The simulation below shows the calculated fire powers according to the model. These values are calculated at the point maximum, locally the fire only reaches these values for a very short period of time and therefore corresponds to the local worst and not to the established power which is at least half as much.


Firing power calculated according to the Firecaster API model. The most important powers are in Yellow.

Nevertheless, we can identify all the areas in yellow that could not have been attacked head-on, being more or less at 12,500 Kw/m, which, according to the logarithmic scale, are still in class 4 "loss of control of easy control – crown fire". Very occasionally, we even have powers of around 20,000 Kw/m corresponding to the highest class, class 5 "crown fire" now called "extreme fire" and which corresponds to the free breaking zones, when the fire resumes a slope just after passing a pass.


  • Estimating the value of the rescued


Thanks to a collaboration with Antoine Belgodere, Pascal Paoli lecturer and researcher at the University of Corsica, we were able, for the very first time in Europe, to analyse the economic value of the area protected by the action of the firefighters, thus underlining the importance of their action.


This analysis is based on a coupling of the simulations carried out by Firecaster and on the method developed in the article "On the Marginal Cost of the Duration of a Wildfire" (Belgodere et al 2023, Journal of Forest Economics).


This method consists of identifying the type of land use on the burned areas (nature, agriculture, campsites, forests, hiking trails, housing, etc.) and estimating the economic value of the elements endangered by the fire. Of course, fires do not destroy all economic value (a burned house does not necessarily have to be completely rebuilt). Therefore, these estimates should not be regarded as an estimate of the cost of the fire, but as an upper bound of that cost.


Estimated cost of burning based on Firecaster simulation

It can be noted that, the fact that the fire was stopped at 200 ha, corresponds to the time of arrival at the D81, if it had crossed the road there would be a very large increase in surface areas due to the surge, as well as an explosion in cost. The cost before the road crossing, around 4 million euros (max marker) is mainly due to the loss of the natural area, its visual impact and restoration, security.

 

Optimizing the understanding of fires through innovation


Aria Firefighting's mission is not limited to fighting flames; we are committed to understanding every fire to constantly improve our fighting and prevention strategies. With our state-of-the-art tools and expertise, we are able to make a significant contribution to the safety of people and ecosystems. We are proud to share this information and learnings with the firefighting professional community and the general public, to strengthen our shared commitment to a safer and more sustainable future.


We would like to thank the Economic Development Agency of Corsica and INIZIÀ - Territorial Incubator of Innovative Companies of Corsica who have been supporting us since the beginning of this project. As well as François Menassé from Helix Corsica who helped us with the data processing.

240 views0 comments

Comments


bottom of page