Brazil Flying Labs & Tech To The Rescue: Building an AI Solution for Forest Fire Assessment

By Brazil Flying Labs

Brazil has experienced historic wildfire seasons, with an estimated 30 million hectares burned in 2024 — an area larger than Italy — representing a 79% increase compared to 2023. These events have left a critical threat to the region's biodiversity and vital water resources.

In the Atlantic Forest (Mata Atlântica) — one of the most biodiverse and most threatened ecosystems in the world — only approximately 24% of the original forest cover remains, with mature, well-preserved forest accounting for an even smaller share. While fire is not the sole driver of degradation, habitat fragmentation and continued vegetation loss significantly increase vulnerability to wildfires, intensifying ecological damage and complicating post-fire recovery.

These figures demonstrate that wildfire monitoring, damage assessment, and rapid-response technologies are no longer optional. They are essential instruments for protecting biodiversity, safeguarding water resources, supporting climate regulation, guiding restoration efforts, and informing public policy in an era of escalating fire risk.

A Solution

Brazil Flying Labs, in partnership with Tech To The Rescue, developed an AI-powered Forest Fire Assessment platform to support environmental authorities and conservation organizations in evaluating wildfire impacts in protected areas across São Paulo state. 

The system was applied to severely affected reserves, including the Luiz Antônio and Jataí Ecological Stations, following the 2024 fires.

Technical Architecture 

The platform integrates Sentinel-2 satellite imagery, spectral indices (NDVI, NBR, and RBR), and AI-driven analysis to:

• Produce interactive fire severity maps
• Calculate burned areas in hectares
• Classify damage levels as light, moderate, or intense
• Enable data export for further analysis and integration into GIS workflows.

The main components of this innovation are:

• API: Developed using Python, Django REST Framework, and Google Earth Engine, enabling spatial queries of protected areas and temporal fire analysis.
• Frontend: A React-based web application for data selection, visualization, and downloads (GeoTIFF and JPEG), with interoperability with GIS platforms such as QGIS.
• Training and Capacity Building: A 35-hour training program covering geoprocessing, remote sensing, Python, and AI, designed to strengthen technical capacity within institutions and local communities.

To ensure transparency, adaptability, and long-term impact, the solution is released as open-source software under the MIT License and is publicly available on GitHub.

Impact

By transitioning to AI-powered open-source tools, we have established a scalable and transparent blueprint that directly addresses the urgent need for damage assessment in São Paulo’s forest reserves. 

This project proves that collaborative innovation is essential for changing the context of disaster response, providing a vital tool for environmental conservation that can be adapted to face future climate challenges.

The project video below presents the development process and key implementation outcomes.

Acknowledgements

We extend our deepest gratitude to: 

• Sonja Betschart, CEO of WeRobotics, for strategic support and facilitating the collaboration with Tech To The Rescue.
• Alice Damasceno (Lenovo), for funding and partnership.
• Daniel Shanklin and Rhea AI, for front-end development expertise and a significant donation.
• UGADS Jundiaí, CREAS Cajamar, EGP Jundiaí, and Iron Mountain, for supporting Python training in local communities — one participant directly contributed to the development of this project.
• The Brazil Flying Labs team: Diego Paolo Ferruzzo Correa, Diogo Dias, Wellington Franklyn, Juliana Berbert, and Reginaldo Cardoso.

As we look toward the future of climate resilience, how do you think open-source AI tools can be further leveraged to empower local communities in protecting their own natural heritage?