Forest Fire Detection Methods

A pinecone made of wax and charcoal won the UK James Dyson Award --- and it might be one of the most elegant wildfire detection systems ever designed. It's called Pyri, and its mechanism is borrowed directly from nature. Certain pine trees in fire-prone regions have evolved a survival strategy called "pyriscence": their pinecones are sealed with resin, and only when a wildfire's heat melts that seal do they release their seeds. The fire that destroys the forest simultaneously plants the next one. Pyri works the same way. When heat melts its wax shell, it releases a saltwater solution whose electrolytes activate internal electronics, broadcasting a radio frequency signal to communication towers tens of kilometers away --- requiring zero maintenance, zero batteries, and zero technical expertise to deploy. Traditional wildfire detection relies on cameras, sensors, and satellites --- expensive systems with limited coverage that are essentially useless in the remote, under-resourced communities most vulnerable to catastrophic fires. Pyri can be scattered across vast, inaccessible terrain from helicopters, costs a fraction of conventional solutions, and if a fire never comes, it simply sits in the forest as a harmless, non-toxic object until it biodegrades. Over the past decade, wildfires have caused an estimated $106 billion in economic losses globally, and extreme wildfire frequency is projected to increase 50% by 2100. A patent application is pending, and the team is moving toward real-world deployment. As co-founder Richard Alexandre put it: "Nature is 100% the ultimate designer."

Two NASA-developed technologies are key components of a new high-resolution sensor for observing wildfires: High Operating Temperature Barrier Infrared Detector (HOT-BIRD), developed with support from NASA’s Earth Science Technology Office (ESTO), and a cutting-edge Digital Readout Integrated Circuit (DROIC), developed with funding from NASA’s Small Business Innovation Research (SBIR) program. Developed with support from NASA’s Earth Science Technology Office (ESTO), the “Compact Fire Infrared Radiance Spectral Tracker” (c-FIRST) is a small, mid-wave infrared sensor that collects thermal radiation data across five spectral bands. Most traditional space-based sensors dedicated to observing fires have long revisit times, observing a scene just once over days or even weeks. The compact c-FIRST sensor could be employed in a SmallSat constellation that could observe a scene multiple times a day, providing first responders data with high spatial resolution in under an hour. In addition, c-FIRST’s dynamic spectral range covers the entire temperature profile of terrestrial wild fires, making it easier for first-responders to detect everything from smoldering, low-intensity fires to flaming, high intensity fires. The need for space-based assets dedicated to wildfire management is severe. During the Palisade and Eaton Fires earlier this year, strong winds kept critical observation aircraft from taking to the skies, making it difficult for firefighters to monitor and track massive burns. Space-based sensors with high revisit rates and high spatial resolution would give firefighters and first responders a constant source of eye-in-the-sky data. c-FIRST leverages decades of sensor development at JPL to achieve its compact size and high performance. In particular, the quarter-sized High Operating Temperature Barrier Infrared Detector (HOT-BIRD), a compact infrared detector also developed at JPL with ESTO support, keeps c-FIRST small, eliminating the need for bulky cryocooler subsystems that add mass to traditional infrared sensors. With HOT-BIRD alone, c-FIRST could gather high-resolution images and quantitative retrievals of targets between 300°K (about 80°F) to 1000°K (about 1300°F). But when paired with a state-of-the-art Digital Readout Integrated Circuit (DROIC), c-FIRST can observe targets greater than 1600°K (about 2400°F). Developed by Copious Imaging LLC. and JPL with funding from NASA’s Small Business Innovation Research (SBIR) program, this DROIC features an in-pixel digital counter to reduce saturation, allowing c-FIRST to capture reliable infrared data across a broader spectral range. Full Article: https://lnkd.in/gJ2MjPW2 #JPL #NASA #cFIRST NASA’s c-FIRST instrument could provide high resolution data from a compact space-based platform in under an hour. (NASA/JPL)

Radar data can often be tricky to interpret - especially when it comes to meteorological targets such as rain, and non meteorological targets such as smoke, in this example. National Weather Service Doppler radar data, particularly reflectivity and correlation coefficient, play crucial roles in identifying and monitoring #wildfires. Here's how each contributes using a real-time situation this afternoon, Thursday 23 October 2025 using the Doppler weather radar from the Tampa Bay National Weather Service based in Ruskin, Florida to sample a burn in a nature preserve in eastern Hillsborough County, south of State Road 60 and west of County Road 39. Radar Reflectivity 📡 🔥 Detection of Smoke and Ash: Reflectivity measures the intensity of the radar signal returned from targets in the atmosphere. During a wildfire, smoke and ash particles can scatter radar waves, resulting in strong reflectivity values. This allows radar systems to identify areas where fires are burning based on heightened reflectivity, which often indicates the presence of particulate matter. Correlation Coefficient (CC) 📡 🔥 Distinguishing Fire Effects: The correlation coefficient measures how similar the returned radar signals are. In the context of wildfire detection, a low correlation coefficient indicates a mix of different particle sizes and shapes, typical for smoke and fire. For instance, the presence of larger smoke particles mixed with smaller water droplets from weather patterns can lead to a lower CC value, differentiating smoke from rain or other meteorological phenomena. By utilizing reflectivity and correlation coefficient data from Doppler radar, meteorologists and emergency responders can gain valuable insights into the presence, movement, and intensity of wildfires and aid in decision making. #EmergencyManagement #Technology #Meteorology #Science

California Forest Fire Analysis Using SAR and Multispectral Data The Eaton and Palisades fires, two of the most destructive wildfires in Southern California's history, have devastated approximately 60 square miles, claiming at least 25 lives and displacing over 88,000 people. To analyze the fire's impact, we used Sentinel-1 SAR data (Log Difference technique) and Sentinel-2 multispectral data (Burn Area Indices via a change detection model in ArcGIS Pro). The analysis utilized pre-fire and post-fire imagery, leveraging band combinations like SWIR, IR, and NIR to highlight affected areas. The Normalized Burn Ratio Index (NBRI) was calculated to map and quantify the burned regions. Post-processing included raster reclassification and polygon conversion to extract precise burned area maps. This workflow not only visualizes fire damage but provides a replicable method for forest fire monitoring in other regions, supporting rapid response and recovery efforts. #RemoteSensing #GIS #ForestFireMonitoring #SAR #MultispectralAnalysis #ArcGIS

From idea → use case: AI fighting wildfires in Germany What surprises me most is how quickly some of these ideas move from concept to reality. A few years ago, “AI detecting wildfires before they spread” sounded like science fiction. Today it’s already happening. 💡 The idea: Detect wildfires within minutes instead of hours or days using sensors, drones, and AI. 🔍 The use case: In Eberswalde (Germany), solar-powered sensors detect early smoke/gas traces. When triggered, an AI-enabled drone (Silvaguard) flies out, uses infrared cameras, and delivers precise location and fire data to emergency responders. This means fires can be contained earlier — saving forests, communities, and lives. 🔥 From idea to use case: this shift feels fast, tangible, and much needed. 🤔 Open question: How can we scale such solutions, and who should drive adoption — startups, governments, or communities themselves? 📖 Source: Associated Press https://lnkd.in/euB3epDd

𝐃𝐞𝐭𝐞𝐜𝐭𝐢𝐧𝐠 𝐖𝐢𝐥𝐝𝐟𝐢𝐫𝐞 𝐃𝐚𝐦𝐚𝐠𝐞 𝐰𝐢𝐭𝐡 𝐏𝐲𝐭𝐡𝐨𝐧, 𝐒𝐚𝐭𝐞𝐥𝐥𝐢𝐭𝐞 𝐃𝐚𝐭𝐚 & 𝐎𝐒𝐌 | #30𝐃𝐚𝐲𝐌𝐚𝐩𝐂𝐡𝐚𝐥𝐥𝐞𝐧𝐠𝐞 (15/30) Detecting wildfire damage from space is an important and impactful use case for remote sensing and satellite data. So, when I saw that Day 15 of the #30DayMapChallenge is themed Fire, I immediately decided to build a small geospatial pipeline to detect wildfire burn scars and assess property damage using geospatial data. Using ESA Sentinel satellite data, NASA FIRMS fire observations, and OSM building data via osmnx, I created a reference map one month before the fire and a burned area map during the wildfire. Then, I computed Normalized Burn Ratio (NBR) differences to detect damage, filtered noisy patches, and mapped these areas interactively against building footprints. As a target area, I focused on the devastating wildfires in the Los Angeles area during the summer of 2025. This methodology shows an example of how wildfire can be quickly and globally, detected - and as you will see, the results, even the efficient steps to contain the wildfire, are visible. Hopefully, with more geospatial intelligence, in the future, we can prevent even more damage caused by natural disasters. Full Python tutorial coming soon: 🔔 𝐖𝐚𝐥𝐤-𝐭𝐡𝐫𝐨𝐮𝐠𝐡 𝐨𝐧 𝐘𝐨𝐮𝐭𝐮𝐛𝐞: https://lnkd.in/dBTUqctW 🔔 𝐂𝐨𝐝𝐞 𝐨𝐧 𝐒𝐮𝐛𝐬𝐭𝐚𝐜𝐤: https://lnkd.in/g3FY3cTP #30DayMapChallenge #WildfireDetection #Geospatial #RemoteSensing #PythonGIS #SentinelHub #NBR #FIRMS #OpenStreetMap #DisasterMapping #SpatialDataScience #EarthObservation #WildfireDamage

Preventing the Next Big Fire: How the Pathfinder Fixed-Wing eVTOL “Drone in a Box” Can Help. As wildfires grow more frequent and severe, proactive strategies are essential. That’s why we at Starling Inc are excited to spotlight how the Pathfinder fixed-wing drone in a box offers a scalable, autonomous solution for wildfire prevention and mitigation — from early detection to rapid response support. Here’s what it brings to the table: -- Continuous airborne monitoring over high-risk areas (terrain, forests, infrastructure) with minimal human intervention. -- Fast hotspot detection, thermal mapping, smoke plume tracking, and change -- detection. -- Persistent presence thanks to automated takeoff/landing and battery swapping, ideal for 24/7 duty cycles. -- Operational safety & scalability: fixed-wing gives greater endurance (did we say 2.5 hours...) and coverage than typical multi-rotors. -- Data integration & situational awareness: feeds can plug into fire operations centers, GIS systems, and decision support platforms. By deploying this kind of technology, agencies can detect emerging fires earlier, allocate resources more intelligently, and intervene before they spread. It’s not a silver bullet — but it’s a powerful tool in the toolkit. If you’re in fire agencies, emergency management, utilities, land management, or climate resilience — We at Starling Inc would love to hear your thoughts on how unmanned systems can become part of the standard toolkit. California Department of Forestry and Fire Protection (CAL FIRE) National Interagency Fire Center National Fire Chiefs Council (NFCC) #WildfirePrevention #Drones #UAS #FireTech #ClimateResilience #EmergencyManagement #WildfireMitigation

🔥 Revolutionizing Fire Surveillance & Safety with #AI: Real-Time Fire & Human Tracking System --- ✨ Key Innovations 1️⃣ Fire Segmentation (YOLO11-Seg + Roboflow) —  🔥 Fire regions are segmented using the Final Fire Segmentation Dataset, powered by Roboflow’s Serverless API and trained on YOLO11 instance segmentation architecture. 2️⃣ Person Segmentation (YOLO11m-Seg by Ultralytics) —  🧍 Local detection pipeline uses YOLO11m-Seg for accurate, real-time person segmentation and motion tracking — ideal for CCTV-style video streams. 3️⃣ EMA Temporal Smoothing —  📈 A custom Exponential Moving Average (EMA) smoother ensures frame-to-frame mask stability for both fire and person tracks — reducing false positives by up to 40%. 4️⃣ Dual Tracking Framework —  🔥 FireTracker stabilizes flame regions over time  🧍 PersonTracker applies centroid + IoU matching to maintain consistent human IDs across frames --- 📊 Results ✅ 90%+ pixel-level segmentation accuracy on complex indoor/outdoor fire scenes ✅ Real-time performance at 25–30 FPS (Full HD 1920×1080) ✅ Temporal smoothing reduces false detections by ~40% ✅ Intuitive visualization with live GIF animation, confidence graphs, and minimaps 🧠 Dataset & Models 📘 Fire Detection: Final Fire Segmentation Dataset – Roboflow  • Architecture: YOLO11 Instance Segmentation  • Hosted via Roboflow Serverless API 🧍 Person Segmentation: YOLO11m-Seg (Ultralytics)  • Locally loaded for frame-by-frame human mask generation  • Integrated with a custom centroid-based temporal tracker --- 🧩 Tech Stack Python | Yolo Group | Roboflow | OpenCV | Numpy Ninja | Pillow | Ultralytics | EMA Smoothing | @Tkinter GUI | Matplotlib | Real-Time Computer Vision | Deep Learning 💡 Strategy Fire Detection via Roboflow YOLO11 Model for pixel-precise segmentation Local YOLO11m-Seg Pipeline for person tracking and motion analysis EMA Temporal Averaging to stabilize predictions Centroid-Based Nearest-Neighbor Tracking for smooth person ID association Visualization Layer with animated GIFs, confidence plots, and proximity minimaps 🚒 Applications 🏭 Industrial & Factory Fire Safety 🏙️ Smart City CCTV Monitoring 🚨 Real-Time Emergency Alert Systems 🚒 Drone & IoT-Based Fire Surveillance 🛰️ Edge AI for Safety Robotics #ArtificialIntelligence #MachineLearning #ComputerVision #DeepLearning #YOLO11 #FireDetection #PersonSegmentation #Roboflow #Ultralytics #YOLOv8 #YOLO11m #SafetyAI #AI #SmartCity #FireSafety #EmergencyResponse #EdgeAI #AIProjects #AIEngineering #TechForGood #Innovation #DataScience #Python #OpenCV #Automation #RealTimeAI #ObjectTracking #Segmentation #AIResearch #ComputerVisionAI #SecurityTech #SmartMonitoring #AIForGood #InnovationInSafety #SoftwareEngineering #DeepLearningAI #AIShowcase #TechCommunity #LinkedInCreators #AIInnovation

My 3rd PhD paper has been published! 📄 It is about mapping burned areas using Sentinel-2 data and Rao's Q-Index. 🔥 📍 Study Areas - Sardinia (Italy). - Thassos (Greece). - Pantelleria (Italy). 🛰️ Sensors Used - Sentinel-2 — spectral data for analysis and mapping. - Copernicus Emergency Management System (#CEMS) — high-resolution data to validate the results. 🗺️ Spectral Indices - Normalized Burn Ratio (NBR). - Mid-infrared Burn Index (MIRBI). - Normalized Difference Vegetation Index (NDVI). - Burned Area Index for Sentinel-2 (BAIS2). ✍🏾 Methodology - Change detection approach based on Rao's Q-Index. - Analysis of Sentinel-2 images pre- and post-fire. - Calculation of spectral indices and application of Rao's Q Index in classical and multidimensional modes. - Validation of results with CEMS data. 📊 Accuracy Results - In classical mode, the highest accuracy was 72% with the MIRBI index. - In multidimensional mode, the highest accuracy was 96% combining MIRBI and NBR. - The combination of MIRBI and NBR consistently achieved the highest accuracy in all study areas. 🚀 Relevance of the Methodology - The combination of spectral indices and the use of Rao's Q Index improves the accuracy in identifying areas affected by fires. - The study highlights the importance of selecting the appropriate indices based on regional characteristics and specific detection needs. 👩🏾💻 Software and Repository - Google Colab with Google Earth Engine (GEE) integration - Python 3.10 - Open repository on GitHub: https://lnkd.in/dmJ4r9Wt 🟢 Conclusion - The combination of MIRBI and NBR indices provides a robust approach to increase the accuracy of fire detection. - Consistent alignment with CEMS data, validating the reliability of the methodology. ____________________ 🔗 Read the article: Tiengo et al., 2025 | https://lnkd.in/dFtsMfDN ____________________ 🍀 Share this post with your network! 📌 Save this post for later!

🛰️🛰️Some basics on application of LiDAR in Wildfire Ecology:🛰️🛰️ LiDAR (Light Detection and Ranging) technology has become a crucial tool in wildfire ecology, providing high-resolution data that improve wildfire risk assessment, fire behavior modeling, and post-fire recovery analysis. 🔥1. Pre-Fire Applications: • Fuel Load and Vegetation Structure Mapping: LiDAR enables precise estimation of forest fuel loads by generating 3D vegetation structure maps, which help assess wildfire risk. • Fire Behavior Modeling: Integrating LiDAR-derived vegetation and topographic data into fire models enhances predictions of fire spread, intensity, and behavior. 🔥🔥2. Active Fire Applications: • Real-Time Fire Monitoring: LiDAR-equipped drones and aircraft help track active fires by providing detailed assessments of fire progression, plume behavior, and fireline intensity. 🔥🔥🔥3. Post-Fire Applications: • Burn Severity Assessment: By comparing pre- and post-fire LiDAR scans, researchers can quantify vegetation loss, soil burn severity, and changes in canopy structure. • Forest Recovery Monitoring: LiDAR data are used to track post-fire regrowth, detecting early signs of vegetation recovery and guiding reforestation efforts. 💻4. Integration with Other Technologies: • Artificial Intelligence (AI) and Machine Learning: AI-powered analysis of LiDAR data improves wildfire risk assessment and automates post-fire damage evaluation. • Multispectral and Hyperspectral Fusion: Combining LiDAR with optical remote sensing enhances fire impact assessment and vegetation health monitoring. Conclusion: LiDAR technology is transforming wildfire ecology by improving pre-fire risk assessment, active fire monitoring, and post-fire recovery studies. As remote sensing and AI technologies advance, LiDAR will play an even more significant role in wildfire management and mitigation. References: • Fuel load estimation with LiDAR: https://lnkd.in/gykv5aHE • LiDAR in fire behavior modeling: https://lnkd.in/gHCcs7R9 • LiDAR for real-time fire monitoring: https://lnkd.in/gQTzX-n5 • Post-fire LiDAR applications: https://lnkd.in/ghzz-ttS • AI and LiDAR in wildfire management: https://lnkd.in/gs4gH6qX #WildfireEcology #LiDAR #RemoteSensing #AI #MachineLearning