Integrating georeferencing and fire anatomy modeling in canyon fire experiments using thermal UAV remote sensing

Background Wildfires profoundly impact ecosystems and human health, making them a critical research focus. Drone-mounted Thermal Infrared (TIR) sensors are widely used for active fire monitoring due to their ability to penetrate smoke and capture real-time fire dynamics. Aims This study addresses the challenges of georeferencing TIR data accurately, caused by limited fire anatomy modeling, imprecise georeferencing, the coarse resolution of TIR imagery and the dynamic nature of fire behavior. Methods We collected a one-hour TIR dataset capturing a fast-moving controlled fire in a California canyon, US, alongside auxiliary data of multispectral and in situ meteorological observations. We developed an enhanced georeferencing method for TIR keyframes that accounts for different stages of fire development, integrating the fire anatomy, topography and fire behavior. The approach optimized accuracy with strategic tie point allocation, TIR temperature preprocessing, fire spread theory and raster transformation tailored to topography. Key results The approach produced a customized dataset of 106 georeferenced TIR images for a one-hour active fire event across fire development. Conclusions The method delivered precise geographic locations for TIR imagery of fast-moving fire, enabling reliable integration with auxiliary data. Implications This approach supports advanced fire behavior analysis when combined with fuel maps, topography, and meteorological datasets.