Towards the use of Remote Sensing for Identification of Building Damage, Destruction, and Defensive Actions at Wildland-Urban Interface Fires

Post-fire remote sensing provides a promising tool for assessing building damage, destruction, and defensive actions from wildland fire. However, limited studies exist to guide image acquisitions. Consequently, we compare remotely piloted aircraft systems and satellite post-fire imagery to ground-based assessments from the 2017 California Tubbs Fire to classify building damage, destruction, and defensive actions in an intermix and interface community. We also geolocate defensive action information from active fire images, videos, and eyewitness accounts. We utilize both manual and object-based classification approaches. Both types of overhead imagery using manual classifications had high kappa statistics ranging from 0.81 to 0.96, indicating almost perfect agreement with ground-based assessments for primary building destruction (e.g., homes). Object-based classifications of destruction had kappa statistics ranging from 0.63 to 0.88 for primary buildings, indicating substantial agreement. Additionally, manual and object-based classifications identified many destroyed secondary buildings (e.g., sheds) missed by ground-based assessments. Occlusions due to canopy cover contribute to lower classification results in the intermix community. All imagery missed significant damage identified in the ground-based assessment. Remotely piloted aircraft systems imagery was superior to satellite imagery in identifying defensive action indicators. Nonetheless, all image types are valuable additions to ground-based assessments of damage, destruction, and defensive actions. Finally, we demonstrate the importance of accounting for defensive actions in assessing building response at wildland-urban interface fires.