Convolving multiple satellite products for near real-time burn scar mapping: case evaluation for top five costliest wildfires in United States
Abstract
The risk of wildfire is growing. Five of the top ten costliest wildfires in the United States happened in 2017 and 2018. Large and rapid post-event insurance claims payouts have posed substantial cash demand on insurers. Therefore, accurate burn area information as prompt as possible is needed for optimal financial arrangements. However, current published burn scar creation methods based on manual inspection or a single satellite image product either fails to deliver the burn scar in a desirable time frame or at the desired granular level.
To serve such practical needs in wildfire risk management, we have developed a new approach in wildfire burn scar creation by combining multiple publicly available high resolution optical and radar satellite image products. Our data sources include Sentinel-1 Synthetic Aperture Radar (SAR) Level-1 Single Look Complex (SLC) and Level-1 Ground Range Detected (GRD), Sentinel-2, and Landsat 8. In addition to Normalized Burn Ratio (NBR) analysis commonly used in burn scar analysis we also use interferometric coherence change. The final burn scar prototype is created by convolving different single satellite image-based burn scars with the National Land Cover Database land use layer. Each single satellite image-based burn scar as well as the final convolved burn scar are validated against our in-house claims data and the Sentinel-2 True Color Image for the top five costliest wildfires in United States, which are Camp Fire, CA (2018), Tubbs Fire, CA (2017), Woolsey Fire, CA (2018), Atlas Fire, CA (2017), and Thomas Fire, CA (2017). The preliminary validation results have shown good detection accuracy for built-up areas. The burn scar prototype in this study is the first attempt in convolving multiple publicly available high-resolution satellite image products and land use information, to our knowledge. The presented burn scar prototype is a substantial improvement in creating near real-time burn scars at a high granular level. The accurate burn scar information would not only serve the damage assessment needs post-fire but also aid fire monitoring and prediction practices during the fire.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFMNH52A..06Z
- Keywords:
-
- 3390 Wildland fire model;
- ATMOSPHERIC PROCESSES;
- 1817 Extreme events;
- HYDROLOGY;
- 4315 Monitoring;
- forecasting;
- prediction;
- NATURAL HAZARDS;
- 4323 Human impact;
- NATURAL HAZARDS