Abstract
Wildfires profoundly affect watershed hydrology by altering vegetation, soil properties, and moisture dynamics. This research advances post-wildfire risk assessment through a four-phase approach. First, satellite remote sensing was used to assess hydrological impacts of major wildfires in California’s Feather River Watershed, which has experienced over 60% burn coverage across three recent fire events. Second, hydrological modeling in the Santa Cruz Creek watershed, affected by the 2007 Zaca Fire, simulated 16 storm events from 2001 to 2024, quantifying changes in peak discharge, soil moisture, and time of concentration, and indicating partial post-fire recovery after five years. Third, a correlation study examines how well pre-fire Enhanced Vegetation Index (EVI) and differenced Land Surface Temperature (dLST) explain vegetation burn severity, using differenced Normalized Burn Ratio (dNBR) across multiple wildfires in the Feather River Watershed. Finally, four methods are used to estimate soil burn severity (SBS) with additional inputs like EVI, dLST, and land cover: (1) dNBR substitution, (2) linear regression, (3) random forest, and (4) dense neural networks (DNN). Overall, this research advances post-fire assessment by integrating remote sensing, hydrologic modeling, and machine learning across multiple fire-affected landscapes. It provides scalable tools for understanding long-term watershed recovery, supporting more informed water resource management in fire-prone regions.
