Machine Learning-Based Wildfire Susceptibility Assessment in California’s WUI
Multi-Source feature selection and analysis
Bibliographic Data
| ID | 22432833 |
|---|---|
| Authors | Jianxing Ren (0009-0001-8129-4164, University of Science and Technology of China), Jun Zhang (0000-0003-1706-1611, University of Science and Technology of China, corresponding author), Weiguo Song (0000-0001-5390-6787, University of Science and Technology of China) |
| Year | 2026 |
| Pages | 106308 |
| Publication date | 2026-07-01 |
| Open Access | Yes |
| Type | ARTICLE |
| Venue | International Journal of Disaster Risk Reduction (JOURNAL) |
| Journal identifiers | ISSN: 2212-4209 |
| Publisher | Elsevier BV (PUBLISHER) |
| DOI | 10.1016/j.ijdrr.2026.106308 |
| OpenAlex | W7168270031 |
| Language | EN |
| References cited | 28 |
The Wildland-Urban Interface (WUI) in California is increasingly threatened by wildfires due to complex socio-environmental interactions. This study aims to quantify WUI wildfire susceptibility and identify the primary non-linear drivers of fire occurrence. We systematically evaluated multi-source covariates using Wasserstein Distance, LASSO regression, Boruta, and Recursive Feature Elimination to construct highly accurate machine learning models for regional susceptibility assessment. We found that the Normalized Difference Moisture Index (NDMI) is the leading environmental driver. It exhibits a critical threshold at -0.21, marking the point where vegetation transitions from a fire-promoting fuel source to an inhibitory factor. Additionally, fire occurrence probabilities were significantly higher in low-density "interface" zones compared to highly integrated "intermix" areas. Anthropogenic variables and local vegetation dryness primarily govern WUI fire risk, highlighting the outsized role of human ignitions in sparsely populated interface regions. These insights support spatially targeted fire management, emphasizing the critical need for localized fuel treatments and active human oversight in vulnerable communities.
Covariate · Dryness · Feature selection · Random forest · Threatened species · Fire Detection and Safety Systems · Fire effects on ecosystems · Injury Epidemiology and Prevention
Physiographically sensitive mapping of climatological temperature and precipitation across the conterminous United States
The Wildland–urban Interface in the United States
Rapid growth of the US wildland-urban interface raises wildfire risk
Development of gridded surface meteorological data for ecological applications and modelling
Evaluation of forest fire risk based on multicriteria decision analysis techniques for Changzhou, China
Wildfire risk for global wildland–urban interface areas
| Highly cited | No |
|---|