Improving Wildfire Prevention: Combining FWI Components, Historical Burn Probabilities, and Multi-Sensor Satellite Data for Better Early Warning Systems in Los Angeles, CA

Wildfires remain a significant challenge in fire-prone regions like Southern California, as evidenced by the ongoing 2024/25 wildfire disaster. This study introduces an innovative methodology for assessing wildfire risk by combining Fire Weather Index (FWI) components, historical burn probabilities, and multi-source meteorological and satellite data, including ERA5 reanalysis, MODIS and Sentinel-2 data.

The methodology includes a decomposition of FWI components — including temperature, wind, humidity, and fuel moisture—and their derived indices: Fine Fuel Moisture Code (FFMC), Duff Moisture Code (DMC), Drought Code (DC), Initial Spread Index (ISI), Build-Up Index (BUI), and the final Fire Weather Index (FWI).  The Fire Weather Index (FWI) meteorological data will be sourced from the Copernicus ERA5 dataset because the ERA5 data provides essential weather information, including wind speed, surface temperature, and relative humidity. These parameters are cross-referenced with MODIS-derived Land Surface Temperature (LST) to validate spatial temperature trends, statistically downscale the derived data, and identify discrepancies that could signal pre-fire anomalies. Additionally, satellite-derived vegetation indices from Sentinel-2 (e.g., NDVI, NDWI, and MSAVI2) are incorporated to evaluate vegetation health and moisture stress. Before the fire, the vegetation states are compared with historical burn probability mapping, constructed using past wildfire records and environmental datasets, to create a comparative framework to assess predicted versus actual fire spread patterns.

The working hypothesis suggests that combining ERA5 meteorological data with satellite-derived indices can provide a deeper understanding of pre-fire conditions, thereby improving early warning capabilities. Preliminary findings suggest that anomalies such as elevated temperatures (from MODIS and ERA5) and vegetation stress (from Sentinel-2) are strong indicators of impending wildfire risks. These patterns highlight the importance of combining meteorological, historical, and satellite-based insights to inform wildfire risk management.

We propose developing an interactive early warning system using Google Earth Engine to operationalise these insights. This system integrates FWI components, ERA5-derived meteorological data, historical burn probabilities, and satellite-based indices into a dashboard for real-time monitoring. The dashboard will be designed to visualise critical thresholds, assess vegetation stress, and analyse fire risk trends. This comprehensive approach empowers proactive decision-making to mitigate the impacts of wildfires and improve overall disaster preparedness.

This study demonstrates the potential of leveraging cross-referenced ERA5, MODIS and Sentinel-2 data, FWI components, and historical probabilities to build a scalable, data-driven framework for wildfire risk assessment in vulnerable regions.