Abstract
Lightning is the most common origin of natural fires, being strongly linked to specific synoptic conditions associated with atmospheric instability, such as dry thunderstorms; dry fuels are required for ignition to take place and for subsequent propagation. The aim was to predict the daily probability of ignition by exploiting a large dataset of lightning and fire data to anticipate ignition over the entire Iberian Peninsula. We trained and tested a machine learning model using lightning strikes (>17million) in the period 2009–2015. For each lightning strike, we extracted information relating to fuel condition, structural features of vegetation, topography, and the specific characteristics of the strikes (polarity, intensity and flash density). Naturally triggered ignitions are typically initiated at higher elevations (above 1000m above sea level) under conditions of low dead fuel moisture (<10–13%) and moderate live moisture content (Drought Code>300). Negative-polarity lightning strikes (−10kA) appear to trigger fires more frequently. Our approach was able to provide ignition forecasts at multiple temporal and spatial scales, thus enhancing forest fire risk assessment systems.
Marcos Rodrigues
Professor of Geographic Information Science
My research interests include forest fires, spatial analysis and forestry.