Management implications of Modeling initial attack success in Catalonia, Spain (Wildfire Conference. Day 2 - Wednesday Nov. 13th, 2019).

By Pau Costa Foundation on

2019 Wildfire Conference. Adressing the Challenges of Bushfire Management

Presentations Notes 2019: DAY 1 (Day 2 - Wednesday Nov. 13th, 2019).

Management implications of Modeling initial attack success in Catalonia, Spain.

Marcos Rodriguez, Fermín Alcasena and Cristina Vega-García (University of Lleida, Spain)


In developed countries in Southern Europe with professionally organised firefighting organizations or agencies, a limited number of fires that escape initial attack (IA) account for the majority of the burned area and damage. The number is usually limited due to the abundant firefighting resources and highly trained crews available for suppression, which usually manage to arrive in time to prevent propagation. However, no agency or organization can warrant an infinite availability of resources and total land and civil protection at all locations and all times, and financial constraints often demand that resources are optimized. Given the stricture of a total fire exclusion policy in populated countries in Southern Europe with high values at risk, but dwindling budgets, the correct identification of locations and times where and when fire escapes may occur is a crucial research and management issue. The factors that determine the chance that a fire escapes and becomes large can be modelled from spatiotemporal variables, though examples of these models are not abundant in the fire literature. A Random Forest model to predict IA success was recently calibrated in the fire-prone region of Catalonia, Spain, using this algorithm on observed outputs of success or failure from georeferenced historical fire ignition locations, response time to fire detection and weather conditions (Rodrigues et al. 2018). The IA model indicated that timely detection, ground accessibility, and aerial support quite determined the broad spatial pattern of fire containment probability, with strong gradients across the region.

Login or register to download this file