Assessing human-caused wildfire ignition likelihood across Europe

Publikation: Beitrag in Buch/Konferenzbericht/Sammelband/GutachtenBeitrag in KonferenzbandBeigetragenBegutachtung

Beitragende

  • Adrián Jiménez-Ruano - , University of Lleida, University of Zaragoza (Autor:in)
  • Marcos Rodrigues Mimbrero - , University of Lleida, University of Zaragoza (Autor:in)
  • Fermín Alcasena Urdíroz - , University of Lleida (Autor:in)
  • Johan Sjöström - , RISE Research Institutes of Sweden (Autor:in)
  • Christopher Marrs - , Juniorprofessur für Umweltfernerkundung (Autor:in)
  • Luís Mário Ribeiro - , University of Coimbra (Autor:in)
  • Palaiologos Palaiologou - , Agricultural University of Athens (Autor:in)
  • Emilio Chuvieco - , University of Alcalá (Autor:in)
  • Pere Joan Gelabert - , University of Lleida (Autor:in)
  • Cristina Vega-García - , University of Lleida (Autor:in)

Abstract

Fire ignition probability is an essential component of most fire risk assessment frameworks. This study, framed within the H2020 project FirEUrisk, features a cohesive modelling approach in a set of representative regions (pilot sites; PS) in terms of fire activity across the European territory. These PS encompass different wildfire regimes in contrasting environmental settings: PS-1 Northern Europe, Kalmar Iän (South-East Sweden); PS-2 Central Europe, Southern Brandenburg and Eastern Saxony (Germany), North Bohemia (Czechia), and Lower Silesia (Poland); PS-3 Central Portugal; PS-4 Barcelona province (Spain); and PS-5 Attica region (Greece). Our main goal was to develop a common approach to model human-caused ignitions at a fine-grained spatial resolution (100 meters). For each pilot site we: (i) ascertain which factors influence ignition, hence, addressing potential differences in driving forces and, (ii) provide a spatial-explicit depiction of the patterns of ignition probability. For that propose, we fitted a Random Forest (RF) model in each PS from historical fire records (compiled by local fire agencies) and geospatial layers for land cover, accessibility, and population related factors. All models attained a high predictive accuracy, with AUCs that ranging from 0.69 (Northern Europe) to 0.89 (Attica Region). In turn, the most relevant explanatory variable was the population density that ranked most influential in four out of the five PS, followed by the fuel type, distance to roads, distance to the WUI, and percent cover of forest and wildlands. These findings are a valuable product to upscale future solutions at regional level (beyond NUTS3-type areas), conduct fire behavior modelling simulations, and enrich the science-based decisions which come from the forest and fire management agents at national and European level.

Details

OriginalspracheEnglisch
Titel2023 8th International Conference on Smart and Sustainable Technologies (SpliTech)
Seiten1-6
ISBN (elektronisch)978-953-290-128-3
PublikationsstatusVeröffentlicht - 2023
Peer-Review-StatusJa

Konferenz

Titel8th International Conference on Smart and Sustainable Technologies
KurztitelSpliTech 2023
Veranstaltungsnummer8
Dauer20 - 23 Juni 2023
Webseite
OrtHotel Elaphusa & Online
StadtSplit - Bol
LandKroatien