Cargando…

Survival analysis and classification methods for forest fire size

Factors affecting wildland-fire size distribution include weather, fuels, and fire suppression activities. We present a novel application of survival analysis to quantify the effects of these factors on a sample of sizes of lightning-caused fires from Alberta, Canada. Two events were observed for ea...

Descripción completa

Detalles Bibliográficos
Autores principales: Tremblay, Pier-Olivier, Duchesne, Thierry, Cumming, Steven G.
Formato: Online Artículo Texto
Lenguaje:English
Publicado: Public Library of Science 2018
Materias:
Acceso en línea:https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5761846/
https://www.ncbi.nlm.nih.gov/pubmed/29320497
http://dx.doi.org/10.1371/journal.pone.0189860
_version_ 1783291599650816000
author Tremblay, Pier-Olivier
Duchesne, Thierry
Cumming, Steven G.
author_facet Tremblay, Pier-Olivier
Duchesne, Thierry
Cumming, Steven G.
author_sort Tremblay, Pier-Olivier
collection PubMed
description Factors affecting wildland-fire size distribution include weather, fuels, and fire suppression activities. We present a novel application of survival analysis to quantify the effects of these factors on a sample of sizes of lightning-caused fires from Alberta, Canada. Two events were observed for each fire: the size at initial assessment (by the first fire fighters to arrive at the scene) and the size at “being held” (a state when no further increase in size is expected). We developed a statistical classifier to try to predict cases where there will be a growth in fire size (i.e., the size at “being held” exceeds the size at initial assessment). Logistic regression was preferred over two alternative classifiers, with covariates consistent with similar past analyses. We conducted survival analysis on the group of fires exhibiting a size increase. A screening process selected three covariates: an index of fire weather at the day the fire started, the fuel type burning at initial assessment, and a factor for the type and capabilities of the method of initial attack. The Cox proportional hazards model performed better than three accelerated failure time alternatives. Both fire weather and fuel type were highly significant, with effects consistent with known fire behaviour. The effects of initial attack method were not statistically significant, but did suggest a reverse causality that could arise if fire management agencies were to dispatch resources based on a-priori assessment of fire growth potentials. We discuss how a more sophisticated analysis of larger data sets could produce unbiased estimates of fire suppression effect under such circumstances.
format Online
Article
Text
id pubmed-5761846
institution National Center for Biotechnology Information
language English
publishDate 2018
publisher Public Library of Science
record_format MEDLINE/PubMed
spelling pubmed-57618462018-01-23 Survival analysis and classification methods for forest fire size Tremblay, Pier-Olivier Duchesne, Thierry Cumming, Steven G. PLoS One Research Article Factors affecting wildland-fire size distribution include weather, fuels, and fire suppression activities. We present a novel application of survival analysis to quantify the effects of these factors on a sample of sizes of lightning-caused fires from Alberta, Canada. Two events were observed for each fire: the size at initial assessment (by the first fire fighters to arrive at the scene) and the size at “being held” (a state when no further increase in size is expected). We developed a statistical classifier to try to predict cases where there will be a growth in fire size (i.e., the size at “being held” exceeds the size at initial assessment). Logistic regression was preferred over two alternative classifiers, with covariates consistent with similar past analyses. We conducted survival analysis on the group of fires exhibiting a size increase. A screening process selected three covariates: an index of fire weather at the day the fire started, the fuel type burning at initial assessment, and a factor for the type and capabilities of the method of initial attack. The Cox proportional hazards model performed better than three accelerated failure time alternatives. Both fire weather and fuel type were highly significant, with effects consistent with known fire behaviour. The effects of initial attack method were not statistically significant, but did suggest a reverse causality that could arise if fire management agencies were to dispatch resources based on a-priori assessment of fire growth potentials. We discuss how a more sophisticated analysis of larger data sets could produce unbiased estimates of fire suppression effect under such circumstances. Public Library of Science 2018-01-10 /pmc/articles/PMC5761846/ /pubmed/29320497 http://dx.doi.org/10.1371/journal.pone.0189860 Text en © 2018 Tremblay et al http://creativecommons.org/licenses/by/4.0/ This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/) , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
spellingShingle Research Article
Tremblay, Pier-Olivier
Duchesne, Thierry
Cumming, Steven G.
Survival analysis and classification methods for forest fire size
title Survival analysis and classification methods for forest fire size
title_full Survival analysis and classification methods for forest fire size
title_fullStr Survival analysis and classification methods for forest fire size
title_full_unstemmed Survival analysis and classification methods for forest fire size
title_short Survival analysis and classification methods for forest fire size
title_sort survival analysis and classification methods for forest fire size
topic Research Article
url https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5761846/
https://www.ncbi.nlm.nih.gov/pubmed/29320497
http://dx.doi.org/10.1371/journal.pone.0189860
work_keys_str_mv AT tremblaypierolivier survivalanalysisandclassificationmethodsforforestfiresize
AT duchesnethierry survivalanalysisandclassificationmethodsforforestfiresize
AT cummingsteveng survivalanalysisandclassificationmethodsforforestfiresize