A carregar...
Publicação
Rede primária de faixas de gestão de combustível: análise espacial à eficácia na contenção de incêndios rurais em Portugal
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 8.55 MB | Adobe PDF |
Orientador(es)
Resumo(s)
A tendência do potencial destrutivo dos incêndios rurais é visivelmente crescente,
resultado da alteração de padrões climáticos, do abandono rural e da desvalorização
da economia florestal. Neste sentido, é inevitável desenvolverem-se esforços que
aumentem a resiliência do espaço rural a estes desastres, ambicionando acima de
tudo a proteção das pessoas e dos seus bens. Neste trabalho perspetiva-se acerca da
funcionalidade da rede primária de faixas de gestão de combustível (RPFGC) através
do desenvolvimento de um método de avaliação da sua eficácia na contenção de
incêndios rurais.
Para derivar esta análise selecionaram-se sete incêndios rurais ocorridos entre
2020 e 2023, onde se verificou interseção entre o fogo e troços de RPFGC. A partir do
método desenvolvido derivou-se um conjunto de dados composto por 1 665
interseções, onde automaticamente se classificou a eficácia do troço de RPFGC na
contenção do incêndio (eficaz/não eficaz). Cada interseção foi posteriormente
caracterizada segundo 11 variáveis independentes (descritoras do comportamento do
fogo, meteorologia, relevo, entre outras) com o intuito de treinar um conjunto de
classificadores automáticos, os quais viriam a servir para determinar os fatores com
maior influência na eficácia da RPFGC.
Os principais resultados indicam que i) a RPFGC contribuiu para a contenção do
perímetro do incêndio em 19% das interseções identificadas; ii) o modelo de
classificação automática com melhor desempenho foi treinado pelo algoritmo random
forest, devolvendo uma exatidão de 92%; e iii) as variáveis que mais influenciam a
eficácia da RPFGC na contenção de um incêndio relacionam-se com o comportamento
estimado do fogo. Este trabalho assume-se como um ponto de partida para a avaliação
de um elemento estruturante da paisagem, contribuindo para o desenho,
planeamento e implementação de medidas de prevenção mais eficazes contra
incêndios rurais.
The trend in the destructive potential of wildfires is visibly increasing, as a result of changing climate patterns, rural abandonment and the devaluation of the forest economy. In this sense, it is inevitable to develop efforts to enhance the resilience of rural areas to these disasters, aiming above all to protect people and their property. This work focuses on the functionality of the fuel break primary network (FBPN) by developing a method for evaluating its effectiveness in containing wildfires. To conduct this analysis, seven wildfires that occurred between 2020 and 2023 were selected, where an intersection between the fire and segments of the FBPN was observed. From the developed method, a dataset of 1,665 intersections was generated, where the effectiveness of the FBPN segment in containing the fire was automatically classified (effective/not effective). Each intersection was subsequently characterized according to 11 independent variables (descriptors of fire behavior, meteorology, terrain, among others) in order to train a set of automatic classifiers, which would then serve to determine the factors with the greatest influence on the effectiveness of the FBPN. The main results indicate that i) the FBPN contributed to containing the fire perimeter in 19% of the identified intersections; ii) the best-performing automatic classification model was trained by the random forest algorithm, achieving an accuracy of 92%; and iii) the variables that most influence the FBPN's effectiveness in containing a fire are related to the estimated fire behavior. This work provides a starting point for evaluating a structural landscape element, contributing to the design, planning and implementation of more effective prevention measures against wildfires.
The trend in the destructive potential of wildfires is visibly increasing, as a result of changing climate patterns, rural abandonment and the devaluation of the forest economy. In this sense, it is inevitable to develop efforts to enhance the resilience of rural areas to these disasters, aiming above all to protect people and their property. This work focuses on the functionality of the fuel break primary network (FBPN) by developing a method for evaluating its effectiveness in containing wildfires. To conduct this analysis, seven wildfires that occurred between 2020 and 2023 were selected, where an intersection between the fire and segments of the FBPN was observed. From the developed method, a dataset of 1,665 intersections was generated, where the effectiveness of the FBPN segment in containing the fire was automatically classified (effective/not effective). Each intersection was subsequently characterized according to 11 independent variables (descriptors of fire behavior, meteorology, terrain, among others) in order to train a set of automatic classifiers, which would then serve to determine the factors with the greatest influence on the effectiveness of the FBPN. The main results indicate that i) the FBPN contributed to containing the fire perimeter in 19% of the identified intersections; ii) the best-performing automatic classification model was trained by the random forest algorithm, achieving an accuracy of 92%; and iii) the variables that most influence the FBPN's effectiveness in containing a fire are related to the estimated fire behavior. This work provides a starting point for evaluating a structural landscape element, contributing to the design, planning and implementation of more effective prevention measures against wildfires.
Descrição
Dissertation presented as the partial requirement for obtaining a Master's degree in Geographic Information Systems and Science, specialization in Geographic Information Systems and Science
Palavras-chave
Sistemas de Informação Geográfica Incêndios rurais Faixas de gestão de combustível Eficácia Comportamento do fogo Geographic Information Systems Wildfires Fuel breaks Effectiveness Fire behaviour