Scholarlink

<p>Wildfires present a significant threat to ecosystems and human life, particularly as global climate change amplifies the likelihood of extreme fire events. This study develops a machine learning-based wildfire susceptibility model for Germany, using data between 2003 and 2023. The primary goal is to identify the dominant wildfire predictors and create monthly susceptibility maps. A Random Forest (RF) algorithm was trained on remote sensing data and a comprehensive set of predictors, including meteorological, terrain, and land cover variables. The results indicate that surface air pressure, elevation, vegetation health, and proximity to urban areas are the most important factors in predicting fire susceptibility. The model achieved 89% accuracy, demonstrating the effectiveness of data-driven approaches in wildfire risk modeling. The monthly susceptibility map for July 2022 highlights northeastern Germany as particularly vulnerable to fire outbreaks. The results offer valuable insights for targeted wildfire prevention and resource allocation, emphasizing the importance of both temporal and spatial dimensions in managing wildfire risks.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Machine learning wildfire susceptibility mapping for Germany

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年4月25日

查看全文
  • Boris Thies

摘要

Wildfires present a significant threat to ecosystems and human life, particularly as global climate change amplifies the likelihood of extreme fire events. This study develops a machine learning-based wildfire susceptibility model for Germany, using data between 2003 and 2023. The primary goal is to identify the dominant wildfire predictors and create monthly susceptibility maps. A Random Forest (RF) algorithm was trained on remote sensing data and a comprehensive set of predictors, including meteorological, terrain, and land cover variables. The results indicate that surface air pressure, elevation, vegetation health, and proximity to urban areas are the most important factors in predicting fire susceptibility. The model achieved 89% accuracy, demonstrating the effectiveness of data-driven approaches in wildfire risk modeling. The monthly susceptibility map for July 2022 highlights northeastern Germany as particularly vulnerable to fire outbreaks. The results offer valuable insights for targeted wildfire prevention and resource allocation, emphasizing the importance of both temporal and spatial dimensions in managing wildfire risks.