Scholarlink

<p>Global changes in climatic parameters have significantly altered fire regimes in Mediterranean ecosystems, increasing wildfire risk through higher temperatures and prolonged dry periods. This study assesses forest fire susceptibility in Antalya Province, Türkiye, using an Explainable Artificial Intelligence (XAI) framework based on the XGBoost algorithm integrated with SHapley Additive exPlanations (SHAP) to interpret model predictions. Two models were developed: a province-wide model and a Wildland–Urban Interface (WUI) model to capture differences between natural and human-influenced landscapes. A total of 33 spatial variables representing topography, vegetation, climate, and anthropogenic factors were used. Both models demonstrated strong predictive performance. The provincial model achieved an accuracy of 0.844 and an AUC-ROC of 0.911, while the WUI model reached an accuracy of 0.826 and an AUC-ROC of 0.891. SHAP results indicated that temperature and land surface temperature are the most influential drivers in both models, followed by NDVI, precipitation, atmospheric pressure, and evapotranspiration. However, scale-dependent differences were evident. At the provincial scale, climatic and biophysical variables dominated susceptibility patterns, whereas in WUI zones, anthropogenic factors such as population density and proximity to infrastructure played a greater role alongside environmental drivers. High susceptibility areas are mainly concentrated along coastal districts extending eastward from the city center, including Muratpaşa, Kepez, Aksu, Serik, and Manavgat. Overall, high and very high susceptibility zones account for approximately 14% of the area in both models, with WUI risk more clustered around settlements and transport corridors. The findings highlight that wildfire susceptibility in Antalya is driven by interacting climatic, ecological, and human factors, with climate dominating at broader scales and human influence increasing in WUI areas. These results support integrated fire management strategies that address both environmental change and anthropogenic pressure.</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

Assessment of fire susceptibility in Wildland—Urban interfaces according to explainable artificial intelligence: a case study of Antalya province, Mediterranean region, Türkiye

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年8月4日

查看全文
  • Mücahit Coşkun,
  • Sohaib K. M. Abujayyab,
  • Onur Canbulat,
  • Nigar Canbulat,
  • Nesrin Sarsıcı Demir,
  • Selime Tut,
  • Kamile Zeren

摘要

Global changes in climatic parameters have significantly altered fire regimes in Mediterranean ecosystems, increasing wildfire risk through higher temperatures and prolonged dry periods. This study assesses forest fire susceptibility in Antalya Province, Türkiye, using an Explainable Artificial Intelligence (XAI) framework based on the XGBoost algorithm integrated with SHapley Additive exPlanations (SHAP) to interpret model predictions. Two models were developed: a province-wide model and a Wildland–Urban Interface (WUI) model to capture differences between natural and human-influenced landscapes. A total of 33 spatial variables representing topography, vegetation, climate, and anthropogenic factors were used. Both models demonstrated strong predictive performance. The provincial model achieved an accuracy of 0.844 and an AUC-ROC of 0.911, while the WUI model reached an accuracy of 0.826 and an AUC-ROC of 0.891. SHAP results indicated that temperature and land surface temperature are the most influential drivers in both models, followed by NDVI, precipitation, atmospheric pressure, and evapotranspiration. However, scale-dependent differences were evident. At the provincial scale, climatic and biophysical variables dominated susceptibility patterns, whereas in WUI zones, anthropogenic factors such as population density and proximity to infrastructure played a greater role alongside environmental drivers. High susceptibility areas are mainly concentrated along coastal districts extending eastward from the city center, including Muratpaşa, Kepez, Aksu, Serik, and Manavgat. Overall, high and very high susceptibility zones account for approximately 14% of the area in both models, with WUI risk more clustered around settlements and transport corridors. The findings highlight that wildfire susceptibility in Antalya is driven by interacting climatic, ecological, and human factors, with climate dominating at broader scales and human influence increasing in WUI areas. These results support integrated fire management strategies that address both environmental change and anthropogenic pressure.