Abstract The objective of the following paper is to identify weather-related thresholds for wildfire danger, i.e., the potential extent of fire, based on local weather conditions. The target area is Greece, a wildfire prone Mediterranean country which experienced on average 2000 wildfires annually over the last two decades. Initially, the Fire Weather Index (FWI) component of the Canadian Fire Weather Index (FWI) System (FWI System) adopted by the European Forest Fire Information System (EFFIS), is evaluated with respect to its wildfire danger predictive ability. Hence, weather and wildfire data at municipality level and on a daily basis, for the period 2000–2016 are exploited. The analysis showed that the FWI thresholds proposed by EFFIS for assessing the level of fire danger in Europe are too low for the case of Greece and, therefore, are not representative of the country's fire weather conditions. Two statistical approaches, cluster analysis and non-linear least-squares regression, are subsequently applied to determine the most appropriate FWI thresholds for discrete levels of wildfire danger. The results are presented in 4 sets of FWI thresholds and they are further evaluated on the basis of verification measures. All sets of FWI thresholds were found to significantly improve the predictability of wildfire danger compared to the EFFIS fire danger class thresholds. In particular, two of them were found to meet selected performance requirements for a balanced predictive performance, namely a reduced overestimation of wildfire danger and increased reliability in danger classification. The results are expected to have significant practical implications for wildfire prevention and risk mitigation strategies implemented by the forest fire control agencies of the country.

中文翻译：

---

地中海地区与天气相关的野火危险阈值：以希腊为例

摘要 以下论文的目的是根据当地天气条件确定与天气相关的野火危险阈值，即火灾的潜在范围。目标地区是希腊，这是一个容易发生野火的地中海国家，在过去的二十年里，该国平均每年发生 2000 起野火。首先，评估欧洲森林火灾信息系统 (EFFIS) 采用的加拿大火灾天气指数 (FWI) 系统 (FWI 系统) 的火灾天气指数 (FWI) 组件，评估其野火危险预测能力。因此，利用了 2000 年至 2016 年期间市级和每日的天气和野火数据。分析表明，EFFIS 提出的用于评估欧洲火灾危险水平的 FWI 阈值对于希腊来说太低了，因此，不代表该国的火灾天气情况。随后应用聚类分析和非线性最小二乘回归两种统计方法来确定最合适的野火危险离散级别的 FWI 阈值。结果显示在 4 组 FWI 阈值中，并在验证措施的基础上进一步评估。与 EFFIS 火灾危险等级阈值相比，所有 FWI 阈值集都被发现显着提高了野火危险的可预测性。特别是，发现其中两个满足平衡预测性能的选定性能要求，即减少对野火危险的高估和增加危险分类的可靠性。