Understanding and predicting forest fires have proved a highly difficult endeavour, which requires extending and adapting complex models used in different fields. Here, we apply a marked point process approach, commonly used in ecology, which uses multiple Gaussian random fields to represent dynamics of Mediterranean forest fires in a spatio-temporal distribution model. Inference is carried out using Integrated Nested Laplace Approximation (INLA) with inlabru, an accessible and computationally efficient approach for Bayesian hierarchical modelling, which is not yet widely used in species distribution models. Using the marked point process approach, intensity of forest fires and dispersion were predicted using socioeconomic factors and environmental and fire-related variables. This demonstrates the advantage of complex model components in accounting for spatio-temporal dynamics that are not explained by environmental variables. Introduction of spatio-temporal marked point process can provide a more realistic perspective of a system, which is of particular importance for a practical and impact-focused worldwide problem such as forest fires.

Supplementary materials accompanying this paper appear online.

事实证明，理解和预测森林火灾是一项非常困难的工作，这需要扩展和适应不同领域使用的复杂模型。在这里，我们应用了生态学中常用的标记点过程方法，该方法使用多个高斯随机场在时空分布模型中表示地中海森林火灾的动态。使用集成嵌套拉普拉斯逼近 (INLA) 和inlabru进行推理，inlabru 是一种易于使用且计算高效的贝叶斯分层建模方法，尚未在物种分布模型中广泛使用。采用标记点过程方法，利用社会经济因素以及环境和火灾相关变量来预测森林火灾的强度和扩散。这证明了复杂模型组件在解释无法由环境变量解释的时空动态方面的优势。引入时空标记点过程可以提供更现实的系统视角，这对于森林火灾等实际且注重影响的全球性问题尤为重要。

本文附带的补充材料出现在网上。