Over the last decade, the use of Machine/Deep learning algorithms and methodologies has found widespread application across various domains in wildfire science, with fire occurrence risk prediction being one of the extensively covered areas. Many algorithms and architectures have been explored to solve the problem as a binary classification task focusing mainly on the model’s classification performance through the standard classification metrics. However, in the context of an application for predicting the fire occurrence risk we have to enhance the trust in the model’s output by deriving a well defined fire susceptibility index, as well as interpretable insights for the model’s predictions when applied to real-world datasets. In this manuscript we propose model agnostic solutions towards (a) the production of a probability-based binning of the model’s prediction distribution for generating a scalar wildfire occurrence risk and (b) of a representative sampling for producing local prediction explanations spanning the whole dataset, to tackle the slow processing times of explainable AI frameworks, when generating vast amounts of local interpretations. The objective is to contribute the development of a reliable application predicting next day’s fire risk, facilitating the adoption of ML-based solutions by users who are not experts in data science.

Under the framework of MedEWSa project, areas with different climatic and physiographic conditions, which at the same time face similar hazardous events, are coupled to form twins so as to exchange meaningful and long-term knowledge as well as best practices. The first of those twins consists of the Attica Region in Greece and Ethiopia. Both areas need to remediate against a growing frequency and intensity of wildfires. The existing service capabilities need to be improved so as to enhance decision support by first responders and policy makers.In this presentation the existing services as well as the ongoing developments of the project will be demonstrated.

Wildfires detection is a major issue for authorities. There are various causes of fire events with the most common being human influence. A fire risk prediction model through the analysis of geo-environmental and climate data is important for early warning and fire management. In this work, a dataset from multiple modalities, including road density, travelers, forest-agriculture interface, burned areas from historical fire events, metrological data, land cover, vegetation indices from data cube, is generated. Artificial intelligence and machine learning models can use this multimodal dataset to improve forest fire management.

This work illustrates an integrated methodology for wildfire vulnerability assessment in selected Wildland Urban Interface (WUI) areas, leveraging Web Services, Remote Sensing, Google Earth Engine (GEE) and GIS techniques. The study generates and fuses physical vulnerability factors (fuel, canopy density, DEM derivatives, and remote sensing indices) and socio-economic elements (population age and density, building materials, land values, and Points of Interest). Thenceforth, fire vulnerability maps were created in a GIS environment incorporating data for the most intense fire period recorded in the past 38 years (1984-2021) in the Attica region. Key municipalities that were investigated include Markopoulo Mesogaias, Lavreotiki, Saronikos, Oropos, and Acharnon. As for the findings, they highlight the need for targeted mitigation and community resilience, aiding in identifying vulnerable regions and managing wildfires in Attica's WUI areas.

In this paper, we handle the problem of next day wildfire prediction via the use of machine learning. In contrast to most works in the relevant literature, we set the problem to its realistic basis, with respect to its large scale, the extreme imbalance in the data distribution, the required high spatial granularity of the predictions and the consideration of the strong spatial correlations inherent in the data. We implement a machine learning workflow that exploits Tree Ensemble and Neural Network algorithms, upon which an extensive hyperparameter search procedure is performed, via cross-validation, in order to select a set of effective models that are expected to generalize well on new data. Our experiments on the whole Greek territory demonstrate the effectiveness of the proposed methodology, rendering it directly applicable to real-world scenarios. Finally, several insights towards further improving the effectiveness of current models are discussed.