Wildfires constitute an extremely serious social and environmental issue in the Mediterranean region, with impacts on human lives, infrastructures and ecosystems. It is therefore important to produce susceptibility maps for wildfire management. The wildfire susceptibility is defined as a static probability of experiencing wildfire in a certain area, depending on the intrinsic characteristics of the territory. In this work, a machine learning model based on the Random Forest Classifier algorithm is employed to obtain national scale susceptibility maps for Italy at a 500 m spatial resolution. In particular, two maps are produced, one for each specific wildfire season, the winter and the summer one. Developing such analysis at the national scale allows for having a deep understanding on the wildfire regimes furnishing a tool for wildfire risk management. The selected machine learning model is capable of associating a data-set of geographic, climatic, and anthropic information to the synoptic past burned area. The model is then used to classify each pixel of the study area, producing the susceptibility map. Several stages of validation are proposed, with the analysis of ground retrieved wildfire databases and with recent wildfire events obtained through remote sensing techniques.

Background Wildfires are a growing threat to many ecosystems, bringing devastation to human safety and health, infrastructure, the environment and wildlife. Aims A thorough understanding of the characteristics determining the susceptibility of an area to wildfires is crucial to prevention and management activities. The work focused on a case study of 13 countries in the eastern Mediterranean and southern Black Sea basins. Methods A data-driven approach was implemented where a decade of past wildfires was linked to geoclimatic and anthropic descriptors via a machine learning classification technique (Random Forest). Empirical classification of fuel allowed linking of fire intensity and hazard to environmental drivers. Key results Wildfire susceptibility, intensity and hazard were obtained for the study area. For the first time, the methodology is applied at a supranational scale characterised by a diverse climate and vegetation landscape, relying on open data. Conclusions This approach successfully allowed identification of the main wildfire drivers and led to identification of areas that are more susceptible to impactful wildfire events. Implications This work demonstrated the feasibility of the proposed framework and settled the basis for its scalability at a supranational level.

Wildfires are a menace which is growing in intensity and spreading in range across all planet’s ecosystems causing devastation on the environment, wildlife, human health, and infrastructure. Most of the damage caused by forest fires is related to extreme wildfire events (EWEs). To foster prevention activities, a thorough understanding of territorial features determining EWEs is crucial in Civil Protection and fire management activities. An approach which learns from past wildfire events providing susceptibility, intensity and hazard maps is presented. This mapping approach leads to the individuation of the main drivers of EWEs and in the zonation of the areas more prone to hazardous and impactful wildfire events. The case study where the mapping approach is applied encompasses thirteen countries of the Eastern Mediterranean and Southern Black Sea basins. The presented results focus on wildfire susceptibility. A Machine Learning approach is pursued, by adopting open data layers as both predisposing factors and past wildfire events. In particular, the role of vegetation continuity in determining the occurrence of EWEs is assessed.

<p>Comparison of Different Algorithms and Vegetation Classes’ Importance Ranking in Wildfire Susceptibility Maps. <br />Wildfire Susceptibility Maps (WSM) and the analysis of the explanatory variables affecting the model’s predictions are innovative tools to support forest protection and management plans. Namely, WSM identify areas subject to wildfire, in terms of relative spatial likelihood, on the base of the observed past events, stored in spatio-temporal inventories, and on the local environmental and anthropogenic properties of an area. Approaches based on Machine Learning (ML) are particularly suited for WSM since they are capable to make predictions on data by modelling the hidden and non-linear relationships between a set of input variables and the output observations.<br />In the present work, Authors continue a research framework developed at local scale for Liguria Region, and lately improved at national scale (Italy), consisting in the implementation of a ML-approach, based on the algorithm Random Forest, allowing to assess the susceptibility to wildfires under the influence of different variables (e.g., land cover, vegetation classes, altitude and its derivatives, nearby infrastructures). In the present study the following improvements are introduced: (i) to evaluate which ML-algorithm performs better in terms of prediction capabilities we compared Random Forest (RF), Multi-layer Perceptron (MLP), and Support Vector Machine (SVM); (ii) to evaluate the impact of different classes of local and neighbouring vegetation on wildfires occurrence we used of a more accurate map of vegetation as input local explanatory variable; (iii) to consider both the spatial and the temporal variability of the burning seasons (summer and winter) we improved the selection of the testing dataset, based on a clustering approach. <br />The output probabilistic predicted values resulting from the different ML-algorithms (RF, MLP, and SVM) allowed to elaborate the seasonal WSMs. Finally, the spatial distribution of the more susceptible areas will be presented. The performance of the three ML-algorithms was assessed by means of the AUC (Area Under the Curve) ROC (Receiver Operating Characteristics), evaluated over the testing dataset. In addition, the variable importance ranking was estimated as by-product of RF, which can handle both the typical numerical variables and native categorical variables (as for the classes of vegetation at pixel level). Vegetation resulted by far to be the most important explanatory variables; the marginal effect of each single class of vegetation was also assessed and the results will be discussed. <br />Reference <br />Trucchia, A.; Izadgoshasb, H.; Isnardi, S.; Fiorucci, P.; Tonini, M. Machine-Learning Applications in Geosciences: Comparison of Different Algorithms and Vegetation Classes’ Importance Ranking in Wildfire Susceptibility. Geosciences 2022, 12, 424. https://doi.org/10.3390/geosciences12110424</p>

No abstract