Probabilistic Analysis of Extreme Wildfire events in Italy Using Data-Cube Technology

Wildfires are a critical component of natural ecosystems, contributing to biodiversity by shaping habitat structures and promoting species adaptation, but also posing significant risks to human life, infrastructure, and air quality. Wildfires can be characterized by both their impact and the drivers of their occurrence. Historical data exploration is essential for researchers to build data-driven models for wildfire risk assessment and also to capture the characteristics of extreme wildfire events (EWE). Such data may include fire perimeter records, weather observations, vegetation types, and topographic details, all of which contribute to understanding the conditions that lead to extreme fire behavior. 

The first step toward achieving this goal involves establishing a comprehensive data-cube that integrates all relevant datasets for wildfire risk assessment. A data-cube framework simplifies data exploration and querying by organizing static and dynamic data (in terms of time varying) in a structured format. The data-cube stores multi-dimensional arrays, allowing for efficient analysis of spatial and temporal variations in complex datasets. Static data (e.g., digital elevation model) represent constant landscape features, while dynamic data (e.g., relative humidity or temperature) capture temporal variations. Cloud storage solutions are vital for managing the high memory requirements of data-cube structures, enabling cheaper storage and open-source availability.  

The primary aim of this study is to utilize available data-cubes to identify the conditions that characterize EWE across historical records. By analyzing spatial and temporal dynamic data related to both wildfire occurrences and predisposing meteorological factors, we want to find patterns and signatures of extreme wildfires. Furthermore, additional datasets from various domains and resolutions will be structured into a similar data-cube format for broader analysis.  

Focus will be on the Italian peninsula, leveraging on climatic data at a 3 km spatial resolution with hourly temporal intervals (Chapter Dataset, https://doi.org/10.25927/0ppk7-znk14) allowing for detailed capture of conditions surrounding extreme wildfire events. The outcomes of this study will contribute to the development of probabilistic risk assessment models, providing valuable insights for wildfire risk management and mitigation strategies. 

Keywords: Extreme Wildfire Events, Probabilistic Wildfire Risk Assessment, Data-Cube, Meteorological indices in Wildfire Risk Assessment