A proposal for a Wildfire Digital Twin Framework through Automatic Extraction of Remotely Sensed Data: the Italian Case Study of the Susa Valley

Wildfires are among the most common forms of natural disaster in many regions. Climate change has increased the probability and severity of wildfires across Europe, fueled by heatwaves and facilitated by drought which is uncommon since the past 500 years. Forest fires in 2017 have been recorded as one of the worst one during the last 30 years for Italy as vast fire occurrences have taken place during summer, adding to the already important events in the autumn season. In 2017, almost the whole Italian National territory suffered a period characterized by very high temperatures and absence of significant rain processes. shaping to become a recurrent problem in the future and creating a need for better integration in the management and information sources, possibly accessible in real time. This thesis proposes a framework based on existing literature for wildfire assessment using Digital Twin (DT) technology combining remote sensing open source data and cloud computing techniques to create virtual representations of real-world environments to model the extent of the damage and help with recovery and reconstruction efforts. The focus of the research is to develop and generate the basic data necessary for the creation of a wildfire assessment DT on the Susa valley such as thematic maps, generated from the processing of Copernicus Sentinel2 imagery, of burned area, burn severity, vegetation recovery index and their relative statistics using GEE platform to automate the process as much as possible. The resulting burned area extracted using the developed methodology demonstrates conformity to the reference sources (such as the European Service of Copernicus Emergency Management Rapid Mapping product) and more accuracy compared to other open source products in GEE. Moreover, the vegetation recovery obtained by the research highlighted the correct trend of recovery over the years. False values have been detected as well, for which complementary methods have been adopted to help with the interpretation. Further outputs of the thesis include the provision of other building blocks for the wildfire assessment DT such as potential stakeholders in the case of Susa Valley and sensors to cover other aspects of the damage assessment. Overall, the adoption of Sentinel2 imagery provided by the GEE platform proves to be beneficial for the context of research, however there are some still open challenges related to dataset unavailability and inaccuracy, or memory management limitations due to hardware storage capacities, additionally to the human expertise and complementary sources such as field observations and complementary usage of sensors, that would be needed for further accuracy and improvement.