דוקטורנטים מספרים - Assaf Shmuel
Machine Learning Applications in Wildfire Prediction and Prevention
Assaf Shmuel (Ph.D. candidate)
Supervisor: Prof. Eyal Heifetz
In recent years, several regions on Earth, such as Australia and the Pacific Coast of the United States, have experienced the largest wildfires in their documented history. It is believed that climate change is and will continue to be a key catalyst of this troubling phenomenon. As deforestation also accelerates climate change, controlling wildfires must be a fundamental environmental goal.
Wildfire forecasting is essential for forest management and firefighting. Forest managers can apply preventive measures such as fire breaks and fuel breaks; however, the resources at their disposal are limited and should therefore be directed to the most urgently-needed locations. Precise wildfire forecasting is a basic requirement of successful forest management strategies.
As in many other fields, big data has the potential to significantly improve wildfire predictions. Satellite observations combined with advanced image processing techniques have recently made global datasets with millions of wildfire observations available. The main goal of our research is to use these datasets to improve wildfire predictions using various Machine Learning (ML) models.
We apply several ML models that predict wildfire behavior based on meteorological conditions, fuel loads, and local fire history. The models are trained on global datasets and produce excellent predictions of wildfire occurrence, magnitude, and duration. It is imperative to work towards prevention and mitigation of disasters of the extreme magnitude that we have seen in recent years. Hopefully, this research and others like it can help to avert the tragic loss of human life and irreplaceable biodiversity in the future.
