Deep Learning for wildfire spread modeling
The thesis aims at developing a Deep Learning model to predict wildfire spread using multimodal and multivariate data.
Requirements
- M.Sc. in Machine Learning, Data Science, Computer Science, Mathematics, Telecommunications, or similar
- Knowledge of Python
- Software development skills
- Basic concepts of data science, concerning data analysis, processing and machine learning
Description
Wildfires are one of the most dangerous natural disasters, causing damage to the environment, infrastructure, and human lives. Predicting how a wildfire will spread is very difficult, because it depends on many factors such as vegetation, weather conditions, terrain, and past fire behavior. Being able to predict wildfire spread in advance is extremely important to support emergency response and risk prevention.
This thesis aims to develop a Deep Learning model to predict wildfire spread using multimodal and multivariate data, including satellite imagery, meteorological variables, land cover information, and historical wildfire data. As ground truth, the work will use a high-quality wildfire spread dataset derived from Copernicus services (ESSD, 2023).
The thesis includes the following activities: (i) Identification and collection of wildfire-related datasets (ii) Short review of existing wildfire spread prediction methods (iii) Data preprocessing and multimodal data integration (iv) Design and training of a Deep Learning model for wildfire spread prediction (v) Evaluation of the model on real wildfire events (vi) BONUS: identification of high-risk or fast-spreading areas