Predicting fire hazards on railroad track embankments with machine learning – Burn ML.
The It4 Forest team is proud to start a new project on fire prediction – an issue with high relevance. In Germany, the number of fires along railways tracks is increasing – as a result of climate change, with sometimes massive consequences for operations and affected infrastructure. Due to a lack of systematic investigations, little is known what causes such fires. As these hazards are expected to further to increase, in-depth knowledge of fire triggers and causes, as well as methodological knowledge of prediction tools, is needed.
The new project Burn ML starting 12/2022 addresses this gap, studies causes and develops methodological knowledge for prediction tools.
The project objectives are:
- to develop a data set for training a model, including data on past slope fires;
- to train a machine learning model for the prediction of fire hazards;
- to test a pilot application of the model with DB Netz AG.
The activities of the project include the spatial identification of past embankment fires by analysing free satellite data on fires and land use and by linking this analysis to weather data from the German WD weather service (DWD) and infrastructure data from DB Netz. The data will then be recombined and analysed to identify potential initiating and contributing factors for embankment fires. Based on this preliminary work, the project team selects an algorithm and trains it to predict such fires. The testing of this algorithm with DB Netz AG and the accompanying discussion with stakeholders complete the implementation of this project.
The project will publish a documented machine learning classifier (in Python), a technical article with findings on triggers and drivers, and a prediction tool for slope fire risks. The results of the project will expand the knowledge of embankment fires and prevention options and enable infrastructure companies to take precautions more efficiently.
The project is implemented by a team from Adelphi and the University of Sustainable Development Eberswalde (HNEE). The project is funded by a grant from the mFund funding programme from the German Federal Ministry of Digital Affairs and Transport (BMDV).