Climate Change AI Hackathon

By: Antonin Villemin Arsalan Larachi Yacine Benouniche

Context:

• Recent major wildfire in the Amazon rainforest
• Difficulty for firefighters to predict the wildfire spread

Our contribution:

• Lack of data for behavior of wildfire spread in history.
• new approach: predict the wind direction / speed to determine the wildfire propagation in the next two hours.

using the api of Canadian Centre for Climate Services we took the data of Fort Nelson from 2017 to 2019. Example of data for the 13th Septembre 2019: Dataset

Demo:

Folder Content:

• final.csv: dataset
• fire_propagation.py: script simulator
• model.ipynb: wind predictor

Challenges we faced:

• Finding the appropriate data, and the simulation of wildfire propagation.
• Implementing the correct prediction model.

Acquired knowledge

• We learned that we may have great ideas but finding the data that will help us realize our project can be tricky.
• We also enriched our knowledge in project management and machine learning.

Built With

• jupyter-notebook
• keras
• pygame
• python