← Back to Explore

Predicting Next-Day Wildfire Spread with Environmental Data and Machine Learning

JSHS · 2024

Overview

Wildfires present a growing challenge, increasingly threatening communities and ecosystems worldwide. The rising wildfire incidents amplify environmental and social risks, emphasizing the need for innovative approaches in fire management. Traditional fire prediction models like FSPro, BehavePlus, FARSITE, and FlamMap have limitations in handling complex wildfire behaviors. AI models, in comparison, hold the potential for enhancements in early detection, real-time support, and wildfire management. This project uses AI to predict next-day wildfire spread. The “next-day” timeframe balances computational efficiency with actionable insights. Two machine learning models, UNET and Random Forest, were compared. Feature analysis was conducted to identify the most important features for both models. The UNET, which could interpret spatial relationships between pixels, outperformed the Random Forest in all metrics except AUC, supporting our first hypothesis. It also outperformed the convolutional autoencoder created by the Google developers of the Next Day Wildfire Spread dataset in precision and recall, highlighting its effectiveness. Our second hypothesis was rejected. Feature importance varied; the previous fire mask was most important for the UNET, but only third most important for the Random Forest. Additionally, the impact of feature addition differed between models, suggesting that focusing on the most informative features could be more effective than collecting all available data. Our models were able to predict a wildfire’s next-day spread patterns, offering a potential improvement over current prediction methods. Further refinement and application could aid firefighters in allocating resources to high-risk areas, leading to more e ffective containment efforts, and enable authorities to make timely evacuation decisions.

Competition history

  • JSHS 2024 Category not listed

Resources

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Source: Junior Science and Humanities Symposium

Save projects to your library

Sign in with Google to keep track of projects you find interesting, organized into folders. An account also raises your daily allowance for “Has this been done?”, and lets you create a key for the MCP server with a much higher limit than anonymous use. Browsing stays public.

Continue with Google