Date of Award


Document Type


Degree Name

Master of Science in Computer Science


School of Computer Science and Engineering

First Reader/Committee Chair

Salloum, Ronald


This report introduces a thorough analysis of wildfire prediction using satellite imagery by applying deep learning techniques. To find wildfire-prone geographical data, we use U-Net, a convolutional neural network known for its effectiveness in biomedical image segmentation. The input to the model is the Sentinel-2 multispectral images to supply a complete view of the terrain features.

We evaluated the wildfire risk prediction model’s performance using several metrics. The model showed high accuracy, with a weighted average F1 score of 0.91 and an AUC-ROC score of 0.972. These results suggest that the model is exceptionally good at predicting the location of wildfire risks and distinguishing between wildfires and non-wildfires.

The model generally demonstrated solid performance but encountered difficulties in certain aspects. There were instances where its risk level predictions diverged from the ground truth data. This discrepancy could stem from the multifaceted factors of wildfire risk prediction, an area impacted by numerous variables. Therefore, to enhance precision and accuracy, the model necessitates additional fine-tuning.

The report also explores using a class imbalance strategy to address the disparities in data distribution among the different classes. We discuss the inherent challenges in predicting wildfire-prone regions, which provides insights into the complexities of wildfire prediction and management.

This study found that deep learning techniques have resulted in a highly accurate prediction of the risk of wildfires. Despite some shortcomings, the model’s predictions aligned closely with the ground truth data. Therefore, this study suggests that deep learning models could effectively manage and prevent wildfires on a large scale.