Student Research Reports

View Research Report

Wildfires pose an increasingly significant threat to natural ecosystems, private property, and public health. Precautionary methods can be taken to mitigate the possibility of a large-scale wildfire. Our research aims to inform citizens of wildfire risk factors on their properties by developing a mobile application, C.S.-F.L.A.R.E (Citizen Science: Fire Likelihood and Risk Evaluation), that analyzes wildfire risk on a local community level based on a culmination of factors that compose a typical wildfire. C.S.-F.L.A.R.E. utilizes YOLOv8, a state-of-the-art image segmentation model to identify the presence of flammable materials in the four cardinal directions at each Area of Interest (AOI) in the GLOBE Observer database. We use Google Teachable Machine to train a machine learning model to classify the downward photo at each AOI by ground moisture. We implement a U-Net model trained with temperature, precipitation, elevation, and slope datasets (Mhawej et al., 2015) from Earth Map to provide users with accurate predictions that are validated by existing wildfire risk assessment datasets. Utilizing our identified fire risk factors and satellite datasets, we have created an algorithm that provides citizens with a quantified wildfire risk rating, while offering comprehensive insights of surrounding risks that can help prevent future wildfires. C.S.-F.L.A.R.E.’s user-friendly application interface, created on Flutter, presents these insights by proposing potential actions to minimize wildfire risk in the pictured area. We hope to integrate this app into local fire departments and environmental agencies so that citizens have a trusted resource to better understand the risk and impact of fire near them.