Machine Learning Applications for Severe Weather Detection and Prediction
Open Access
Author:
Bradley, Kyle J
Area of Honors:
Data Sciences
Degree:
Bachelor of Science
Document Type:
Thesis
Thesis Supervisors:
James Z Wang, Thesis Supervisor John Joseph Hannan, Thesis Honors Advisor
Keywords:
machine learning data science weather
Abstract:
Weather prediction using machine learning attempts to use historical weather data to make predictions about the future. Especially with recent advancements in computer vision algorithms, many visual forms of weather data can be examined under this machine learning framework. Most importantly, increasing the “lead time” of accurate weather predictions is vital in the space of weather forecasting. In this work, the effectiveness of making predictions on simulated data is examined. This thesis shows that there is promise to detection of storms in this simulated setting. The model was able to detect storms accurately in 87% of simulated images. This result shows that making predictions on simulated weather data is an avenue that can be explored further to increase prediction lead time. Given these results and the ability to use a more refined model, it could be possible that accurate predictions can be made using a similar process more than 48 hours in advance.
