3.16 Machine-Learning Classification of Blowing Snow in Complex Terrain from Webcam Imagery

Monday, 13 July 2020: 3:15 PM
Virtual Meeting Room
Sarah McCorkle, Univ. of Wyoming, Laramie, WY; and Z. J. Lebo, B. Geerts, R. Capella, E. M. Collins, R. Cox, A. Lyons, M. Brothers, and T. Alcott

Handout (49.3 MB)

Blowing snow causes a safety hazard for many drivers with its ability to create limited visibilities, which is especially a problem along I-80 in Wyoming during the winter months. For Wyoming alone, parts of I-80 have been closed 54 times from October 1, 2019 to February 29, 2020, according to Wyoming Department of Transportation (WYDOT). Blowing snow is difficult for forecasters to predict ahead of time, as much is still unknown about the favorable conditions to create potentially dangerous blowing snow, especially for the complex mountainous terrain of Wyoming along I-80. The main objectives of this work are to improve the prediction of blowing snow by constraining the conditions under which it occurs. To do this, a machine-learning model is created to verify blowing snow events via image classification using neural networks. The model is trained on WYDOT camera images along I-80, classifying them in terms of blowing snow intensity. The goal of this model is to accurately classify blowing snow intensities upon the model seeing a new, real-time WYDOT camera image, which can aid forecasters in identifying blowing snow events in an automated fashion. Moreover, the model can also be supplemented with collocated meteorological data to understand the conditions under which blowing snow is more prevalent and most intense.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner