Speaker: Sonya Rauschenbach, PhD Student, Atmospheric Science, University of California, Davis

Topic: Prototyping an Image-based Coastal Fog Detection Network Using Camera Locations Along the Oregon Coast

In the Mediterranean climates of the Pacific Northwest, USA, where summers are characterized by high temperatures and low precipitation, summertime coastal fog provides moisture and shading to local ecosystems. Due to the complex and dynamic nature of coastal fog formation, quantifying fog’s role in local weather and water cycles requires a high spatial resolution dataset. Image data from cameras capturing the visible spectrum in red, green and blue bands are common along the coast, for example commercial camera networks that monitor beach and surfing conditions, and traffic cameras that look over coastal highways. We have collected a database of ~75,000 images at seven locations along the Oregon coast, six of which have data for the summers of 2022-2024 and one that has data from 1992-2019. These images present an opportunity to analyze spatial patterns of fog formation, however, there first needs to be a reliable methodology for fog detection using image data. We have tested a few different machine learning classification algorithms on 8,000 labeled photos and are able to classify when an image contains fog with ~85% precision. This study provides insight on which methodology can reliably detect fog and be applied to webcams globally. We will present the results from different ML techniques for fog classification and show initial results of spatiotemporal patterns of fog along the Oregon coast.