[M-IS15] Mountain Science

[MIS15-07] An automated method for alpine weather monitoring using time-lapse cameras and machine learning

*Ryotaro Okamoto¹, Hiroyuki Oguma², Takashi Hamada³ (1.University of Tsukuba, 2.National Institute for Environmental Studies, 3.Nagano Environmental Conservation Research Institute)

Keywords:Alpine weather, Machine learning, Time-lapse camera

Alpine ecosystems are one of the most vulnerable to ongoing climate change. Monitoring meteorological changes of alpine ecosystems are important for conserving the ecosystem and understanding the mechanism of ecosystems' responses to climate change. Local and temporal fogs are characteristic meteorological events in alpine regions. Foggy, cloudy weather might affect the behavior of animals and the photosynthesis of plants. However, few quantitative studies have been conducted due to the difficulties of long-term visual observation. In this study, we are developing an automatic method for classifying weathers from time-lapse cameras to grasp the spatio-temporal patterns of alpine fogs and clouds. Using ground-based time-lapse cameras has several benefits. 1. unlike airborne or satellite images, we can easily distinguish fogs and clouds. 2. the monitoring cost is meager. During the procedure, we split an image into patches and then classified the weather of each patch using a neural net. We used a pre-trained neural net for better performance. Applying the developed method to the existing time-lapse cameras, we expect to learn about the spatio-temporal distribution of clouds and fog in the alpine ecosystems.