Abstracts Geography

Testing and Applying Methods to Assess Snow Interception at a Subalpine Site

by Natasha Harvey

Snow interception and the sublimation of intercepted snow are important processes in subalpine forests. Direct interception measurements are difficult beyond the point scale of a single suspended tree, resulting in the use of indirect methods to assess interception, including eddy covariance instruments, time-lapse photography, and models. Here timelapse photography was leveraged to train a convolution neural network. These methods were tested over a single winter month (January 2019) at the Niwot Forest AmeriFlux site in the Colorado Front Range, with results compared to labeled images and a common statistic (balanced accuracy) calculated. The most promising methods during testing were then applied to winter 2018/2019 (1st October 2018 – 31st May 2019). Interception occurred between 40% and 76% of winter days, based on the convolution neural network and eddy covariance measurements, respectively, with model simulations giving results between these values. Model simulations also suggested that around 60% of snowfall was intercepted. Slightly under a quarter of total snowfall was lost to sublimation over the winter, with 16% of snowfall sublimating from intercepted snow. All methods provided insight into the process of interception in a subalpine forest but did present challenges, including the need to limit latent heat flux measurements of sublimation to when the forest was not transpiring and the inability to monitor snow interception at night with red-green-blue imagery.