Large-Scale Volumetric Imaging and Tracking of Natural Snowfall

The interactions between large collections of settling snowflakes and various turbulence intensity levels over the air column make snow precipitation difficult to forecast. Characterizing the multi-scale spatial distribution and transport of snowflakes is crucial for understanding the spatial modulations of snow deposition and interpreting remote sensing signals. We perform three-dimensional tracking of natural snowflakes falling in the atmospheric surface layer in the Swiss Alps. We utilize a novel super-resolution field imaging system that combines 16 cameras arranged in 2-by-2 arrays to image a volumetric domain of 10x10x10 cubic meters. Analysing the point-cloud data using Voronoi tessellation, we find a predominance of clusters and voids compared to a random Poisson process. Leveraging the simultaneous tracking of millions of falling snowflakes, we then identify and track snow clusters from regions of relatively high snow concentration. We discuss the specific challenges in processing such large-scale data, particularly the need to discern snowflake images of different sizes through the large depth of field.