Imaging the Orientation Dynamics of Snow in Freefall from a Hovering Microscopy Platform

From single crystal formation high in the atmosphere down to precipitating snowfalls at ground level, no snowflake takes the same path through the air column. During descent, snow crystals grow, aggregate, break, and rime into graupel while interacting with the surrounding air. Among the well-studied effects of temperature and humidity super-saturation, the specific role of the various turbulence activities throughout the atmosphere remains elusive. In this work, we utilize a novel, flexibly deployable, airborne microscopy platform mounted on an uncrewed aerial vehicle for in-situ imaging of snowflakes up to 120 Meters above ground level during their most ‘turbulent end-of-lifetime’ as they descend through the atmospheric surface layer. Our platform mounts an Infinity K2 DistaMax long-range microscope combined with powerful pulsed LED illumination and a LI-550 TriSonica Mini sonic anemometer for wind characterization on a DJI Matrice 600 Pro hexacopter capable of carrying a 5.5 Kilogram payload. The resolving power of the optical systems allows us to collect 38-Micrometer diffraction-limited high-resolution imagery of snowflakes in freefall at 3 Meters distance, well outside of the drone’s aerodynamic envelope in hovering flight. We will present the first data captured at the start of the 2025 snow season, performed at a professional meteorological field site for cross-validation. Running our system at a 10 Hertz acquisition frequency, we collect a large data sample of 1’500 snowflakes using an online image acceptance and rejection over a relatively small observation volume of approximately 30 cubic Centimeters. Further refining our data sample to 200 best in focus snowflakes, initial data analyses reveal a large variety in snowflake morphology, including dendrite crystals, aggregates, and apparent riming. Extracting morphological metrics of size, aspect ratio, orientation angle, and complexity, we then sort the data and plot statistical distributions. In particular, our data reveals a predominance in horizontal fall orientation, which we discuss in relation to the wind vector.