Unsupervised classification of snowflake images using a generative adversarial network and <i>K</i>-medoids classification

Abstract

Abstract. The increasing availability of sensors imaging cloud and precipitation particles, like the Multi-Angle Snowflake Camera (MASC), has resulted in datasets comprising millions of images of falling snowflakes. Automated classification is required for effective analysis of such large datasets. While supervised classification methods have been developed for this purpose in recent years, their ability to generalize is limited by the representativeness of their labeled training datasets, which are affected by the subjective judgment of the expert and require significant manual effort to derive. An alternative is unsupervised classification, which seeks to divide a dataset into distinct classes without expert-provided labels. In this paper, we introduce an unsupervised classification scheme based on a generative adversarial network (GAN) that learns to extract the key features from the snowflake images. Each image is then associated with a distribution of points in the feature space, and these distributions are used as the basis of K-medoids classification and hierarchical clustering. We found that the classification scheme is able to separate the dataset into distinct classes, each characterized by a particular size, shape and texture of the snowflake image, providing signatures of the microphysical properties of the snowflakes. This finding is supported by a comparison of the results to an existing supervised scheme. Although training the GAN is computationally intensive, the classification process proceeds directly from images to classes with minimal human intervention and therefore can be repeated for other MASC datasets with minor manual effort. As the algorithm is not specific to snowflakes, we also expect this approach to be relevant to other applications.

Arctic observations and numerical simulations of surface wind effects on Multi-Angle Snowflake Camera measurements

“…Variables derived from the highresolution images include those describing a hydrometeor's size, shape, fall orientation, and approximate riming degree (Garrett et al, 2012;Garrett and Yuter, 2014;Garrett et al, 2015). As these hydrometeor properties are crucial for accurate numerical modeling and microwave scattering calculations, the MASC has been used at various polar and midlatitude locations to constrain microphysical characteristics (e.g., Grazioli et al, 2017;Dunnavan et al, 2019;Jiang et al, 2019;Vignon et al, 2019), improve radar-based estimates of snowfall rates (Gergely and Garrett, 2016;Cooper et al, 2017;Schirle et al, 2019), automatically classify hydrometeors (Praz et al, 2017;Besic et al, 2018;Hicks and Notaroš, 2019;Leinonen and Berne, 2020;Schaer et al, 2020), reconstruct particle shapes (Notaroš et al, 2016;Kleinkort et al, 2017) and size distributions (Cooper et al, 2017;Huang et al, 2017;Schirle et al, 2019), and as ground truth comparisons for radar measurements (Bringi et al, 2017;Gergely et al, 2017;Matrosov et al, 2017;Kennedy et al, 2018;Oue et al, 2018;Matrosov et al, 2019). Unlike more common precipitation gauges, the wind velocity field in the proximity of the MASC has not been simulated for various surface wind speeds, directions, or turbulence kinetic energies (TKEs).…”

“…Studies of hydrometeor behaviors using the MASC have shown, somewhat surprisingly, that frozen hydrometeor fall speeds are only weakly dependent on their size or shape, particularly under conditions of high turbulence intensity (Garrett and Yuter, 2014). Prior studies had shown a much stronger dependence but had theoretically assumed or experimentally arranged for falling hydrometeors to settle in still air (Locatelli and Hobbs, 1974;Böhm, 1989). MASC measurements led to a hypothesis that snow "swirls" in turbulent air in a manner that spreads particle fall speeds to both higher and lower values (Garrett and Yuter, 2014) -an effect shown in prior work to be non-negligible in turbulent flows (Nielsen, 2007).…”

Reconstruction of the mass and geometry of snowfall particles from multi-angle snowflake camera (MASC) images

“…More recently, accurate and high-resolution depictions of snowflakes could be obtained with imagers like the snow video imager/particle image probe (Newman et al, 2009) or with the multi-angle snowflake camera (MASC; Garrett et al, 2012). The availability of actual images has promoted the development and rapid improvement of several automatic hydrometeor classification techniques (Grazioli et al, 2014;Gavrilov et al, 2015;Praz et al, 2017;Leinonen and Berne, 2020) adapted to the data of these sensors. While the accuracy of the measurements of fall velocity provided by those instruments is often hampered by wind and turbulence (Nešpor et al, 2000;Garrett and Yuter, 2014;Fitch et al, 2021), the added value in terms of microphysical characterization is significant.…”

“…In this article we present a method, based on a generative adversarial network (GAN), to retrieve the three-dimensional distribution of mass of individual snowflakes using as input the two-dimensional triplet of images collected by a MASC. GANs are nowadays finding application in the field of environmental and atmospheric sciences (e.g., Leinonen and Berne, 2020;Leinonen et al, 2021) thanks to their versatility, and their ability to perform 3D reconstruction of images has been already explored, for example, in the medical field (Yang et al, 2017). The GAN presented here is trained on a set of simulated snowflakes (generated using the technique of Leinonen et al, 2013;Leinonen and Moisseev, 2015;Leinonen and Szyrmer, 2015;Karrer et al, 2020) and evaluated on 3D-printed 1 : 1 scale snowflake replicas repeatedly dropped into the MASC sampling area.…”

Numerical simulations and Arctic observations of surface wind effects on Multi-Angle Snowflake Camera measurements

“…Variables derived from the high-resolution images include those describing a hydrometeor's size, shape, fall orientation, and approximate riming degree (Garrett et al, 2012;Garrett and Yuter, 2014;Garrett et al, 2015). As these hydrometeor properties are crucial for accurate numerical modeling and microwave scattering calculations, the MASC has been used at various polar and mid-latitude locations to constrain microphysical characteristics (Garrett et al, 2012;Garrett and Yuter, 2014;Garrett et al, 2015;Grazioli et al, 2017;Kim et al, 2018;Dunnavan et al, 2019;Jiang et al, 2019;Kim et al, 2019;Vignon et al, 2019), improve radar-based estimates of snowfall rates (Gergely and Garrett, 2016;Cooper et al, 2017;Schirle et al, 2019), automatically classify hydrometeors (Praz et al, 2017;Besic et al, 2018;Hicks and Notaroš, 2019;Leinonen and Berne, 2020;Schaer et al, 2020), reconstruct particle shapes (Notaroš et al, 2016;Kleinkort et al, 2017) and size distributions (Cooper et al, 2017;Huang et al, 2017;Schirle et al, 2019), and as ground truth comparisons for radar measurements (Bringi et al, 2017;Gergely et al, 2017;Matrosov et al, 2017;Kennedy et al, 2018;Oue et al, 2018;Matrosov et al, 2019). Unlike more common precipitation gauges, the wind velocity field in the proximity of the MASC has not been simulated for various surface winds speeds, directions, or turbulence kinetic energies (TKE).…”