Presentation Information
[ACC45-P03]Development of an Automated Snow Crystal Imaging System and Preliminary Analysis of the 2025–2026 Winter Observations
*Keita Sumiya1, Souya Kuramoto1, Kazue Suzuki1, Hidehiko Suzuki1 (1.Meiji University)
Keywords:
Snow Crystal
The crystal habit of snowflakes reflects their growth environment, i.e., upper-atmospheric meteorological conditions. Since Bentley’s early photographic documentation and Nakaya’s laboratory reproduction of artificial snow crystals leading to the Nakaya diagram [1], followed by refinements such as the Kobayashi diagram [2], snow crystal morphology has been systematically related to temperature and supersaturation. These studies suggest that upper-atmospheric conditions may be inferred from observed snow crystal habits.
The objective of this study is to develop a compact belt-conveyor-type imaging system capable of automatically and continuously recording snow particle habits and sizes, and to establish an observational methodology for estimating upper-atmospheric conditions from the acquired data.
Existing direct observation systems include (1) the Multi-Angle Snowflake Camera (MASC) [3], (2) the Kanto Snow Crystal Project [4], and (3) belt-conveyor-type snowfall imaging systems [5, 6]. Our group has also conducted in-air imaging observations of falling snow crystals using a high-sensitivity camera [7]. However, these approaches face challenges such as difficulty in (1) capturing particles with different fall velocities, (2) observational bias toward beautiful crystals, and (3) operational limitations, including fixed installations or the need for periodic reconfiguration and post-event data retrieval.
To address these issues, we developed a compact and portable system designed for stable operation and automated data acquisition. The system aims to record snow particles and their size distributions comprehensively and objectively. The ultimate goal is to estimate upper-atmospheric conditions through future machine-learning-based classification of snow crystal habits.
The current system measures 48 × 22 × 20 cm (38 cm in height when mounted). It consists of an aluminum frame with insulated housing and includes a camera, LED illumination, belt conveyor, sensors (temperature, humidity, pressure, thermocouple), heater, brush mechanism, control boards (Raspberry Pi and slee-Pi), and a power supply (Fig. 1). Snow crystals are transported at fixed intervals for imaging and subsequently removed by brushes, enabling continuous automated observation. Image data are uploaded sequentially to a NAS, allowing remote monitoring when power and network access are available. The system can be rapidly deployed when snowfall is forecast.
The system development was reported at the 2025 MSJ Autumn Meeting [8]. Since January 2026, field observations have been conducted in Kanagawa, Shizuoka, and Fukui Prefectures. Examples of acquired images are shown in Fig. 2. Portability enabled documentation of regional variability, while continuous imaging captured temporal changes in dominant crystal habits (e.g., transitions from needle-like to dendritic crystals) and variations in snowfall intensity.
This presentation reports operational challenges identified during field observations and introduces the acquired dataset together with its preliminary analysis. The analysis will focus on dominant crystal habit distributions at each observation time and SWE-based precipitation intensity.
[1] Nakaya, U., 1955, Journal of the Faculty of Science, Hokkaido University. Ser.2, Physics,4(6), 341-354.
[2] Kobayashi, T., 1957, Low Temperature Science, Series A (Physical Sciences), 16, 1-26 (in Japanese).
[3] T. J. Garrett, et al., 2012, Atmos. Meas. Tech., 5, 2625–2633.
[4] Araki, K., 2018, Seppyo (Journal of the Japanese Society of Snow and Ice), 80(2), 115–129 (in Japanese).
[5] Motoyoshi, H., 2017, NIED News No.199 (in Japanese).
[6] Shimada, W., et al., 2024, Japanese Society of Snow and Ice (in Japanese).
[7] Kuramoto, S., et al., 2024, 2024 MSJ Autumn Meeting (in Japanese).
[8] Sumiya, K., et al., 2025, 2025 MSJ Autumn Meeting (in Japanese)
The objective of this study is to develop a compact belt-conveyor-type imaging system capable of automatically and continuously recording snow particle habits and sizes, and to establish an observational methodology for estimating upper-atmospheric conditions from the acquired data.
Existing direct observation systems include (1) the Multi-Angle Snowflake Camera (MASC) [3], (2) the Kanto Snow Crystal Project [4], and (3) belt-conveyor-type snowfall imaging systems [5, 6]. Our group has also conducted in-air imaging observations of falling snow crystals using a high-sensitivity camera [7]. However, these approaches face challenges such as difficulty in (1) capturing particles with different fall velocities, (2) observational bias toward beautiful crystals, and (3) operational limitations, including fixed installations or the need for periodic reconfiguration and post-event data retrieval.
To address these issues, we developed a compact and portable system designed for stable operation and automated data acquisition. The system aims to record snow particles and their size distributions comprehensively and objectively. The ultimate goal is to estimate upper-atmospheric conditions through future machine-learning-based classification of snow crystal habits.
The current system measures 48 × 22 × 20 cm (38 cm in height when mounted). It consists of an aluminum frame with insulated housing and includes a camera, LED illumination, belt conveyor, sensors (temperature, humidity, pressure, thermocouple), heater, brush mechanism, control boards (Raspberry Pi and slee-Pi), and a power supply (Fig. 1). Snow crystals are transported at fixed intervals for imaging and subsequently removed by brushes, enabling continuous automated observation. Image data are uploaded sequentially to a NAS, allowing remote monitoring when power and network access are available. The system can be rapidly deployed when snowfall is forecast.
The system development was reported at the 2025 MSJ Autumn Meeting [8]. Since January 2026, field observations have been conducted in Kanagawa, Shizuoka, and Fukui Prefectures. Examples of acquired images are shown in Fig. 2. Portability enabled documentation of regional variability, while continuous imaging captured temporal changes in dominant crystal habits (e.g., transitions from needle-like to dendritic crystals) and variations in snowfall intensity.
This presentation reports operational challenges identified during field observations and introduces the acquired dataset together with its preliminary analysis. The analysis will focus on dominant crystal habit distributions at each observation time and SWE-based precipitation intensity.
[1] Nakaya, U., 1955, Journal of the Faculty of Science, Hokkaido University. Ser.2, Physics,4(6), 341-354.
[2] Kobayashi, T., 1957, Low Temperature Science, Series A (Physical Sciences), 16, 1-26 (in Japanese).
[3] T. J. Garrett, et al., 2012, Atmos. Meas. Tech., 5, 2625–2633.
[4] Araki, K., 2018, Seppyo (Journal of the Japanese Society of Snow and Ice), 80(2), 115–129 (in Japanese).
[5] Motoyoshi, H., 2017, NIED News No.199 (in Japanese).
[6] Shimada, W., et al., 2024, Japanese Society of Snow and Ice (in Japanese).
[7] Kuramoto, S., et al., 2024, 2024 MSJ Autumn Meeting (in Japanese).
[8] Sumiya, K., et al., 2025, 2025 MSJ Autumn Meeting (in Japanese)