4:30 PM - 4:45 PM
[AAS08-10] Characteristics of Hydrometeors in various climatic zones and heights of convection
Keywords:Cloud microphysics, Artificial intelligence
Solid Hydrometeors constitute clouds above 0°C height, and it is known that the growth habits of particles change with temperature and water vapor content (Nakaya 1954, Kobayashi 1961). Since cloud particles are closely related to precipitation, it can be inferred that the distribution and types of cloud particles differ depending on climatic zones and the heights of convection.Therefore, the purpose of this study is to clarify the differences in the characteristics of solid hydrometeors and their factors that depending on climatic zones and the heights of cloud convection from the results of cloud particle observations in the Republic of Palau, Okinawa Prefecture, and Hokkaido, Japan.
Hydrometeor Video sondes (HYVIS , Murakami and Matsuo 1990, Orikasa and Murakami 1997) were used at Rikubetsu Town, Hokkaido (RKB), Republic of Palau (PLW), and Yonaguni Island, Okinawa Prefecture (YNG). Nine HYVIS were launched at Rikubetsu-cho, Hokkaido, Japan from February 8 to 28, 2011 aiming to get hydrometeors in solid phase. Seven were launched at Palau from May 28 to July 1, 2013, for tropical cumulonimbus, and 10 were launched at Yonaguni Island, Okinawa, Japan from June 7 to July 2, 2023 for Baiu fronts. In this study, we extracted nine from RKBs, six from PLWs, and one from YNG. For RKB and PLW observations, we used an AI-based objective Hydrometeor classification system (Yoshimura et al., 2022 Spring Meeting of Meteorological Society) for all observed frames, and only microscopic images for every 10 images from YNG observations. The AI classified particles into eight categories: plate, column, column-plate, bullet, supercooled water, frozen droplets, aggregates which is overlapped or sticked particles, and irregulars that not classified in any of the above categories, then the results were compared. The case of YNG is a convective precipitation system with weak precipitation at the observation point. In this case, supercooled water droplets with a few concentrations of nearly 70,000[1/m3] were observed in the lower 5-5.5 km, plate were observed at an altitude of about 10 km, and irregularly shaped particles were observed up to 13 km. None of columns or bullets were observed. The area from the ground surface to an altitude of 3 km was very wet. In the PLW observations, the number concentration was considerably higher than that of the YNG, ranging from 300000 to 2300000[1/m3] at maximum. Moreover, 7-8 types of particles were observed. The highest altitude at which particles were observed was quite high, ranging from 13 to 15 km. In the RKB observations, there were many columnar particles, and some of them were large enough to be clearly seen in a close-up camera.
The AI-based objective hydrometeor classification system uses as training data the observation results from the Republic of Palau for solid particles and from Okinawa Island for supercooled waters, but the YNG system showed fewer misdetections / misclassifications than the visual verification. On the other hand, the RKB observations had about twice as many misclassifications as those of YNG. Because the particle sizes of supercooled droplets are smaller than those of Palau and Okinawa Island, they were often misidentified as plates due to the unclear contours and contrasts of overlapping supercooled waters. It can be inferred that there are differences in the shape and color between the particles captured as training data and those observed by RKB, but the specific differences and their factors will be investigated in detail in the future
Hydrometeor Video sondes (HYVIS , Murakami and Matsuo 1990, Orikasa and Murakami 1997) were used at Rikubetsu Town, Hokkaido (RKB), Republic of Palau (PLW), and Yonaguni Island, Okinawa Prefecture (YNG). Nine HYVIS were launched at Rikubetsu-cho, Hokkaido, Japan from February 8 to 28, 2011 aiming to get hydrometeors in solid phase. Seven were launched at Palau from May 28 to July 1, 2013, for tropical cumulonimbus, and 10 were launched at Yonaguni Island, Okinawa, Japan from June 7 to July 2, 2023 for Baiu fronts. In this study, we extracted nine from RKBs, six from PLWs, and one from YNG. For RKB and PLW observations, we used an AI-based objective Hydrometeor classification system (Yoshimura et al., 2022 Spring Meeting of Meteorological Society) for all observed frames, and only microscopic images for every 10 images from YNG observations. The AI classified particles into eight categories: plate, column, column-plate, bullet, supercooled water, frozen droplets, aggregates which is overlapped or sticked particles, and irregulars that not classified in any of the above categories, then the results were compared. The case of YNG is a convective precipitation system with weak precipitation at the observation point. In this case, supercooled water droplets with a few concentrations of nearly 70,000[1/m3] were observed in the lower 5-5.5 km, plate were observed at an altitude of about 10 km, and irregularly shaped particles were observed up to 13 km. None of columns or bullets were observed. The area from the ground surface to an altitude of 3 km was very wet. In the PLW observations, the number concentration was considerably higher than that of the YNG, ranging from 300000 to 2300000[1/m3] at maximum. Moreover, 7-8 types of particles were observed. The highest altitude at which particles were observed was quite high, ranging from 13 to 15 km. In the RKB observations, there were many columnar particles, and some of them were large enough to be clearly seen in a close-up camera.
The AI-based objective hydrometeor classification system uses as training data the observation results from the Republic of Palau for solid particles and from Okinawa Island for supercooled waters, but the YNG system showed fewer misdetections / misclassifications than the visual verification. On the other hand, the RKB observations had about twice as many misclassifications as those of YNG. Because the particle sizes of supercooled droplets are smaller than those of Palau and Okinawa Island, they were often misidentified as plates due to the unclear contours and contrasts of overlapping supercooled waters. It can be inferred that there are differences in the shape and color between the particles captured as training data and those observed by RKB, but the specific differences and their factors will be investigated in detail in the future