Presentation Information
[O12-P60]Development of a workflow for forecasting the sea of cloud at Takeda Castle
*Kanon Ichimura1 (1. Ikeda Senior High School Attached to Osaka Kyoiku University)
Keywords:
sea of cloud formation conditions,sea of cloud predictions,Takeda Castle Ruins
[ Purpose of the research ]
Sea of cloud, which is fog that forms in mountainous areas or basins, appears like a sea of cloud when viewed from above or from higher elevations. It is one of the powerful tourism resource, but the fog can also cause traffic disruptions. Thus, predicting the occurrence of fog enables us to mitigate traffic problems and to revitalize the tourism industry. Several portal sites provide sea of cloud forecasts for Takeda Castle Ruins, which is famous for its sea of clouds (Asago City; 2026, Mitsubishi Motors; 2025). However, these forecasts are limited to the autumn, which is known as sea of cloud season, and weekends. Based on these backgrounds outlined above, this researchaims.
(1) To develop a workflow that summarizes the conditions for sea of cloud formation in order to improve prediction accuracy
(2) To conduct sea of cloud predictions throughout the year
(3) To examine recent changes in sea of cloud formation conditions under different scenarios.
[ Materials and Methods ]
The research area was set to the vicinity of Takeda Castle Ruins. The analysis period covered approximately three years, from September 25, 2022 to June 30, 2025. For fog (sea of clouds) prediction, which was conducted at 22:00 on the previous day, the following data were used: the surface weather chart at 18:00 on the previous day, the forecast weather chart for 9:00 on the day, meteorological data from Asago City, radar images of rain clouds over the Kinki region, and weather warnings and alarms for Hyogo Prefecture. Based on these data, a workflow summarizing the conditions for the occurrence of the sea of clouds at Takeda Castle Ruins was developed. The occurrence of the sea of clouds was confirmed using live camera and in-situ observations. The workflow was developed starting in 2023 and it was improved each year. In particular, sea of cloud predictions were conducted for one year (from July 1, 2024 to June 30, 2025) using both the 2024 version of the workflow (ver.2024) and the 2025 version (ver.2025), which classifies conditions into three types: (1) Radiative cooling type (2) Warm air advection type, and (3) Precipitation type. The results were then evaluated (Fig.1).
[ Results and Discussion ]
The prediction results indicate that the performance of ver. 2025 was improved compared with ver. 2024 except for False Alarm Ratio. Accuracy rate increased from 91.4% to 94.4%, Miss Rate decreased from 35% to 18%, and Probability of detection improved from 65% to 82%, while False Alarm Ratio increased from 4% to 7%. The better performance of ver. 2025 compared with ver. 2024 would be originated from the adding “Warm air advection type and Water vapor type” classification. Our analyses also elucidate that sea of clouds occurs throughout the year, with particularly high frequency from September to November, when influenced by high-pressure systems and typhoons (Fig2).Furthermore, our analyses revealed that the Radiative cooling-type from September to November, decreased 12.4% over the past three years (Fig 2).It might be influenced by global warming, which has shortened the autumn season, as well as by the reduced frequency of passing high-pressure systems.
In the future, we should try to improve the workflow further. For this aim, the introduction of conditions such as precipitation and temperature drops in quantitative terms, as well as analyzing the cloud altitude, which are not currently considered, has potential for further improvement. The in-situ measurement of meteorological data at different elevations is required. Furthermore, it is necessary to generalize the findings obtained in this study by conducting analyses over different time periods.
[ Acknowledgments ]
This work is supported by SEEDS program of The University of Osaka (Professor Yousuke Sato, Professor Tomoo Ushio) for advice of this study.
[ References ]
Asago city, 2026, "ASAGO CITY PORTAL SITE ASABURA”
https://www.asabura.jp/unkaiforecast (2026)
MITUBISHI MOTORS, 2025, ”THE WEEKEND EXPLORER”
https://www.mitsubishi-motors.co.jp/special/weekend-explorer/unkai/17.html (2026)
Sea of cloud, which is fog that forms in mountainous areas or basins, appears like a sea of cloud when viewed from above or from higher elevations. It is one of the powerful tourism resource, but the fog can also cause traffic disruptions. Thus, predicting the occurrence of fog enables us to mitigate traffic problems and to revitalize the tourism industry. Several portal sites provide sea of cloud forecasts for Takeda Castle Ruins, which is famous for its sea of clouds (Asago City; 2026, Mitsubishi Motors; 2025). However, these forecasts are limited to the autumn, which is known as sea of cloud season, and weekends. Based on these backgrounds outlined above, this researchaims.
(1) To develop a workflow that summarizes the conditions for sea of cloud formation in order to improve prediction accuracy
(2) To conduct sea of cloud predictions throughout the year
(3) To examine recent changes in sea of cloud formation conditions under different scenarios.
[ Materials and Methods ]
The research area was set to the vicinity of Takeda Castle Ruins. The analysis period covered approximately three years, from September 25, 2022 to June 30, 2025. For fog (sea of clouds) prediction, which was conducted at 22:00 on the previous day, the following data were used: the surface weather chart at 18:00 on the previous day, the forecast weather chart for 9:00 on the day, meteorological data from Asago City, radar images of rain clouds over the Kinki region, and weather warnings and alarms for Hyogo Prefecture. Based on these data, a workflow summarizing the conditions for the occurrence of the sea of clouds at Takeda Castle Ruins was developed. The occurrence of the sea of clouds was confirmed using live camera and in-situ observations. The workflow was developed starting in 2023 and it was improved each year. In particular, sea of cloud predictions were conducted for one year (from July 1, 2024 to June 30, 2025) using both the 2024 version of the workflow (ver.2024) and the 2025 version (ver.2025), which classifies conditions into three types: (1) Radiative cooling type (2) Warm air advection type, and (3) Precipitation type. The results were then evaluated (Fig.1).
[ Results and Discussion ]
The prediction results indicate that the performance of ver. 2025 was improved compared with ver. 2024 except for False Alarm Ratio. Accuracy rate increased from 91.4% to 94.4%, Miss Rate decreased from 35% to 18%, and Probability of detection improved from 65% to 82%, while False Alarm Ratio increased from 4% to 7%. The better performance of ver. 2025 compared with ver. 2024 would be originated from the adding “Warm air advection type and Water vapor type” classification. Our analyses also elucidate that sea of clouds occurs throughout the year, with particularly high frequency from September to November, when influenced by high-pressure systems and typhoons (Fig2).Furthermore, our analyses revealed that the Radiative cooling-type from September to November, decreased 12.4% over the past three years (Fig 2).It might be influenced by global warming, which has shortened the autumn season, as well as by the reduced frequency of passing high-pressure systems.
In the future, we should try to improve the workflow further. For this aim, the introduction of conditions such as precipitation and temperature drops in quantitative terms, as well as analyzing the cloud altitude, which are not currently considered, has potential for further improvement. The in-situ measurement of meteorological data at different elevations is required. Furthermore, it is necessary to generalize the findings obtained in this study by conducting analyses over different time periods.
[ Acknowledgments ]
This work is supported by SEEDS program of The University of Osaka (Professor Yousuke Sato, Professor Tomoo Ushio) for advice of this study.
[ References ]
Asago city, 2026, "ASAGO CITY PORTAL SITE ASABURA”
https://www.asabura.jp/unkaiforecast (2026)
MITUBISHI MOTORS, 2025, ”THE WEEKEND EXPLORER”
https://www.mitsubishi-motors.co.jp/special/weekend-explorer/unkai/17.html (2026)