Presentation Information
[AAS08-P46]Estimation of fog formation and aerosol concentrations by analyzing camera images taken from Mt. Unzen
*Nanami Sumiyoshi1, Masafumi Ohashi1, Tomoki Nakayama2 (1.Graduate School of Science and Engineering, Kagoshima University, 2.Faculty of Environmental Science, Nagasaki University)
Keywords:
aerosol particles,fog,image processing,time-laps camera
Hygroscopic growth of aerosol particles and formation of cloud/fog droplets play important roles in the hydrological cycle; therefore, understanding their occurrences and characteristics is essential. In situ observations of fog droplets in mountainous regions have the advantage of providing continuous data with high temporal resolution. However, due to cost and technical constraints, long-term continuous observations of fog formation processes and properties using high-precision instruments have been conducted at a limited number of sites. In contrast, image acquisition using fixed-point cameras can be conducted easily with low cost across multiple locations. An image analysis technique based on quantitative image degradation index is one of the potential techniques to estimate fog appearance and aerosol concentrations in the atmosphere from camera image. This study aims to develop a method for estimating fog appearance and aerosol concentrations by deriving an image degradation index through comparisons between image analysis results and in situ observational data obtained simultaneously at Mt. Unzen in Nagasaki Prefecture.
Images were acquired every 10 minutes using a time-lapse camera installed at the Unzen–Myoken Observatory (altitude: 1300 m) in Unzen City, Nagasaki Prefecture. In addition, a cloud particle sonde (CPS) was installed at the same site to measure number density and related parameters. In this study, the data obtained between March 2022 and March 2024 were used for analysis. Among the captured images, those in which objects behind the handrail of the observatory were not visible (hereafter referred to as fog images) provided information only about the fog appearance at the observatory, whereas images in which the shore in Amakusa (approximately 30 km from the observatory) was visible (hereafter referred to as non-fog images) were influenced by aerosol loading between the observatory and the shore in Amakusa. We classified the images into fog and non-fog images using the sum of RGB pixel values.
For fog images, we quantified image degradation using the visibility of the handrail at the observatory. The Canny edge detection method with a Gaussian filter size of 5 × 5 pixels was applied to the handrail region of the image, and the amount of detected edges was used as an image degradation index. The correlation between this index and the number of cloud droplets measured by the CPS was found to be high in spring and autumn seasons with Pearson’s correlation coefficients (r) exceeding 0.7.
For non-fog images, we quantified image degradation using the visibility of the shore in Amakusa. The degradation index was defined as the mean Sobel gradient magnitude within a specified region including the shore of the image. For comparison, in-situ observation data on suspended particulate matter (SPM) concentrations at two sites (Shimabara station in Nagasaki prefecture and Amakusa station in Kumamoto prefecture) of the Atmospheric Environmental Regional Observation System of the Ministry of the Environment, Japan, were used. Correlation analysis between SPM concentration and image degradation index revealed seasonal correlation coefficients ranging from 0.41 to 0.78 at Shimabara station, whereas those at Amakusa station remained between 0.31 and 0.41. The results suggest that the SPM data at Shimabara station are more representative of the temporal variation of aerosol particles between the shore in Amakusa and the observatory than those at Amakusa station.
Images were acquired every 10 minutes using a time-lapse camera installed at the Unzen–Myoken Observatory (altitude: 1300 m) in Unzen City, Nagasaki Prefecture. In addition, a cloud particle sonde (CPS) was installed at the same site to measure number density and related parameters. In this study, the data obtained between March 2022 and March 2024 were used for analysis. Among the captured images, those in which objects behind the handrail of the observatory were not visible (hereafter referred to as fog images) provided information only about the fog appearance at the observatory, whereas images in which the shore in Amakusa (approximately 30 km from the observatory) was visible (hereafter referred to as non-fog images) were influenced by aerosol loading between the observatory and the shore in Amakusa. We classified the images into fog and non-fog images using the sum of RGB pixel values.
For fog images, we quantified image degradation using the visibility of the handrail at the observatory. The Canny edge detection method with a Gaussian filter size of 5 × 5 pixels was applied to the handrail region of the image, and the amount of detected edges was used as an image degradation index. The correlation between this index and the number of cloud droplets measured by the CPS was found to be high in spring and autumn seasons with Pearson’s correlation coefficients (r) exceeding 0.7.
For non-fog images, we quantified image degradation using the visibility of the shore in Amakusa. The degradation index was defined as the mean Sobel gradient magnitude within a specified region including the shore of the image. For comparison, in-situ observation data on suspended particulate matter (SPM) concentrations at two sites (Shimabara station in Nagasaki prefecture and Amakusa station in Kumamoto prefecture) of the Atmospheric Environmental Regional Observation System of the Ministry of the Environment, Japan, were used. Correlation analysis between SPM concentration and image degradation index revealed seasonal correlation coefficients ranging from 0.41 to 0.78 at Shimabara station, whereas those at Amakusa station remained between 0.31 and 0.41. The results suggest that the SPM data at Shimabara station are more representative of the temporal variation of aerosol particles between the shore in Amakusa and the observatory than those at Amakusa station.