13:45 〜 14:00
[PPS05-11] Characterization of the Martian dust devils observed by InSight: Machine learning-based classification
キーワード：火星、インサイト、ダストデビル
NASA’s InSight (Interior Exploration using Seismic Investigations, Geodesy and Heat Transport) has conducted seismic and meteorological observations (e.g., pressure, temperature, wind speed) on Mars for more than three years (Feb 2019 – Dec 2022) (e.g., Banerdt et al., 2020). Dust devils (wind whirls with dust lifting) are one of the most common meteorological phenomena observed on Mars (e.g., Martinez et al., 2017; Spiga et al., 2021), strongly linked with the local atmospheric conditions (e.g., Banfield et al., 2020; Spiga et al., 2018). Thereby, these events are useful to better understand Martian meteorology. On the other hand, they cause the ground deformation, enabling us to investigate the subsurface rigidity structure.
From the seismic and pressure data retrieved by InSight, we can see various characteristics of pressure drops and ground motions. Although various types of signals associated with dust devils have been observed in previous studies, the differences between them have not been elucidated yet.
The objective of this study is to identify how pressure drop events can be classified based on their features. From the existing pressure drop list (Spiga et al., 2021), we extracted all available events (~13,000) and classified by machine learning. As a result, we found that there were two types of events; one with sharp pressure drop (Type-I) and the other with gradual fluctuations (Type-II). Looking at the underlying characteristics, it turned out that Type-I events were likely to keep a higher correlation (correlation coefficient > 0.8) with the ground motion up to 2 Hz while Type-II events barely show a high correlation above 1 Hz. No strong correlation was found with other weather elements ¬such as wind direction, wind speed, and temperature. The previously proposed models of a moving local low-pressure (e.g., Ellehoj et al., 2010; Vatistas et al., 1991; Lorenz et al., 2015) suggest that the parameters that determine the shape of pressure fluctuation are the advection velocity and the distance between the spacecraft and the dust devil. The former is closely related to the wind speed and wind direction. Since we have not confirmed a strong correlation between the occurrence timing of each type of dust devils and the wind conditions, it is possible that the difference between clusters may be caused by the distance from the observation point rather than the advection speed. It is suggested that the events at a greater distance have lower correlations due to various noises before their effects reach the observation points.
In this presentation, we present the classification results and the characteristics of each type of event.
References:
・Banerdt et al. (2020), Nat. Geo., 13, 183–189.
・Banfield et al. (2020), Nat. Geo., 13, 190–198.
・Spiga et al. (2018), Space Sci. Rev., 214, 1-64.
・Martinez et al. (2017), Space Sci. Rev., 212, 295–338.
・Ellehoj et al. (2010), JGRE, 115, E00E16.
・Vatistas et al. (1991), Exp Fluids, 11, 73–76.
・Lorenz et al. (2015), Bull. Seismol. Soc. Am., BSSA-S-15-00169
From the seismic and pressure data retrieved by InSight, we can see various characteristics of pressure drops and ground motions. Although various types of signals associated with dust devils have been observed in previous studies, the differences between them have not been elucidated yet.
The objective of this study is to identify how pressure drop events can be classified based on their features. From the existing pressure drop list (Spiga et al., 2021), we extracted all available events (~13,000) and classified by machine learning. As a result, we found that there were two types of events; one with sharp pressure drop (Type-I) and the other with gradual fluctuations (Type-II). Looking at the underlying characteristics, it turned out that Type-I events were likely to keep a higher correlation (correlation coefficient > 0.8) with the ground motion up to 2 Hz while Type-II events barely show a high correlation above 1 Hz. No strong correlation was found with other weather elements ¬such as wind direction, wind speed, and temperature. The previously proposed models of a moving local low-pressure (e.g., Ellehoj et al., 2010; Vatistas et al., 1991; Lorenz et al., 2015) suggest that the parameters that determine the shape of pressure fluctuation are the advection velocity and the distance between the spacecraft and the dust devil. The former is closely related to the wind speed and wind direction. Since we have not confirmed a strong correlation between the occurrence timing of each type of dust devils and the wind conditions, it is possible that the difference between clusters may be caused by the distance from the observation point rather than the advection speed. It is suggested that the events at a greater distance have lower correlations due to various noises before their effects reach the observation points.
In this presentation, we present the classification results and the characteristics of each type of event.
References:
・Banerdt et al. (2020), Nat. Geo., 13, 183–189.
・Banfield et al. (2020), Nat. Geo., 13, 190–198.
・Spiga et al. (2018), Space Sci. Rev., 214, 1-64.
・Martinez et al. (2017), Space Sci. Rev., 212, 295–338.
・Ellehoj et al. (2010), JGRE, 115, E00E16.
・Vatistas et al. (1991), Exp Fluids, 11, 73–76.
・Lorenz et al. (2015), Bull. Seismol. Soc. Am., BSSA-S-15-00169