17:15 〜 18:45
[PEM11-P04] Seasonal and universal time variations of Pi2 activity analyzed with a machine learning approach
キーワード：Pi2地磁気脈動、Wp指数、エコーステートネットワーク、非定常ポアソン過程
It is widely accepted that geomagnetic activities have seasonal and universal-time (UT) variations. Such variations were conventionally examined by statistical analysis of geomagnetic activity indices such as aa, Dst, and AE indices. Our recent study developed an approach to analyze the response of event occurrence rate to various solar wind conditions, seasons, and UT conditions by training a machine learning model called echo state network. In this study, we focus on the seasonal and UT changes in the occurrence frequency of Pi2 events identified from Wp index and analyze them using echo state network.
There are two well-known factors for explaining the seasonal and UT variations: the Russell-McPherron effect and the equinoctial effect. The Russell-McPherron effect is derived from the statistical property of the solar wind magnetic field in the GSM coordinates. When the solar wind parameters, season, and UT are used as input variables to the machine-learning model, it learns the Russell-McPherron effect as a response to the solar wind variation. On the other hand, the equinoctial effect is due to the seasonal and UT variation of the direction of the Earth's dipole axis, and the model learns it as the effect of the seasonal and UT variation independent of the solar wind variations. We can therefore distinguish between the Russell-McPherron effect and the equinoctial effect by the analysis with the machine-learning model. We examine the contribution of each effect to Pi2 occurrence frequency. The result shows that the seasonal and UT variations of Pi2 occurrence frequency are mostly attributed to the equinoctial effect. However, the pattern of the seasonal and UT variations of the Pi2 occurrence rate as identified with the Wp index is slightly different from the well-known pattern of the equinoctial effect. This might suggest that the UT variation of the Wp index is controlled by another effect such as the UT dependence of the location of geomagnetic observatories which are used for the calculation of the Wp index.
There are two well-known factors for explaining the seasonal and UT variations: the Russell-McPherron effect and the equinoctial effect. The Russell-McPherron effect is derived from the statistical property of the solar wind magnetic field in the GSM coordinates. When the solar wind parameters, season, and UT are used as input variables to the machine-learning model, it learns the Russell-McPherron effect as a response to the solar wind variation. On the other hand, the equinoctial effect is due to the seasonal and UT variation of the direction of the Earth's dipole axis, and the model learns it as the effect of the seasonal and UT variation independent of the solar wind variations. We can therefore distinguish between the Russell-McPherron effect and the equinoctial effect by the analysis with the machine-learning model. We examine the contribution of each effect to Pi2 occurrence frequency. The result shows that the seasonal and UT variations of Pi2 occurrence frequency are mostly attributed to the equinoctial effect. However, the pattern of the seasonal and UT variations of the Pi2 occurrence rate as identified with the Wp index is slightly different from the well-known pattern of the equinoctial effect. This might suggest that the UT variation of the Wp index is controlled by another effect such as the UT dependence of the location of geomagnetic observatories which are used for the calculation of the Wp index.