Nowadays, it is so important in saving our economic activity and evading the disasters caused by terrestrial electromagnetic effects to predict both temporal and spatial scales of the geomagnetic disturbances based on in-situ solar wind observations. Recently, Neural Network (NN) is one of the notable techniques for the predictions of the magnetospheric activities. However, NN has a problem referred to as ‘black box’, which is difficult to extract which solar wind parameters are the most important for prediction. In this study, we examine a significant relationship between Kp index, which represents the magnetospheric activity, and the solar wind conditions based on an interpretable neural network: ‘Potential Learning (PL)’. A feature of the PL is to make a network that can understand the input variables by learning the “input potentialities”, which are indices calculated using the variances of the solar wind parameters as input variables. In this study, we investigate the magnetospheric activity profile when the Interplanetary Magnetic Field (IMF) oriented southward (Bz < 0). As the input solar wind data, we utilize the two components of the magnetic field (Bx, By) in GSE, and solar wind flow speed, and number density during 20 years between 1999 and 2018. Furthermore, we divide the associated values of Kp into two groups (targets): ‘Kp = 6- to 9 (positive target)’ and ‘Kp = 0 to 1+ (negative target)’. Because the data number of positive target was smaller than that of negative target, the negative target samples are randomly selected so that the data numbers of both targets become equal. Based on the PL neural network, we obtain two important results; 1) the solar wind plasma flow speed might have the most influential in the increase of the Kp index, and 2) as the secondary influential parameter for the Kp increase, the solar wind proton density is considered. In the presentation, we will discuss feasibility of the application to the prediction of the magnetospheric activity based on the solar wind parameters.

Predicting geomagnetic conditions based on in-situ solar wind observations allows us to evade disasters caused by large electromagnetic disturbances originating from the Sun to save lives and protect economic activity. In this study, we aimed to examine the relationship between the Kp index, representing global magnetospheric activity level, and solar wind conditions using an interpretable neural network known as potential learning (PL). Data analyses based on neural networks are difficult to interpret; however, PL learns by focusing on the “potentiality of input neurons” and can identify which inputs are significantly utilized by the network. Using the full advantage of PL, we extracted the influential solar wind parameters that disturb the magnetosphere under southward Interplanetary magnetic field (IMF) conditions. The input parameters of PL were the three components of the IMF (Bx, By, -Bz(Bs)), solar wind flow speed (Vx), and proton number density (Np) in geocentric solar ecliptic (GSE) coordinates obtained from the OMNI solar wind database between 1998 and 2019. Furthermore, we classified these input parameters into two groups (targets), depending on the Kp level: Kp = 6- to 9 (positive target) and Kp = 0 to 1+ (negative target). Negative target samples were randomly selected to ensure that numbers of positive and negative targets were equal. The PL results revealed that solar wind flow speed is an influential parameter for increasing Kp under southward IMF conditions, which was in good agreement with previous reports on the statistical relationship between the Kp index and solar wind velocity, and the Kp formulation based on the IMF and solar wind plasma parameters. Based on this new neural network, we aim to construct a more correct and parameter-dependent space weather forecasting model.