Following substorm auroral onset, the active aurora region usually expands poleward toward the poleward auroral boundary. Such poleward expansion is often associated with a bulge region that expands westward and forms the westward travelling surge (WTS). In this paper we show all-sky imager and Poker Flat Advanced Modular Incoherent Scatter Radar (PFISR) radar observations of two surge events to investigate the relationship between the surge and flow from the polar cap. For both events, we observe auroral streamers, with an adjacent flow channel consisting of decreased density and cool electron temperature plasma flowing equatorward. This flow channel appears to impinge and lead/feed surge formation, and to stay connected to the surge as it moves westward. Also, for both events, streamer observations indicate that, following initial surge development, similar flows led to explosive surge enhancements. The observation that the streamers connected to the auroral polar boundary and that the flow channels consisted of low density, low electron temperature plasma indicates that the impinging plasma came from the polar cap. For both events, the altitude variations of F region plasma within the surges are related with aurora emission and the poleward/equatorward flow, and the surges develop strong auroral streamers that initiate along the poleward auroral boundary when contacted with flow from the polar cap. These results suggest that the polar cap flow channels play a crucial role in auroral surges by feeding low entropy plasma into surge initiation and development, and also playing an important role in the dynamics within a surge.

BepiColombo is a joint mission of ESA and JAXA to the planet Mercury, which consists of two spacecraft, which are the MPO and Mio. The mission made its first planetary flyby, which is the only Earth flyby, on 10 April 2020, during which several instruments collected measurements. In this study, we analyze MPO magnetometer observations of Flux Transfer Events (FTEs) in the magnetosheath and the structure of the subsolar magnetopause on the dawnside of the magnetopause (Y_GSM ~ - 4 R_E). The magnetosheath plasma beta was high with a value of ~ 8. At the start around the magnetopause crossing, the IMF was southward with a clock angle ~ -150 degrees, and one FTE-type flux rope was observed. And then the clock angle increased to -100 degrees and stayed around 10 minutes, and no FTE-type flux ropes were observed in this period. The clock angle then decreased from ~ -100 degrees to ~ -150 degrees again and following this change, several FTEs appeared. These FTEs traveled northward indicating that the magnetopause X-line was located southward of the spacecraft. Most of the FTEs were in ion-scale, <10 s duration, suggesting that they were newly formed. Only one large-scale FTE, ~ 20 s, was observed. It was made up of two successive bipolar signatures in the normal magnetic field component, which is evidence of coalescence at a secondary reconnection site. Further analysis demonstrated that the reconnection rate of the secondary reconnection associated with the coalescence site was ~ 0.14. While this investigation was limited to the MPO MAG observations, it confirms that the magnetic shear angle controls the magnetopause reconnection on the Earth’s dayside magnetosphere. In addition, it strongly supports a key feature of dayside reconnection discovered in the MMS, the growth of FTEs through coalescence at secondary reconnection sites.

The formation of flux ropes (FRs) is an integral part of the magnetotail dynamics that have important consequences for the onset and subsequent evolution of reconnection in the tail. Global hybrid simulations and in-situ observations have shown that earthward-moving flux ropes can undergo “re-reconnection” with the near-Earth dipole field in the region earthward of the Near Earth Neutral Line to create DFs-like signatures. Recently, Poh et al., [2019] analyzed MMS observation of fields and plasma signatures associated with the encounter of an ion diffusion region ahead of an earthward-moving FR. The signatures of this re-reconnection event were: (i) +/- B_Z reversal, (ii) -/+ bipolar-type B_Y variations as part of the quadrupolar Hall magnetic fields, (iii) northward Alfvénic electron outflow jet, (iv) presence of Hall electric field and intense currents, and (v) positive J·E’. Their analysis suggests that the MMS spacecraft encountered the ion diffusion region but misses the electron diffusion region and estimated that the FR would have fully dissipated by X_GSM ~ -15 R_E. In this study, we conducted a statistical survey of signatures associated with the encounter of “re-reconnection” X-line events preceding observations of earthward-moving flux ropes using five years of MMS magnetic field and plasma measurements. We also compared our observations of re-reconnection signatures and analysis results with simulation results from the BATS-R-US MHD with Embedded PIC (MHD-EPIC) numerical model to understand the flux rope dissipation process and the kinetic physics of re-reconnection from a realistic global perspective. Through correlation analysis, we further demonstrated that there is clear relationship between the dissipation of earthward-moving FRs, global ionospheric current systems and substorm activity in the auroral zone, consistent with earlier observations.

  The magnetic field induced transition (MIT) is the mixing of atomic states with the same parity and same magnetic quantum numbers of different J-values under the action of an external magnetic field, opening a new transition channel. The strength of MIT spectral lines are related to the square of the magnetic field and the energy level interval between the mixed energy levels. The newly discovered Fe X MIT can be used for the diagnosis of the coronal magnetic field and has attracted the attention of some researchers. However, the mixing energy level interval is very small of Fe X MIT, and the intensity of the spectral line is sensitive to the magnetic field and plasma density. The uncertainty introduced in the current research is large. We propose to use SH-HtscEBIT to carry out the Fe X MIT experiment in the ground laboratory. Accurately measure the energy level interval of the mixed energy levels, and simulate in the laboratory that the electrons obey the Maxwell distribution to achieve a plasma environment similar to that in the corona. At the same time, the Grasp2018 and FAC program packages are used to calculate the atomic parameters. Combing the experimental results to give the calibration curve of the relationship between spectrum line intensity and magnetic field intensity to reduce the systematic error used in the measurement of the coronal magnetic field and provide reliable support for the diagnosis of the coronal magnetic field.

In the Earth’s magnetosphere, plasma of ionospheric origin such as singly-charged oxygen ions, O+, are transported to and energized in the plasma sheet and the inner magnetosphere. The energized ionospheric plasma plays a key role in the inner magnetospheric dynamics. However, it remains open questions how and where the ionospheric plasma gain energy predominantly. In order to better understand transport and energization of ionospheric plasma, the present study aims to determine global characteristics of O+ ions and other ionospheric ion species for a wide range of energy up to ~180 keV/e. We utilize a long-term dataset with ion composition obtained from the LEP-i and MEP-i instruments on board the Arase satellite which has observed the inner magnetosphere and near-Earth magnetotail for >4 years. A particular focus is on spatial distributions of He+ and O+ fluxes and moments. Their energy dependences may provide a clue as to the relative significance of O+ ion supply and acceleration/heating in the enhancements of the ring current total pressure. 

During geomagnetic active times, such as geomagnetic storms, large amounts of energy would be released into the Earth’s magnetosphere and change the ring current intensity. Previous studies showed that the significant enhancement of ring current was related to geomagnetic storms, while few studies have examined substorm effects on the ring current dynamics. In this study, we examine the ring current variation during non-storm time (SYM-H > -50 nT) substorms, especially during super-substorms (AE > 1000 nT). We perform a statistical analysis of the ring current plasma pressures and number flux of various ion species under different substorm conditions based on Van Allen Probe observations. The plasma pressure and ion fluxes of ring current increased dramatically during super-substorms, while little change can be observed for substorm with AE < 1000 nT. The results shown in this study indicate that the non-storm time super-substorm may also have a significant contribution to the ring current.

Electromagnetic ion cyclotron (EMIC) waves are believed to play a crucial role in the dynamics of the Earth’s magnetosphere. It has been widely accepted that plasma compositions can influence the growth rate of EMIC waves in the inner magnetosphere, but how warm O+ ions change the wave growth rate is not well known. In this study, we investigate the impact of ring current O+ concentration on the EMIC growth rate during a specific storm when the O+ ion flux significantly increases. We calculate the growth rate of EMIC waves by using physics-based ion distributions output from the Ring current-Atmosphere interactions Model with Self-Consistent magnetic (B) and Electric (E) fields (RAM-SCBE) model. The percentage of warm O+ is experimentally varied to examine how the maximum EMIC growth rate changes over time and how the global distribution of the maximum EMIC growth rate is affected. We found that the maximum growth rate of H-band appears in the dusk-to-midnight sector near the plasma pause, while O-band is excited at a slightly outer region. The maximum growth rate of He-band is closely related to the cold plasma density. With the increase of warm O+ density, the maximum growth rate of H-band and He-band EMIC wave is reduced, while that of O-band EMIC wave is increased, and the region with this wave excitation is widened. Such variation in the maximum EMIC growth rate implies a potential impact on the associated wave-particle interactions and change of the decay rate in the ring current.

We study cyclotron resonant interaction of relativistic electrons in Earth’s radiation belts with electromagnetic ion-cyclotron wave packets corresponding to satellite observations by Van Allen Probe B.  We focus on proton band rising-tone EMIC wave packets. Using an automatic algorithm, we distinguished 31 isolated wave packets and obtained their amplitude and frequency profiles. The maximum wave amplitude is about 1.2-1.6 nT. We assume that EMIC wave packets are generated near the equator, propagate along the geomagnetic field line and then dissipate in the He⁺  gyroresonance region located father from the equator than the spacecraft. As a result of interaction with the considered wave packets, electrons with energies of 1.5-9 MeV can effectively precipitate into the loss cone. The main nonlinear effects affecting the precipitation are the Lorentz force direct influence on particle phase (force bunching) and nonlinear shift of the resonance point. Force bunching blocks precipitation for electrons with low pitch angles, up to 20 degrees. While the precipitation is blocked from low pitch angles, it can occur in a quasi-linear regime from a noticeable range of higher pitch angles (10-30 degrees). This corresponds to maximum precipitating fluxes, close to the limiting value of the strong diffusion regime, and particles with energies 2-5 MeV. Particle trapping by the wave field is not effective due to short packet lengths. The wave packet amplitude modulation leads to a significant change of precipitated particles energy spectrum during short time. Generation of higher frequencies at the packet trailing edge near the equator and dissipation of lower frequencies in the He⁺ gyroresonance region at the leading edge also influence precipitation. For short time intervals, the approximation of each local wave packet amplitude maximum by a Gaussian amplitude profile and a linear frequency drift gives a satisfactory description of the resonant interaction. 

Wave-particle interactions in the Earth's magnetosphere can lead to an efficient exchange of energy between waves and different particle populations, including the energetic electrons in the outer Van Allen radiation belt. Seven years of systematic in situ measurements of whistler mode chorus have been collected by the Van Allen Probes Electric and Magnetic Field Instrument Suite and Integrated Science's Waves instrument. In its continuous burst mode, this instrument recorded selected multicomponent waveforms with a duration of 6 seconds and with a sampling frequency of 35 kHz. We use this data set to analyze the fine structure of whistler mode chorus elements, showing peak instantaneous amplitudes at a level of a few hundred picotesla, and exceptionally reaching up to a few nT. We show that the instantaneous wave vector directions within the peaks of this subpacket structure can turn by tens of degrees. We investigate properties of wave amplitudes and wave vector directions using a large number of events, and we analyze statistics of these crucial parameters determining the nonlinear wave-particle interactions.

We used ionograms acquired from Taiwan and Peru in order to recover signals of ionospheric E and F layers. Three different models CNN, DeepLabV3+ & SA-UNet were used for semantic segmentation of ionogram images. For the Peru data, the ground truth is binary. In the Taiwan data, 7 kinds of layers were labeled in order to distinguish various signals from E and F layers. SA-UNet can predict both data better, it got good IoU scores from Peru data and the most important Es and F2 layers of Taiwan data. We also found that even if we avoided labeling the signal which went into the noise, SA-UNet could still predict the critical frequencies in the regions of strong noise. The work was performed in cooperation with NCU AI group: Chang Y.-C., Dmitriev A., Hsieh M.-C., Hsu H.-W., Huang G.-H., Li Y.-H., Lin C.-H., Lin Y.-C., Mendoza M., Tsai L.-C., Tsogtbaatar E.

We analyze ionogram images from Taiwan which contain signals reflected from the ionospheric E, and F layers. However, determination of the layers is very difficult because of very strong noise. Studying statistical properties of ionogram images can help us realize characteristics of each layer. There are three main statistical properties: percentage of each layer in the data set, pixel position and distribution of particular layers, and signal to noise ratio (SNR). First of all, determining the percentage of each layer in the data set shows the difference between the ionospheric layers. Besides we create a Golden Set which has a similar percentage of E layer and F layer that means we get similar weighting when we train a model. Next, determining the pixel position and distribution of particular layers show the characteristic of each layer, and also define the boundary by pixel position which can remove some wrong prediction pixels. Finally, we compared the Taiwan data with other datasets such as ionograms acquired from Peru. It was found that the Taiwan data has very low SNR, in contrast to the Peru data with higher SNR. From modeling the data by Densely Connected Convolutional Networks (DenseNet), it was found positive correlation between the accuracy of DenseNet prediction and SNR. This fact means that the Taiwan data are more complicated than Peru data for the DenseNet model. The work was performed as part of NCU AI group: Chang Y.-C., Dmitriev A., Hsieh M.-C., Hsu H.-W., Huang G.-H., Li Y.-H., Lin C.-H., Lin Y.-C., Mendoza M., Tsai L.-C., Tsogtbaatar E. 

Abstract: The Energetic Particle Instrument (EPI), proposed by the Peking university (PKU) team, consists of three dual double-ended foil/magnet semiconductor telescopes. The foil telescope side utilizes three layers of semiconductors, respectively, with a 300-micron thickness, 500-micron thickness and 500-micron thickness to measure electrons at energies from 20 keV to 1 MeV. The front 300-micron thick detector is also coated with a 5-micron thick polymide foil to stop protons below 400 keV but leave the electron spectrum essentially unchanged. This paper describes the design and GEANT4 simulations of the electron detection side of the PKU EPI.

With the continuous development of astronautical technology, in addition to the near-space exploration, deep space exploration has also gradually become the mainstream. The Faraday Cup, as one of the main instruments for measuring the plasma, with the advantage of mature technology, simple structure and strong reliability, deserves more attention. In addition to the plasma investigation, Faraday Cup can also help to reveal the challenges inherent to the space environment, such as spacecraft charging and total ionizing dose. Our team propose a Faraday Cup equipment for deep space exploration. It was designed to investigate the solar wind plasma, with performance requirements as follows: energy range of 100eV~8keV for ions and 100eV~1.5keV for electrons; energy resolution ΔE/E<20%; 360° x 120° FOV; time resolution of 4s for a full ion scan and 2s for a full electron scan. We hereby introduce the technical characteristics of the Faraday Cup. Its advantages and applications in deep space exploration mission will also be discussed.

Recent observations by the Extreme Ultraviolet Imager (EUI) on board Solar Orbiter have revealed prevalent small-scale transient brightenings in the quiet solar corona termed "campfires". To understand the generation mechanism of these coronal brightenings, we constructed a self-consistent and time-dependent quiet-Sun model extending from the upper convection zone to the lower corona using a realistic three-dimensional radiation magnetohydrodynamic simulation. From the model we have synthesized the coronal emission in the EUI 174 passband. We identified several transient coronal brightenings similar to those in EUI observations. The size and lifetime of these coronal brightenings are mostly 0.5-4 Mm and ~2 min, respectively. These brightenings are generally located at a height of 2-4 Mm above the photosphere, and the local plasma is often heated above 1 MK. By examining the magnetic field structures before and after the occurrence of brightenings, we concluded that these coronal brightenings are generated by component magnetic reconnection between interacting bundles of magnetic field lines or neighboring field lines within highly twisted flux ropes. Occurring in the coronal part of the atmosphere, these events generally reveal no obvious signature of flux emergence or cancellation in photospheric magnetograms. These transient coronal brightenings may play an important role in heating of the local coronal plasma.

High-resolution solar observations show the complex structure of the magnetohydrodynamic (MHD) wave motion. We apply the Proper Orthogonal Decomposition (POD) and Dynamic Mode Decomposition (DMD) techniques to identify the dominant MHD wave modes in two sunspots, with circular and elliptical cross-sectional shape, using the intensity time series. The POD technique was used to find modes that are spatially orthogonal, whereas the DMD technique identifies temporal orthogonality. Here we show that the combined POD and DMD approaches can successfully identify both sausage and kink modes in a sunspot umbra with an approximately circular cross-sectional shape.

X-rays and ultraviolet rays associated with solar flares reach the Earth and stimulate an ionization of neutral particles in the daytime ionosphere, which changes the ionospheric current in a very short time, and this effect can be observed as drastically changing geomagnetic field perturbations [Campbell, 2003]. There are two types of magnetic field perturbations: positive SFE, in which the magnetic field becomes stronger than just before a solar flare, and negative SFE*, in which the magnetic field becomes weaker [Yamazaki et al., 2009].Rastogi et al. (1996) stated that the positive variation of the magnetic field (SFE) through the equatorial region is a result of the enhanced equatorial electrojet (EEJ), while the negative variation (SFE*) in the morning and evening side of dip-equator is a result of the enhanced counter electrojet (CEJ).However, generation of SFE* around local noon dip equator is also reported [e.g., Rastogi et al., 2003]. Yamazaki et al. (2009) reported two unique SFE*events. They suggested that this phenomenon may be caused by an increase in electrical conductivity in the lower E layer due to X-class flares and the penetration of a westward electric field into the magnetic equatorial region due to the northward turning of the IMF Bz. Whereas, Rastogi et al. (2013) examined the same event and proposed that SFE* was enhanced partial CEJ by solar flares. However, reports on SFE* are limited, so the generation mechanism of the unique SFE* is not well known. Moreover prior studies have reported only case studies, and quantitative analysis methods for SFE* have not been established. The purpose of this study is to understand the generation mechanism of unique SFE* around noon time by analyzing the geomagnetic variations due to solar flares. In this presentation, we will report the methods and the results of our analysis.

Coronal mass ejections (CMEs) are the largest-scale eruptive phenomena in the solar system. Associated with enormous material ejections and energy releases, CMEs have an important impact on the solar-terrestrial environment. CMEs can also occur on other stars and will greatly influence the habitability of the orbiting exoplanets around the hosting stars. Therefore, the detection of stellar CMEs and how stellar CMEs affect the space environments are indispensable when evaluating the habitability of exoplanets. Observationally, solar CMEs could result in a blue-wing enhancement of spectral line profiles. However, it remains unclear whether we can detect stellar CMEs through the blue-wing enhancement in the stellar spectra. In this work, based on spectral line profiles observed in the solar chromosphere and corona during CMEs, we synthesize the Sun-as-a-star spectral profiles during CME eruptions. Furthermore, we make quantitative assessments of the impacts of the area of stellar CMEs and the instrument spectral resolution on the detectability of stellar CMEs. These results will be useful for designing future instruments and planning future observations for the detection of stellar CMEs.

Extreme space weather events (Dst < -500 nT) are rare but low-frequency but high-impact hazard for the human society, due to its increasing dependency on the technological infrastructure.  Among them, the Carrington event in 1859 is arguably considered one of the most extreme events in the observational history and considered much more extreme than any space weather events in the space age. However, this left a large gap between this superstorm and other known storms. Here, we show current efforts for the storm reconstruction and an overview of the reconstructed extreme space weather events after 1859 in terms of their intensity estimates, as well as their current problems and reservations. So far, we identify 6 extreme events beyond the intensity threshold of Dst < -500 nT and analyze these storms with the mid-latitude magnetic observations and the equatorward boundary of the auroral oval. These reconstructions show that the Carrington event is certainly one of the most extreme events, but not likely unique.

Large solar energetic particle (SEP) events present significant hazards for both manned and robotic space missions. The most intense, extreme SEP events are quite rare, which makes estimating their characteristics and occurrence probability a challenging task. Observations of extreme SEP events have also large uncertainties because very high particle fluxes commonly saturate particle instruments. In this report, we discuss on recent results concerning the distribution of extreme SEP events based on both historical data and more recent events. In general, the shape of the tail of the SEP event distribution is not known but by fitting different functions to the available data, e.g., power law or Weibull distributions, we can obtain estimates for the intensity and fluence of 100-year and 1000-year SEP events. We also present results concerning recent extreme SEP events such as the 2012 July 23 backside eruption for which we have multi-spacecraft observations.

The Super Dual Auroral Radar Network (SuperDARN) is a network of high-frequency (HF) radars in the high- and mid-latitude regions. At present more than 35 SuperDARN radars are deployed in both hemispheres for studying dynamics of the ionosphere and upper atmosphere. The network is operated under international cooperation by about 10 countries. The original purpose of the network was to monitor the high-latitude ionosphere and upper atmosphere. On the other hand, over the last approximately 15 years, SuperDARN has expanded into the mid-latitude regions. With radar coverage that now extends continuously from auroral to sub-auroral and mid-latitudes, a wide variety of new scientific achievements have been made. In this presentation, a review of the accomplishments of the mid-latitude SuperDARN will be presented. Future directions of the SuperDARN, including the expansion of the SuperDARN field of view into even lower latitudes to monitor severe space weather events, will also be discussed.

The plasma drift along the geomagnetic field lines is known as the field-aligned plasma drift. The field-aligned drift in the equatorial topside ionosphere transports plasma from one hemisphere to the other, and it can be the effect of the changes in the neutral winds, plasma density, and temperature in the two hemispheres. Using C/NOFS satellite observations, this paper studies the disturbance field-aligned plasma drifts in the equatorial topside ionosphere during eight geomagnetic storms in 2011-2015. During all six storms occurred in the solstices, the disturbance field-aligned plasma drift is from winter to summer hemisphere especially in the morning-midnight local time sector and the disturbance is stronger in June solstice. The two storms occurred at equinoxes have very little effect on the field-aligned plasma drift. Using the plasma temperature data from DMSP satellites and GPS-TEC, it is suggested that the plasma density gradient seems likely to cause the disturbance winter-to-summer plasma drift while the role of plasma temperature gradient is opposite to the observed plasma drift.

In this talk, we introduce our recent applications of deep learning to solar and space weather data. Our major applications are (1) generation of solar farside magnetograms and global field extrapolation, (2) generation of modern satellite images from Galileo sunspot drawings, (3) improvement of global IRI TEC maps, (4) one-day forecasting of global TEC maps, (5) flare classification and visual explanation, and (6) forecasting solar X-ray profiles. (1) We successfully generate farside solar magnetograms from STEREO/EUVI images using a deep learning model (pix2pix) for image translation method.  Our global field extrapolation based on AI-generated data are more consistent with observations than before in view of active regions and coronal holes. (2) We successfully generate modern satellite data  from Galileo sunspot drawings using a deep learning model. (3) We have successfully made an improvement of IRI global TEC maps by a deep learning model. AI-generated TEC maps are much more consistent with the corresponding IGS TEC maps than the IRI ones. (4) We have successfully made a one-day forecasting model of global TEC maps by using deeplearning. For training, we use two sets of input data: the present-day global TEC map and one-day difference map. Our model well generates TEC values over equatorial anomaly regions and is better than the previous models.(5) We have successfully developed three flare forecast models based on novel deep learning methods (CNN) using only full-disk magnetograms. Our models are better than the previous models in view of statistical scores. Visual explanation methods for a deep learning flare classification model show that our model considers the vicinity of polarity inversion lines more important than the rest area of active regions. (6) We successfully forecast solar X-ray profiles using deep learning methods (seq2seq with LSTM and attention). Our methods are much better than conventional regression methods.

We present the inversion model of three-dimensional (3-D) coronal density distribution from solar coronagraphic images using a deep learning method. The observed intensity arises from Thomson scattering of photospheric light by coronal electrons, and thus in principle, it can be directly converted into the total number of electrons along the line of sight. However, its inversion into the 3-D electron distribution is non-trivial and requires heavy numerical calculations. We use an image-to-image translation method, the so-called pix2pixHD, and train the model to translate observed images (model input) into the electron density distribution images (target). We have obtained error-free training sets by using the results of MHD numerical simulations. The input and target images are given in the form of a synoptic map. Each independent model has been trained for six selected heights (2.0, 2.2, 2.5, 4.0, 6.0, and 12.0 solar radii), and consequently the individual models return the electron density synoptic maps according to the kind of input images. To train effectively, we use the byte-scaling images. The mean absolute error of the generated byte-scaling images is 6%. The error in electron density is slightly greater than that of byte-scaling ones and is about 10%. We demonstrate that this method is essentially the same as the traditional tomography for coronagraph data while it is much simpler and easier to use, so that our method and the results can be used more widely for scientific purposes.

Corona mass ejection (CME) is a phenomenon in which a large amount of plasma is ejected suddenly from the sun. When the discharged plasma of high energy reaches the earth, it causes the disturbance of the geomagnetism sphere, and the failure of artificial satellite and large-scale power failure can occur. Therefore, it is important to minimize the effect in advance by predicting CME. It has been proven that the existence of characteristic S-shaped loop structure (sigmoid) observed in the solar corona image is related to CME. Hence, new progress of the space weather forecast field can be expected by the construction of CME predictor which noticed the sigmoid structure. We designed a CME predictor using machine learning of various features extracted from the solar corona image. The CME predictor is a neural network designed by learning the relation between the feature set and the occurrence of CME in time series data of solar flare events. The features are brightness, size, cumulative angle change, longest wire length, and contour length on the shape of a sigmoid, and so on. Regarding some of these features, we designed a novel algorithm to extract them from a sigmoid. We investigated the performance of the CME predictors by various combinations of the set of the features, and we found the best combination for the CME predictor with high predictive performance. Our CME predictor achieves 0.798 for True Skill Statistics (TSS) and 0.899 for Accuracy. This result shows that our proposed CME predictor is effective for the prediction of CME.

An operational Dst index prediction model is developed by combining empirical and artificial neural network models. Artificial neural network algorithms are widely used to predict space weather conditions. While they require a large amount of data for machine learning, large-scale geomagnetic storms have not occurred sufficiently for the last 20 years. Conversely, the empirical models are based on the numerical equations that are derived from human intuition and is therefore safe to extrapolate for large storms. In this study, we distinguish between Corona Mass Ejection (CME) driven and Corotating Interaction Region (CIR) driven storms, estimate the minimum Dst values, and derive an equation for describing the recovery phase. The combined Korea Astronomy and Space Science Institute (KASI) Dst Prediction (KDP) model achieved better performance contrasted to Artificial Neural Network (ANN) model. This model could be used practically for space weather operation by extending prediction time up to 24 hours and updating the model output at every hour.