Coronal mass ejections (CMEs) are eruptions of magnetized plasma structures generated by solar eruptive phenomena such as flares. CMEs occasionally reach Earth and cause various disturbances of the space environment, where a vast amount of social infrastructure is in operation. Therefore, forecasting the arrival time of CMEs is an important topic in the field of space weather research.
Many studies have attempted to forecast CME arrival using empirical models (e.g. Gopalswamy et al. 2001), and global magnetohydrodynamic (MHD) simulations of the heliosphere such as ENLIL (Odstrcil 2003), SUSANOO (Shiota et al. 2014; Shiota and Kataoka 2016), and EUFORIA (Pomoell and Poedts 2018). These simulations calculate the background solar wind from magnetic field data of the solar surface and empirical models of the solar wind speed (Wang and Sheeley 1990, Arge and Pizzo 2000, Arge et al. 2004). Then, the simulations approximate the CMEs as simple structures such as cones or spheromaks. The modeled CMEs are placed at the inner boundary, which is approximately a few tens of solar radii (Rs), and their propagation is simulated to 1 AU.
The accuracy of CME arrival-time forecasts has been statistically validated in previous studies (e.g., Wold et al. 2018), which suggest an error of more than 10 h for current MHD simulations. There are many possible causes of this arrival-time error. Ambiguity of the modeled CME parameters is the first candidate. Many studies use white-light coronagraph images, which are typically provided by the Large Angle and Spectrometric Coronagraph (LASCO: Brueckner et al. 1995) onboard the Solar and Heliospheric Observatory (SOHO), to detect CMEs and derive characteristics such as CME speed, width, and tilt angle for input into CME models (Wold et al. 2018 and references therein). However, these observations typically contain errors due to the projection effect (Temmer et al. 2009). In fact, some studies have suggested that forecasting accuracy can be improved by deriving the input parameters from multiple spacecraft observations (Millward et al. 2013) or by using the heliospheric imager (HI) onboard Solar TErrestrial Relations Observatory (STEREO) satellites (Mostl et al. 2014). There are many other potential reasons for the observed arrival-time error, such as interaction between CMEs and background solar wind, accuracy of the simulated background solar wind, and CME-CME interactions during propagation (e.g., Möstl et al. 2014; Millward et al. 2013.; Lee et al. 2013).
Turbulence contained in the solar wind plasma can scatter radio emission from extra-galactic radio sources, which is known as interplanetary scintillation (IPS, Hewish et al. 1964). Rapidly propagating CMEs sweep the background solar wind, forming dense regions in front of the CMEs. These regions can significantly scatter radio emissions. Hence, IPS observations can be used to detect CMEs propagating in interplanetary space (e.g., Tokumaru et al. 2003; 2005; Manoharan 2006; Glyantsev et al. 2015; Johri and Manoharan 2016). IPS data are also used as additional information for the inner boundary of global MHD simulations (Jackson et al. 2015; Yu et al. 2015).
Iwai et al. (2019) developed a new CME forecasting system that combines MHD simulations (SUSANOO-CME; Shiota and Kataoka 2016) with IPS observations at 327 MHz. They used this system to investigate a CME generated by an X9.3 flare on September 6, 2017. After comparing IPS data with that estimated by multiple CME simulations under different initial conditions, they found that the CME simulation that best estimates the IPS observation can more accurately predicted the CME arrival time on Earth.
The accuracy of arrival-time forecasts using real-time MHD simulations can be improved by incorporating IPS data because the IPS can observe CMEs in the interplanetary space, whereas their statistical accuracy under the real-time operation should be evaluated to include the IPS based simulation to the space weather forecasting system. Therefore, the purpose of this study is to evaluate the performance of the SUSANOO-CME model combined with IPS data for real-time forecasting. To do this, we simulate the arrival times of CMEs using MHD simulations with and without IPS data, and compare their accuracy.