The Complexity and Forecasting Approach of Solar Eruptions

Author:Xie Yu east

Supervisor:Wei Fengsai, feng Xueshang


Degree Year:2018





Solar eruptions,such as flares and coronal mass ejections,accompanied by huge amount of high-energy particles and magnetized plasma,will bring great disturbances to the interplanetary space.When arrived earth,those particles and plasma interacted with earth’s magnetosphere,ionosphere and thermosphere,which resulted in hazardous spaceweather events.For a better prediction of such events,a spaceweather forecasting platform based on better understanding of the properties of flares and CMEs is necessary and of great importance.We analysis the M9.3 class flare and following high-speed CME on August 4th,2011 to uncover the mechanism of this unique twisting ejection to a better understanding of the mechanism of such eruptions.Then we analyzed the evolution of magnetic helicity and sunspots in three different addictive regions to get a clearer view of the pre-flare magnetic activities,and to give some possible clues for flare prediction.At last,after combing through current machine learning study on space physic,we apply such method on prediction geo-effectiveness of interplanetary coronal mass ejections,providing a new method to approach a better prediction of Dst index.On 03:55 UT,August 4th,2011,the active region NOAA AR 12261 gave a GOES M9.3 class flare,after which a high-speed halo CME.A clear highly twisted flux rope could be found in multiple wavelength bands of SDO/AIA.After the flare,a tornado-like structure was found on the west leg of the formal flux rope.Also,a double helical untwisting movement could be found in the STEREO/COR1 observation of the counterpart of the CME.We calculated the photosphere plasma velocity and magnetic helicity flux using vector magnetograms form SHARPs.From these calculations,we found that the twisting movements were almost only found on the west leg.With accumulating of magnetic free energy transported mainly by twisting movement from the photosphere,the flux rope uplift,reconnected to open field lines near the west leg and ignited the flare and the CME.To get a clearer view of the eruption process,we employed the CESE-MHD-NLFFF model to simulate this eruption.This simulation well reproduced the eruption phenomena from the twist motion of the flux rope to the final eruption.We also recognized magnetic cloud signatures in the in-situ solar wind data from WIND satellite.Based on these analyses,we concluded that such eruptions with only one foot-point fixed on solar surface can produce magnetic-cloud structures.Then,we analyzed three active regions,namely NOAA AR 11261,NOAA AR 1283 and NOAA AR 11429 using vector magnetogram data from SDO/HMI SHARPs and sunspots data from SDO/HMI-Debrecen sunspots database.There are multiple GOES M class or X class flares accompanied with high-speed CMEs(linear speed large than 400 km/s).These active regions were on the earth side,near the sun center when the flares initiated.We calculated magnetic helicity flux and accumulated magnetic helicity in these three active regions while separating the helicity flux form shearing movement and the helicity flux from emerging flux.We found that shearing helicity injection plays an important role in trigging solar flares.Shearing helicity flux and total helicity flux will see a steep increase and the following decrease before most flares.We also applied WGmmethod to track sunspots movement.We found a similar approaching/receding trends in the distances of barycenters of sunspots with opposite polarities and the total/shearing helicity flux.This emphasizes the importance of the accumulation of magnetic free energy generated by shearing motions and could give some clue to the flare forecast.Considering the complexity of solar activities,new methods are needed to give a better forecast of solar eruptions and their geoeffectiveness.Firstly we gave a brief review of current applications of machine learning method on analysis and prediction of solar eruptions and their geoeffectiveness.Then we applied Support Vector Machine on data based on Richardson ICME list and in-situ observations to make a prediction on whether an ICME would bring moderate geomagnetic storm(Dst<-50 nT)or not,which gave a favorable performance with a precision larger than 0.8 and a True Skill Statistics about 0.7.Besides,based on the Fisher Score of selected features,we compare the importance of different ICME parameters in moderate or strong(Dst<-100 nT)geomagnetic storms.