An Automatic Detection Technique for Prominence Eruptions and

An Automatic Detection Technique for Prominence Eruptions and Surges using SDO/AIA Images S. Yashiro 1, 2, N. Gopalswamy 2, P. Mäkelä 1, 2, S. Akiyama 1, 2, and A. Sterling 3 (1) The Catholic University of America, Washington, DC 20064; (2) NASA/GSFC, Greenbelt, MD 20771; (3) NASA/MSFC, Huntsville, AL 35812 Email: Seiji. Yashiro@nasa. gov; Nat. Gopalswamy@nasa. gov; Pertti. Makela@nasa. gov; Sachiko. Akiyama@nasa. gov; Alphonse. C. Sterling @ nasa. gov Background Method Prominence eruptions (PEs) are important for a clear understanding of coronal mass ejections (CMEs) because they are the same phenomena at different heights from the Sun. Making their catalog is an essential task because many researches can be performed based on it (e. g. , Gopalswamy et al. 2003). Nobeyama Radio Heliograph 1(No. RH) have detected prominence eruption routinely from 1992 (Shimojo et al. 2010), but the data coverage is limited (~8 hours/day) because of the ground observation. Solar Limb Prominence Catcher and Tracker (SLIPCAT 2; Wang et al. 2010) have detected PEs observed by STEREO/SECCHI 304 A wavelength. We present an automatic technique to detect and characterize eruptive events (EEs), e. g. prominence eruptions and surges, using SDO/AIA 304 Å images. 1: http: //solar. nro. nao. ac. jp/norh/ 2: http: //space. ustc. edu. cn/dreams/slipcat/ Validation • Because automated methods would detect artifacts and miss real PEs, we visually examined detected PEs appeared in high latitude. • Out of 75 detections, we found that 58 (or 77%) were true PEs, 13 (or 17%) were artifacts. The remaining 4 (or 5%) were multiple counts of a single event. • There is no high latitude PE detected by No. RH and/or SLIPCAT but missed by our method. We did not find any missing PEs. • Left plots shows latitudes of automatically detected PEs in No. RH (top), STEREO/EUVI 304 Å (middle), and SDO/AIA 304 Å images (bottom). • The high latitude activity in the northern and southern hemisphere started in 2010 and 2012, respectively. This north-south asymmetry is seen in all three methods/observations, but promptly seen in our method. 1) We use AIA near-real-time synoptic data (1024 x 1024@3 min cadence) provided by JSOC at Stanford University. 2) The SDO 304 Å images are polar-transformed for easy handling of the outward motion of eruptive events (EEs) and for saving computer resources. 3) The transformed images are divided by a background map, which is determined as the minimum intensity of each pixel during 24 hours. 4) The EEs are defined as a region in the ratio maps with pixels having a ratio >2. Because a stationary prominence has relatively high background, the prominence is detected only when it moves. 5) Pattern recognition is performed to separate different EEs at different locations. 6) In successive images, two EEs with more than 50% of pixels overlapping are considered to be the same EE. 7) If the height of an EE increases monotonically in 5 successive images, we consider it as a reliable eruption. Summary & Future Works • Our technique detected 1428 prominence eruptions and 1921 surges from 2010 May to 2012 December. • We confirmed that our technique detected all high latitude PEs. • However we found that 17% of the detections are artifacts. • The artifact are caused by inclusion of bad images. In order to minimize the number of the artifacts we plan to customize the parameters of the PE detecting programs using Machine Learning Technique. • The PE catalog will be available at the CDAW Data Center (http: //cdaw. gsfc. nasa. gov/) and HEK (Heliophysics Events Knowledgebase) operated at Lockheed. Acknowledgments This research was supported by NASA grant NNX 10 AL 50 A and References Gopalswamy, N. , M. Shimojo, W. Lu et al. , Ap. J, 586, 562, 2003, . Shimojo, M. , T. Yokoyama, A. Asai, et al. , PASJ, 58, 85, 2006 Wang, Y. , H, Cao, J. Chen, et al. , Ap. J, 717, 973, 2010