Automated Discovery of Short Duration Solar Radio Bursts in Murchison-Widefield Array (MWA) Data

Tuesday, 16 December 2014
Brian Timar1, Victor Pankratius2, Colin Lonsdale2, Divya Oberoi3, Roger J Cappallo2 and Lynn Matthews2, (1)University of California Berkeley, Berkeley, CA, United States, (2)MIT Haystack Observatory, Westford, MA, United States, (3)Tata Institute for Fundamental Research, National Center for Radio Astronomy, Bangalore, India
Low-frequency radio observations of the Sun with the MWA have revealed a previously unknown class of weak radio events, with durations on the order of 1 second or less, and frequency widths of a few MHz. This radio phenomenon is not well-understood, and insight generation is difficult due to the large volume of data produced by the MWA at rates of several terabytes per hour. To address this situation, we developed a new approach for the detection, characterization, and classification of such events, as well as for the well-known Type III flares. Our technique consists of a pipeline of processing steps that starts with background noise estimation and subtraction. Radio events are then isolated algorithmically using region-growing techniques, wavelet decompositions, and thresholding. Physical parameter metadata for each event are then extracted and stored in a database. Scientists can query these data, filter events based on specified properties, and generate statistics and plots for exploratory studies. Our toolset is the first to empower MWA solar scientists with such computational intelligence in order to enhance their ability to interpret large numbers of short-lived events in voluminous MWA data. Computer vision approaches on solar images obtained from optical, x-ray, and infrared instruments are thus complemented by detections of phenomena in the radio frequency domain.