Pre-processed Magnetograms derived from SDOBenchmark Dataset

This solar flares dataset was part of a Masters dissertation by Mangaliso Mngomezulu. It was derived from https://i4ds.github.io/SDOBenchmark/, (accessed: December 05, 2023). It contains magnetogram derivatives with morphological closing and opening operations applied on them to enhance the umbra and penumbra isolating them from the background and eliminating fine grain pixel variations. It was produced to investigate the effects of using the original raw magnetograms in contrast to these preprocessing methods when predicting the catergory of an upcoming flare in given previous magnetogram observations (image series to class). The umbra and penumbra are enhanced motivated by their theorized partial indicativeness of an upcomning flare. In the dissertation, regions of interest (ROIs) are further derived from the magnetogram derivatives to explore the impact of focusing on the largest umbra and or penumbra in the magnetogram. The ROIs are not included here to keep the dataset lightweight but the dissertation provides an overview of how they were obtained. File naming structure: < AR_yyyy − mm − ddT hhminsms_class > which indicates the active reggion number, year, month, day, hours time, minutes, milliseconds, flare category respectively. The flare catergories are based on the Geostationary Operational Environmental Satellites (GOES) system. The dataset provides a catalogue with magnetogram derivatives and associated classes. Users must be careful about AR overlaps and timestamps of the observations (some are collected within narrow time differences in the same AR and they may give results that need to be interpreted with caution.