Generation of He I 1083nm Images from SDO/AIA Images by Deep Learning

Abstract:

In this study, we use “pix2pix” model, one of deep learning models based on conditional Generative Adversarial Networks (cGAN) to generate He I 1083nm images from Solar Dynamic Observatory (SDO)/Atmospheric Imaging Assembly (AIA) images. The Mauna Loa Solar Observatory (MLSO)/Chromospheric Helium-I Imaging Photometer (CHIP) images are used as target data of the models. We make two models, one with only SDO/AIA 193 Å images as input and the other with SDO/AIA 193 and 304 Å as input. We use the data from 2011 to 2012 except for April, August and December for training and the remaining one for test. The results are a little better than when we divide the data chronologically. The major results of our study are as follows. First, the models successfully generate He I 1083 nm images. Second, the model with two inputs shows slightly better results than that with one input in terms of metrics, correlation coefficient (CC) and root mean squared error (RMSE). CC and RMSE of the model are 0.88 and 16.32, respectively. Third, generated images show well not only active regions but also coronal holes. This work is meaningful in that our model can produce He I 1083 nm images with higher cadence without data gaps, which is useful for studying the time evolution of chromospheric structures and coronal holes.

This work was supported by Institute for Information & communications Technology Promotion(IITP) grant funded by the Korea government(MSIP) (2018-0-01422, Study on analysis and prediction technique of solar flares).