Generate Radioheliograph Image from SDO/AIA Data with Machine Learning Method

TL;DR

This paper presents a neural network model that generates microwave radioheliograph images from EUV data, enhancing solar activity studies by filling observational gaps with high consistency results.

Contribution

A novel neural network approach that converts EUV images from SDO/AIA into radioheliograph images, enabling continuous solar observation data synthesis.

Findings

Generated images are consistent with actual radioheliograph data.

Model fills observational gaps due to limited radio telescope operation time.

Supports studies of microwave and EUV emission relationships.

Abstract

The radioheliograph image is essential for the study of solar short term activities and long term variations, while the continuity and granularity of radioheliograph data is not so ideal, due to the short visible time of the sun and the complex electron-magnetic environment near the ground-based radio telescope. In this work, we develop a multi-channel input single-channel output neural network, which can generate radioheliograph image in microwave band from the Extreme Ultra-violet (EUV) observation of the Atmospheric Imaging Assembly (AIA) on-board the Solar Dynamic Observatory (SDO). The neural network is trained with nearly 8 years of data of Nobeyama Radioheliograph (NoRH) at 17 GHz and SDO/AIA from January 2011 to September 2018. The generated radioheliograph image is in good consistency with the well-calibrated NoRH observation. SDO/AIA provides solar atmosphere images in multiple EUV wavelengths every 12 seconds from space, so the present model can fill the vacancy of limited observation time of microwave radioheliograph, and support further study of the relationship between the microwave and EUV emission.