Compression method for solar polarization spectra collected from Hinode SOT/SP observations

Abstract

The rapidly increasing volume of observational solar spectral data poses challenges for efficient and accurate analysis. To address this issue, we present a deep learning-based compression technique using the deep autoencoder (DAE) and 1D-convolutional autoencoder (CAE) models, developed for use on the Hinode SOT/SP data. This technique focuses on compressing Stokes I and V polarization spectra from sunspots in addition to the quiet Sun, offering a wider and more efficient avenue for spectral analyses. Our findings reveal that the CAE model surpasses the DAE model in reconstructing Stokes profiles, exhibiting enhanced robustness and achieving reconstruction errors close to the observational noise level. The proposed technique is demonstrated to be effective in compressing Stokes I and V spectra from both the quiet Sun and sunspots, highlighting its potential for transformative applications in solar spectral analyses, including the identification of unique spectral signatures.