Compression Method for Solar Polarization Spectra Collected from Hinode SOT/SP Observations

TL;DR

This work tackles the challenge of exploding solar spectropolarimetric data volumes by applying deep autoencoder-based compression to Hinode SOT/SP observations, focusing on Stokes I and V for both quiet Sun and active regions. It compares a deep autoencoder (DAE) and a 1D-convolutional autoencoder (CAE), finding that the CAE delivers more stable performance and reconstructs spectral shapes with residuals at the level of observational noise, especially when using a 28-node bottleneck that balances compression with fidelity. The study implements rigorous data balancing, normalization, and evaluation across four targeted line-core regions, showing robust reconstruction even under varying training-set balance. The results demonstrate the viability of two-dimensional polarimetric spectral compression for solar physics applications, including anomaly detection and cross-comparison with simulations, while outlining future work to incorporate full Stokes Q/U and broader disk positions for universally applicable models.

Abstract

The complex structure and extensive details of solar spectral data, combined with a recent surge in volume, present significant processing challenges. To address this, we propose a deep learning-based compression technique using deep autoencoder (DAE) and 1D-convolutional autoencoder (CAE) models developed with Hinode SOT/SP data. We focused on compressing Stokes I and V polarization spectra from the quiet Sun, as well as from active regions, providing a novel insight into comprehensive spectral analysis by incorporating spectra from extreme magnetic fields. The results indicate that the CAE model outperforms the DAE model in reconstructing Stokes profiles, demonstrating greater robustness and achieving reconstruction errors around the observational noise level. The proposed method has proven effective in compressing Stokes I and V spectra from both the quiet Sun and active regions, highlighting its potential for impactful applications in solar spectral analysis, such as detection of unusual spectral signals.

Figures (12)

• Figure 1: Sample profiles of Stokes parameters corresponding to spatial positions marked in red are provided for (a) quiet Sun, (b) pore, and (c) sunspot core in the FoV image.
• Figure 2: Model architectures for (a) DAE and (b) CAE. Blue and orange blocks represent the input and output (true and reconstructed spectra) of the models, while the encoder, decoder, and bottleneck are respectively depicted in yellow, pink, and green blocks. At each layer name, one index signifies the shape of the layer, while two indices denote the number of filters and the kernel size.
• Figure 3: Snapshot of dataset and its partitioning for model training (Versions A to E), validation, and testing.
• Figure 4: Degree of balance (DoB) in histograms for the five different training sets.
• Figure 5: Evaluation areas of Stokes profiles. Colored shaded regions indicate the calculations of RMSD for left and right cores of both Stokes parameters ($LLC_{I}$, $RLC_{I}$, $LLC_{V}$, $RLC_{V}$).