Authors: Tania Varesano (CU Boulder / SWRI), Don Hassler (SwRI), Delores Knipp (CU Boulder)

The Spectral Imaging of the Coronal Environment (SPICE) instrument aboard the Solar Orbiter mission provides high-resolution extreme ultraviolet (EUV) spectral data of the Sun’s transition region and corona, offering unprecedented insights into solar dynamics. This project leverages deep learning (specifically Siamese Neural Networks, a self-supervised approach) to automate and streamline the identification and classification of spectral features in SPICE data.

The goal is to correlate these features with plasma fractionation and flare eruption, thereby enhancing the understanding of solar activity and contributing to improved space weather forecasting. The dataset consists of SPICE rasters from the SPROUTS campaign. The trained model aims to detect subtle spectral similarities and extract clusters (using HDBSCAN) that can be mapped back onto the original intensity map, indicating regions where the elemental composition is similar.

First results showed interesting patterns of similar composition at different locations along transition region loops and active region cores, suggesting that this approach may uncover previously unrecognized trends in plasma fractionation.