Multimodal Flare Forecasting with Deep Learning

TL;DR

This study employs deep learning as a purely data-driven approach to compare the predictive capabilities of chromospheric and coronal UV and EUV emissions across different wavelengths with those of photospheric line-of-sight magnetograms, and introduces a deep-learning architecture suited for extracting temporal features from full-disk videos.

Abstract

Solar flare forecasting mainly relies on photospheric magnetograms and associated physical features to predict forthcoming flares. However, it is believed that flare initiation mechanisms often originate in the chromosphere and the lower corona. In this study, we employ deep learning as a purely data-driven approach to compare the predictive capabilities of chromospheric and coronal UV and EUV emissions across different wavelengths with those of photospheric line-of-sight magnetograms. Our findings indicate that individual EUV wavelengths can provide discriminatory power comparable or better to that of line-of-sight magnetograms. Moreover, we identify simple multimodal neural network architectures that consistently outperform single-input models, showing complementarity between the flare precursors that can be extracted from the distinct layers of the solar atmosphere. To mitigate potential biases from known misattributions in Active Region flare catalogs, our models are trained and evaluated using full-disk images and a comprehensive flare event catalog at the full-disk level. We introduce a deep-learning architecture suited for extracting temporal features from full-disk videos.

Figures (5)

• Figure 1: Video Local Event Neural System (VideoLENS) Diagram. The architectures of the C3D and Local-Timeseries blocks are detailed in Table \ref{['tab:SolVideoLENS']}. The blue initial block represents the original receptive field of the time series shown in dark orange. The final local predictions are derived from these features, providing event predictions that are localized to their respective original receptive fields.
• Figure 2: ROC AUC of various models. Efn-input denotes the EfficientNetV2-S model trained on individual frame input. EfnFuse-inputs refers to a logistic regression model utilizing features extracted by EfficientNetV2-S models trained on each of the distinct inputs. VideoLENS represents the video-based models.
• Figure 3: Models' TSS. Models' labels are the same as the ones described in \ref{['fig:figAUC']}.
• Figure 4: Models' HSS. Models' labels are the same as the ones described in \ref{['fig:figAUC']}.
• Figure 5: Models' MCC. Models' labels are the same as the ones described in \ref{['fig:figAUC']}.