cs.CVNov 15, 2026

Spatiotemporal Reconstruction of Auroral Precipitations with Neural Fields

Authors: Nazar Misyats, Mizuki Fukizawa, Yoshimasa Tanaka, Satoshi Ikehata

Organizations: University of British Columbia, Canada · National Institute of Polar Research, Japan · National Institute of Informatics, Japan

Abstract

Auroras provide remote optical observations of energetic electrons precipitating into the upper atmosphere, and whose energy distribution encodes key information about large scale magnetospheric processes. Recovering the underlying electron precipitation field from ground-based camera images of auroras remains a challenging inverse problem. Existing auroral tomography techniques are often limited to simplified static settings, discrete spatial and energy grids, or strong assumptions on the electron energy spectrum. We propose to model the dynamic auroral electron precipitation field via a temporally conditioned neural network. The state of the precipitation field at each frame is represented by a per-frame latent code driven by learned discrete dynamics. A shared decoder maps horizontal positions and latent states to electron energy spectra. The predicted flux is converted to multi-wavelength volume emission rates and rendered into camera images accounting for delayed emissions. We evaluate the method on synthetic dynamic auroral sequences and real multi-view auroral videos, showing that the proposed representation produces temporally coherent precipitation reconstructions and image projections consistent with the observations.

Explore similar work

Jun 24, 2026

Does Aurora Encode Atmospheric Structure? Latent Regime Analysis and Attribution

ML foundation models are able to emulate atmospheric dynamics accurately and efficiently but operate as opaque ``black boxes''. We investigate the internal representations of the Aurora model using spatially pooled PCA and layer-wise relevance propagation (LRP). We find evidence that Aurora's latent space is primarily organized by seasonal cycles, whereas extreme storm events do not form a linearly separable cluster. LRP indicates that the model attends to features consistent with the 3D vertical structure of the Great Storm of 1987. Perturbation tests show masking relevant regions degrades forecasts 3.31×3.31\times more than random masking. These findings suggest that Aurora learns meteorological coherence and vertical structure without explicit instruction.
Sep 24, 2026

Image Fidelity is Not Field Fidelity: Joint Thermodynamic Reconstruction and Error Localization in Neural Tomography

Neural fields for scientific tomography are optimized from 2D images, but the actual quantity of interest is often a latent 3D physical field. Because the forward map is many-to-one, low 2D image error need not certify a correct 3D field. Moreover, the latent field is not directly supervised during training, and its error cannot be evaluated against truth at deployment. We develop CoroNeRF to jointly optimize 3D electron density and temperature fields directly from multiview, multiline intensities through a differentiable atomic-emission renderer. Using solar coronal tomography as a controlled testbed, we evaluate physical-field recovery and test whether cross-seed instability provides a ground-truth-free-at-inference indicator of local physical-field error. We underscore the following two observations. (i) Image fidelity is not field fidelity: spectral ablations show that limited-channel reconstructions can fit their available observations well while recovering substantially worse fields, whereas evaluation on a common richer probe exposes the discrepancy. (ii) Cross-seed instability ranks local physical-field error across tested matched-model conditions, supported by sparsification and physical signal-strength controls. Seed-deviation projections provide complementary directional validation, but shared forward-model mismatch can still produce incorrect cross-seed consensus. These results characterize joint thermodynamic recovery and the usefulness and limits of seed-based error localization in a controlled, single-scene solar tomography testbed.
Jul 3, 2025

NeuralDMD: Interpretable Neural Representation of Dynamics from Sparse and Noisy Measurements

Many challenges in scientific imaging involve solving ill-posed inverse problems, where the goal is to recover spatio-temporal fields from indirect, noisy, and highly sparse measurements - often without access to ground truth data or reliable simulators. To address this challenging scenario, we present NeuralDMD, an interpretable, untrained (per-instance) reconstruction framework that combines neural implicit representations with Dynamic Mode Decomposition (DMD) to reconstruct continuous spatio-temporal dynamics directly from measurements. NeuralDMD parameterizes DMD modes as continuous neural fields, and imposes a low-rank linear dynamics prior with spectral time evolution to enforce temporal continuity. This formulation enables both forecasting under sparsity, and yields interpretable modes and spectra. We find that NeuralDMD outperforms baselines on a wide variety of tasks: from weather data assimilation from sparse station observations to interferometric (Fourier domain) observations of Sagittarius A, the black hole at the center of our galaxy. Moreover, NeuralDMD remains stable when extrapolating into the future. While this framework is most naturally suited to linear dynamics, we show that it can be applied to nonlinear regimes, though with extrapolation performance that degrades with increasing nonlinearity. Together, these results show that NeuralDMD enables interpretable reconstruction and forecasting of spatio-temporal dynamics directly from sparse and indirect measurements without relying on numerical simulators or training data.