Does Aurora Encode Atmospheric Structure? Latent Regime Analysis and Attribution

Emma Kasteleyn, Ana Lucic

Keywords: Foundation Models, Weather Forecasting, Explainable AI (XAI)

TL;DR: Latent analysis suggests that Aurora organizes representations by seasonal cycles and learns the 3D vertical structure of storms despite being a data-driven model.

Abstract:

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 $3.31\times$ more than random masking. These findings suggest that Aurora learns meteorological coherence and vertical structure without explicit instruction.

Anonymization: This submission has been anonymized for double-blind review via the removal of identifying information such as names, affiliations, and identifying URLs.

Style Files: I have used the style files.

Submission Number: 71