Towards mechanistic understanding in a data-driven weather model: internal activations reveal interpretable physical features

• URL: http://arxiv.org/abs/2512.24440v1
• Date: Tue, 30 Dec 2025 19:50:30 GMT
• Title: Towards mechanistic understanding in a data-driven weather model: internal activations reveal interpretable physical features
• Authors: Theodore MacMillan, Nicholas T. Ouellette,
• Abstract summary: We adapt tools from interpretability research in Large Language Models to analyze intermediate computational layers in GraphCast.<n>We uncover distinct features on a wide range of length and time scales that correspond to tropical cyclones, atmospheric rivers, diurnal and seasonal behavior, large-scale precipitation patterns, specific geographical coding, and sea-ice extent.
• Score: 0.0
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Large data-driven physics models like DeepMind's weather model GraphCast have empirically succeeded in parameterizing time operators for complex dynamical systems with an accuracy reaching or in some cases exceeding that of traditional physics-based solvers. Unfortunately, how these data-driven models perform computations is largely unknown and whether their internal representations are interpretable or physically consistent is an open question. Here, we adapt tools from interpretability research in Large Language Models to analyze intermediate computational layers in GraphCast, leveraging sparse autoencoders to discover interpretable features in the neuron space of the model. We uncover distinct features on a wide range of length and time scales that correspond to tropical cyclones, atmospheric rivers, diurnal and seasonal behavior, large-scale precipitation patterns, specific geographical coding, and sea-ice extent, among others. We further demonstrate how the precise abstraction of these features can be probed via interventions on the prediction steps of the model. As a case study, we sparsely modify a feature corresponding to tropical cyclones in GraphCast and observe interpretable and physically consistent modifications to evolving hurricanes. Such methods offer a window into the black-box behavior of data-driven physics models and are a step towards realizing their potential as trustworthy predictors and scientifically valuable tools for discovery.

Related papers

• Stable Long-Horizon Spatiotemporal Prediction on Meshes Using Latent Multiscale Recurrent Graph Neural Networks [0.0]
  We propose a deep learning framework for predicting full temperature histories directly on meshes.<n>The framework maintains over thousands of time steps and generalizing across heterogeneous geometries.<n>Experiments on simulated powder bed fusion data demonstrate accurate and temporally stable long-horizon predictions.
  arXiv  Detail & Related papers  (2026-02-20T11:22:47Z)
• A Physics-guided Multimodal Transformer Path to Weather and Climate Sciences [59.05404971880922]
  Many problems in meteorology can now be addressed using AI models.<n>Data-driven algorithms have significantly improved accuracy compared to traditional methods.<n>We propose a new paradigm where observational data from different perspectives are treated as multimodal data and integrated via transformers.
  arXiv  Detail & Related papers  (2025-04-19T04:31:35Z)
• Learning Physically Interpretable Atmospheric Models from Data with WSINDy [0.0]
  We show that an algorithm can learn effective atmospheric models from both simulated and assimilated data.<n>Our approach adapts the standard WSINDy algorithm to work with high-dimensional fluid data of arbitrary spatial dimension.
  arXiv  Detail & Related papers  (2025-01-01T06:03:07Z)
• Analyzing Deep Transformer Models for Time Series Forecasting via Manifold Learning [4.910937238451485]
  Transformer models have consistently achieved remarkable results in various domains such as natural language processing and computer vision. Despite ongoing research efforts to better understand these models, the field still lacks a comprehensive understanding. Time series data, unlike image and text information, can be more challenging to interpret and analyze.
  arXiv  Detail & Related papers  (2024-10-17T17:32:35Z)
• Comparing and Contrasting Deep Learning Weather Prediction Backbones on Navier-Stokes and Atmospheric Dynamics [41.00712556599439]
  We compare and contrast the most prominent Deep Learning Weather Prediction models, along with their backbones. We accomplish this by predicting synthetic two-dimensional incompressible Navier-Stokes and real-world global weather dynamics. For long-ranged weather rollouts of up to 365 days, we observe superior stability and physical soundness in architectures that formulate a spherical data representation.
  arXiv  Detail & Related papers  (2024-07-19T08:59:00Z)
• Generalizing Weather Forecast to Fine-grained Temporal Scales via Physics-AI Hybrid Modeling [55.13352174687475]
  This paper proposes a physics-AI hybrid model (i.e., WeatherGFT) which generalizes weather forecasts to finer-grained temporal scales beyond training dataset.<n>Specifically, we employ a carefully designed PDE kernel to simulate physical evolution on a small time scale.<n>We also introduce a lead time-aware training framework to promote the generalization of the model at different lead times.
  arXiv  Detail & Related papers  (2024-05-22T16:21:02Z)
• ClimaX: A foundation model for weather and climate [51.208269971019504]
  ClimaX is a deep learning model for weather and climate science. It can be pre-trained with a self-supervised learning objective on climate datasets. It can be fine-tuned to address a breadth of climate and weather tasks.
  arXiv  Detail & Related papers  (2023-01-24T23:19:01Z)
• Leveraging the structure of dynamical systems for data-driven modeling [111.45324708884813]
  We consider the impact of the training set and its structure on the quality of the long-term prediction. We show how an informed design of the training set, based on invariants of the system and the structure of the underlying attractor, significantly improves the resulting models.
  arXiv  Detail & Related papers  (2021-12-15T20:09:20Z)
• Model discovery in the sparse sampling regime [0.0]
  We show how deep learning can improve model discovery of partial differential equations. As a result, deep learning-based model discovery allows to recover the underlying equations. We illustrate our claims on both synthetic and experimental sets.
  arXiv  Detail & Related papers  (2021-05-02T06:27:05Z)
• Physics-Integrated Variational Autoencoders for Robust and Interpretable Generative Modeling [86.9726984929758]
  We focus on the integration of incomplete physics models into deep generative models. We propose a VAE architecture in which a part of the latent space is grounded by physics. We demonstrate generative performance improvements over a set of synthetic and real-world datasets.
  arXiv  Detail & Related papers  (2021-02-25T20:28:52Z)
• Deducing neighborhoods of classes from a fitted model [68.8204255655161]
  In this article a new kind of interpretable machine learning method is presented. It can help to understand the partitioning of the feature space into predicted classes in a classification model using quantile shifts. Basically, real data points (or specific points of interest) are used and the changes of the prediction after slightly raising or decreasing specific features are observed.
  arXiv  Detail & Related papers  (2020-09-11T16:35:53Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.