AI Paper Atlas

Research map / Probabilistic and scientific modeling / Dynamical systems

LoPhyDA: Low-Rank Tensor and Physics Gradient Guided Diffusion for Atmospheric Data Assimilation

Danyang Peng, Yang Chen, Yunlong Zhou, Xiaotong Yuan

ICML 2026 · Probabilistic and scientific modeling › Dynamical systems · approach: Data assimilation from sparse observations

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add low-rank tensor completion prior and physics gradient guidance (ablation)latent diffusion assimilation (LDA) · ERA5, 1% observationsMSE: 0.0415 -> 0.0336Improved
add geostrophic-balance physics gradient guidance only (ablation)LDA · ERA5, 1% observationsMSE: 0.0415 -> 0.0388Improved
add low-rank tensor prior only (ablation)LDA · ERA5, 1% observationsMSE: 0.0415 -> 0.0365Improved
add physics guidance on top of low-rank prior (ablation)LDA-LoR · ERA5, 1% observationsWRMSE z500: 65.3315 -> 58.8376Improved
add physics guidance alone (ablation)LDA · ERA5, 1% observationsWRMSE (second WRMSE column, t850): 1.4726 -> 1.4964No clear effect
use low-rank tensor completion instead of Kriging interpolation (ablation)LoPhyDA · ERA5, 1% observationsMSE: 0.0391 -> 0.0336Improved
use low-rank tensor completion instead of learnable encoder (ablation)LoPhyDA · ERA5, 1% observationsMSE: 0.0352 -> 0.0336Improved
LoPhyDA over best baseline AlignDA on real GDAS observationsLoPhyDA · GDAS real observationsMSE: 0.0413 -> 0.0385Improved
LoPhyDA over best baseline PhyDA on simulated 1% observationsLoPhyDA · ERA5, 1% observationsMAE: 0.1225 -> 0.1183Improved
LoPhyDA vs AlignDA on 15-day cyclic assimilationLoPhyDA · ERA5 15-day cycling, fixed observation locationsRMSEImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy