EO-WM: A Physically Informed World Model for Probabilistic Earth Observation Forecasting

arXiv 2026 61.8 method, application

TLDR

EO-WM is a physically informed video diffusion world model for probabilistic Earth observation forecasting with diagnostic benchmarks for weather response.

Reasoning

The paper introduces a novel physically informed conditioning framework that separates baseline and anomaly signals, and includes diagnostic benchmarks for evaluating weather-response behavior. However, it is domain-specific to Earth observation and does not address interactive or reinforcement learning settings.

Read-first score

Read-first score 61.8, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 36.

Methodology quality 18%

100

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=baseline,benchmark,evaluation,experiment,metric

Recency 6%

100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Citation impact 18%

93.4

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.93388316

Topical relevance 29%

51.4

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Reproducibility 18%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=github

Citation velocity 12%

Citation velocity estimates citations per publication-year to reduce old-paper bias. velocity=0.00

Field roles

FoundationFrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 438.

Keyword Scores

world model

world dynamics prediction

video world model

generative world model

world simulator

interactive world model

model-based reinforcement learning world model

Deep Analysis

Innovations

Viewing Earth Observation forecasting as a partially observed, weather-driven world modeling problem with probabilistic forecasting
Physically informed conditioning framework that separates climatological baseline, weather anomalies, and cumulative physical stress signals via distinct conditioning pathways
Two diagnostic benchmarks: Extreme Summer Benchmark for severity-aware prediction of vegetation degradation and Seasonal Matched-Pair Benchmark for testing response fidelity under changed weather forcing

Methodology

EO-WM is a video diffusion transformer for multispectral Earth Observation forecasting. It incorporates a physically informed conditioning framework that represents meteorological forcing through a climatological baseline, weather anomalies, and cumulative physical stress signals, with baseline and anomaly processed via distinct pathways and anomalous forcing accumulated over time to capture sustained heat and drought stress. Evaluation uses standard pixel-level metrics alongside the two newly introduced diagnostic benchmarks.

Key Results

EO-WM reduces the error in predicted NDVI decline amplitude by a relative 5.63% and improves directional hit rate by a relative 7.80%, while remaining competitive on standard pixel-level metrics.

Limitations

Forecasting remains inherently uncertain due to sparse observations and unobserved land-surface states
Evaluation is limited to two diagnostic benchmarks; real-world operational performance and generalization to other regions or conditions are not assessed
The method's weather-response behavior is tested only on specific extreme and seasonal scenarios, which may not cover all possible meteorological forcings

Resources

Paper