Yang et al. (2025) Local Off‐Grid Weather Forecasting With Multi‐Modal Earth Observation Data

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Journal of Advances in Modeling Earth Systems
• Year: 2025
• Date: 2025-12-01
• Authors: Qidong Yang, Jonathan Giezendanner, Daniel Civitarese, Johannes Jakubik, Eric Schmitt, Anirban Chandra, Jeremy Vila, Detlef Hohl, Chris Hill, Campbell Watson, Sherrie Wang
• DOI: 10.1029/2025ms005207

Research Groups

A multi-modal transformer model is proposed to downscale large-scale gridded weather forecasts to precise, off-grid near-surface locations by integrating local historical observations, significantly improving prediction accuracy for high-stakes applications.

Objective

• To develop an end-to-end multi-modal transformer model that combines gridded forecasts with local historical observations to produce accurate, off-grid near-surface weather predictions.

Study Configuration

• Spatial Scale: Northeastern United States (at station-level locations).
• Temporal Scale: Various lead times (for urgent applications).

Methodology and Data

• Models used: Multi-modal transformer model.
• Data sources: Local historical weather observations (e.g., wind, temperature, dewpoint) from weather stations; Gridded forecasts (from machine learning models or numerical weather prediction systems).

Main Results

• The proposed model outperforms a range of data-driven and non-data-driven off-grid forecasting methods.
• Direct input of station data provides a marked improvement in local weather forecasting accuracy.
• Prediction error is reduced by up to 80% compared to pure gridded data-based models.

Contributions

• Proposes a novel end-to-end multi-modal transformer architecture for downscaling gridded weather forecasts to off-grid locations.
• Demonstrates the critical value of integrating local historical observations with large-scale gridded forecasts for enhanced near-surface weather prediction accuracy.
• Provides a method to bridge the gap between large-scale weather models and precise, location-sensitive forecasts for urgent applications like wildfire management and and renewable energy generation.

Funding

Citation

@article{Yang2025Local,
  author = {Yang, Qidong and Giezendanner, Jonathan and Civitarese, Daniel and Jakubik, Johannes and Schmitt, Eric and Chandra, Anirban and Vila, Jeremy and Hohl, Detlef and Hill, Chris and Watson, Campbell and Wang, Sherrie},
  title = {Local Off‐Grid Weather Forecasting With Multi‐Modal Earth Observation Data},
  journal = {Journal of Advances in Modeling Earth Systems},
  year = {2025},
  doi = {10.1029/2025ms005207},
  url = {https://doi.org/10.1029/2025ms005207}
}