Global Sat Metar: Satellite and Weather Station Data Patches

Global Sat Metar is a dataset of dense global satellite imagery paired with sparse global METAR station observations. The data is rasterized onto a 3600×1800 grid and sliced into 128×128 spatial patches with a 7-frame hourly temporal context. Created by meteolibre-dev, it is designed as a pre-training corpus for weather foundation models.

Use Cases

• Pre-training weather foundation models based on paired satellite and ground-truth observations.
• Developing models for spatial-temporal weather forecasting based on hourly satellite imagery patches.
• Training models for data fusion tasks based on co-registered satellite and METAR station data.
• Benchmarking geospatial deep learning methods based on the structured 128×128 patch format.

Strengths

• Data is paired, combining satellite imagery with ground-truth station observations.
• Spatial structure is defined with a 128×128 patch size and a 3600×1800 global grid.
• Includes temporal context with 7 consecutive hourly frames.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

meteolibre-dev

Collection Method

Likely involves processing and rasterizing raw satellite (GMGSI) and METAR station data.

Freshness

Last updated 2026-06-14 10:20:02; freshness should be verified.

Geography

Global coverage