COSMO Ensemble Simulations for Weather and Climate Model Verification

An ensemble of 5-day weather simulations produced with the COSMO 5.09 model for a model verification study. The data includes 600-member and 200-member ensembles run on GPU and CPU versions in single and double precision. It was created by Christian Zeman at ETH Zurich for the corresponding research paper.

Use Cases

• Detecting differences between model executables based on ensemble output.
• Verifying the impact of numerical precision (single vs. double) on simulation results.
• Comparing computational performance of GPU versus CPU model versions.
• Studying ensemble spread and uncertainty in weather forecasts.

Strengths

• Includes a large 600-member ensemble, which likely provides robust statistical power.
• Provides direct comparisons between GPU and CPU model versions.
• Allows for analysis of precision effects with single and double precision runs.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset is split across two parts, requiring users to locate and combine files.

Provenance

Source

ETH Zurich

Collection Method

Simulations from the COSMO 5.09 weather and climate model.

Time Range

5-day simulation period