Hailfinder: A Bayesian Network for Severe Weather Forecasting

56-node Bayesian network designed for forecasting severe hail. The model, authored by B. Abramson, J. Brown, W. Edwards, A. Murphy, and R. L. Winkler, contains 66 arcs and 2656 parameters, representing complex probabilistic relationships. It was published in the International Journal of Forecasting in 1996.

Use Cases

• Benchmarking Bayesian network inference algorithms based on its 56-node, 66-arc structure.
• Testing structure learning methods on a known, domain-specific network.
• Studying probabilistic reasoning for weather forecasting applications.
• Educational demonstrations of complex causal models in meteorology.

Strengths

• Well-documented provenance with a clear 1996 academic citation.
• Defined network structure with 56 nodes and 66 arcs, providing a concrete benchmark.
• Parameters (2656) and average Markov blanket size (3.54) are explicitly quantified.

Limitations

• Column-level documentation and sample data are unavailable, obscuring variable semantics.
• Row count and file formats are unknown, limiting suitability assessment for some tasks.
• Last update date is unknown; the model reflects knowledge from its 1996 publication.

Provenance

Source

B. Abramson, J. Brown, W. Edwards, A. Murphy, and R. L. Winkler.

Collection Method

Likely constructed by domain experts for weather forecasting.

Time Range

Publication date is 1996.

Freshness

Publication date is 1996; freshness for contemporary applications is unverified.

Geography

null