Weather-associated construction injury patterns and severity using data-driven Bayesian networks and risk matrix
Purpose Weather variability is increasingly recognized as an important factor for construction safety; however, quantitative evidence on its associations with construction injury patterns and severity remains limited. This study characterizes weather-associated patterns in reported construction injury records and their severity and translates probabilistic model outputs into exploratory relative-priority categories. Design/methodology/approach Occupational injury records from the US Occupational Safety and Health Administration (OSHA) were integrated with hourly weather observations from OpenWeather. Two data-driven Bayesian Network (BN) models were developed to characterize conditional dependencies among weather variables, worker demographics, temporal factors, and injury outcomes. Model 1 evaluated weather associations with construction injury patterns, while Model 2 assessed weather associations with injury severity using inverse-frequency class weighting. A Risk Matrix (RM) mapped BN-derived outputs into four exploratory relative-priority levels. Findings Higher temperatures and lower humidity were associated with greater construction-sector representation among reported occupational injury records, where rainfall and stronger winds were associated with higher class-weighted severity scores. Worker age, gender, time of day, region, and season were also associated with variation in construction record patterns and severity. The integrated BN-RM framework distinguished relative model outputs across different weather conditions. Practical implications The proposed framework may support exploratory weather-responsive safety review and planning, subject to further validation using exposure-based and independent data. Originality/value This study provides a data-driven BN-RM framework that integrates high-resolution weather observations with OSHA injury records to characterize weather-associated construction injury patterns and translate probabilistic inference into interpretable, exploratory risk categories for weather-responsive safety management.
Authors
- Wenlin Li (ORCID: https://orcid.org/0009-0008-7646-2849)
- Peiyi Lv (ORCID: https://orcid.org/0000-0003-2560-8664)
- Siyuan Song (ORCID: https://orcid.org/0000-0002-6444-3914)
Institutions
- University of Maryland Eastern Shore (US)
- Arizona State University (US)
Publication Details
- Journal
- Engineering Construction & Architectural Management
- Published
- 2026-11-20
- DOI
- https://doi.org/10.1108/ecam-02-2026-0303
- Primary Topic
- Occupational Health and Safety Research
- Type
- article
- Field-Weighted Citation Impact
- 0.00