Non-parametric weather generator for modelling construction operations: Comparison with the parametric approach and evaluation of construction-based impacts

Weather generators are frequently used to provide a realistic environment to model weather-sensitive construction operations. Although a parametric weather generation approach is commonly used to construct such generators, this approach is associated with drawbacks such as the assumption that weather variables are normally distributed. In this paper, a simplified weather generation approach, based on the bootstrap technique, is proposed. An experiment comparing both approaches is conducted. A three-stage model evaluation has been applied to outputs from both generators. The evaluation includes: (1) an evaluation of both generators' assumptions, (2) a comparison of both generators' outputs with historical records, and (3) an evaluation of imperfection effects associated with the generators' outputs when applied on weather-sensitive construction models. Results indicate that although both generators are reliable for generating daily synthetic weather series and in modelling construction operations, non-parametric weather generators demonstrate a better performance when modelling time-scale flexibility (hourly/daily) in construction operations.