On selecting weather data sets to estimate a building design's robustness to climate variations

The integration of techniques for uncertainty and sensitivity analysis in building performance simulation (BPS) has a number of potential benefits related to design. It allows assessing the accuracy of performance predictions; it can be used to provide concept specific design guidance, and it enables a robustness assessment of the design proposal to different future climate scenarios. The later is considered here. The problems associated with using climate data sets as input to sampling based uncertainty and sensitivity analysis techniques are; (1) these represent time series data with history, and (2) when used as reference data sets, are purpose bound. To address the problems a typical office room is exposed to measured historic weather files, projected future weather data and a derived artificial reference weather data set representative for the period and location. Its response is compared using peak cooling load as criterion for the buildings robustness. It is found that the individual artificial reference data sets are not suited to predict the peak cooling load and its uncertainty band, as they were created for the prediction of a specific performance metrics and for specific building types. However scenario based multi-year future weather data sets show the potential to be successfully used with sampling based uncertainty and sensitivity analysis techniques