Disaggregation of climatic variables using a novel stochastic approach and its application in building performance simulation studies

Daily weather data, both observed and synthesized, are available for most locations worldwide but not at the sub-daily temporal scale, which is desirable for building performance analysis. In this context, we introduce a novel stochastic methodology based on pattern mapping for disaggregation of daily data into sub-daily (hourly) time resolution. The methodology has been tested and validated for three weather variables required for energy simulation (temperature, relative humidity and solar radiation). This approach yields better results than existing methodologies for the selected climate variables. Process efficiency is maintained even with the limited input data. Its stochastic nature enables time-invariant pattern mapping to generate future weather files. The annual 'space conditioning energy' was simulated to assess building performance accuracy using weather files for the hottest, coldest, average and test reference years (TRY) from disaggregated and observed datasets. Results show that the pattern mapping disaggregation methodology accurately downscales daily to sub-daily data.