RI05: Weather data for daylight modelling (2022) (pdf)

The quality of weather data underpins the accuracy of any building simulation analysis. Realistically representing the boundary conditions of a building design is essential to devise effective solutions for when the building is constructed and occupied. As provider of widely used weather files for the UK, CIBSE is keen to maintain high quality standards for the data it distributes by revising and updating these files to meet the requirements of the latest building simulation methods.

This document presents findings from the Weather Data for Daylight Modelling project conducted at Loughborough University. The project aimed at understanding the sensitivity of climate-based daylight modelling (CBDM) to solar radiation models and at suggesting possible improvements to the quality of solar data offered in weather files. Models used in the current version (2016) of CIBSE weather files were compared with high quality measured irradiances and with other models used by international agencies. A new approach was proposed, based on the use of measured global horizontal irradiance from the Met Office network and combined with separation and luminous efficacy models to derive all irradiance and illuminance components (i.e. direct normal and diffuse horizontal). In addition to

daylighting, the project investigated the influences of existing and proposed models on other types of building simulations, such as those required for overheating assessments.

Current CIBSE weather files were found to provide reliable results for annual daylight assessments that use climate-based daylight metrics such as useful daylight illuminance, daylight autonomy and total annual illumination. Generally speaking, evaluations that consider longer time frames (i.e. several months or a year) and that are influenced by global illumination (i.e. both direct and diffuse) are less prone to errors in the underlying irradiance/illuminance series. On the other hand, evaluations based on direct illuminance (e.g. annual sunlight exposure or glare analyses) or that investigate more accurately specific moments of the year (e.g. on overheating risk) are directly influenced by the accuracy of the irradiance data, thus of the applied solar radiation model. In such cases, CIBSE weather

files were found to overpredict the frequency of low irradiance instances (0 to 400 W/m2) and underpredict the frequency of high irradiance instances (700 to 900 W/m2). This tends to lead to overpredictions of annual sunlight exposure results and underpredictions of the overheating risk when evaluated against the criteria presented in CIBSE TM59: Design methodology for the assessment of overheating risk in homes (2017). The newly proposed approach (termed RSO) showed a much better performance but is reliant on the availability of measured global horizontal irradiance for all desired locations. It is expected that satellite data will play a prominent role in filling the gaps in ground measurements time series and in estimating solar radiation at a higher spatial resolution than ground

weather networks.

The more detailed the analysis, the more we need to make sure that the underlying data are reliable and up to that level of accuracy. Past solar radiation models were created to answer questions about annual energy consumption and their uncertainty was proportionate to that need. Now that we are querying our building simulation models for more and more detailed results, we need adequate, high

quality weather files to guarantee robust evaluations. This document provides a pathway to reach that high quality in solar radiation data and outlines some of the outstanding issues in weather files for building performance simulation.