An Analysis of Nonovercast Sky Luminance Models Against Hong Kong Data

Daylighting is an important issue in modern architecture that has been characterized by the use of curtain walls in buildings. Nonovercast skies, including clear and partly cloudy days, are essential because they may occur more frequently for places such as in equatorial regions and the tropics. Better understanding of nonovercast sky luminance distribution is vital to estimate the dynamic variation in daylight illuminance as sky condition and solar position change. This paper presents the work on the evaluation of six clear sky and three partly cloudy sky models against three-year (1999–2001) measured Hong Kong sky luminance data. The general features and characteristics for the models were described and assessed. The nonovercast sky conditions were identified using the ratio of zenith luminance $(Lz)$ to diffuse illuminance $(Dv)$ and the ratio of global illuminance $(Gv)$ to the extraterrestrial illuminance $(Ev)$⁠. Subsequent interpretations of the clear skies into high and low turbid types were conducted in conjunction with the cloud cover (CLD) and the luminous turbidity $(Tv)$⁠, and partly cloudy skies were further subdivided into thin and thick cloud modes using sunshine hour (SH) and global irradiance (GSI). A statistical analysis of the models revealed that the Gusev model (i.e., CIE (Internal Commission on Illumination) polluted sky No. 13) and the model by Chen et al. (1999, “Luminance Distribution Model of Intermediate Skies,” Zhaom Ing Gong Chen Xuebao, 10(1), pp. 59–63 (in Chinese)) developed using artificial neural network (ANN) theory with the measured data in Chongqing, China (⁠$29.6degN$ and $106.5degE$⁠) showed the best predictions for sky luminance at this location under the clear and partly cloudy sky conditions, respectively.