This paper presents semi-empirical models for estimating global, direct and diffuse illuminances under clear sky conditions in Thailand. The major components of the atmosphere causing attenuation are written explicitly. Rayleigh scattering, ozone absorption and aerosol attenuation, and the actual attenuation by each component is determined by a multilinear regression against measured data. Atmospheric constituents on aerosol optical depth and total ozone column are obtained from AERONET (Aerosol Robotic Network) and the OMI/AURA satellite respectively. The formulation of the models are based on global and diffuse illuminance collected at four sites in Thailand. The model coefficients are developed using three years of hourly data (2008-2010) and validated against an independent data set for 2011. Global illuminance can be predicted with a mean bias difference (MBD) and root mean square difference (RMSD) of 3.5 klx and 5.9 klx, respectively, with corresponding values for diffuse illuminance being 0.3 klx and 3.3 klx, and for direct illuminance the model can be predicted with MBD and RMSD of 2.3 klx and 8.8 klx. The performance of the proposed models is also compared with that other models reported in literature. It is found that the proposed models perform better than the others, due to more explicit treatment of atmospheric attenuation of the models. (C) 2013 Elsevier B.V. All rights reserved.