Pergamon PII: S0038-092X (96)00126-0 Solar Energy Vol. 57, No. 6, pp. 433-443, 1996 Copyright© 1997 ElsevierScienceLtd Printed in Great Britain. All rights reserved 0038-092X/96 $15.00 +0.00 PERFORMANCE OF GLOBAL TO DIRECT/DIFFUSE DECOMPOSITION MODELS BEFORE AND AFTER THE ERUPTION OF MT. PINATUBO, JUNE 1991 F. J. OLMO,* F. J. BATLLES** and L. ALADOS-ARBOLEDAS *t * Dpto de Fisica Aplicada, Universidad de Granada, 18071 Granada, Spain ** Dpto de Fisica Aplicada, Universidad de Almeria, 04120 Almeria, Spain (Received 7 December 1995; revised version accepted 30 May 1996) (Communicated by Richard Perez) Abstrae~Different decomposition models have been developed to estimate hourly direct and diffuse solar irradiance from hourly global irradiance measurements. Besides clearness index kt, other geometric and meteorological variables have been tested as predictors of hourly diffuse fraction, k, or hourly beam transmittance index, kb, by means of piecewise correlations. The recent volcanic eruption of Mount Pinatubo (June 1991 ) was responsible for important modifications in the solar radiation fluxes. In this article we address the effect of the volcanic aerosols on the performance of some well-known empirical models, using a data set including different radiometric and meteorological variables. The study has revealed that the more pronounced performance changes affected the higher k t intervals associated with the lower levels of cloud cover. After the eruption, all the models present a general tendency to overestimate the direct component and underestimate the diffuse component. Copyright © 1997 Elsevier Science Ltd. 1. INTRODUCTION There is an increasing need for hourly values of solar radiation incident on inclined and oriented surfaces to study the transient processes of solar energy systems. Nevertheless, because most of the observations are made on horizontal sur- faces, methods are needed to compute radiation on tilted surfaces. Estimating procedures usually require the beam and diffuse components of global radiation, but unfortunately these two quantities are seldom measured, and thus esti- mating methods are often used to extract beam and diffuse components from other available variables. There are two categories of correlations avail- able in the literature that predict the diffuse and direct irradiances. These are atmospheric transmittance models and decomposition of global irradiance models. The atmospheric transmittance models (e.g. Gueymard, 1993) require detailed information about the atmo- spheric conditions, i.e., cloud cover information, atmospheric turbidity, moisture content; in con- trast, the decomposition models try to estimate the diffuse or direct irradiances from the meas- urement of global horizontal irradiance (e.g., Orgill and Hollands, 1977; Erbs et al., 1982; Skartveit and Olseth, 1987; Maxwell, 1987; Reindl et al., 1990; Perez et al., 1991). In this way, correlations between dimensionless num- *Author to whom correspondence should be addressed. bers such as kt, clearness index (global radiation/extraterrestrial horizontal radiation), and k, diffuse fraction (diffuse radiation/global radiation), or kb, hourly beam transmittance index (direct radiation/extraterrestrial radia- tion), have been extensively used. Different authors (Orgill and Hollands, 1977; Erbs et al., 1982; Reindl et al., 1990) have developed diffuse fraction correlations for hourly intervals following the pioneering work by Liu and Jordan (1960) for daily values. Nevertheless, the use of k t as a unique independent variable is the origin of high errors in the hourly diffuse fraction estimation. These errors are explained by the wide range of diffuse fraction values encountered for a given value of clearness index, k t. A dependence of hourly diffuse irradi- ance on other variables besides kt, such as solar elevation, water vapour content, atmospheric turbidity and climatic variables have been sug- gested by different authors (e.g., Garrison, 1985; Skartveit and Olseth, 1987; Reindl et al., 1990; Camps and Soler, 1992; Chendo and Maduekwe, 1994). Maxwell (1987) and Perez et al. (1991) have developed correlations between kb and kt, including solar elevation dependence. The model developed by Perez et al. (1991) also uses information on the inter- hourly variability of solar global irradiance and the hourly dew point temperature. Anyway the decomposition models are based on empirical correlations of measured global 433