Assessment of some global solar radiation parameterizations

In spite of their practicability, most classical models are not versatile but rather restrictive in their application. Consequently, their applicability for a particular location depends largely on validation against actual measurements. Global solar radiation parameterizations have been evaluated in this study for a lowland and a mountain site. Tested models were broadly categorised as cloud-based (Kasten) and sunshine-based ( 4 Angstr5 om–Prescott, Garg and Garg, Sivkov). Data sets utilised for the evaluation extended from 1991 to 1994. Adjustable parameters in the models were determined. Observed monthly mean values of solar radiation G and those estimated using Kasten model agreed within 2.5% for the lowland site and 13% for the mountain site. Root mean square errors of estimated hourly values of G using Kasten model appreciated signi=cantly with fractional cloud cover N (particularly for N ¿ 4 octals). For the study sites as well as other locations examined here, 4 Angstr5 om–Prescott coe?cients did not show a distinctive trend with respect to season, geographical co-ordinate or altitude. Monthly mean values of G estimated using 4 Angstr5 om–Prescott model agreed with observation within 2.5% for the lowland site and 3.4% for the mountain site. The e@ect of air mass, latitude and water vapour terms on the 4 Angstr5 om–Prescott relation has also been investigated. In general, 4 Angstr5 om–Prescott as well as Garg and Garg models yielded the least RMSE (¡ 0:047) for the study sites and are thus recommended for estimating G for an arbitrary location. c © 2002 Elsevier Science Ltd. All rights reserved.