Towards increasing availability of the Ångström–Prescott radiation parameters across China: Spatial trend and modeling

Abstract

Application of the Ångström–Prescott (A–P) model, one of the highly rated global solar radiation (Rs) models based on sunshine, is often limited by scarcity of model parameters. Increasing availability of its parameters in the absence of Rs measurements is an effective way to overcome this problem. Although the A–P parameters have been extensively studied and calibrated worldwide, models relating them to easily obtainable variables are less investigated. Furthermore, it is uncertain whether the existing models are applicable under Chinese climate. Using a two-step approach and data from 80 sites covering three agro-climatic zones in China, this paper focuses on spatial modeling of the A–P parameters in order to increase their availability and consequently to extend the applicability of the model. It was found that the countrywide average was a=0.18, b=0.56 and (a+b)=0.75, and that a had the largest spatial variability followed by b and (a+b). Values of a and (a+b) increased from east towards west while b showed no regular trend. The influence of various variables on the parameters was investigated, finding that a was mostly affected by P (annual average precipitation) followed by n/N‾ (annual average daily sunshine fraction), Z (altitude), T (annual average daily temperature), ϕ (longitude) and φ (latitude), while b was only influenced by Z, and the sum (a+b) mostly by Z followed by T, ϕ, P and n/N‾. Parameter models of varying complexity were proposed and validated. The three-variable model with Z, n/N‾ and ϕ for a and Z, φ and ϕ for (a+b) performed the best. However, the best- and the poorest-performing model obtained Rs prediction differed insignificantly from the calibrated parameters. Particularly, both Z and P alone could be used with confidence to predict these parameters, implying a large potential for simplifying the current parameter models. We therefore recommend choosing the simplest one-variable models in applications. Meanwhile, we question the necessity of the earlier developed complex models. Variation in model for b affected little on Rs prediction, while accurate modeling for a was more important. This questions the current approach of modeling in b: are efforts for model of b aiming at improving its prediction worthwhile? Could b be considered as a constant across whole China? These questions deserve further investigation. The findings in this study using large datasets have both theoretical implication in guiding the research direction of the A–P model and practical implications in increasing parameter availability and thus in facilitating a wider application of the A–P model.