Assumptions of the Classical Equation

There is a set of assumptions that are necessary for the validation of Angstrom

equation. These assumptions can be given as follows,

1. The model parameters are assumed invariant with time on the average as if the same sunshine duration appears on the same days or months of the year in a particular location,

2. Whatever the scatter diagram of H versus S, the regression line is automatically fitted leading to constant a and b estimates for the given data. In fact, these co­efficients depend on the variations in the sunshine duration during any particular time interval. Since sunshine duration records have inherently random variability so are the model parameters, but in practice they are assumed as constants,

3. Angstrom approach provides estimations of the global solar irradiation on hori­zontal surfaces, but unfortunately it does not give clues about global solar irradi­ation on a tilted surface because diffuse and direct irradiations do not appear in the Angstrom model separately.

4. In this approach many meteorological factors are ignored such as the relative hu­midity, maximum temperature, air quality, and elevation above mean sea level. Each one of these factors contributes to the relationship between H and S and their ignorance causes some errors in the prediction and even in the model iden­tification. For instance, classical Angstrom equation assumes that the global solar irradiation on horizontal surfaces is proportional to the sunshine duration only. The effects of other meteorological variables always appear as deviations from the straight line fit on any scatter diagram all,

5. The physical meanings of the model coefficients are not considered in most of the applications studies but only the statistical linear regression line is fitted and parameters estimations are obtained directly. The regression method does not provide dynamic estimation of the coefficients from available data, and

6. Statistically linear equations have six restrictions such as the normality, linearity, conditional distribution means, homoscedascity (variance constancy), autocorre­lation and lack of measurement error.

Updated: August 2, 2015 — 12:07 pm