Daily solar irradiance data collected from 144 surface stations across Australia over a 23. year period is analyzed to determine patterns in cloud absorption. Two statistical models are developed and fitted to the estimated cloud absorption at each station. The first model, which is equivalent to a non-parametric mean model, is a regression that uses variables of year and pentad (5-day) for each surface station. The second model clusters the surface stations using a factor analysis (FA) on the locational variable, while retaining the year and pentad (or alternatively, month) variables. This identifies seven factors, and a regression model is then developed for each of these, with year and month (or pentad) as the variables. The first model is able to fit into the radiation observations well especially those in the tropical region in which the seasonal variation is clear. The FA for establishing the second model determines seven dominant factors (with the eighth unimportant one ignored). These seven factors have distinct geographic distributions in each of which a unique factor is the dominant component. Examination of the seasonal cycle and interannual variation of estimated cloud absorption based on this model indicates three major regional groups in term of the estimated annual cloud cover. The southeast region forms a group with the largest cloud cover percentage, the group consisting of central east, northeast and central west Australia has medium percentage, and that with central, central north and northwest Australia as members has the lowest percentage. These three geographic groups have distinct distribution compared with the climate regimes in Australia, and the relationship with the origins of cloud systems is analyzed. The implications of these statistical models to the short- and long-term estimation of solar energy availability based on ground observations and their utilities are discussed.