Estimation of solar radiation in Australia from rainfall and temperature observations

This study evaluated the accuracy of several models for estimating daily solar radiation (Q) across Australia. Data from 39 sites were taken from MetAccess. The sites were grouped into four climatic zones. Coefficients of nine models, including two newly developed formulae, were fitted for the 39 sites. Correlation coefficients (R2) between estimated Q and measured Q and root mean squared error (RMSE) associated with the estimates were calculated. It was concluded that models which expressed rainfall as a binary quantity (1 for rainfall>0; 0 for rainfall=0) performed better than those using amount of rainfall (precipitation in mm). The best performed models, one using temperature data only, one using rainfall data only and one using rainfall and temperature were further evaluated. A newly developed model that included temperatures (minimum and maximum) and rain-day information proved the best method. Using the coefficients for estimates of Q at the same sites, averaged R2 was 0.68, 0.74 and 0.79 and RMSE was 3.24, 3.05 and 2.89 MJ m−2, for the model using rainfall only, temperature only and both rainfall and temperature, respectively. When Q was estimated using the coefficients from other sites, estimation of Q within eastern and southern zones were more reliable than in northern and central zones. In general, the model using both temperature and rainfall was the best model for either estimated Q for the site where coefficients of the model were developed at the site or imported from other sites. However, in some tropical coastal sites of Australia, such as Broome and Darwin, estimates of Q based on rainfall only can be more reliable when the coefficients were imported from other sites. It was concluded that radiation within a similar climatic region could be well estimated with no local recording, regardless of distance between sites.