Improving Modeled Solar Irradiance Historical Time Series: What is the Appropriate Monthly Statistic for Aerosol Optical Depth?

Under clear-sky conditions, direct normal irradiance (DNI) is particularly sensitive to the aerosol optical depth (AOD). It is found that the daily AOD is nearly always log-normally distributed, resulting in significant differences between the monthly mean, median and mode values. Monthly AOD data are still used extensively in radiation modeling. This study thus investigates whether the monthly mean AOD should still be selected as the proper statistic to provide the necessary aerosol input to solar radiation models that are used for hourly irradiance predictions. To solve this issue, daily AOD data from nine sunphotometric sites in various climates are analyzed, and their monthly frequency distributions are obtained to derive the usual statistics. Simulations of cloudless-sky DNI are obtained with the REST2 radiative model, alternatively using individual daily data (providing reference results) and monthlyaggregated statistics of AOD as inputs. Results show that the monthly mean AOD yields too low predicted DNI, whereas the monthly mode results in too high DNI values. The median value is found close to the optimum in general. Some scatter exists, however, so that this finding may not be valid at all sites or for all months. The ratio between the monthly median and mean AOD also varies largely, with a typical average of ≈0.8. The practical outcome of this study is that the existing DNI predictions made by NREL in the past may be too low over the USA (where monthly mean AOD values were used), and may be too high over India (where the monthly mode was used).