Broadband atmospheric transmittance is a key component in many broadband clear-sky solar radiation models, whose performance is inevitably bound to the goodness of their atmospheric transmittance parameterizations. A key element in the development of these parameterizations, that has been traditionally overlooked, is the spectral integration to the broadband range. The related scientific literature reveals that two different spectral integration approaches have been proposed, but they have been vaguely evaluated despite the central role of broadband transmittance. This work discusses such approaches, proposes new ones that mitigate current issues and conducts an unprecedented worldwide evaluation study that analyzes the direct normal irradiance evaluation error that is directly attributable to the choice of spectral integration approach. The study makes use of more than 7 million atmospheric transmittance spectra and a 1-year hourly simulation across a 2-degree spatial grid worldwide. The results show that the independent integration scheme, which is the one virtually used always, results in deviations as high as 40 W/m in dry and clean, and in humid and turbid environmental conditions. Overall, while the error stays within 20 W/m for 90% of the cases worldwide using the independent scheme, it is reduced to only 1.5 W/m using one of the alternative integration schemes here proposed. Hence, as a matter of conclusion, it is recommended that any broadband clear-sky solar irradiance model intended for global applications considers a higher-performance spectral integration scheme for atmospheric transmittance, such as the ones proposed here.
