Solar reflectance can vary with the spectral and angular distributions of incident sunlight, which in turn depend on surface orientation, solar position and atmospheric conditions. A widely used solar reflectance metric based on the ASTM Standard E891 beam-normal solar spectral irradiance underestimates the solar heat gain of a spectrally selective "cool colored" surface because this irradiance contains a greater fraction of near-infrared light than typically found in ordinary (unconcentrated) global sunlight. At mainland US latitudes, this metric R(E891BN) can underestimate the annual peak solar heat gain of a typical roof or pavement (slope <= 5:12 [23 degrees]) by as much as 89W m(-2), and underestimate its peak surface temperature by up to 5 K. Using RE891BN to characterize roofs in a building energy simulation can exaggerate the economic value N of annual cool roof net energy savings by as much as 23%. We define clear sky air mass one global horizontal ("AMIGH") solar reflectance R(g,0), a simple and easily measured property that more accurately predicts solar heat gain. R(g,0) predicts the annual peak solar heat gain of a roof or pavement to within 2 W m(-2), and overestimates N by no more than 3%. R(g,0) is well suited to rating the solar reflectances of roofs, pavements and walls. We show in Part II that R(g,0) can be easily and accurately measured with a pyranometer, a solar spectrophotometer or version 6 of the Solar Spectrum Reflectometer. (C) 2010 Elsevier Ltd. All rights reserved.