This study defines albedo and reflectance in the context of solar applications. It examines the main sources of surface albedo data that can be used to help solar irradiance modeling efforts, such as undertaken by the National Renewable Energy Laboratory (NREL) over the Americas, for instance, per the National Solar Radiation Database (NSRDB). The MODIS-based albedo database, as used as input to derive historical time series of solar irradiance in the current version of the NSRDB, is compared to other sources of global gridded data over the Americas in general and some specific areas in particular. The particular role of snow cover is analyzed. It is found that, in the current MODIS-based NSRDB implementation, snow-impacted areas have been assigned a too high albedo in general because of the lack of consideration for the antagonistic effect of surface roughness, among other factors. Other albedo databases, such as CMSAF, MINES-ParisTech or the ERAS reanalysis, do not display such high albedos at latitudes above approximate to 40 degrees N. The gap-free and snow-free MODIS MCD43GF product is found affected by a long-term downward trend over some areas and out-of-phase seasonal effects over other areas, in comparison with MERRA-2 reanalysis data in particular. Comparisons with actual albedo measurements at the arid Desert Rock site in Nevada also show that albedo data derived from land surface models are affected by systematic bias. Five albedo databases are intercompared for the area around Ivanpah Playa, California, which has great importance in both remote sensing and solar applications. Various aspects of the proper usage of albedo data in solar radiation modeling are discussed, regarding calculations of both the backscattering process and reflections on a tilted surface: spatial resolution, temporal resolution, angular effects, and spectral effects. An important distinction is made between the use of albedo data at approximate to 41cm resolution for backscattering calculations and at approximate to 10 m resolution for transposition calculations to tilted planes, which might additionally require spectral information for monofacial or bifacial PV applications. It is shown that the tilted calculations usually performed in solar resource assessments are far from realistic in the case of most large-scale PV applications, for which a more suitable methodological approach is described. For such needs, the limitations of the current sources of high-resolution reflectance data are discussed. The impact of surface albedo estimation errors on the estimation of backscattered irradiance over snowy areas is evaluated. It is found non-negligible, and even large under cloudy conditions.