We derive spectral and broadband surface albedo for global arid regions from data collected by the Moderate Resolution Imaging Spectroradiometer (MODIS) aboard the Terra spacecraft, at 1 km spatial resolution for 2001. MODIS data show considerable spatial variability both across various arid regions of the globe (from the bright deserts of northern Africa and the Arabian peninsula to substantially less reflective American and Asian deserts) and within regions (variability related to soil and rock types). For example, over arid northern Africa and the Arabian peninsula, albedo in the visible broadband varies by a factor of over 2, from the brightest sand sheets to the darkest luvisols. Few, if any, global and regional land-atmosphere models capture this observed spatial variability in surface albedo over arid regions. We suggest a scheme that relates soil groups (based on the United Nations Food and Agriculture Organization (FAO) soil classification) to MODIS-derived surface albedo statistics. This approach allows for an efficient representation in climate and weather forecasting models of the observed spatial and temporal variability in surface albedo over global deserts. Observed variability in albedo was reduced to a small (1–13, depending on the region) number of soil-related classes (end-members) that could be used in climate models. We also addressed the temporal evolution of albedo during 2001 over global deserts. Regions/soils of stable albedo with very low temporal variability were identified. For other regions/soils, temporal signals in albedo were related to ephemeral inundation with water or variations in sample size.