The origin, formation and evolution of volcanic sands are less well known than the formation of the much more common quartz-rich sand sheets. Combining active volcanism and a cold climate, Iceland is covered for about 21% of its surface by sandy areas. The sands were analyzed in detail at two sites and results reveal their diverse origins. The first site is Dyngjusandur, located north of Vatnajökull, and the second site is the Lambahraun area, located south of Langjökull. At both sites, the sand origin is determined from field observations (wind directions from ventifacts), chemical and mineralogical analyses of rocks and sands. At Dyngjusandur, the sand is dominated by glass grains, a situation typical of sand plains in Iceland. Hyaloclastite ridges presently buried beneath Vatnajökull are the dominant source of the sand, and only large size plagioclase crystals (0.5 cm) in sands seem to be derived from the lava flows. Hyaloclastite ridges were crushed by glaciers and mechanically eroded sediments were washed out by melt-water onto flood plains. The sand chemical composition is spatially homogeneous and similar to the average composition of neighboring sub-aerial lava flows, reflecting efficient mixing of distinct sources below the glacier. The presence of sand north of Dyngjujökull can be taken as a way to explore the average chemical composition of non-exposed volcanic material beneath the glacier. In the case of Lambahraun, prevailing winds indicate several potential sources of sand at the north of the sand sheet. Comparison of chemical and mineralogical analyses of sands and rock samples helped to refine the exact origin. In contrast with the first site, the sand is dominated by crystals and is chemically consistent with a mixture of material derived from the lava flows of Eldborgir and Skersli shield volcanoes. Analysis of the contact between the lava flows and the glacier reveals that basaltic sand grains formed as the result of recent advances of the glacier abrading the rocks. The direct interaction of glacial and fluvio-glacial activity with basaltic plains appears to be necessary to produce a large amount of sands in a relatively short period of time (<4000 years). This site appears to be an excellent natural laboratory for further studies concerning the sand evolution and physical sorting processes in basaltic material, which have important implications for understanding aeolian processes on Mars. Copyright © 2011 John Wiley & Sons, Ltd.