Atmospheric hydrogen peroxide (H2O2) measurements at Summit, Greenland, in May--June, 1993 exhibited a diel variation, with afternoon highs typically 1--2 parts per billion by volume (ppbv) and nighttime lows about 0.5 ppbv lower. This variation closely followed that for temperature; specific humidity exhibited the same general trend. During a 17-day snowfall-free period, surface snow was accumulating H2O2, apparently from nighttime cocondensation of H2O and H2O2. Previous photochemical modeling (Neftel et al., 1995) suggests that daytime H2O2 should be about 1 ppbv, significantly lower than our measured values. Previous equilibrium partitioning measurements between ice and gas phase (Conklin et al., 1993) suggest that air in equilibrium with H2O2 concentrations measured in surface snow (15--18 μM) should have an H2O2 concentration 2--3 times what we measured 0.2--3.5 m above the snow surface. A simple eddy diffusion model, with vertical eddy diffusion coefficients calculated from balloon soundings, suggested that atmospheric H2O2 concentrations should be affected by any H2O2 degassed from surface snow. However, field measurements showed the absence of either high concentrations of H2O2 or a measurable concentration gradient between inlets 0.2 and 3 m above the snow. A surface resistance to degassing, that is, slow release of H2O2 from the ice matrix, is a plausible explanation for the differences between observations and modeled atmospheric profiles. Degassing of H2O2 at a rate below our detection limit would still influence measured atmospheric concentrations and help explain the difference between measurements and photochemical modeling. The cumulative evidence suggests that surface snow adjusts slowly to drops in atmospheric H2O2 concentration, over timescales of at least weeks. The H2O2 losses previously observed in pits sampled over more than 1 year are thought to have occurred later in the summer or fall, after the May--July field season. ¿ American Geophysical Union 1995