Modeling snow accumulation and melt is important for determining how watersheds that span the rain-to-snow transition zone will respond to changes in climate. The spatial distribution of snow cover as well as the timing and magnitude of snowmelt are simulated in a 1.8 km^2 semi-arid watershed that spans the rain-to-snow transition zone. The spatial distribution of changes in surface water inputs (SWI, the sum of melt and rain) and snow-water equivalent (SWE) that occur during rain-on-snow events are examined in a wet and dry year with similar annual air temperature (averages differ by ~1°C). Our results show that the spatial distribution of SWE and the contribution of SWI throughout each event are largely influenced by (1) initial conditions of SWE prior to the event, (2) wind redistribution of snow, (3) elevation, (4) aspect, and (5) vegetation. Basin average peak SWE for the wet year is ~1/3 of the mean annual precipitation, while during the dry year basin average peak SWE is < 1/5 of the mean annual precipitation. Although these two years have similar runoffs produced during their peak streamflow, they exhibit different spatial patterns of snowpack evolution, largely controlled by precipitation magnitude. In the dry year, "deep" snowpacks (defined as those with SWE exceeding 100 mm) on south-facing slopes persist ~1/3 as long as on north-facing slopes. In the wet year, deep snowpacks persist much longer on both north- and south-facing aspects, but the difference is more marked on south-facing slopes. More frequent low-snowfall years in the future may lead to lower, less persistent snow storage, significantly affecting water resource availability, its spatial distribution, and thermal and biogeochemical processes at both soil and watershed scales.