High spatial variability in snowmelt phenomena was observed in the experimental watershed of the upper Cordevole (7 km 2), in the Dolomites. Snowmelt depends, at a point scale, on available energy, which in turn is heavily affected by slope, aspect and shading effects of the site. A distributed hourly model of snowmelt was set up as a geographical information system (GIS) module. The model accumulates snowfall for each raster element (20 m × 20 m) using a temperature threshold. It simulates snowmelt using air temperature and a radiation index consisting of daily average values of clear sky radiation that have cumulated at selected dates since 21 December. It should be underlined that the only relevant calibration parameter of the model is the combined snowmelt factor, which is unique for the whole basin. The clear sky radiation maps were computed for a selected period of the year, based on the watershed digital terrain model and the solar path sampled at very short intervals. When run for a season, the model produces snow water equivalent maps at given dates. The model was validated, with satisfactory results, by comparing these maps with 60 snow covered area surveys and related water equivalent measures collected in six snowmelt seasons from 1986 to 1991. The classical temperature index approach to snowmelt modelling does not allow the full variability over the basin to be taken into account. Besides, it seems important to point out that this fully distributed approach allows us to avoid the use of snow covered area (SCA) depletion curves, the definition of which still proves a troublesome task.