Abstract. Snow, whilst not a dominant feature of Britain's maritime climate, can exert a significant influence on major floods through its contribution as snowmelt. Flood warning systems which fail to take account of melting snow can prove highly misleading. Selected results of a study on methods for improved snowmelt forecasting using trail catchments in upland Britain are presented here. Melt models considered range from a temperature excess formulation, with the option to include wind and rain heating effects, to a full energy budget melt formulation. Storage of melt in the pack is controlled by a store with two outlets, allowing slow release of water followed by rapid release once a critical liquid water content is reached. For shallow snow packs, a partial cover curve determines the proportion of the catchment over which snow extends. The melt, storage and release mechanisms together constitute the PACK snowmelt module which provides inputs to the catchment model. Either a lumped or distributed catchment model can be used, configured to receive snowmelt inputs from elevation zones within the catchment; a PACK snowmelt module operates independently within each zone and its inputs are controlled by appropriate elevation lapse rates. Measurements of snow depth and/or water equivalent, from snow cores or a snow pillow, are assimilated to correct for a lack of direct snowfall measurements needed to maintain a water balance during snowfall. The updating scheme involves operating a PACK module at the measurement site (the "point model") in parallel to PACK modules in the catchment model, with point model errors being transferred using a proportioning scheme to adjust the snowpack water contents of the catchment model. The results of the assessment of different model variants broadly favour the simpler model formulations. Hourly automatic monitoring of water equivalent using the snow pillow can help in updating the model but preferential melting from the pillow can be a problem. The energy budget melt formulation proves useful in understanding the energy components of melt typical of upland Britain. It reveals that, during the main melt phase, melt can occur in almost equal measure by sensible heat exchange and by latent heat of condensation, as warm air near saturation in cloud condenses on the snowpack; net radiation makes a negligible contribution. This provides a physical explanation for the success of the simple temperature excess approach to snowmelt estimation.