In order to test the ability of a physically based water and heat model to predict surface runoff and pipe discharge, adaptations were made to an agricultural field in the north of Sweden. A five-year period was selected, including observations of meteorological data, frost in the soil and discharge. Basic model requirements on soil properties, i.e., the water retention curve and the saturated hydraulic conductivity, were available from a previous investigation. Unsaturated conductivity was estimated from the water retention curve and by assuming a substantial influence of macro pores in the subsoil. Snow properties and thermal soil properties were adjusted to obtain a reasonable agreement with observed frost depths for areas with barley and with grass leys. Surface runoff was the dominating part of the total runoff, especially during conditions of frozen soil. The simulated discharge agreed well with the general partitioning between surface runoff and pipe discharge but discrepancies occurred in their temporal patterns. A probable explanation of these discrepancies was that the model did not account for the enhanced spatial heterogeneity in water flow through snow and in partially frozen soil.