A physically based SVAT-model was rested with soil and snow physical measur
ements, as well as runoff data fr om an 8600 m(2) catchment in northern Swe
den in order to quantify the influence of soil frost on spring snowmelt run
off in a moderately sloped, boreal forest. The model was run as an array of
connected profiles cascading to the brook. For three winter seasons (1995-
98) it was able to predict the onset and total accumulation of the runoff w
ith satisfactory accuracy. Surface runoff was identified as only a minor fr
action of the total runoff occurring during short periods in connection wit
h ice blocking of the water-conducting pores. Little surface runoff, though
, does not mean that soil frost is unimportant for spring runoff. Simulatio
ns without frost routines systematically underestimated the total accumulat
ed runoff. The possibility of major frost effects appearing in response to
specific combinations of weather conditions were also tested. Different sce
narios of critical initial conditions for the winter, e.g, high water satur
ation and delayed snow accumulation leading to an increased frost penetrati
on, were tested. These showed that under special circumstances there is pot
ential for increased spring runoff due to soil frost. Copyright (C) 2001 Jo
hn Wiley & Sons, Ltd.