Simulation of snowmelt in a subarctic spruce woodland: Scale considerations

Subarctic woodlands comprise stands of spruce trees with varying degrees of openness, giving rise to large contrasts in melt rates within the forest. The spatial variability of the changing snow depth during a melt season was investigated at three scales (2, 4 and 16 m), using an example from a site in Yukon, Canada, where the computation of snowmelt takes into account the differential rates within the woodland. During the melt period, the mean d aily snow depth decreases but the variability increases as continued ablati on leads to greater unevenness of the snow cover. At the three scales of re presentation, increasing the grid size results in a reduction in the standa rd deviation and the skewness of depth distribution. The blurring of snow c over pattern at the larger scales is due to a loss in information, consider ed as the absolute value of the difference in snow depth calculated at two scales for the same location. This loss increases as the snow depth becomes more variable during the melt season. Knowledge of the scale-induced infor mation loss is relevant to the modelling of snowmelt that exhibits large sp atial variations.