SNOWMELT MODELLING AND SCALING ISSUES IN A FORESTED SUBARCTIC ENVIRONMENT

Abstract

Few studies considerthe influenceoftreemicroclimatology on the snowmelt energy balance offorests. Hence, two models are produced to study the spatial distribution ofmelt patterns, from scales ranging from a single tree to an entire forest. The first model is presented forthe simulation ofsnowmelt under a spruce tree in the subarctic, making use of physical processes associated with melt, canopy geometry, and empirical functions and coefficients obtained from the field. This model allows detailed study ofthe effect ofa tree on the micro-pattern ofsnow ablation. Simulated results compare well with the daily snow depths. Radiation represents the principal component ofthe total melt energy and the tree canopy enhances the long wave radiation balance for the snow surface. Strong asymmetry inmeltrates among different aspects within and beyond the tree canopy issimulated and can he explained by the changing intensities ofthe snowmelt processes. The second model is presented to simulate snowmelt in a subarctic woodland, using a Geographic Information System coupled with a simplified numerical snowmelt model to express the spatial distribution ofsnow depth and the pattern of changing tree shadows during the day. The woodland is distinguished into several zone types, including openings in the sun and in the shade, zones beneath the tree canopy and the tree trunks. Data obtained at an open site are transposed to each zone for the calculation of melt rates. The experimental slope is subdivided into 2x2m2 grid cells, each with different fractional areas occupied by various zone types. Melt rate at each cell is obtained by weighting the zonal melt with these fractional areas. Despite some limitations, the model provides a spatial dimension to snowmelt in the woodland and the mean melt values thus obtained greatly improve the representation ofthe forestmelt conditions conventionally obtainedby calculationsforsingle points. This model run is repeated at 4x4m2, 16x16m2, and 80x64m2 grid-cell sizes to consider the amount ofinformation lost or preserved when upscaling. The upscaledmodel performs well at all scales as the mean snow depth values are preserved. The results also suggest that beyond the 4x4m2 scale, many local-scale melt features are eliminated and the standard deviation and skewness ofsnow depth distribution are modified.