In this thesis, the distribution of snow over first-year (FYI), multiyear (MYI) and rubble (RI) sea ice were measured at 15 sites sampled during two years of field research in the Canadian Arctic. A geostatistical technique known as the variogram was used to model the statistical pattern of the snow distribution. The variogram examines the spatial continuity of a regionalized variable and how this continuity changes as a function of distance and direction. Results indicate that the variogram provided a good estimate of the type and change of spatial dependence on the snow depths over the various ice types. Over FYI, the regular smooth ice topography produced a periodicity in the snow drifts that was best estimated using a wave theoretical variogram in combination with a gaussian model. The more irregular ice topography characteristic of MYI and RI produced a more irregular snow drift pattern. The most appropriate models were a combination of the spherical and gaussian variogram models. Geometric anisotropy was present in all 15 sites, indicating a directional trend in the spatial continuity of the snow distribution patterns which was attributed to the prevailing wind vector during depositional storm events. These distribution models were used to estimate the spatial dispersion of the photosynthetically active radiation (PAR) transmittance through the snow and ice covers on the arctic oceans. This application of the models illustrated the importance of snow distribution on the transmittance of PAR. Plots of the transmittance for each ice type were produced. (Abstract shortened by UMI.)