Time–space linear regression analysis of the snow cover in a pre-Alpine semi-forested catchment

Abstract

AbstractSub-grid variability of the snow cover is an important issue with regard to catchment runoff or mesoscale meteorological modeling. Here, an evaluation is presented of spatial snow measurements conducted on 5 days in winter 1998/99 and 2 days in winter 1999/2000 in a 0.7 km2 Swiss pre-Alpine catchment. Snow-depth data were analyzed with two different linear regression models, one including altitude, terrain slope, terrain aspect and canopy density, and one using altitude and simple land-use indicators. For the single measurement date/the first model was somewhat superior to the indicator model. The error term of the regression models showed only weak spatial dependence. Finally, a time-space linear regression model describing both the temporal development and the spatial distribution of the snow cover was fitted with the measurements of the first winter and tested with measurements of the second winter. The validation showed a satisfactory match between measurements and models in late December, but a slight overestimation of the measurements by the models in early April. In view of the models’ability to reproduce the snow-depth patterns satisfactorily at this rather detailed scale, it was concluded that such regression models might be a suitable tool to treat sub-grid variability of snow depth in larger-scale models.