Validating and Improving Snow Modeling in Forests

Forested watersheds constitute one of the toughest challenges for snow measurements and estimates. Forested terrain often coincides with higher elevations, and forests cover vast expanses of the global terrestrial environment, making them crucial headwater areas. Unfortunately, our understanding of their snow accumulation and meltwater production is lacking. Sparse automated and manual snow measurement networks leave many gaps in watersheds, especially in areas of high relief. Remote sensing observations are often thwarted by the intervening canopy and its transient snowpack, especially during warmer periods in spring where canopy is thick, and canopy controls on snow are important. Our approach to snow estimates in unforested landscapes has primarily involved SnowModel (Liston and Elder 2006) and assimilating snow data from ground observations or remotely-sensed datasets (Liston and Hiemstra 2008). In forested areas, this approach is less successful. Most daily datasets (e.g., MODIS and passive microwave) are too coarse. At coarse resolution, visible snow-cover products are susceptible to misclassification. To improve our modeling abilities, we are using snow patterns developed from higher resolution (and more sporadic) information. Intermittent LANDSAT data collected over many years (1983-present; 30 m resolution) yields improved model accuracy in forested areas. Increasingly, high-resolution (0.5-2.6m) satellite imagery (e.g., WorldView) can be used to map snow cover needed to validate the model and improve its accuracy in this challenging land cover class.