Predicting Snow Density

Abstract

Snow density is an important measure in hydrological applications. It is used to convert snow depth to the snow water equivalent (SWE). A model developed by Sturm et al. (2010) predicts the snow density by using snow depth, the snow age and a snow class defined by the location. In this work the model is extended to include seasonal weather variables and variables concerning the location. The model is tested and fitted for 4040 Norwegian snow depth and densities measurements in the period $1998-2011$. A Bayesian modeling framework is chosen. To do inference a Markov Chain Monte Carlo method with Gibbs sampler is used, and cross-validation is used for model evaluation. The final model improved the snow density predictions for the Norwegian data compared to the model of Sturm et al. (2010). In addition year specific measurements are performed in different areas, and included in the model by using random effects. The associated reduction in the prediction error is computed, indicating a significant improvement by utilizing information of annual snow measurements.