Item

Abstract

In alpine regions, snow melt controls on freshwater availability, and the spatial variability of snow cover is large. Operational monitoring of snow water resources typically deploys snowpack models. However, snowpack models often struggle to capture the natural variability, because of shortcoming in either the meteorological forcing or the model itself. Data assimilation offers a way to bring models closer to reality, but reliable in-situ observations of the snowpack are not spatially exhaustive. In this situation, the performance of spatialized data assimilation typically suffers from two issues: equifinality between improved model results due to different adjustments, and a lack of robustness with respect to exceptional outliers in scarce observations. In this work, we assess the potential of a Particle Filter (PF) to address both issues at the same time. To this end, we set up an ensemble of snowpack simulations [TJ1] with FSM accounting for meteorological and snowpack modelling uncertainties. The PF is applied to a dense network of snow depth stations covering entire Switzerland. Bi-weekly snow water equivalent and total density measurements are available on some of these (snow pit stations). We first assess the ability of our PF application to address equifinality by validating alternative simulations against independent multi-parameter observations. Then, we use a new inflation technique to mitigate the effect of outlier observations when propagating information across the whole network. This data assimilation framework is compared against an operational version of the snow model to assess its potential in a fully spatialized setting over the whole Switzerland.