Near-real time assimilation of glacier observations in Switzerland using a particle filter

Abstract:

Glaciers are important indicators of climate change and they provide essential services such as water supply. This is why public interest in their near real-time mass balance is high, especially in countries such as Switzerland where the population lives close to glaciers. However, when interest is peaking during summer, in situ observations of glacier mass balance are scant. As a result, modeled near real-time mass balances can become remarkably uncertain, making robust statements impossible. In the project “CRAMPON” (Cryospheric Monitoring and Prediction Online) we (1) reduce mass balance uncertainty by assimilating frequent in situ and remote observations, and (2) develop a nowcasting platform that makes frequent mass balance information available to the broad public.

To obtain the near real-time glacier status, we run an ensemble of four mass balance models driven with uncertain meteorological data and uncertain parameters. As supporting observations, we use in situ point mass balances and satellite observations. In situ data stem from seven on-ice cameras distributed on three Swiss glaciers, which take images of marked ablation stakes providing daily point mass balances. Remote observations include melting season broadband albedo and snow lines from Sentinel-2 imagery. To assimilate both observations into the model ensemble, we designed a custom particle filter in which the models that perform poorly over a period of time can recover at a later stage. This allows evaluating individual model performance and parameter evolution over time.

In a first phase with only camera observations, we show that assimilated mass balances outperform reference models with mean model parameters and parameters tuned on mid-season observations (skill scores 0.71-0.98; one outlier). A leave-one-out cross-validation between camera stations reveals differences in predicted and observed cumulative mass balance smaller than 8%. A comparison to the seasonal glacier-wide mass balances evaluated in the framework of the Glacier Monitoring Switzerland (GLAMOS) initiative shows good agreement.

Our contribution will present the work flow of our assimilation system, show the added value of frequent observations for glacier mass balance nowcasting, and give insights into a nowcasting web platform presently being established.