Propagating information from snow observations with CrocO ensemble data assimilation system: a 10-years case study over a snow depth observation network

Abstract

Abstract. The mountainous snow cover is highly variable at all temporal and spatial scales. Snowpack models only imperfectly represent this variability, because of uncertain meteorological inputs, physical parameterisations, and unresolved terrain features. In-situ observations of the height of snow (HS), despite their limited representativeness, could help constrain intermediate and large scale modelling errors by means of data assimilation. In this work, we assimilate HS observations from an in-situ network of 295 stations covering the French Alps, Pyrenees and Andorra, over the period 2009–2019. In view of assimilating such observations into a spatialised snow cover modelling framework, we investigate whether such observations can be used to correct neighbouring snowpack simulations. We use CrocO, an ensemble data assimilation framework of snow cover modelling, based on a Particle Filter suited to the propagation of information from observed to unobserved areas. This ensemble system already benefits from meteorological observations, assimilated within SAFRAN analysis scheme. CrocO also proposes various localisation strategies to assimilate snow observations. These approaches are evaluated in a Leave-One-Out setup against the operational deterministic model and its ensemble open-loop counterpart, both running without HS assimilation. Results show that intermediate localisation radius of 35–50 km yield a slightly lower root mean square error (RMSE), and a better Spread-Skill than the strategy assimilating all the observations from a whole mountain range. Significant continuous ranked probability score (CRPS) improvements of about 13 % are obtained in the areas where the open-loop modelling errors are the largest, e.g. the Haute-Ariège, Andorra and the Extreme Southern Alps. Over these areas, weather station observations are generally sparser, resulting in more uncertain meteorological analyses, and therefore snow simulations. In-situ HS observations thus shows an interesting complementarity with meteorological observations to better constrain snow cover simulations over large areas.

Using snow depth observations to provide insight into the quality of snowpack simulations for regional-scale avalanche forecasting

“…While precipitation biases play an important role in these difference, errors in modelled snow depth are influenced by many processes in snowpack models including the parameterization of snow density (Helfricht et al, 2018), handling of precipitation type, and simulating settlement processes. While this simple precipitation-adjustment method is not the state of the art in data assimilation (Cluzet et al, 2022;Winstral et al, 2018), it provides a simple approach to analyse the impact of precipitation errors in different regions.…”

“…The precipitation-adjustment method applied in this study is likely insufficient for an operational model system, as it assumes a constant bias over time and oversimplifies the causes of snow depth errors. More advanced data assimilation methods have recently been suggested for snowpack models (Largeron et al, 2020), including methods presented by Winstral et al (2018) and Cluzet et al (2022) that use similar snow observation networks in Europe. Cluzet et al (2022) found assimilating snow depth observations improved simulations in areas of France with relatively sparse observations but that the density of snow observations was correlated with the density of precipitation observations.…”

“…More advanced data assimilation methods have recently been suggested for snowpack models (Largeron et al, 2020), including methods presented by Winstral et al (2018) and Cluzet et al (2022) that use similar snow observation networks in Europe. Cluzet et al (2022) found assimilating snow depth observations improved simulations in areas of France with relatively sparse observations but that the density of snow observations was correlated with the density of precipitation observations. For large regions in western Canada, sparse and inconsistent observations would pose a challenge to implementing data assimi-lation methods, especially considering snow depth changes over short timescales were poorly resolved with the upscaled observations (Fig.…”

“…Many of these datasets have limited application for real-time monitoring because they are point-scale observations and require dedicated data collection campaigns. Quéno et al (2016) and Vionnet et al (2019) performed regional-scale comparisons of modelled snow depths with spatially distributed point observations, and Winstral et al (2018) and Cluzet et al (2022) have recently developed systems to assimilate snow depth observations into snowpack models. Real-time model verification systems would be valuable for avalanche forecasters learning how to interpret snowpack model output.…”

The Multiple Snow Data Assimilation System (MuSA v1.0)

“…Due to the harsh environmental conditions that often prevail in snow-dominated areas, in situ monitoring of the snow-E. Alonso- González et al: MuSA: the Multiple Snow Data Assimilation System (v1.0) pack based on automatic devices and weather stations is both costly and logistically challenging. In addition, due to the spatiotemporal variability in the snowpack (López-Moreno et al, 2011), even dense monitoring networks may suffer from a lack of representativeness (Molotch and Bales, 2006;Cluzet et al, 2022). Yet, estimating SWE spatial distribution is important to make accurate predictions of snowmelt runoff in alpine catchments.…”