CrocO_v1.0: a particle filter to assimilate snowpack observations in a spatialised framework

Abstract. Monitoring the evolution of snowpack properties in mountainous areas is crucial for avalanche hazard forecasting and water resources management. In situ remotely sensed observations provide precious information on state but usually offer limited spatio-temporal coverage bulk or surface variables only. particular, visible–near-infrared (Vis–NIR) reflectance can about are by terrain shading clouds. Snowpack modelling enables estimation any physical variable virtually anywhere, it affected large errors uncertainties. Data assimilation offers a way to combine both sources propagate from observed non-observed areas. Here, we present CrocO (Crocus-Observations), an ensemble data system able ingest observation (applied as first step height snow (HS) Vis–NIR reflectances) spatialised geometry. uses simulations represent uncertainties particle filter (PF) reduce them. The PF prone collapse when assimilating too many observations. Two variants were specifically implemented ensure that observational propagated space while tackling this issue. global algorithm ingests all available with iterative inflation errors, klocal localised approach performing selection assimilate based background correlation patterns. Feasibility testing experiments carried out identical twin experiment setup, synthetic HS reflectances only one-sixth simulation domain. Results show compared against runs without assimilation, analyses exhibit average improvement equivalent continuous rank probability score (CRPS) 60 % 40-member 20 CRPS 160-member ensemble. Significant improvements also obtained outside These promising results open possibility real context.