Because snow melt-water supplies a significant component of the annual water budget in many areas, the knowledge of snowpack dynamics is of critical importance to several real-time applications such as agricultural production, water resource management, flood prevention and hydropower generation. With the aim of improving the hydrological predictions in snow-dominated areas, an increasing interest focuses on the combined used of different sources of information (namely ground-based measurements and remotely sensed observations) by assimilating observed data within models. Several data assimilation (DA) techniques with different degrees of complexity have been developed and are currently employed. Generally, the research community agrees on the superior performance of the multivariate DA with respect to the univariate one. This study intends to assess the feasibility of a multivariable DA scheme for snow modelling through two of the most widely-used sequential ensemble-based DA techniques, namely the Ensemble Kalman Filter (EnKF) and the Particle Filter (PF). A dual purpose firstly aims at identifying and overcoming the most constraining limitations in implementing multivariate DA schemes within a snow module. Secondly, the goal is to analyze the main differences in the effectiveness of the two selected DA techniques in consistently updating the snowpack states. The main weaknesses and strengths of both EnKF and PF schemes are investigated to assess their suitability to be operationally effective for real-time hydrological applications. The modelling system consists of a newly developed multilayer snow energy-balance model coupled with a multivariable DA scheme. The system is tested at three Alpine sites: Torgnon (Italy), Col de Porte (France) and Weissfluhjoch (Switzerland). Both DA configurations are analyzed in order to assess their performances in assimilating snow-related in-situ measurements and the resulting impact on snow model predictions for hydrological purposes, under changing local conditions. Both DA techniques prove to be suitable for being implemented within a multivariable DA scheme, since they allow to take into account different sources of uncertainty. With respect to the EnKF, the PF technique generally better handles the model nonlinearities and it allows to maintain the internal physical consistency of each ensemble member.