A New Lagrangian Adaptive Multi-Layer Snow (LAMS) Scheme for Modelling Snowpack Evolution

Snow cover dynamics have profound impacts on the global climate system, hydrological cycle, and freshwater resources. Traditional snowpack models embedded in land surface models often employ layered or Euler scheme, which face challenges such as numerical diffusion and imprecise interface tracking when handling snow layer evolution, phase changes, and mass transport processes and may not be sufficient for snowpack simulation. Here we present a Lagrangian, layer-tracking snow model—LAMS (Lagrangian Adaptive Multi-Layer Snow model)—that explicitly tracks the physical state changes of each snow layer, capturing the evolution of snow layers during compaction, freeze-thaw cycles, and rainfall/snowfall events. The model uses an adaptive layer management algorithm, which automatically adds new layers during snowfall events and removes layers that have melted away, thereby efficiently capturing the dynamic layered structure of the snowpack. Each snow layer, treated as an independent unit, evolves in thickness, density, liquid water content, and ice content under coupled mass and energy balance equations. Because LAMS is not constrained by a fixed layer count, it naturally handles snowfall-driven accretion and melt-driven ablation without remeshing artifacts, mitigating numerical diffusion and improving interface fidelity. Across multiple SnowMIP sites, LAMS reproduces snow depth, snow-water equivalent (SWE), and internal temperature profiles with high skill. Although a few existing snowpack-specialized models such as SNOWPACK and CROCUS also employ Lagrangian theory, this paper proposes a distinct framework based on coupled governing equations and an explicit numerical method. This approach can be more readily integrated into land surface models (LSMs) while still providing robust performance across a wide range of snowpack characteristics. The model offers a substantial improvement over simpler schemes, such as the two-layer snow module in VIC (Variable Infiltration Capacity) model. The VIC scheme's simplified representation of energy transfer and phase changes often responds too rapidly and strongly to energy inputs, which can systematically overestimate snowmelt. By contrast, the vertically resolved, layer-tracking LAMS framework provides a more physically consistent depiction of snowpack evolution, resulting in more accurate simulations of snow accumulation and melt. Coupling LAMS with land-surface or hydrological models will enable more accurate assessments of snow dynamics and water-resource impacts under climate change.