Seasonal Snow Simulation: Sensitivity to Initialization, Parameterization, and Forcing

Abstract

Snow cover is a critical element of the Earth's climate system, influencing energy exchange and water storage. This study examines the sensitivity of seasonal snow simulations to initialization, parameterization, and external forcing to assess the influence on snow representation in seasonal simulations. Using the TERRA Standalone (TSA) land surface model with ERA5 reanalysis forcing, we investigate the effects of perturbations in initial snow water equivalent (SWE), albedo (parameterization), and precipitation (forcing) affect snow dynamics across Eurasia over ten winters.  A +10% perturbation in snow albedo during 3-month spring simulations leads to a median increase in snow cover area of 24%, which closely matches interannual variability.  Similarly, a +10% perturbation in precipitation during early winter simulations leads to a median of 28% increase in total snow mass, more than three times the interannual variability. Sensitivity analysis shows that seasonal snow simulations are most sensitive to precipitation perturbations, especially during the melt season, while initialization effects persist throughout the winter. The results show that snow cover during the melt season is strongly influenced by albedo and precipitation perturbations. The simulations show that overestimations have a stronger impact than underestimations. Mid- and late-winter initializations effectively reduce snow bias from precipitation and albedo perturbations. In addition, improved albedo parameterization can improve snow simulation accuracy during the melt season. Spatial analysis identifies snow margin zones as particularly sensitive to albedo and precipitation perturbations. These findings underscore the importance of accurate precipitation forcing and snow parameterization to ensure reliable seasonal snow simulations for climate forecasts and projections.

Simulation Setup

TERRA Standalone (TSA)

We use the TERRA Standalone (TSA) land surface model. TERRA is the land surface component of the weather and climate models COSMO (Consortium for Small-scale Modelling) model and the ICON (ICOsahedral Non-hydrostatic) model.

Thermal processes are represented by solving the heat conduction equation in the soil and snow layers, which can further lead to freezing/thawing of the soil or melting of the snow. These processes are represented in eight fixed layers down to 21.87 m, with the bottom layer being the so-called climate layer with a temperature prescribed to the climate mean near-surface temperature, and an additional snow layer if present. Hydrological processes are represented by solving the Richards equation to account for soil water transport, evapotranspiration, rain interception and infiltration, and runoff. They are considered only within the first six layers down to 2.43 m, with downward gravitational transport at the lower boundary.

Here we use a standalone version of TERRA (TSA), which can be forced with atmospheric datasets besides COSMO or ICON, such as reanalysis or station data. There are some notable snow-specific features of the model: TSA uses a single layer to represent snow. Snow aging is considered only for density and therefore also for thermal conductivity. However, snow albedo aging is not considered, but is assumed constant at 0.7 in TSA. In contrast, the TERRA model embedded in COSMO/ICON considers snow albedo aging. The albedo in TSA depends on snow cover, soil type, and vegetation cover. For precipitation, TSA assumes snowfall when the near-surface temperature is below 0 °C, and rainfall when above. It does not account for liquid water in the snowpack or for liquid water retention by the snow, resulting in direct runoff and soil infiltration in the case of rain on snow.

Experiments

We simulated ten winters from 2004/2005 to 2013/2014 with initial soil conditions and hourly atmospheric forcing from the ERA5 reanalysis. The initial soil conditions include SWE, snow density and temperature (if snow is present), soil temperature and moisture for each layer, water intercept by the vegetation canopy, and skin temperature. The forcing includes hourly near-surface specific humidity, temperature and horizontal winds, surface pressure, skin temperature, precipitation, net longwave and shortwave radiation, and albedo. The albedo and skin temperature are only needed to derive the downward radiation fluxes from the forcing. We use a rotated longitude-latitude grid with a spacing of 0.25° × 0.25° (similar to the ERA5 forcing) covering northern and central Eurasia, including the Tibetan Plateau. It should be noted that the ERA5 data for initialization (e.g., SWE) and forcing (e.g., precipitation) are not fully consistent with each other due to the applied land data assimilation, i.e., if there is a bias in snowfall, the snow depth is adjusted considering observations.

There is no continuous run over the ten years; each winter simulation from 2004/2005 to 2013/2014 is started again in October and runs until the end of May, and each experiment is also started with shorter lead times, i.e., every following month. For the perturbation of the initial conditions, we perturb the initial SWE by ±10% compared to the reference, separately for a positive and a negative perturbation. The same is done for albedo and precipitation, but with permanent perturbations throughout the simulation. In addition, we repeated all experiments with ±20% perturbations to roughly cover the range of uncertainties. Due to the constant albedo in TSA, the values for the snow albedo are 0.56, 0.63, 0.7, 0.77, or 0.84, according to the prescribed relative change. We calculated the median, minimum, and maximum values for each simulation over the ten winters and will focus on these statistics in the following analysis.