Abstract. This study explores the influence of implementing a multi-layer snow scheme on the climatological bias within a seasonal forecast system. Traditional single layer snow schemes in land surface models, particularly those utilising a composite snow-soil layer, often inadequately represent the insulating effect of snowpack, resulting in cold and warm biases during winter and snowmelt seasons, respectively. By contrast, multi-layer snow schemes improve the simulation of energy exchange between the land surface and atmosphere by realistically capturing snowpack thermal processes. To examine this impact, two sets of LSM offline experiments are conducted, using either a single-layer or a multi-layer snow scheme. Results show that the multi-layer configuration better reproduces the observed Northern Hemisphere snow seasonality. To further assess the role of snow insulation in coupled forecast systems, two sets of experiments with the Global Seasonal Forecast System (GloSea) version 6 are compared over 24 years (1993–2016) corresponding to the incorporation of single- (G6single) and multi-layer (G6multi) snowpack schemes. In G6multi, the onset of snowmelt is delayed by approximately 1–2 weeks, postponing springtime evaporation, slowing soil moisture depletion, and improving the memory of soil moisture. Increased soil moisture enhances the partitioning of available energy into latent heat flux, thereby promoting evaporative cooling and suppressing excessive water-limited land–atmosphere coupling. The improved model fidelity of land-atmosphere interactions, particularly over mid-latitude regions, mitigate near-surface warming biases across the entire diurnal period and reduce the sensitivity of atmospheric conditions to land surface variability. The model performance in simulating precipitation is also improved with the increase in precipitation occurrence over snow-covered regions. Above all, this study demonstrates the value of implementing a multi-layer snowpack scheme in seasonal forecast models, not only during the snowmelt season but also for the subsequent summer season, for model fidelity in simulating temperature and precipitation along with the reality of land-atmosphere interactions.
Seo et al. (Wed,) studied this question.