Cold climates, complex hydrology: can a land surface model accurately simulate deep percolation?

Abstract. Cold regions present unique challenges for land surface models in simulating deep percolation or potential groundwater recharge. Previous model evaluation efforts often overlooked these regions and did not account for various sources of uncertainties influencing model performance. This study uses high-resolution integrated lysimeter measurements to comprehensively assess the performance of the Soil, Vegetation, and Snow (SVS) land surface model in a cold climate. SVS performs well in the daily snow depth simulation, with a correlation coefficient (r) greater than 0.94 and a mean bias error (MBE) smaller than 3.0 cm for most of the simulation period. The newly implemented soil-freezing scheme simulates the near-surface soil temperature reasonably well (r: 0.89), with a slight cold bias (MBE: −0.8 °C). However, the results show that SVS is limited in matching the temporal dynamics of deep percolation (daily timescale). In addition, it significantly underestimates deep percolation (r: 0.35, MBE: −0.8 mm d−1) and near-surface soil moisture (MBE: −0.058 m3 m−3) during cold months. This is likely to be related to the model's inability to represent frozen-soil infiltration and preferential flow. These limitations must be addressed to make SVS a reliable tool for simulating deep percolation in cold environments. The findings highlight the importance of a comprehensive model evaluation to identify key deficiencies and to guide future model development efforts to improve hydrological simulations in cold regions. Such improvements lead to more informed decision-making regarding groundwater resource management in a changing climate.