This study integrated spatially distributed field observations and soil thermal models to constrain the impact of frozen ground on snowmelt partitioning and streamflow generation in an alpine catchment within the Niwot Ridge Long-Term Ecological Research site, Colorado, USA. The study area was comprised of two contrasting hillslopes with notable differences in topography, snow depth and plant community composition. Time-lapse electrical resistivity surveys and soil thermal models enabled extension of discrete soil moisture and temperature measurements to incorporate landscape variability at scales and depths not possible with point measurements alone. Specifically, heterogenous snowpack thickness (~0–4 m) and soil volumetric water content between hillslopes (~0.1–0.45) strongly influenced the depths of seasonal frost, and the antecedent soil moisture available to form pore ice prior to freezing. Variable frost depths and antecedent soil moisture conditions were expected to create a patchwork of differing snowmelt infiltration rates and flowpaths. However, spikes in soil temperature and volumetric water content, as well as decreases in subsurface electrical resistivity revealed snowmelt infiltration across both hillslopes that coincided with initial decreases in snow water equivalent and early increases in streamflow. Soil temperature, soil moisture and electrical resistivity data from both wet and dry hillslopes showed that initial increases in streamflow occurred prior to deep soil water flux. Temporal lags between snowmelt infiltration and deeper percolation suggested that the lateral movement of water through the unsaturated zone was an important driver of early streamflow generation. These findings provide the type of process-based information needed to bridge gaps in scale and populate physically based cryohydrologic models to investigate subsurface hydrology and biogeochemical transport in soils that freeze seasonally.