Under strong warm advection, sensible and latent heat fluxes may provide larger energy for surface snowmelt than does net radiation flux. With these thermally stable conditions, the height of the first model level may be well above the surface-layer depth and thus outside the range of applicability of the surface-layer similarity theory on which the models' surface thermal flux computation is based. This situation can strongly affect the magnitude of simulated surface thermal fluxes and snowmelt. To explore this issue, the impact of selected heights of the first model level on the simulated surface fluxes and snowmelt under stable surface stratification conditions was investigated. Simulations using a mesoscale atmospheric model considering two extreme contrasts in surface roughness were performed. Setting the first model level to 3 or 10 m, which typically was within the stable surface layer, yielded nearly the same contribution of simulated surface turbulent thermal fluxes to snowmelt. When the first model level height was set at about 40 m, as is used in many regional model simulations, it exceeded the depth of the stable surface layer over the snow cover. The surface turbulent thermal flux contribution in this case was smaller (by about 40%), with a directly proportional effect on the snowmelt. Pending observational support, results presented in this study imply that setting a model's lowest level to 10 m or less will likely improve simulated snowmelt accuracy during warm advection.