Clear-sky stable boundary layers with low winds over snow-covered surfaces. Part 2: Process sensitivity

This study evaluates the relative impact of snow-surface coupling, long-wave radiation, and turbulent mixing on the development of the stable boundary layer over snow. Observations at three sites are compared to WRF single-column model (SCM) simulations. All three sites have snow-covered surfaces but are otherwise contrasting: Cabauw (Netherlands, grass), Sodankylä (Finland, needle-leaf forest) and Halley (Antarctica, ice shelf). All cases are characterized by stable, clear-sky, and calm conditions. Part 1 of this study determined the optimal SCM forcing strategy. In this study, the process intensities from that reference are perturbed to study their relative significance and to assess which process could be responsible for the most optimal agreement between model and observation. The analysis reveals a large variability in the modelled atmospheric state and surface parameters. Overall, the modelled gradients of temperature and moisture are under-estimated but decreasing the process intensities improves this. The impact is strongest with reduced mixing, though this then causes the model to overestimate the near-surface wind speed. To study the surface energy balance terms, we use so-called ‘process diagrams’. The achieved variation between the sensitivity runs indicates the model sensitivity to each process. The overall sensitivity is similar for the three sites but the relative offsets in the position of the sensitivity runs with respect to the observations differ, hampering general recommendations for model improvement. Furthermore, sometimes a meaningful interpretation of observations is troublesome, which hampers the comparison with model results. Radiation is relatively more important at Cabauw and Sodankylä, whilst coupling plays a more important role at Halley. The sensitivity analysis is performed with two boundary-layer schemes (MYJ, YSU). YSU generates larger, more accurate gradients of atmospheric temperature and humidity, while wind speeds are predicted better with MYJ. The behaviour of an increase in 2 m temperature with decreased mixing is most obvious with YSU.