Towards a General Snow Thermal Conductivity Scheme in Land Models

In high latitude regions where snow-cover persists through much of the winter, snowpack affects soil temperature, permafrost, and carbon dynamics. During winter, snow acts as a strong insulator preventing energy release from the soil to the atmosphere and attenuating the penetration of cold temperatures into the ground. Thus, deeper snowpack is often associated with warmer subsurface soil temperatures. This insulation function of snow is attributed to the small but highly variable thermal conductivity of seasonal snowpack, which is determined by the layered and microstructural arrangement of ice, air and liquid water. In land models, snow thermal conductivity is usually represented by an empirical parameterization derived from limited in-situ or laboratory observations. These observations and the resulting parameterizations are most applicable to areas with snow climatology similar to the observed conditions and thus lack the general representativeness that is desired for use in a global land model. Using a number of snow parameterizations, we characterize land model uncertainty in simulating subsurface soil temperature at cold region stations where snowpack and soil temperature observations are available. Despite a close agreement between simulated and observed snowpack states, different snow thermal conductivity parameterizations produce substantial differences in simulated soil temperature and, in regions with permafrost, active layer thickness. By prescribing the model upper boundary conditions using observed snow characteristics and thus reducing the uncertainty associated with snow thermal mass, we identify and map the optimal snow thermal conductivity parameterizations that reduce soil temperature errors at global northern latitude sites. Using local site climatology associated with the model errors, we relate the most appropriate snow thermal conductivity scheme to snow climatology characteristics. In such a manner, we present a framework that can inform the optimal selection of snow thermal conductivity parameterization for land model simulations at regional to global scales. Validation using an independent set of in-situ observations reveals that the newly assembled snow parameterization scheme improves our ability to simulate subsurface soil temperature and thus permafrost dynamics.