Abstract. Before operational use or for decision making, models must be validated, and the degree of trust in model outputs should be quantified. Often, model validation is performed at single locations due to the lack of spatially-distributed data. Since the analysis of parametric model uncertainties can be performed independently of observations, it is a suitable method to test the influence of environmental variability on model evaluation. In this study, the sensitivities and uncertainty of a physically-based mountain permafrost model are quantified within an artificial topography consisting of different elevations and exposures combined with six ground types characterized by their hydraulic properties. The analyses performed for all combinations of topographic factors and ground types allowed to quantify the variability of model sensitivity and uncertainty within mountain regions. We found that modeled snow duration considerably influences the mean annual ground temperature (MAGT). The melt-out day of snow (MD) is determined by processes determining snow accumulation and melting. Parameters such as the temperature and precipitation lapse rate and the snow correction factor have therefore a great impact on modeled MAGT. Ground albedo changes MAGT from 0.5 to 4°C in dependence of the elevation, the aspect and the ground type. South-exposed inclined locations are more sensitive to changes in ground albedo than north-exposed slopes since they receive more solar radiation. The sensitivity to ground albedo increases with decreasing elevation due to shorter snow cover. Snow albedo and other parameters determining the amount of reflected solar radiation are important, changing MAGT at different depths by more than 1°C. Parameters influencing the turbulent fluxes as the roughness length or the dew temperature are more sensitive at low elevation sites due to higher air temperatures and decreased solar radiation. Modeling the individual terms of the energy balance correctly is hence crucial in any physically-based permafrost model, and a separate evaluation of the energy fluxes could substantially improve the results of permafrost models. The sensitivity in the hydraulic properties change considerably for different ground types: rock or clay for instance are not sensitive while gravel or peat, accurate measurements of the hydraulic properties could significantly improve modeled ground temperatures. Further, the discretization of ground, snow and time have an impact on modeled MAGT that cannot be neglected (more than 1°C for several discretization parameters). We show that the temporal resolution should be at least one hour to ensure errors less than 0.2°C in modeled MAGT, and the uppermost ground layer should at most be 20 mm thick. Within the topographic setting, the total parametric output uncertainties expressed as the standard deviation of the Monte Carlo model simulations range from 0.1 to 0.5°C for clay, silt and rock, and from 0.1 to 0.8°C for peat, sand and gravel. These uncertainties are comparable to the variability of ground surface temperatures measured within 10 m × 10 m grids in Switzerland. The increased uncertainties for sand, peat and gravel is largely due to the high hydraulic conductivity.
Modeling the Spatial Distribution of Snow Cover in the Dudhkoshi Region of the Nepal Himalayas
