In mountain areas, most things vary strongly over short distance - examples of this are: elevation, ground material and the amount of sunshine. Similarly, the mountain cryosphere (snow, glaciers, permafrost) is subject to strong variability: A southern slope may have blossoming flowers, while at 20m distance a northern slope is snow-covered, or, horizontal ground at 2600m elevation can have a grass cover while less than 1 km away, horizontal ground at 3600m is glacier covered. This variability is remarkable, because similar gradients in flat terrain would stretch over distances of many hundreds or thousands of km. Techniques for estimating conditions at the land surface often use computer models that represent the physics of important processes. Here, simulating land surface conditions in mountains requires a fine spatial resolution to capture the strong lateral variability. The fine resolution, however, comes at a price: it requires a lot of computer resources and limits such model simulations to small areas, only. In CRYOSUB, we will evaluate a technique to represent the most important part of the variability of land surface in mountain areas with a strongly reduced amount of computations. Distributed models calculate points at a regular spacing (e.g., every 25m). In this project we will use a so-called lumped model that exploits similarity between places that are alike. This means that only few points with certain characteristics are computed and all other points in between will be interpolated. As an example: steep south-facing rock at 2700m is likely to behave in a predictable way based on simulations for 2500m and 3000m south facing situations. Also a lumped model can have different resolutions - in the previous example, one could e.g., additionally calculate at 2750m. This approach may save a factor of 1,000-100,000 of computing resources compared to distributed models. This opens the possibility to quantify and investigate phenomena in mountains using resources for different purposes: covering a larger area, having results quicker, or quantifying uncertainty. This new method will have a benefit (saved computations) and a cost (lost quality of the simulation). We will use one high-resolution distributed model results and measurements as a baseline and compare the cost and benefit of diverse model resolutions, both with the new lumped and the distributed method. This method has the potential to improve our ability to quantify the impact of climate change in mountain areas and help answer questions such as: Where is permafrost? Where will it thaw the fastest? How will the snow cover on south slopes change in the coming decades?