Land surface models operate on coarse grid cells that typically have sizes of many kilometers. However, many land-surface processes (e.g. evaporation, surface energy balance, permafrost occurrence) are controlled by the small-scale topography within each grid cell. This "subgrid" variability can now be included in some of the models through so-called "tiling", i.e. the land-surface processes are simulated not only once for a grid cell, but for several "subgrid tiles" which can for example feature different expositions and altitudes.
In this thesis, you will investigate how one can determine the "optimal set of tiles" and their properties for a given grid cell. You will work with high-resolution DEMs and then apply clustering techniques to determine e.g. the ten most typical pairs of slope, aspect and altitude within the grid cell (in case ten tiles are used in the model). With this, one would for example be able to represent the effects of subgrid topography in a grid cell where the south slopes in reality are systematically steeper than the north slopes (like in many areas in Mongolia).
The goal of the thesis is to compile an algorithm that has the coarse grid cell outlines and a high-resolution DEM as input and outputs the tile properties slope, aspect and altitude for a given number of tiles for each grid cell.
The work can be extended by including additional factors like land cover in the tiling scheme, and/or using the CryoGrid land surface model to demonstrate the effect of the tiling scheme on simulations of the land-surface processes. We also have field data sets from e.g. Norway, Svalbard, or Mongolia that can be used for validation. For the thesis, programming skills (Python, Matlab, or similar languages) are required.
While this thesis sounds and is rather technical, you will be working on one of the biggest limitations of today's land surface models which negatively affects the performance of e.g. weather forecasts and climate model projections.