Soil properties vary within centimeters, which is not captured by state-of-the-art land-surface models due to their kilometer-scale grid. This mismatch can lead to systematic errors when simulating the exchange of energy, water, and greenhouse gases between the land and atmosphere—collectively referred to as “aggregation error.”</p> <p>To quantify the potential aggregation error of soil temperature, we developed the two-dimensional pedon-scale geophysical soil model DynSoM-2D, which has a spatial resolution of 10 cm. We applied DynSoM-2D at a permafrost-affected, non-sorted circle site using three different setups: (i) a homogeneous soil profile representing a typical land surface model, compiled by averaging the heterogeneous soil inputs; (ii) the actual heterogeneous soil profile of a typical non-sorted circle; and (iii) the heterogeneous soil profile including lateral heat fluxes.</p> <p>Our results show that DynSoM simulates warmer soil temperatures when heterogeneous soil properties are considered, with this warming becoming even more pronounced and consistent across the domain when lateral heat fluxes are included. By aggregating grid cells, we traced the aggregation error back to the spatial distribution of organic matter, which nonlinearly alters soil thermal and hydrological properties, leading to the observed differences between simulations. In our case, the heterogeneity-induced warming led to a deepening of the active layer and an extension of the snow-free period, both of which can strongly alter ecosystem dynamics, while having only a minor effect on soil-atmosphere heat exchange on an annual basis.