The interface between atmosphere and land surface plays an important role for both atmospheric dynamics and land surface conditions. Land surface processes such as infiltration, evapotranspiration, and sensible heat exchange significantly affect the simulation results of large-scale climate models as well as regional atmospheric models. Therefore, land surface processes should be modeled appropriately. However, land surfaces are spatially inhomogeneous at the spatial scales of vegetation canopies to watersheds or even continents. In order to consider the spatial heterogeneity of land surface properties, a stochastic upscaling method for land surface process modeling for heterogeneous landscapes was developed by the authors. In the model, the Fokker-Planck equation (FPE) as a probabilistic model for land surface processes was formulated in a one-dimensional probability domain for land surface temperature. In the present study the upscaled land surface process model was applied to watersheds in order to verify its applicability and validity.