This report proposes improvements of the land surface model (LSM) used in the turbulence-resolving Dutch Atmospheric Large-Eddy Simulation model (DALES). Important changes include the infiltration of precipitation, the parametrization of the soil hydraulic functions and the formulation of the soil water extraction by vegetation roots. The performance of the improved LSM version is validated using offline simulations and observations from the CESAR meteorological observatory in Cabauw (Netherlands), during the year of 2015. An optimal parameter set for this location set is obtained from both satellite retrievals and in-situ observations. In particular, the seasonal dependence of both the Leaf Area Index (LAI) and the roughness lengths for heat and momentum is prescribed to the model.

The offline simulations determined that the improved version of the LSM is able to model the surface fluxes during the year of 2015 with an accuracy of 10 W/m2. This accuracy was found to be strongly dependent on the chosen land surface parameters such as the LAI, the roughness lengths, the soil textures and the vegetation root profiles. Furthermore, the lower boundary for soil water transfer, set to free drainage, was found to cause too much vegetation stress after an extended simulation time.

Finally, the LSM is tested in a fully-coupled mode during three different meteorological events observed at the CESAR observatory. Firstly, stratocumulus simulations confirm the importance of the surface evaporation on the development of the stratocumulus layer, which is considerably reduced compared to simulations that assume the surface to be saturated with water. Secondly, the growth of the clear convective boundary layer was found to be greatly affected by the partitioning of the surface fluxes, as the sensible heat flux plays a key role in entrainment. Although the modelled sensible heat flux was found to be in agreement with the observations, the growth of the boundary layer was still substantially underestimated. Lastly, it was found that it is possible to model the main development stages of radiation fog in DALES, despite the fact that the modelled fog layer appears to be too persistent.