Improved Representation of Land-surface Heterogeneity in a Non-hydrostatic Numerical Weather Prediction Model

This study focuses on the relevance of accurate surface parameters, in particular soil moisture, and of parameterizations for heterogeneous land surfaces, for the prediction of sensible and latent heat fluxes by a mesoscale weather forecast model with horizontal grid resolution of 7 km. The analysis is based on model integrations for a 30-day period, which are compared both to flux measurements obtained from the LITFASS-2003 field experiment and to high-resolution-model (1-km grid spacing) results. At first, the relevance of improved parameter sets and input data compared to usual operational practice for an accurate prediction of near-surface fluxes is shown and discussed. It is demonstrated that an observation-based land-surface assimilation scheme leads to an improved soil moisture analysis, which is shown to be essential for the realistic simulation of surface fluxes. Secondly, the implementation of two efficient parameterization strategies for subgrid-scale variability of the surface, the mosaic and the tile approach, is presented. Using these methods, the simulations are in better agreement with measurements than simulations with simple aggregation methods that use effective surface parameters. Integrations with the mosaic approach reproduce high resolution simulations very well and more accurately than simulations with the tile method. Finally, the high resolution simulations are analyzed to justify and discuss the approximations underlying both methods.