The land surface tiling is a solution to represent the natural and anthropogenic heterogeneity of the Earth's surface in numerical weather and climate prediction models. This is computationally inexpensive compared to running a higher resolution land-atmosphere model and represents an alternative to the so-called effective-parameters approach that aims at obtaining the main features of the process (e.g. bulk canopy temperature, grid-box average evaporation or momentum fluxes). While the tiling approach is widely adopted in modern schemes, as thought to provide a better compromise for the physical realism of the surface processes, experimental work showing the benefits and shortcomings is still pretty limited. An attempt to quantify experimentally potential and limitations involved with this method is proposed over several field-sites with contrasted land-use and where dedicated observations (e.g. eddy-covariance flux-towers) were made available. In particular errors in specifying the land use, due to natural or anthropogenic disturbances of the vegetation state (e.g. natural fires and deforestation) or to simplifications and inaccuracies in the land surface scheme (e.g. negligence of inland water-bodies), will be estimated. Other approaches alternative to the tiling will be discussed and preliminary results will be presented.