Atmospheric mesoscale numerical models are commonly used not only for research and air quality studies, but also for other related applications, such as short-term weather forecasting for atmospheric, hydrological, agri- cultural and ecological modelling. A key element to produce faithful simulations is the proper representation of the soil parameters used in the initialization of the corresponding mesoscale numerical model. The Regional Atmospheric Modeling System (RAMS) is used in the current study. The model code has been updated in order to permit the model to be initialized using a heterogeneous soil moisture and temperature distribution derived from land surface models. Particularly, RAMS has been adapted to incorporate the Global Land Data Assimilation System (GLDAS) dataset for the initialization of the corresponding soil parameters. The results obtained using this heterogeneous initialization are compared to the model results obtained by the default homogeneous RAMS initializations. A series of numerical experiments have been conducted for a 7-days period over eastern Spain within the 2011 summer season. The selected period covers different typical summer atmospheric situations from the region of study. Ground data from two FLUXNET stations, together with the measurements registered by a portable weather station, located over the region of study, and other permanent weather stations are used for the result assessment. Incorporating the GLDAS product in the initialization of RAMS has been found to remarkably improve the representation of surface sensible weather parameters. On the other hand, significant differences are still observed in the proper simulation of the surface parameters when the model is applied to well vegetated areas in comparison to those obtained over poor and/or sparsely vegetated regions. Considering the better agreement found in this latter case, we have performed several sensitivity tests regarding land-surface- atmosphere coupling with the aim of improving the original results over well vegetated areas.