Boundary Layer Processes before the Initiation of Convection over Land in Semi-Arid Conditions analysed in Large-Eddy Simulations

During the African Monsoon Multiscale Analysis (AMMA) campaign that took place in 2006 in West Africa, a case of daytime development of deep convection above Niamey was particularly well described by several observational datasets on 10 July 2006 (Lothon et al.,MWR, 2011). This case-study, representative of convective situations encountered in semi-arid conditions, is characterized by an important growth of the convective boundary layer reaching more than 2000m, a long-lasting shallow convection phase persisting for several hours prior to a late afternoon initiation of deep convection. From these observations, a set-up for a Large-Eddy simulation (LES) was built. A simulation at 200m resolution over a 100km-wide domain was performed which represents satisfactorily most of the observed features (Couvreux et al., QJRMS, 2012). A statistical analysis of the boundary-layer structures is carried out using distributions and joint-distributions of thermodynamic and dynamic variables. It highlights the large thickness of the entrainment zone, which is characteristic of this area. In this particularly deep boundary layer, the potential temperature and water vapor mixing ratio are not well-mixed in the upper part of the boundary as revealed by their double peak distribution. Clouds have been identified individually and the evolution of their characteristics and links with the subcloud layer is analyzed as a function of their timing within the transition from shallow to deep convection. Eventually, the cold pools that appear at the end of the simulation have also been sampled. This analysis underlines significant variability of humidity within the cold pools, in particular, they are not always drier than the environment. The interest of this set-up is that it can be used for single-column models as well, which allows direct comparisons of explicit and parameterized simulations. Here, a single-column version of a mesoscale model is used to evaluate the transport realized in the boundary layer using tracer-based diagnostics implemented in the LES.