The next-generation GEOS-5 data assimilation system developed by NASA's Global Modeling and Assimilation Office contains several new components. One of these is the GEOS-5 model, a newly developed general circulation model based on the flux-form semi-Lagrangian dynamical core by Lin and Rood, combined with a new physics parameterization package. The physics part of the model in particular is undergoing continued development and tuning. In the initial version of the model, the boundary layer parameterization was based on a simplified version of the Lock scheme (Lock et al 2000). For example, the non-local term in the equation for the temperature flux is neglected. Several prognostic quantities - e.g. precipitation rates, low level cloud distribution and cloud height - are highly dependent on the boundary layer parameterization used. Tests have been carried out with a more recent implementation of the Lock scheme, which includes the use of a more appropriate turbulent velocity scale, both for the calculation of the temperature excess and for the formulation of the eddy diffusivity for the surface driven turbulence. The new implementation also includes a term representing non-local fluxes of heat and moisture, with the aim of providing a more realistic representation of the boundary layer and thereby indirectly also of other related model variables. Changes in the moisture tendency can have a large impact on the availability of the water for the convection scheme. In this presentation, we provide an overview of the primary impact of the boundary layer parameterization, and we show the effect of two different schemes on the mean low-level cloud field and precipitation rates.