Exploring variability of model results in the GEWEX Atmospheric Boundary Layer Study (GABLS)

Within the GEWEX Atmospheric Boundary Layer Study (GABLS), inter-comparison studies are organized for column models using a prescribed surface temperature tendency. It appears that the results for both the first and second GABLS show a significant variability in the surface fluxes and atmospheric profiles, despite the relatively simple boundary condition. The variability is directly related to the different parameterizations of the different models. However, due to the known interdependence of the surface temperature and surface sensible heat flux, a prescribed surface temperature as a boundary condition may be too simplified. In this paper we investigate to what extent the degree of variability among the models is influenced by not allowing for surface feed-back. To this end we use the set up of the second GABLS inter-comparison case and study three diurnal cycles of the boundary layer over land. Firstly, we use our state-of-the-art, first order closure model and vary the parameters in the turbulence scheme for both stable and unstable conditions in a reasonable range. As such 16 model runs are performed, all with prescribed surface temperature tendency as inspired by the observations. Secondly, the exercise is repeated, but then using an interactive prognostic equation for the surface temperature. It appears that with the prescribed surface temperature, the variability of the surface fluxes and the nighttime profiles as seen within the second GABLS case can also be reproduced just be taking a single model. Contrary, the daytime profiles show smaller variability. The model variability of the atmospheric temperature profile (when integrated over the model domain) shows an exponential growth in time. In the coupled case, the temperature profile variability is a factor 2 smaller than with prescribed surface temperature, although wind speed variability is slightly increased. For a lower geostrophic wind forcing the variability between the models decreases strongly.