Single-column model intercomparison for a stably stratified atmospheric boundary layer

In the frame of GABLS (GEWEX Atmospheric Boundary Layer Studies) of the WMO, a stably stratified case based on Arctic observations was proposed as a test case for a number of models, namely Large-Eddy models and single-column turbulence parameterisations.

Eighteen models have participated in the intercomparison. Among them the turbulence schemes of some of the main weather-forecast and climate centers of the world. Furthermore, many research turbulence schemes have submitted results.

The LES intercomparison (shown in another presentation by Beare et al) provides statistics that can be used as references for the single-column models. Basically the case has a wind-driven stably stratified PBL, with a PBL height of about 170 m, where a inversion separes it from the free atmosphere aloft. The temperature difference between the surface and the free atmosphere is about 3 K. The wind is above geostrophic at the middle of the inversion and returns to the geostrophic at about 230 m, with turbulence until this height.

The models are preliminarily classified according to their order of closure, which ranges from first order models (that includes most of the operationally running schemes) to TKE and k-epsilon schemes.

The results of the models show a very large dispersion. The main differences occur at the PBL top, where the strength of the inversion and the level of the return of the wind to the geostrophic values vary a lot between models. The vertical extension of the turbulence is also very different at that layer, since there are models that do not allow turbulence there by construction.

The mixing coefficient approach computed by local values seems to be a good solution for the interior of the PBL, since most of the models using this option are able to generate profiles parallel to the LES output. The models show small sensitivity to the spatial and temporal discretisation.

Finally a sensitivity exercise has been made, allowing some models to vary the characteristics of their schemes, and almost every model has been able to reproduce the LES results with small changes in their proposals.