Wind turning across the marine atmospheric boundary layer

The directional wind turning across the boundary layer in short-range forecasts made as part of the recent ERA-40 reanalysis project is compared with radiosonde data from oceanic stations in the North Atlantic. Significant errors are identified, particularly in cases of warm advection in which it is found that the observed wind turning is systematically underestimated. Similar problems are found in the higher-resolution operational forecasts, and also in forecasts made with the independent Met Office model. Furthermore, comparisons of forecast wind direction with scatterometer data give consistent results over the North Atlantic and indicate that the same problems are present over other oceans. A large fraction of the cases with the largest errors are shown to have stable boundary layers. In these cases the modelled boundary layers are typically too deep, and sensitivity tests have shown that more realistic structures are obtained by switching to a stable boundary-layer parametrization based on Monin-Obukhov similarity. Some possible problems with the convective boundary layer parametrization in highly baroclinic conditions are also identified.