Abstract
The wave propagation properties of short internal gravity waves (wave length λ_??_10km) are very different from those of long ones (λ-100km) owing to nonhydrostatic effects. Considering this observation, we parameterize the orographic gravity wave drag (GWD) in two ways. The major difference between two schemes is in the vertical partitioning of drag forcing, i, e., one weighs mainly in the stratosphere (type A) and the other in the troposphere (type B). We apply them to a global numerical weather prediction model and study their impacts on medium-range forecasts.
These two schemes individually reduce systematic forecast errors and their combination achieves the best forecast skill. In the troposphere, the impacts of these two schemes, including time evolutions, are very similar to each other. Hence, the troposphere is considered to be insensitive to the vertical partitioning of GWD at least within a medium-range time scale. In the case of the type A scheme, most of the drag forcing given to the lower-stratospheric mean-flow is rapidly transferred downward and contributes to change in the tropospheric circulations. In the stratosphere, the impacts of the type A scheme is much larger than those of the type B, especially in a short-range time scale. The stratospheric impacts of the type B scheme gradually increase after a few days, possibly corresponding to a time scale required for the vertical propagation of Rossby waves excited in the lower troposphere. The improvement of forecasts is evident in the zonal-mean fields. Although the type A scheme almost eliminates the westerly errors in the stratosphere, there still remain westerly errors in the troposphere. The type B scheme effectively reduces the remaining tropospheric errors. This may justify a need of tropospheric drag forcing, like the type B scheme.
