Cut-off low pressure systems over southern Australia: a numerical modelling study and sensitivity experiments

A series of numerical model simulations of a cut-off low over southeastern Australia on 1 to 2 September 1997 is described. The aim was to determine t he model configuration and variables that are vital to achieving a skillful forecast of the intensity of the cut-off low and of its associated precipi tation. The first step was a control run which used only the archived real- time initial and boundary conditions. The control predictions of sea-level pressure (SLP) and precipitation amounts and distribution were evaluated by both subjective and objective measures of skill. The predictions were rate d as good to very good, but it was also noted that the forecast central pre ssure of the cut-off low was too weak, the position of the low was located too far to the west, and that the precipitation totals were generally under -predicted. Sensitivity experiments were then carried out to examine the role of the mo del configuration and a range of meteorological variables in determining th e skill of the numerical simulation. The impact on the control forecast of. horizontal and vertical model resolution; the topography; the sea-surface temperature (SST); the upper-level jet; the surface fluxes of momentum and heat, and the use of an improved, explicit cloud microphysics scheme, were assessed. The simulations of the locations and intensities of both the SLP and the precipitation fields of the cut-off low pressure system were found to be sensitive, in varying degrees, to almost all the sensitivity variable s. The most important factors were found to be the horizontal resolution, t he SST patterns, the strength of the upper-level jet, the surface momentum fluxes, and whether an explicit or a parametrised moist convection scheme w as used. Clearly, careful attention must be given to representing variables that are found to contribute significantly to the skill of forecasts, as a dequately as possible in numerical models.