P2.6 USING A PERTURBATION METHODOLOGY IN A MESOSCALE ATMOSPHERIC MODEL TO ASSESS THE VARIABILITY OF A FLASH FLOOD RAINFALL AMOUNT IN A WATERSHED UNDER CURRENT CLIMATE CONDITIONS M.S. Speer 1 and L.M. Leslie 2 1 National Severe Storms Laboratory, Norman, Oklahoma. 2 School of Meteorology, The University of Oklahoma, Norman, OK 1. INTRODUCTION Then, the perturbation methodology is applied to two other flood events and the results are compared with the November 1996 storm. There is considerable uncertainty about the frequency and intensity of extreme events under current climate conditions, especially rainfall amounts that can produce severe flash flooding. On 23 November, 1996 Coffs Harbour, a city on the Australian east coast with a population of approximately 30,000, experienced a flash flood following intense, short duration rainfall rates in excess of 100 mm/hr. Speer and Leslie (2000) carried out a series of quantitative precipitation forecast (QPF) model simulations that provided the motivation for this study, namely, to estimate extreme precipitation totals over the Coffs Harbour Creek catchment. 2. METHODOLOGY In attempting to estimate the uncertainty associated with single model forecasts, a number of weather centres routinely produce deterministic and ensemble forecasts of meteorological variables, including precipitation, out to as far as two weeks in advance. These centres include the European Centre for Medium Range Weather Forecasting (EC) (Molteni et al. 1996), the National Center for Environmental Prediction (NCEP) (Toth and Kalnay 1993), The United Kingdom Meteorological Office (UKMO), The US Fleet Numerical and Oceanographic Center (FNMOC) and the Bureau of Meteorology, Australia. Other institutions such as Florida State University (FSU) also routinely produce ensemble precipitation forecasts (Krishnamurti et al. 2000). Examples of the ensemble methodologies employed are the so- called ‘breeding’ method used at NCEP and the singular vector approach used at the EC. However, in attempting to produce a possible forecast spread of rainfall for a single event such as the Coffs Harbour flash flood, a simple scenario approach is appropriate, and possibly preferable, because the important mechanisms that produced the heavy rainfall causing the flash flood are known (Speer and Leslie 2000). The methodology in this study is to focus on generating a range of input variables to produce a distribution of forecast rainfall amounts. The key input variables identified by Speer and Leslie (2000) were: sea-surface temperatures (SST); sea level pressure gradients in the trough over the ocean adjacent to Coffs Harbour; and the strength of the low level (900 hPa) winds along the coast, just south of Coffs Harbour. The distribution of rainfall totals is subject to a very simple cluster analysis in which the rainfall totals are grouped into three categories consisting of totals within plus and minus one standard deviation for two of the groups and the remainder in the third group. In Australia, the frequency and intensity of rainfall have been assessed for locations where rainfall gauges have long been in place (Bureau of Meteorology, 1987). Locations with long-term reliable rainfall records are predominantly clustered around population centres. However, rainfall that can cause an extreme flash flood in or close to a population centre often originates from data sparse regions upstream of the main impact area. Such areas may be as close as ten kilometres or less from the area affected. Intense rainfall is focussed by processes also acting on small horizontal scales. In the case of the November 23, 1996 flash flood that affected Coffs Harbour, a 24 hour rainfall total of 168 mm was recorded at Coffs Harbour Airport. The airport is closer to the coast than is the city, being less than one kilometre from the ocean. In contrast, at two locations just three to four kilometres further west on higher ground, 24 hour totals exceeded 500 mm. In this study our main aim is to quantify the maximum possible rainfall totals over the catchment, within the accepted range of values of key variables. For Coffs Harbour, the primary variables are known to be surface temperature (SST), wind strength, and the central pressure and central pressure gradients of the weather systems responsible. First, a perturbation methodology is used to generate a distribution of predicted rainfall total for the Coffs Model Harbour flash flood rainfall of 23 November, 1996. The model used here is the same as that used by Speer and Leslie (2000). It is The University of New South Wales HIRES model. This model is a hydrostatic model formulated in terms of the advective form of the primitive equations for momentum, mass, moisture, and thermal energy --------------------------------------------------------------------- Corresponding author address: Milton S. Speer National Severe Storms Laboratory, Norman, OK, 73069; email: milton.speer@noaa.gov