There are various sources of forecast uncertainty in the modeling of fine scale weather phenomena such as supercell thunderstorms. These include the uncertainty associated with the initial and boundary conditions, and uncertainty arising from model deficiencies. Model deficiencies can be due to the various assumptions and parameterisations used in the model physics. Our study is driven by the important role played by model physics in determining the resulting forecasts, and the need for model diversity in short-range ensemble forecasting systems. Here, we assess the variability in single model forecasts of a supercell by investigating the sensitivity to parameters within the ice physics representation used in the non-hydrostatic HIRES model, currently run at The University of Oklahoma. For supercell thunderstorms, the ice microphysics is extremely important in representing the various phases of water vapour that may lead to the development of large hail. Using archived operational data, our case study examines the devastating supercell hailstorm of 14 April 1999 in Sydney, Australia, at a horizontal resolution of 1 km.

The methodology adopted is to test the sensitivity of the intercept slope for graupel and graupel density. It is recognized that there is a wide range of possible values for these two parameters. Collection efficiencies are relatively well specified for collisions of cloud droplets, rain drops, and graupel/hail either with themselves or with each other. Therefore, it was decided to restrict the testing of collection efficiencies to the collection efficiency of snow for cloud water (E_sc), as it is generally believed to be the most poorly understood.

A total of ten model forecasts (including control) was performed that included four different, though possible, values of intercept slope for graupel and three for graupel density, and, also, two for E_sc. The control forecast used values of graupel intercept slope and graupel density representative of the subtropical latitude of Sydney. The control forecast was very skilful in terms of the track and the amount of accumulated hail from the supercell over its 5 ½ hour duration. There was also an indication of a bifurcation in development in the control forecast. The observed track took a right turn away from the coast and weakened over the sea having reached its most intense phase when passing over Sydney's coastal suburbs. The results for graupel intercept slope and graupel density showed sensitivity to the amount of accumulated precipitation, the strength of updrafts and downdrafts and surface temperature changes in a manner consistent with supercell dynamics. Furthermore, the bifurcation present in the control also was sensitive to graupel intercept slope and graupel density. In future case studies the consistency check of the collection efficiency of snow for cloud water will continue to be performed, as the overall sensitivity to collection efficiency depends on the distribution of liquid cloud zones, which in turn could vary widely with the type of weather system being investigated.