Storm surge is a destructive and life threatening natural hazard associated with tropical cyclone (TC) events with the Queensland (Qld) coastline being particularly vulnerable. In a disaster management context, understanding the full range of possible scenarios that may arise from an impending event is critical for informed evacuation decision making. Tropical cyclones are difficult to forecast in terms of track and intensity, and storm surge magnitude is highly sensitive to the specific direction and location of peak cyclonic wind forcing in interaction with the local bathymetry. Here we investigate the optimal design of the next-generation storm surge forecasting system using a sensitivity analysis approach. The storm surge forecasting system, QSurge consists of a web-based probabilistic hazard mapping interface drawing from a large database of pre-simulated events. The aim of this paper is identify the appropriate increments of various tropical cyclone parameters necessary to capture the full range of possible storm surge outcomes whilst minimising computational run time. Sensitivity modelling experiments were conducted, consisting of simulating a range of TC parameter values within a parametric wind field model and then simulating storm surge using the Coral Sea MIKE21 hydrodynamic model established for this study and implemented on high performance computing. Detailed analysis of the results was undertaken for the study sites of Cairns, Cardwell and Townsville in North Queensland, and used to determine the optimal range of values for the various cyclone parameters. The number of scenarios necessary to complete a full ensemble database for landfall locations between Cooktown and the Whitsundays was initially estimated as 2,165,240. This was reduced to a set of 664,335 scenarios using polynomial relationships between peak storm surge magnitude and TC parameters such as central pressure and radius of maximum winds. Such a simulation set is achievable using high performance computing and given the optimised run time of the numerical model.