Megathrust earthquakes that occur repeatedly along the plate interface of subduction zones can cause severe damage due to strong ground motion and the destructive tsunamis they can generate. We developed a set of scenario earthquakes to evaluate tsunami hazards and tsunami early warning systems for such devastating earthquakes. Although it is known that the slip distribution on a fault strongly affects the tsunami height distribution in near-field coastal areas, the slip distribution of future earthquakes cannot be exactly predicted. One way to resolve this difficulty is to create a set of scenario earthquakes in which a set of heterogeneous slip distributions on the source fault is stochastically generated based on a given slip probability density function (SPDF). The slip distributions generated in this manner differ from event to event, but their average over a large ensemble of models converges to a predefined SPDF resembling the long-term average of ruptures on the target fault zone. We created a set of SPDF-based scenario earthquakes for an expected future Mw 8.2 Tonankai earthquake in the eastern half of the Nankai trough, off southwest Japan, and computed the ensuing tsunamis. We found that the estimated peak coastal amplitudes among the ensemble of tsunamis along the near-field coast differed by factors of 3 to 9, and the earliest and latest arrivals at each observation site differed by 400 to 700 s. The variations in both peak tsunami amplitude and arrival time at each site were well approximated by a Gaussian distribution. For cases in which the slip distribution is unknown, the average and standard deviation of these scenario datasets can provide first approximations of forecast tsunami height and arrival time and their uncertainties, respectively. At most coastal observation sites, tsunamis modeled similarly but using a uniform slip distribution underpredicted tsunami amplitudes but gave earlier arrival times than those modeled with a heterogeneous slip distribution. Use of these earlier arrival times may be useful for providing conservative early warnings of tsunami arrivals. Therefore, tsunami computations for both heterogeneous and uniform slip distributions are important for tsunami disaster mitigation