Abstract
Prediction interval for a successive occurring extreme is derived. The well-known confidence intervel for a return level of precipitation has been sometimes, or frequently in our mind, misused by misunderstanding what the interval stands for. Return period and encounter probability, which has been employed in many previous literatures, are indirect to handle our target against the extremes. Not only predicting a future extreme but anticipating the past extraordinary events, such as the daily amount of precipitaion in the Tokai heavy rain can be also examined in the view of statistical test by the derived prediction distribution function, in which the degree of experience plays an important role.
