STATISTICAL ESTIMATES OF PMP VALUES

The article discusses two statistical methods to estimate PMP values, the Hershfield and the NERC methods. Neither method offers any explana tion why the PMP values can be calculated by the use of unbounded stat istical distributions, but both methods include the use of envelope cu rves that are not independent of the region. Bounded data that fits an unbounded distribution must deviate from the distribution for high re turn periods and tend to a limiting value, and then there exists, a li miting reduced variate that can be used to find the PMP value. When th e distribution is EV1, the limiting reduced variate can be defined by a mapping transformation, or by cutting off the distribution. It is sh own that when Hershfield or NERC methods are used, the limiting reduce d variate is included in the PMP values and can be separated from regi onal parameters. It is suggested that the limiting reduced variate, th at depends solely on return period, may more easily be transferred bet ween regions than the other parameters. This may be a great help in fi nding PMP values in regions where observations are not extensive enoug h to define limiting return periods with necessary certainty. A case s tudy with data from Iceland demonstrates, that using the limiting redu ced variate, similarities emerge in the Icelandic data and the NERC PM P that justify the acceptance of the NERC method.