ABSTRACT
ABSTRACT: The most straightforward method for estimating flood risk at a particular point where a gauge station is located is to adjust a cumulative probability distribution function to the recorded annual maximum flows. Unfortunately this method may give rise to highly variable flood quantile estimators for high return periods. To increase reliability of the quantile estimator it can used the best statistical model and/or to increase the amount of information. The high number of distribution functions used in Hydrology is a demonstration more efforts must be done in the theoretical basis of the statistical model selection. Using historical or palaeoflood censored data can increase the information length at the gauge station. However, care must be taken with the information error and the underling stationary hypothesis. Few statistical models have been developed to introduce the non-stationarity of the long term flooding process.A very important additional output of these types of models will be their use for the future flood hazard estimation in a climate change framework.
