Value of Historical Flood Information in Partial Duration Series and Annual Maximum Series Frameworks

This paper illustrates the value of historical flood and paleoflood information when flood frequency analysis is performed in partial duration series (PDS) and annual maximum series (AMS) frameworks. The PDS analysis considers the Generalized Pareto distribution for exceedances over a threshold with the Poisson arrival model. This corresponds to a 3-parameter Generalized Extreme Value (GEV) distribution in a AMS framework. The simple relationship between the two models is quite convenient. When the shape parameter is set to zero (2-parameter models), one obtains the exponential/Poisson model for the PDS, and the Gumbel distribution for the AMS framework. The value of historical information with partial duration series and annual maximum series data is explored using the relatively new Generalized Maximum Likelihood estimators (GMLEs). Historical flood information can be of tremendous value depending on the magnitude of the perception threshold for the historical and paleoflood information, and the length of the historical record length.