Sensitivity analysis of flood frequency to the length of the data recording period

This study examined the sensitivity of flood-frequency estimates to the length of hydrometric records using the annual maximum series (AMS) and partial value series (PVS) approaches at 15 hydrometric stations in the Karkheh Basin, Iran. Flood frequency analysis is a fundamental tool for understanding and managing the flood risks, and numerous statistical models have been developed to estimate the flood characteristics. Among these, the Partial Value Series (PVS) model and the Annual Maximum Series (AMS) model are widely applied.

The results demonstrate that record length is a major source of uncertainty in floodfrequency analysis, particularly when estimates are extrapolated to long return periods. In general, extending the available record length improved the stability of the estimated distribution parameters and design-flood quantiles. The coefficient of variation generally decreased as the record length increased, indicating a reduction in relative statistical uncertainty. In contrast, skewness and kurtosis showed less systematic behavior among stations, especially for short records, reflecting their greater sensitivity to a limited number of extreme events and to the local hydrological characteristics of each catchment.