Research papersFlood frequency analysis for a braided river catchment in New Zealand: Comparing annual maximum and partial duration series with varying record lengths

- Several technical challenges surrounding flood frequency analysis are discussed.
- Quantile estimates of PDS are compared to those of historical AMS.
- A recently developed formula is used to translate the PDS into the annual domain.
- The formula simplifies the use of the PDS.
- The advantage of the partial duration model over the annual maximum model is confirmed.

This study addresses technical questions concerning the use of the partial duration series (PDS) within the domain of flood frequency analysis. The recurring questions which often prevent the standardised use of the PDS are peak independence and threshold selection. This paper explores standardised approaches to peak and threshold selection to produce PDS samples with differing average annual exceedances, using six theoretical probability distributions. The availability of historical annual maximum (AMS) data (1930-1966) in addition to systemic AMS data (1967-2015) enables a unique comparison between the performance of the PDS sample and the systemic AMS sample. A recently derived formula for the translation of the PDS into the annual domain, simplifying the use of the PDS, is utilised in an applied case study for the first time. Overall, the study shows that PDS sampling returns flood magnitudes similar to those produced by AMS series utilising historical data and thus the use of the PDS should be preferred in cases where historical flood data is unavailable.