Regional flood frequency analysis for hydrologically diverse regions in New Zealand

To reduce flood risk, it is critical to accurately estimate design floods, which are associated with specific annual exceedance probabilities. This study aims to develop a Regional Flood Frequency Analysis (RFFA) approach that clusters hydrologically diverse catchments into more homogeneous groups, thereby improving the reliability of design flood estimates. Traditional regionalisation often fails due to high inhomogeneity. The RFFA approach was applied and evaluated using 363 catchments in New Zealand. It was found that incorporating climate zones and catchment characteristics improved homogeneity. Cluster analysis based on catchment attributes was applied to delineate homogenous regions. The two-parameter Log-Normal and Pearson 3 distributions were identified as dominant regional probability distributions. The Generalised Additive Model and Index Flood L-moment approach were used to estimate regionalised design floods. Model performance, assessed with Jackknife Resampling, showed significantly smaller error estimates than previous RFFA studies in New Zealand. This approach provides region-specific design values for flood risk management in both gauged and ungauged catchments.