The paper describing our method for estimating extreme river flows anywhere on Earth.
The characterization of flood behavior in data poor regions has been receiving considerable attention in recent years.
In this context, we present the results of regional flood frequency analyses (RFFA) conducted using a global database of discharge data. A hybrid-clustering approach is used in conjunction with a flood-index methodology to provide a regionalized discharge estimates with global coverage. The procedures are implemented with varying complexity, with results indicating that catchment area and average annual rainfall explain the bulk of variability in flood frequency; a split-sample validation procedure revealed median errors in the estimation of the 100 year flood to be around 56%. However, far larger errors were also found, with performance varying between climate regions and estimation of the index-flood found to be the dominant source of uncertainty. Moreover, the RFFA procedure is utilized to provide insights on the statistical characteristics of floods across different climates and catchments.