Comparative analysis of statistical and catchment modelling approaches to river flood frequency estimation

Abstract

The paper compares results from two approaches to the quantification of river flood frequencies applicable nationally in Britain. One approach uses both the flood peak and event-based methods of the Flood Estimation Handbook (FEH) of current water industry practice and the other approach is a recently developed set of continuous simulation techniques using parameter-sparse modelling of catchment flood runoff response. The methods were applied to over a hundred sites in Britain, treated as if without flow data, although such observations existed and were used only for testing purposes. Errors of ≤20% in peak flows at ungauged sites are currently very good in this hydrologically challenging context; errors of up to around 35% may have to be contended with in flood management practice. The results from the FEH statistical method reinforce its established role in peak-flow estimation. On the basis of the aspects that have been tested here, the emerging continuous simulation approaches show considerable potential to offer good performance for peaks and flow time series. The errors associated with the FEH unit hydrograph approach reflect the additional challenge it incorporates of ungauged rainfall estimation in addition to ungauged discharge.