Flood frequency analysis and discussion of non-stationarity of the Lower Rhine flooding regime (AD 1350-2011): Using discharge data, water level measurements, and historical records

- Alternative inferior data types can reduce flood frequency analysis uncertainty.
- Flood frequency analysis can be biased using both short and long flood records.
- The Rhine flooding regime contains multi-decadal alternations in flooding intensity.
- The occurrence of extreme events shows no clear relation with general flooding intensities.

SummaryAccurate estimates of the recurrence time of extreme floods are essential to assess flood safety in flood-prone regions, such as the Lower Rhine in The Netherlands. Measured discharge records have a limited length and are, in general, poorly representing extremes, which results in considerable uncertainties when used for flood frequency analysis. In this paper, it is shown how alternative discharge monitoring stations along the Rhine, measurements of water levels, and historical records can be used to increase data availability. Although pre-processing and the conversion of data types into discharge estimates introduces extra uncertainty, the added value of this data in flood frequency analysis is considerable, because extending record length by including slightly less-precise data results in much better constrained estimates for the discharges and recurrence intervals of extreme events. Based on results obtained with the Generalised Extreme Value (GEV) distribution, it was concluded that large floods of the last century are presumably rarer than previously considered using shorter data series. Moreover, the combined effect of climatic and anthropogenic-induced non-stationarities of the flooding regime is more easily recognised in extended records. It is shown that non-stationarities have a significant effect on the outcomes of flood frequency analysis using both short and long input data series. Effects on outcomes of dominant multi-decadal variability are, however, largely subdued in the longer 240-year series.