Extending the flood record : assessing the uncertainty and viability of palaeoflood data

This study assesses the uncertainty and viability of palaeoflood records in relation to the British database, which is a collection of radiocarbon dated geomorphological fluvial deposits used to infer the flood-frequency record during the Holocene. There are different forms of evidence used to interpret flood-frequency records and there are inherent uncertainties associated with both the data used to for analysis and the method of data analysis used. Previous studies, which have used summed probability distribution functions, have failed to show how sensitive the shape of the curve is to characteristics of the data used and to the radiocarbon calibration curve. This study firstly applies sensitivity analysis testing to the British database using the summed probability distribution methodology. This study also discusses the potential to apply a robust quality control protocol to the British database to verify the 14C ages currently available in line with geochronology studies that apply 14C dating. Sub-datasets of the British database were created based on the following criteria: number of samples per site, sample material, archaeological context and likely association to a flood event and analysed using summed probability distribution functions. Statistical indicators were used to show how similar the sub-datasets were to the unfiltered British database. This study identifies that statistically the most reliable results are generated when five or more samples from a single site location are analysed. Secondly, an alternative technique is used to analyse the data: Lomb-Scargle spectral analysis to test the data for cyclicities. Spectral analysis is used to identify cyclicities within the British database and the residual Δ14C data to identify cyclicities between the two datasets to determine if any cycles present are probable or a result of the radiocarbon calibration process. The results from this study impact researchers using summed probability distribution functions to interpret environmental and climatic data in any field.