A comparison of dating models for reservoir sediment records using mass accumulation, radioisotopes, and Bayesian statistics

Abstract Paleolimnological investigations of reservoirs (i.e. human-made lakes) are becoming more common, establishing a need to explore dating approaches for sediment records from reservoirs. Traditional dating methodologies such as Constant Rate of Supply (CRS), 137 Cs, and Mass Accumulation Rate (MAR) have been utilized for sediment records from reservoirs for decades. However, with the advent of newer Bayesian approaches, such as the Bacon and Plum models, ambiguity persists for researchers when selecting a particular dating model. This study aims to quantify and compare the output differences among various dating models for the same sediment records and provide a discission-making framework based on practical constrains, rather than to validate which model is most accurate. Here we investigated the differences among age model outputs (dates and sedimentation rates) generated with CRS, MAR, Plum, and Bacon for three reservoirs in the Southeastern United States. Though they are all viable age models, their utility varies with research question, application, and financial cost. MAR and Bacon produced similar linear age models. The sedimentation rate for MAR was a static value throughout the record, while Bacon’s had more variable and significantly larger sedimentation rate than MAR for each reservoir studied. Both MAR and Bacon are more practical when there is a limited budget or less time for analysis. CRS and Plum produced similar non-linear age models, with CRS and Plum having the smallest and second smallest region of error, respectively. Similarly, CRS and Plum sedimentation rates exhibited aligned variations of sedimentation rates for two of the three sediment records analyzed. CRS and Plum dates and sedimentation rates were more similar when sediment records had a larger excess 210 Pb total inventory. We found the smallest differences among model outputs, < 5% difference based on a RMSE normalized ‘record age ratio’, when models utilized the same input data (i.e. top–bottom dates for MAR with Bacon and radioisotope for CRS with Plum). Generally, CRS and Plum would be preferred when there is ample time and funds for the lengthy radioisotope analysis and its higher costs. If there are additional known dates to incorporate into a model, we encourage the use of Plum.