LAUSR

Abstract Saltmarsh-based reconstructions of relative sea-level (RSL) change play a central role in current efforts seeking to quantify the relationship between climate and sea-level rise. The development of an accurate chronology is pivotal, since errors in age–depth relationships will propagate to the final record as alterations in both the timing and magnitude of reconstructed change. A range of age-depth modelling packages are available but differences in their theoretical basis and practical operation mean contrasting accumulation histories can be produced from the same dataset. We compare the performance of five age-depth modelling programs (Bacon, Bchron, Bpeat, Clam and OxCal) when applied to the kinds of data used in high resolution, saltmarsh-based RSL reconstructions. We investigate their relative performance by comparing modelled accumulation curves against known age–depth relationships generated from simulated stratigraphic sequences. Bpeat is particularly sensitive to non-linearities which, whilst maximising the detection of small rate changes, has the potential to generate spurious variations, particularly in the last 400 years. Bacon generally replicates the pattern and magnitude of change but with notable offsets in timing. Bchron and OxCal successfully constrain the known accumulation history within their error envelopes although the best-fit solutions tend to underestimate the magnitude of change. The best-fit solutions of Clam generally replicate the timing and magnitude of changes well, but are sensitive to the underlying shape of the calibration curve, performing poorly where plateaus in atmospheric 14 C concentration exist. We employ an ensemble of age-depth models to reconstruct a 1500 year accumulation history for a saltmarsh core recovered from Connecticut, USA based on a composite chronology comprising 26 AMS radiocarbon dates, 210 Pb, 137 Cs radionuclides and an historical pollen chronohorizon. The resulting record reveals non-linear accumulation during the late Holocene with a marked increase in rate around AD1800. With the exception of the interval between AD1500 and AD1800, all models produce accumulation curves that agree to within ∼10 cm at the century-scale. The accumulation rate increase around AD1800 is associated with the transition from a radiocarbon-based to a 210 Pb-dominated chronology. Whilst repeat analysis excluding the 210 Pb data alters the precise timing and magnitude of this acceleration, a shift to faster accumulation compared to the long-term rate is a robust feature of the record and not simply an artefact of the switch in dating methods. Simulation indicates that a rise of similar magnitude to the post-AD1800 increase (detrended increase of ∼16 cm) is theoretically constrained and detectable within the radiocarbon-dated portion of the record. The absence of such a signal suggests that the recent rate of accumulation is unprecedented in the last 1500 years. Our results indicate that reliable (sub)century-scale age-depth models can be developed from saltmarsh sequences, and that the vertical uncertainties associated with them translate to RSL reconstruction errors that are typically smaller than those associated with the most precise microfossil-based estimates of palaeomarsh-surface elevation.