Veerle Vink, Creating a chronology:
Holocene History of Environment, Vegetation and Human Settlement on Catta Ness, Lunnasting, Shetland.
By K. D. Bennett, S. Boreham, M. J. Sharp and V. R. Switsur. (1992)
Shetland consist of a group of sparsely inhabited islands, lying in the north-east Atlantic Ocean about 78km north-east of Orkney (England) and about 350 km west of Norway. Holocene history of the environment, vegetation and human settlement on Catta Ness (Lunnasting, Shetland) are investigated using C14-dated Holocene lake sediments. One of the results from this study is that the first sign of human activity on the island were as seen at c. 4500 yr BP.
Dating original paper
Six samples of lake sediment were taken for C14-dating, each associated with particular horizons identified from the pollen and sediment analyses. No calibration of the C14 ages has been done. The relationship between age and depth for these dates was close to linear, with a correlation coefficient of 0.995 for an overall sediment accumulation rate of 28.8 years/cm.
Table 1. Radiocarbon dates from the sediments of Dallican Water. Ages are based on the zero year of AD 1950. The uncertainty of the age determination is given for each date. Depths are measured from the water surface: the lake was 390 cm deep at the coring point
Alternative chronologies
Since the proportion of radioactive carbon atoms in the earth’s atmosphere has not been constant over time, a calibration curve is needed to map between radiocarbon and calendar ages. Six C14-dates are used to calibrate the ages using the ‘Bchron’ package in R with the INTCAL13 iteration. The medians of the calibrated dates have high probabilities that they encompass their true age. Therefore, median calibrated ages were used in further calculations. Because the thickness of the samples was between 6 and 12 centimetres, the median from each depth is used in further calculations. A linear regression was constructed from the calibrated dates using R (version 1.0.153). The relationship between age and depth for these dates is close to linear, with a correlation coefficient of 0.992 for an overall sediment accumulation rate of 32.5 years/cm (figure 1).
Moreover, calibrated ages are used to construct a Bayesian ‘Bacon’ age-depth model. Modelled ages are presented in figure 2.
Figure 1. Age-depth model from a linear prediction based on the median calibrated ages of six 14C dates (blue dots) from the sediments of Dallican Water (Bennett et al., 1992). Black line is the fitted model (age ~ depth), and the dashed lines are 95% confidence intervals. Depths are measured from the water surface: the lake was 390 cm deep at the coring point.
Figure 2. Bayesian ‘Bacon’ age model based on the calibrated ages with the known errors of six 14C dates (blue dots) from the sediments of Dallican Water (Bennett et al., 1992). The grayscale on the model represents the likelihood, where the darker the gray, the more likely the model is of running through that section. The dashed red line represents the mean age output by the model. Depths are measured from the water surface: the lake was 390 cm deep at the coring point.
Best chronology
Age-depth models are only meaningful and useful when calculated using calibrated radiocarbon dates. Because Bennett et al. did not use calibrated radiocarbon dates, but C14-dates instead, the alternative Bayesian ‘Bacon’ chronology is more meaningful. Moreover, the Bayesian age-depth model is more meaningful than the linear regression of the calibrated ages, because the linear regression assumes that sedimentation rate is constant throughout the core while the Bayesian ‘Bacon’ model does not have this assumption. Moreover, Bayesian age modelling combines data with prior information, which increases the reliability of the model.
Implications of the new chronology for the interpretation of the original publication
Using the Bayesian age-depth model instead of the original linear regression model has some impact on the interpretations of the findings of the original publication. As we can see in the comparison of the Bayesian age depth model with the original model (figure 3), most of the depths correspond with older ages in the Bacon model in comparison with the original age-depth model. Bennett et al. found the first signs of human activity at at 4500 yr BP. This corresponds with a depth of 535 cm. In the Bayesian ‘Bacon’ model this depth corresponds with an age of 5100 cal yr BP. This suggest that humans entered the landscape of Catta Ness 600 years earlier than previously thought.
Figure 3. The differences between prediction of the original model (Bennett et al., 1992) with prediction from a Bayesian ‘Bacon’ age model.