Every few years, the radiocarbon calibration curve used to determine the calendar dates of almost all 14C measurements gets updated. The last recalibration was in 2013. Called IntCal13, it was based on 7,019 raw data points. This year, a major revamp – one of the biggest since its inception – has taken place and the new IntCal20 now takes into consideration more than 12,900 measurements. As with previous versions, the separate curves for samples from the Southern Hemisphere (SHCal20) and from marine reservoirs (Marine20) have also been updated. In this month’s ‘Science Notes’, we break down the history of these curves and dissect some of IntCal20’s new ‘features’.
The first IntCal was published in 1998. It marked the first time that researchers had come together to form an internationally standardised curve – previously there were several different ones available, each with its own specific biases and idiosyncracies. This original curve was primarily underpinned by dendrochronological data – it remains so today – allowing dating up to 11,857 calBP (‘Before Present’). To extend the range up to 24,000 years BP, a small number of uranium-thorium dated corals and foraminifera from annually layered sediments (that is, varves) were included. This older part of the curve was not especially robust, however, so the first update in 2004 (IntCal04) included more non-tree-ring datasets to allow the end date to be lengthened to 26,000 BP. It also utilised a random walk model (RWM) for curve construction.
It was not until 2009 that the curve saw some major changes to the dating range. Between updates, the non-tree-ring datasets had become much more reliable, so for IntCal09 more of these measurements were included, lengthening the dating range massively to 50,000 BP. IntCal13, the latest update before now, further strengthened this oldest part of the curve with the inclusion of more coral and foraminifera data, as well as data from plant macrofossils from varves and speleothems (or cave deposits).
IntCal20 includes all of these innovations and much more. As with previous updates, many new data points have been added to the entire span of the curve. In particular, new tree-ring dates have extended the dendrochronologically based part of the curve to 13,910 calBP, while additional non-tree-ring data now allows samples as old as 55,000 calBP to be dated.
One of the biggest improvements seen in this new curve is the addition of more precise measurements that have become available due to the new generation of Accelerator Mass Spectrometry (AMS) machines, which are able to deliver dates at a much higher precision than older AMS models. In fact, large sections of the timescale have been redated, mostly at annual resolution for tree-rings, using this new AMS technology to build in even more precision. Despite a lot of new single year tree-ring data, though, the more recent end of IntCal20 has changed very little.
IntCal20 also includes, for the first time, Miyake events, which are narrow (that is, 1-2 year) spikes of increased 14C production observed in tree-rings. While these historical spikes can actually widen the possible calibrated date ranges, for ordered sequences of radiocarbon dates these spikes help pinpoint archaeological dates more precisely.
The statistical methodology behind the curve has also been updated. The RWM used since IntCal04 has now been replaced with Bayesian splines, which form a curve by connecting together ‘knots’ that lie at specific calendar ages. As the authors of IntCal20 note, ‘this switch… provides several benefits, allowing us to more accurately represent many of the unique aspects of the radiocarbon data and to test robustness to data and model assumptions.’ While this is a major in-built shift, for the most part the average user’s experience will remain much the same.
Overall, this new curve will significantly improve the accuracy of radiocarbon dating. But, while remarkable progress has been made, the part of the curve that extends beyond the tree-ring data still remains less precise, with many more variables to take into consideration. Additionally, it is important to note that even within the more-robust dendrochronological part of the curve there is still the possibility of error, whether due to an offset in a particular lab’s measurement methodology or within the sample itself (such as seasonal or regional effects; see CA 341). And while additional uncertainty is included in the curve to account for this variability, it may not be accounted for in individual samples. Usually, though, this margin will only affect the results by a decade or two. With each update, we get a little bit closer to absolute dating.
Publications regarding IntCal20 can be found in the latest issues of Radiocarbon.
This article appears in issue 366 of Current Archaeology. To find out more about subscribing to the magazine, click here.
