On Applications of Space–Time Modelling with Open-Source 14C Age Calibration

Abstract

In archaeology, the meta-analysis of scientific dating information plays an ever-increasing role. A common thread among many recent studies contributing to this has been the development of bespoke software for summarizing and synthesizing data, identifying significant patterns therein. These are reviewed in this paper, which also contains open-source scripts for calibrating radiocarbon dates and modelling them in space and time, using the R computer language and GRASS GIS. The case studies that undertake new analysis of archaeological data are (1) the spread of the Neolithic in Europe, (2) economic consequences of the Great Famine and Black Death in the fourteenth century Britain and Ireland and (3) the role of climate change in influencing cultural change in the Late Bronze Age/Early Iron Age Ireland. These case studies exemplify an emerging trend in archaeology, which is quickly re-defining itself as a data-driven discipline at the intersection of science and the humanities.

Sparse Radiocarbon Data Confound Culture-Climate Links in Late Pre-Columbian Amazonia

“…The solid lines in Figure 3 are the bootstrapped confidence interval means and are somewhat misleading. The overlap (or lack thereof) of envelopes is the appropriate visual metric with KDEs in this case [13], which here largely illustrate a lack of separation in time between subsets of the data. The KDEs in the Llanos de Moxos region (Lomas and Zanjas) and the Central Amazon (Paredão and Guarita) counter-indicate a "rise and fall" associated to an adaptive cycle: the envelopes of the "intensive" phases increase at the points where the relevant climatic events start, and persist for over a century each.…”

“…Kernel bandwidth is therefore set at 25 years. While increasing the bandwidth much beyond this value somewhat narrows the confidence intervals (a documented trade-off between uncertainty and precision [13]), it was found that larger values rapidly obscure any semblance of patterning in the radiocarbon data. The high sensitivity of the KDEs to this balancing act serves to underscore the real limitations of the dataset: on one hand, large bandwidths shorten the KDE date ranges to the point of being meaninglessly brief, while on the other, small bandwidths greatly inflate confidence intervals.…”

“…The analysis has thus far presumed, for illustrative purposes, that the observed patterns are robust and invariant to sample sizes. This is not necessarily the case; aggregate analysis of radiocarbon usually operates with hundreds, if not thousands, of dates per sample, particularly where relatively short-lived processes such as those advanced in De Souza et al are concerned; see McLaughlin [13] for a case study of the Black Death. The paper now turns to alternate methods that are more suited to shedding light on this issue directly.…”

“…At time of writing, summed probability distributions are virtually the standard method in archaeological research for visualising assemblages of calibrated radiocarbon dates [2,6,7]. They also have widely acknowledged limitations, such as sensitivity to calibration curve effects and downplaying the uncertainties associated to the archaeological sampling procedure itself [13][14][15]. Of relevance to the period under examination here is the presence of centennial-scale plateaus in the IntCal13 and SHCal13 radiocarbon curves.…”

“…Of relevance to the period under examination here is the presence of centennial-scale plateaus in the IntCal13 and SHCal13 radiocarbon curves. An emerging alternative to SPDs that accompanies rcarbon are composite kernel density estimates (KDEs) [13], which simultaneously minimise noise potentially introduced by calibration effects and directly estimate uncertainties. With adequately-resolved radiocarbon data and a kernel bandwidth parameter tailored to analytical needs, KDEs can detect short-term events that SPDs struggle to separate from noise.…”