New Research In
Physical Sciences
Social Sciences
Featured Portals
Articles by Topic
Biological Sciences
Featured Portals
Articles by Topic
- Agricultural Sciences
- Anthropology
- Applied Biological Sciences
- Biochemistry
- Biophysics and Computational Biology
- Cell Biology
- Developmental Biology
- Ecology
- Environmental Sciences
- Evolution
- Genetics
- Immunology and Inflammation
- Medical Sciences
- Microbiology
- Neuroscience
- Pharmacology
- Physiology
- Plant Biology
- Population Biology
- Psychological and Cognitive Sciences
- Sustainability Science
- Systems Biology
Rapid climate change did not cause population collapse at the end of the European Bronze Age
Edited by B. L. Turner, Arizona State University, Tempe, AZ, and approved October 14, 2014 (received for review May 7, 2014)

Significance
The impact of rapid climate change on humans is of contemporary global interest. Present-day debates are necessarily informed by paleoclimate studies in which climate is often assumed, without sufficient critical attention, to be the primary driver of societal change. Using new methods to analyze paleoclimatic and archeological datasets, we overturn the deterministic idea that population collapse at the end of the northwestern European Bronze Age was caused by rapid climate change. Our work demonstrates the necessity of high-precision chronologies in evaluating human responses to rapid climate change. It will be significant for geoscientists, climate change scientists, and archeologists.
Abstract
The impact of rapid climate change on contemporary human populations is of global concern. To contextualize our understanding of human responses to rapid climate change it is necessary to examine the archeological record during past climate transitions. One episode of abrupt climate change has been correlated with societal collapse at the end of the northwestern European Bronze Age. We apply new methods to interrogate archeological and paleoclimate data for this transition in Ireland at a higher level of precision than has previously been possible. We analyze archeological 14C dates to demonstrate dramatic population collapse and present high-precision proxy climate data, analyzed through Bayesian methods, to provide evidence for a rapid climatic transition at ca. 750 calibrated years B.C. Our results demonstrate that this climatic downturn did not initiate population collapse and highlight the nondeterministic nature of human responses to past climate change.
Past population collapse in many parts of the world has been attributed to the direct effects of rapid climate change. Key case studies on the collapse of Anasazi (1) and Mayan (2) civilizations have attracted considerable public interest due to the concerns over the threat of climate change to contemporary populations. Recent paleoenvironmental studies have identified a major climate shift across much of northwestern Europe toward the end of the Bronze Age (3, 4). This has been associated with socioeconomic collapse in Ireland (5), northern Britain (6), and central and western Europe (7), and the expansion of Scythian culture into Europe and eastern Asia (8).
In northwestern Europe, the eighth century calibrated years (cal.) B.C. sees the transition from the Late Bronze Age to the Early Iron Age. Whereas evidence for Late Bronze Age settlement and craft production is widespread, it is notoriously elusive for Early Iron Age communities in many parts of northwestern Europe (5, 9⇓–11), suggesting a reduction in population levels. At the same time the international exchange networks required to support bronze-based economies appear to break down. To what extent might these changes be linked to the environmental downturn implied by the paleoclimate data?
Context
Ireland forms an important study region for examining the relationships between human populations and past climate change. Firstly, the widespread occurrence of peat bogs, containing tephra layers for precise dating and correlation, enables the creation of robust terrestrial climate histories (12, 13) (Figs. 1 and 2 and Table S1). Secondly, the period from around 1995 to 2008 saw an enormous upsurge in archeological activity in Ireland, fueled by the unprecedented growth of the Celtic Tiger economy. Development-led excavations generated enormous quantities of new data, including 14C dates from a broad range of landforms and environments across the island (Fig. 1). The archeologically untargeted nature of this work means that this 14C dataset is unbiased by the interests and preoccupations of archeologists to a degree that is unique globally (14).
Map of Ireland showing relevant archeological sites and locations of peatlands used to derive the climate proxy data (Table S1). 1, Dead Island; 2, Derragh; 3, Garry; 4, Glen West; 5, Lough Lurgeen; 6, Moyreen; 7, Owenduff; 8, Slieveanorra; 9, Sluggan. The linear distributions visible in many of the archeological sites reflect their discovery through road and pipeline schemes. Basemap courtesy of Esri, Inc.
Proxy climate data from Irish peatlands for the period 1200 B.C.–A.D. 400. Humification data and testate amoeba-based water table reconstructions are shown. Tephra layers, used for correlation and dating, are illustrated. The combined data are shown alongside the total modeled chronological error (from the Bacon models) (Fig. S2). LOWESS models [smooth = 0.02 (blue) and 0.1 (red)] illustrate the major common features of the proxy climate compilation (errors are a 95% bootstrap range on the 0.02 model). Humification data are expressed as residuals; the reconstructions are expressed as water table depth below the peat surface and are based on the transfer function of Charman et al. (42). HU, humification; WT, water table reconstruction.
There have been previous attempts to correlate archeologically defined levels of human activity in prehistoric Ireland with paleoclimatic proxy records, comparing summed 14C dates of archeological activity with a climate index inferred from bog oak population dynamics (15). However, bog oak population dynamics have been shown to be an unreliable record of past climate (16). There have been other attempts to link rapid climate change to human societal changes (8); however, establishing clear causal links between humans and climate change is difficult due to the problems of establishing an accurate, precise chronological framework (17).
Summed probability functions (SPFs), which involve the summing of large numbers of 14C dates, have been used as a proxy for human population in numerous recent studies (18, 19), although their reliability has sometimes been questioned on the grounds that artificial peaks and troughs introduced by the calibration process irretrievably distort the real patterns (20, 21). It has been demonstrated that real patterns can be observed in SPFs through statistical analysis with randomly generated data (14), as well as comparison with dendrochronologically dated phases (17). Use of these methods enables us to observe genuine fluctuations in human activity and compare these directly with the paleoclimate data.
Results and Discussion
We examined 2,023 archeological radiocarbon dates from Ireland spanning the period 1200 cal. B.C. to cal. A.D. 400 (14) (Dataset S1). For the period 1200 to 500 cal. B.C., the results (Fig. 3) demonstrate clear patterns in relative levels of human activity that appear to reflect demographic fluctuations (14). There appears to be a distinct peak in human activity in Ireland at around 1050–900 cal. B.C., followed by steady decline to around 800 cal. B.C., and a rapid fall to 750 cal. B.C. Our analysis (14) shows that the pattern in the archeological SPFs cannot be explained by random variations (Fig. 3). This analysis is supported by the regional pollen records which identify a peak in farming activity in the late 11th century B.C., followed by a decrease during the 9th to early 8th centuries B.C. (22, 23). To evaluate whether or not climate change instigated the observed demographic changes, we analyze high-resolution proxy climate data from Ireland alongside the archeological evidence (Fig. 3). A major, rapid climate deterioration (shift to much wetter conditions) is registered in testate amoeba-based water table reconstructions and humification records from peatlands in Ireland, and has been precisely dated to ca. 750 cal. B.C. (24). At the best-dated site (Glen West bog, County Fermanagh), the start of this shift has been constrained using a Bayesian age-depth model to 748 cal. B.C. (maximum probability) or 786–703 cal. B.C. (modeled range). The timing of the shift has been linked by tephrochronology to several other peatland sites in Ireland (25) (Fig. 2) and independently dated by 14C to 791–429 cal. B.C. (Derragh Bog, County Longford) (Fig. S1). The precise tephra-framed replication of this event in several different peatlands suggests that it is a reliable widespread response of peatland hydrology to a rapid increase in precipitation and/or decrease in temperature, rather than related to internal peatland dynamics (26). A climate shift at ca. 800–750 cal. B.C. is also seen across northwestern Europe and is potentially the most profound climatic shift of the Mid- to Late Holocene before the Little Ice Age (3, 27, 28).
(A) SPF of the archeological data with SPFs based on simulated series of dates in 14C and calendar years. This illustrates clearly that the features in the archeological SPF cannot be the product of random variations. (B) Running correlation coefficient between the archeological and simulated SPFs. (C) LOWESS models (smooth = 0.02 and 0.1) illustrating the major features of the proxy climate compilation (errors are a 95% bootstrap range on the span = 0.02 model) (Fig. 2). (D) The archeological SPF minus both simulated series (in A) compared with the LOWESS models.
Comparison of the archeological and paleoclimate data demonstrates that the decline in population at the end of the Bronze Age began more than a century before the climatic downturn of the mid-eighth century B.C. (Fig. 3). Therefore, the decline can be categorically disassociated with the climate downturn. To explain the end of the Bronze Age we must look instead toward socioeconomic factors. Bronze-based economies relied on complex, long-distance trade networks to bring together the raw materials necessary for bronze production. Control of these networks appears to have formed the basis of social power in Bronze Age Europe and promoted the development of complex, hierarchical social structures (29). It has long been argued that the widespread availability of iron ores fatally undermined these social structures by democratizing access to metals (30). The adoption of iron technology thus made redundant the long-established networks that underpinned Late Bronze Age society. Resultant social destabilization may well be the cause of the population collapse at the end of the Bronze Age. Against the background of contemporary debates it is easy to view climate as the primary driver of past cultural change. Such assumptions need to be critically assessed using high-precision chronologies to guard against misleading correlations between unrelated events.
Materials and Methods
14C dates were collected by collating the published and unpublished literature for prehistoric Ireland and by written requests for information to all active commercial and research-based archeological groups in Ireland (31). A total of 2,023 14C (Dataset S1) dates meeting appropriate quality thresholds (32) were used to create an SPF, with a further 78 being excluded as they did not meet the threshold, and a further 157 excluded as not enough information to judge their quality was available. The graphics in this paper represent the period of interest from 1200 to 500 cal. B.C. (graphic representation of the full dataset is provided in ref. 14). Because few sites have produced substantial numbers of dates, and few are phased in the conventional sense, it was neither necessary nor practical to adopt the approach of Collard et al. (33) of summing dates for particular phases to avoid intersite biases in the overall quantities of dates. We generated random simulations of calendar and 14C ages as null hypotheses to test the summed archeological dates against. We used a distributionless random number generator [using R (34)] and the same number of dates as in the archeological dataset (n = 2,023). For the calendar year simulation, a random series of calendar years was calculated and converted to simulated 14C determinations in OxCal (35). Each simulated determination was attributed a random error between 20 and 80 y. We used the running correlation method of Armit et al. (14) to examine periods of correlation and noncorrelation between the random simulations and the archeological data. In addition we subtracted the random simulations from the archeological data (14).
Previous Holocene paleoclimate studies in Ireland, including those based on lake and speleothems (36, 37), lack the chronological resolution needed for examining human–environmental relations at centennial scales. It has also been suggested that narrow ring events in bog oaks signify extreme, rapid environmental change at 2345, 1628, 1159, and 207 cal. B.C. and cal. A.D. 540 (38); however, the climatic meaning of these has yet to be determined. Paleoclimate reconstructions from peatlands reflect past changes in the length and intensity of the summer water deficit, most probably controlled by summer precipitation in oceanic northwestern Europe (39). In Ireland, published high-resolution records based on peat humification and testate amoeba-based water table reconstructions were chosen for this investigation (40, 41). Water table reconstructions were carried out using the European transfer function (42) for Dead Island, Derragh, Glen West (high-resolution section only) and Slieveanorra bogs. These records are dated by 14C, tephrochronology, and spheroidal carbonaceous particles. Tephra layers provide a robust chronological framework for precise comparison and correlation of the records in time (25). The chronologies and associated errors for the water table reconstructions were modeled using Bacon, an age-depth model based on piecewise linear accumulation (43), where the accumulation rate of sections depends to a degree on that of neighboring sections. The total chronological error (difference between maximum and minimum probability ages at 95%) associated with each depth (in all of the above sites) was calculated from the model. These records have centennial to multidecadel chronological resolution.
The water table data were standardized to z scores, and then combined and ranked in chronological order (i.e., by maximum age probability as modeled by Bacon). Locally weighted scatterplot smoothing (LOWESS; smooth = 0.02 and 0.1) (44) was calculated (Dataset S2). Polynomial regressions in a neighborhood of x were fitted with the following:
Where Wki(x) denoted k-nearest neighbors weights (45), bootstrapping was used (999 random replicates) to calculate 95% error ranges on the high-resolution LOWESS function (smooth = 0.02 and 0.1). To retain the structure of the interpolation, the procedure used resampling of residuals rather than resampling of original data points. It was found that interpolation to annual interval made little difference to the overall shape of the LOWESS function. This function represents a statistical compilation of the peatland water table records and models the intersite events. It can be used as an exploratory tool to assess overall trends in the data but the interpretation of wet and dry shifts is based on a consideration of the individual site records.
Acknowledgments
We thank Peter Langdon for providing published paleohydrological data. We thank Antony Blundell, Andrew Wilson, Carl Heron, and Lindsey Büster for comments on an earlier version of the manuscript. Illustrations were prepared by Rachael Kershaw and Emily Fioccoprile. The authors thank all the individuals and organizations that facilitated access to data: Aegis Ltd.; Arch-Tech Ireland Ltd.; Archaeological Consultancy Services Ltd.; Theresa Bolger, Emmett Byrnes, and Mary Cahill (National Museum of Ireland); Judith Carroll (Judith Carroll and Co. Ltd.); Kerri Cleary (University College Cork); Rose Cleary (University College Cork); Sarah-Jane Clelland (University of Bradford); Gordon Cook (The Scottish Universities Environmental Research Centre); Gabriel Cooney (University College Dublin); Cultural Development Services Ltd.; Vicky Ginn (Queen’s University Belfast); Department of the Environment, Northern Ireland Environment Agency; Marion Dowd (Institute of Technology, Sligo); Ed Bourke (Department of Environment, Heritage and Local Government); Eachtra Archaeological Projects; George Eogan; James Eogan (National Roads Authority Ireland); Margaret Gowen (Margaret Gowen and Co. Ltd.); Rubicon Archaeology Ireland; Irish Archaeological Consultancy Ltd.; Mary Henry Archaeological Services Ltd.; Thomas Kerr (Queen’s University Belfast); Jan Lanting (Faculteit Archeologie, Bio-Archaeology, Leiden University); Chris Lynn; Phil MacDonald (Centre for Archaeological Fieldwork, Queen’s University Belfast); James McDonald (Chrono, Queen’s University Belfast); Martin Reid (Department of Environment, Heritage and Local Government); Mayo County Council; Finbar McCormick (Queen’s University Belfast); Conor McDermott (University College Dublin); Jim Mallory; Tiernan McGarry; Sarah Milliken; the Moore Group; Matt Mossop; National Roads Authority Ireland; Northern Archaeological Consultancy Ltd.; Ellen O’Carroll, Trinity College Dublin; Aidan O’Connell (Archer Heritage); John Ó Néill; Aidan O’Sullivan (University College Dublin); Paula Reimer (Queen’s University Belfast); Matt Seaver (University College Dublin); Rónán Swan (National Roads Authority Ireland); TVAS Ireland Ltd.; Hans van der Plicht (Leiden University); Fintan Walsh (Irish Archaeological Consultancy Ltd.); Richard Warner; and Peter Woodman. The research forms part of the Mobility, Climate and Culture: Re-Modelling the Irish Iron Age project, funded by the British Academy. Preliminary data collection was funded by the Irish Heritage Council.
Footnotes
- ↵1To whom correspondence should be addressed. Email: i.armit{at}bradford.ac.uk.
Author contributions: I.A., G.T.S., and K.B. designed research; I.A., G.T.S., and K.B. performed research; M.B. contributed new reagents/analytic tools; I.A., G.T.S., K.B., and G.P. analyzed data; and I.A., G.T.S., and K.B. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1408028111/-/DCSupplemental.
References
- ↵
- ↵
- ↵.
- van Geel B,
- Buurman J,
- Waterbolk HT
- ↵
- ↵.
- Raftery B
- ↵
- ↵.
- Maise C
- ↵
- ↵.
- Haselgrove C,
- Pope R
- Haselgrove C,
- Pope R
- ↵.
- Haselgrove C,
- Pope R
- Fontijn D,
- Fokkens H
- ↵.
- Gerritsen FA
- ↵
- ↵
- ↵
- ↵
- ↵.
- Swindles GT,
- Plunkett G
- ↵
- ↵.
- Gamble C,
- Davies SWG,
- Richards M,
- Pettitt P,
- Hazelwood L
- ↵
- ↵
- ↵
- ↵.
- Molloy K
- ↵
- ↵.
- Swindles GT,
- Plunkett G,
- Roe HM
- ↵
- ↵.
- Swindles GT,
- Morris PJ,
- Baird AJ,
- Blaauw M,
- Plunkett G
- ↵.
- Brown T
- ↵.
- Yeloff D,
- van Geel B,
- Broekens P,
- Bakker J,
- Mauquoy D
- ↵.
- Haselgrove C,
- Pope R
- Needham S
- ↵.
- Childe VG
- ↵.
- Becker K,
- Armit I,
- Swindles G
- ↵.
- Becker K,
- Armit I,
- Eogan J,
- Swindles G
- ↵
- ↵.
- R Development Core Team
- ↵.
- Bronk Ramsey C
- ↵.
- McDermott F,
- Mattey DP,
- Hawkesworth C
- ↵.
- Schettler G,
- Romer R,
- O’Connell M,
- Molloy K
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵.
- Cleveland WS
Citation Manager Formats
Sign up for Article Alerts
Article Classifications
- Physical Sciences
- Environmental Sciences
- Social Sciences
- Anthropology