2
Dendrochronology and Lessons Learned from Irish Examples
Abstract
The increasing number of sites and artefacts being dated by dendrochronology highlights the success rates with which this technique can be applied in Irish wetland archaeological situations. Completeness of samples with respect to sapwood (the outer band of rings below the bark which represent the last years of the tree’s life) is the limiting factor for dating precision and subsequent archaeological interpretation. Examples of the dating method and its associated limitations are discussed.
Introduction
When archaeologists supported the construction of an oak chronology in Ireland, the exercise was primarily aimed at the calibration of the radiocarbon time-scale. If a long chronology could be constructed by overlapping the annual ring patterns of Irish oaks with those of timbers of the same species from Medieval buildings, from archaeological sites and from natural, mostly bog, sources, then a standard chronology would exist against which radiocarbon could be tested. In fact, what happened was that between 1970 and the mid-1980s a year-by-year oak chronology was constructed in Ireland which extended back to 5289 BC (Pilcher et al. 1984). The chronology was continuous and precisely dated so that samples of exactly known calendar age could be supplied to radiocarbon laboratories. Moreover, the oak chronology was replicated by step-wise comparisons from Ireland through to England to parallel, independent, oak chronologies in Germany. Thus, by 1986 there existed not only a replicated tree-ring chronology, but a definitive Irish radiocarbon calibration curve dating back to 5200 BC that could be used to convert radiocarbon dates from Irish sites and deposits to estimates of real age (Pearson et al. 1986). It is worth remembering that the calibration curve is, by definition, precisely dated, so that it is possible to view changes in the atmospheric concentration of radiocarbon at any point in the last seven millennia.
By the time the calibration results were published, it was already known that the oak chronology could be used to directly date any long-lived oak specimens that were obtained from natural bog or lake contexts, or from archaeological sites and buildings. Thus, for the last two decades a list of well-dated archaeological sites and artefacts has been developed in the Palaeoecology Centre, Queen’s University Belfast (Fig. 2.1).
Figure 2.1: Cumulative plots of the numbers of tree-ring dated Irish archaeological sites through time. The discontinuous nature of the distribution is apparent.
In this chapter we wish to look in a little detail at the dating process and at how even dendrochronological dates have to be interpreted depending on the completeness of the samples. We will review issues concerning the importance of sapwood and how it is the limiting factor when it comes to answering questions about cause and effect in the past. We will draw the distinction between the issue of dating an individual site and the patterns of dates arrived at when numerous site dates are combined, as in Figure 2.1. As should be obvious from Figure 2.1, the dates are not random but are bunched in time, with clear construction episodes separated by long periods from which few or no archaeological oak timbers appear to survive. Alongside this picture of human activity in the past Irish landscape, some efforts have been made to understand environmental information contained in the tree-ring records. In combining these different lines of enquiry we have been privy to the first clear glimpse of what may have been going on in the Irish past, at high chronological resolution.
Practicalities of Dendrochronology
It is fair to say that dendrochronology only seems easy in theory. The method has to be kept under some rigorous control otherwise unreliable results may start ‘leaking’ into the literature. Perhaps the simplest way to explain this process is to start by outlining an ideal dating situation and then working down from there.
Imagine we are working in a small area such as the north of Ireland. Within that area we can expect the oak tree-ring signal–that is, the signal to which most of the oaks growing in the area are attempting to respond–to be fairly uniform. If we take a selection of oaks from different sites within the north of Ireland and build a master chronology, that chronology is highly ‘stable’. The easiest way to demonstrate what is meant by ‘stable’ is to repeat the exercise. Thus, we can take another fresh selection of oaks and build a new oak chronology for the same general area. We have done this and the result is two totally independent master patterns that are as similar in terms of year-by-year detail as to be effectively identical. In scientific terms this means, in our ideal situation, that we work in an area where it is possible to saturate the signal–i.e. there is only one master signal producing one master chronology and it is a fair reflection of what an average oak was doing in terms of year-by-year growth. One obvious question is whether this chronology is ideal, adequate, inadequate, or even useless, for the whole island of Ireland. Here we have to fall back on empirical evidence. We know that for many periods we can date oak timbers from anywhere in Ireland with a high degree of reliability. However, there seem to be other periods when this ideal situation breaks down. These could be periods when the master chronology is not very well replicated, or where all the sites in the chronology tend to come from a small area (for an example of this, see Baillie and Brown 1991; 1998), or where some alteration in climate means that different parts of the island are temporarily in different zones as far as the tree-ring signal is concerned. There are other potentially compounding factors; for example, the differences between bog-grown and land-grown timbers. Most of the prehistoric chronology was constructed of bog oaks that had grown on the surfaces of peat bogs. These trees are somewhat different in their growth response from land-grown timbers that have been preserved on archaeological sites. Figure 2.2 shows a section of pure Irish bog oak chronology plotted against a section of archaeological oak chronology. While the curves are very similar, there are some notable differences.
This observation implies that on some occasions it may prove difficult to date sections of archaeological chronology against a pure bog-oak chronology, especially if there is a significant distance factor involved. A good example of this would be the first Corlea track to be dated (Raftery 1990). Initially, only one oak timber was available from this Co. Longford site and it failed to date against the Belfast (northern) master chronology. However, as soon as even four Corlea ring patterns were averaged together the resultant Corlea site chronology cross-matched immediately against the Belfast master at 148 BC.
This latter point brings out the difference between single ring patterns and replicated site-master chronologies. By averaging several individual ring patterns together from a single site we move the signal away from the single tree situation and closer to the saturated master described above. There is no exact rule concerning how many timbers have to be averaged together to ensure that a site chronology will date against a master chronology. Experience suggests that a mean chronology of three or more timbers, however, will stand a good chance of dating successfully. One reason for the rather tentative nature of these statements is that there are other factors that need to be taken into account. One critical issue is chronology length. There is no doubt, in our experience, that individual ring patterns and even site masters must have 100 rings or more to stand a reasonable chance of successful, and reliable, dating. Workers elsewhere have different experiences but in Ireland we would advise extreme caution when trying to date any short ring patterns.
Figure 2.2: Although the year-to-year detail in bog oak and archaeological chronologies is very similar, it is clear that in this mid-twelfth century episode bog-oak growth was more severely affected.
We can summarise the above by saying that where multiple, long, ring patterns can be built up into site chronologies, or where very long (say >150 years) individual tree-ring patterns are available then there is a high probability of successfully cross-dating material against the available master chronologies in Ireland. This has been particularly well demonstrated by the high level of success in dating trackways and other wetland structures from the Bronze and Iron Ages.
Moving From General to Specific Issues
We have indicated that oak ring patterns can be dated with a high degree of success in Ireland (indeed the same can be demonstrated across northern Europe). However, establishing the exact calendar dates of every ring in a tree-ring pattern is not the same as providing a useful date for an archaeologist, building historian or palaeoecologist. This is because of a fundamental issue in dendrochronology, namely, that the most useful date that a dendrochronologist can produce is the date of the final growth ring–the death date. Whether it is the construction date of an archaeological site, the date of the storm that blew down a bog oak, or the date of the rise in water level that killed a lake-side pine, the only date that allows proper interpretation is the death date of the tree. This brings us to the issue of sapwood. Many tree species have a clearly visible band of rings immediately below the bark, called the sapwood to distinguish it from the inert and usually darker coloured heartwood. The presence of living cells in the sapwood allows it to be distinguished from the inner heartwood. In any dendrochronological exercise, the presence of complete sapwood allows the researcher to establish an absolutely precise calendar record of events via the study of dates associated with the death of trees.
In real life situations, timbers preserved under various conditions will have suffered various amounts of sapwood loss; in extreme cases, heartwood rings will also be missing through rot or woodworking practises. These situations, where samples are effectively incomplete, mean that many dendrochronological dates have associated errors because an allowance for missing sapwood has to be added to the date of the last existing ring. In extreme cases, where there is no evidence for sapwood on a sample, the dendrochronologist is forced to add a sapwood allowance together with a cautionary statement that ‘the actual felling date may be in or after the estimated felling range’. In such cases, the dendrochronological date serves only to provide a terminus post date. The dendrochronologist may be able to give the exact calendar date of the last existing growth ring, but that date could be years, decades, or in extreme cases, centuries prior to the actual death date of the tree. Since it is the death date that is important when examining archaeological or environmental activity, these terminus post dates come a very poor second to the exact dates conferred by samples with complete sapwood. It is therefore beholden on anyone wanting tree-ring dates to try, as far as possible, to acquire samples with their sapwood intact.
Sapwood Estimation and Preservation
The early tree-ring work at Belfast included measurement of the number of sapwood rings on a range of oak samples. The mean value obtained for Irish oak was 32 ± 9 rings (at one standard deviation, 32 ± 18 at two standard deviations); no evidence exists that this range needs to be altered, although it is now known that the number of sapwood rings varies considerably across Europe (Hillam et al. 1987). In addition to estimating sapwood ring numbers, the early work also proposed that, depending on sapwood presence/absence, several different qualities of dendrochronological dates can be envisaged. These are:
Type A–Where complete sapwood is present and a ‘death date’ can be given.
Type B–Where some trace of sapwood is present and the addition of a sapwood allowance onto the date of the last heartwood ring provides a good estimate of felling range.
Type C–Where there is no trace of sapwood, heartwood rings may be missing and a terminus post date as noted above is provided.
Because dendrochronologists have always advised archaeologists that complete sapwood is needed to obtain a precise felling date for an oak sample, wetland archaeologists have provided almost ideal samples for studying issues related to sapwood preservation. Recent examination of samples with sapwood has provided some interesting statistics.
Table 2.1: The numbers of grouped archaeological sites from the prehistoric period with a breakdown of the types of dating quality as indicated by complete sapwood (Type A), some sapwood (Type B) and no sapwood (Type C).
If we look at the period from around 1500 BC to around 100 BC, the Palaeoecology Centre has dated oak timbers from about 65 archaeological sites; as shown in Figure 2.1, these cluster into clear groupings through time. For the group of sites c. 1600–1400 BC there are 22 phases of activity evident from 21 sites. From these phases, only five sites provided samples with complete sapwood; 11 sites provided timbers with some sapwood, or the heartwood-sapwood boundary, present; six sites provided timbers with no sapwood present, including some with heartwood rings missing. So, almost one third of Middle Bronze Age sites produced samples that provided the poorest quality, Type C, dating; less than one quarter produced ideal dating. Equivalent numbers for the Later Bronze Age and Iron Age groups of sites are provided in Table 2.1. Given the fact that many archaeologists know the importance of sapwood, and should hence have been trying to maximize the number of samples with complete sapwood, it is possible to use the information in Table 2.1 to speculate on whether woodworking practises had changed through time or if sites of some periods were exposed to the elements for longer periods before being buried in wetland contexts. For example, how else might we explain the larger number of samples from the c. 1200–900 BC grouping that exhibit no sapwood, as compared with the c. 1600–1400 BC group?
Referring to Table 2.1, we would have expected, considering the good conditions for preservation in Irish peat bogs, that we might have observed a greater proportion of samples with sapwood. In some ways this presents a rather depressing picture. It is important to recognise that dendrochronological dating raises expectations; archaeologists are expecting calendar dates from dendrochronological exercises; palaeoecologists are also expecting to be able to answer questions at calendar resolution. For example, in the tenth century BC we would want to know if there were more discrete episodes of construction activity within the century, i.e. we are asking questions that require mostly Type A dates. We want to know if the building activity is responsive to a changing environment, as possibly reflected in short-term environmental changes deduced from both tree-ring patterns and archaeological information. Unfortunately, with only one fifth of sites producing actual felling dates, the question cannot be answered, even though at face value we have 28 tree-ring dated sites for study.
Interestingly, if we look at dates for a set of Early Christian period horizontal mills we tend to get a slightly better preservation of sapwood. Out of the 32 sites dated, eight have complete sapwood and produce Type A dates, while seventeen produce Type B dates. Overall, it seems that if archaeologists want answers involving real dating precision, they need to push up the proportion of samples capable of giving Type A dates. Unfortunately, because even Type B and C dates seem adequate in most archaeological contexts (because they are better than most other available dating methods) most archaeologists are satisfied with such dates at a site level. However, when larger questions are posed involving archaeological activity in the context of precisely-dated environmental change, or regional site construction, the problems and inadequacy of Type B and C dates become fully evident. In other words, archaeologists must always strive to get the best and most refined dates possible if their sites are ever to fit into the world of real chronology that is currently being developed by dendrochronologists and ice-core workers, among others.
Late and Post Medieval House Structures and Sequences
Recently Brown (2002) has published a complete list of Irish buildings that have provided dendrochronological dates. In these later periods, we can look at some examples of sites in which the timbers have complete sapwood present on most, or all, of the samples. We can contrast relatively simple wood exploitation in some structures with more complex use patterns in others. A good example of this contrast in the Post Medieval period would be to compare results on timbers from Rathfarnham Castle, Dublin, with those from Cultra House, Co. Down.
Rathfarnham Castle, Dublin
From documentary evidence, it was believed that Archbishop Loftus, having acquired the land forfeited to the crown in the 1570s by the rebellion of James Eustace (Viscount Baltinglass), built a massive castle at Rathfarnham in the 1580s; most likely in 1583. In 1996, the then Office of Public Works requested that a dendrochronological exercise should be carried out on the roof timbers from the castle. As a result of the request, nine oak timbers from the roof and three other timbers from floor joists at Rathfarnham Castle were cored (hollow tubular drills are used to extract pencil sized cores and these ring records are supplemented where possible with overlapping sections cut through surviving sapwood). Two sash pulleys were also made available for examination.
While the primary reason for this exercise was to test the historical record, it needs to be remembered that dendrochronological dating, despite the comments about Type A dating above, does not give the actual building date for a structure; it merely gives the felling or death date closest to the building date. Thus there is no expectation that dendrochronology, even on complete samples, will yield the documentary date. Timbers have to be transported and could be stockpiled for building; even seasoning may occur. As a result, the expectation is that dendrochronological dates falling in the few years before the documentary target date would be acceptable evidence that the historical documentation was broadly correct.
However, so refined is this type of dating that it is necessary to give even more detail. All the timbers sampled from Rathfarnham Castle were oak, Quercus robur or Quercus petraea. This species of tree is ring porous and in transverse section exhibits large spring vessels at the start of the year’s growth, with much smaller vessels and thick walled fibres later in the growing season. Leaf opening is at the end of April but the bulk of early-wood synthesis is in May; summer growth is finer-celled, represented by small summer vessels (Baillie 1982). Thus, when a sample has complete sapwood present the final ring can be examined to see how much annual ring growth had taken place before the tree was felled, i.e. some suggestion of ‘season’ is possible with complete samples. If only the spring vessels are complete, felling is in the early summer of that year. If the ring appears complete with both spring vessels and summer wood, then felling will have been later in that year or early in the following year.
Table 2.2: Details of the individual samples from Rathfarnham Castle, Dublin. The letters after the number of rings indicate if the dating was of Type A, B or C.
Of the 12 samples taken from the main timbers at Rathfarnham, eight had complete sapwood; two had some sapwood present, while two had neither sapwood nor the heartwood-sapwood boundary present. With eight of the 12 samples providing complete sapwood a study can be made of the probable season when the trees were felled, see Table 2.2.
For the eight samples with complete sapwood, a remarkably consistent picture emerges. All eight were felled in the year from early summer 1583 to spring 1584 at the latest. However, it is most likely that they were felled in the period bracketed by early summer to autumn 1583, given that they represent a population intended for use in a prestige building. The other samples are not inconsistent with the 1583 dating. For example, the last existing ring in Sample Q9182 grew in the year 1581. As the last ring could not be identified as the final ring under the bark, a sapwood allowance has to be added to the date of the last heartwood ring. This produces a date range of 1594 ± 9 years. However, detailed examination of the sample indicates that in reality only the outer few rings of sapwood are damaged and it is reasonable to suggest that felling was probably consistent with the main 1583 phase.
Just to labour the issue of interpretation, Sample Q9179 has sapwood out to the year 1561 but it appears incomplete. The best estimate of the felling range for this sample can be stated as ‘1563 ± 9 but definitely after 1561’. This may seem sufficiently prior to the target date of 1583 to suggest that this was a re-used timber. However, we would caution that the one standard deviation (± 9, giving a 68% probability) value for Irish oak sapwood is a statistical estimate that will only be correct two thirds of the time. If we apply the more reliable two standard deviation range of ± 18 years (giving 95% probability), then the estimated felling range extends to 1581 suggesting that this sample may in fact be part of the c. 1583 assemblage. Sample Q9187 represents Type C dating where it is likely that heartwood rings have been removed during woodworking. Thus, its seemingly early date range of ‘1544 ± 9 or later’ could easily be explained as the result of woodworking rather than re-use of a genuine older timber.
Overall, a perfectly reasonable interpretation of this group of samples is that they are all part of the original group of trees felled to build Rathfarnham Castle. The complete samples allow us to propose that the trees were felled in the six months after April 1583 or just possibly into early 1584. As seasoning is not believed to be a serious issue for building timbers in Ireland in the Later Middle Ages, it is likely that the house was indeed built in 1583 or 1584. The very fact that so many of the timbers retain their sapwood is suggestive that the timbers were not seasoned as this would probably have contributed to sapwood degradation or loss.
This then allows an additional refinement. By viewing the correlation values of the Rathfarnham ring patterns against a range of Irish regional master chronologies it is possible to attempt some localisation of the timber source. When this was done, it appeared most likely that the bulk of the trees were collected from in and around the Dublin area. However, there is just a hint in some correlations involving Samples Q9184M, Q9186 and Q9187 that they may have been collected from an area closer to Belfast than Dublin. Such indications are suggestive rather than definitive, though experience shows remarkable consistency in location exercises (Baillie 1995a).
Cultra House, Co. Down
We can compare the excellent single-phase dating at Rathfarnham with another exercise carried out at Cultra House, Co. Down. In this case, the Historic Monuments and Buildings Branch of the Environment and Heritage Service for Northern Ireland, Department of the Environment, requested a dendrochronological study to be carried out on the roof of the house. In this case ten timbers were available for sampling (Table 2.3).
In this exercise, seven of the ten samples had sapwood present, with four having complete sapwood. This should have provided sufficiently tight chronological information to confirm the dating of the roof structure from this house. However, the results indicated a more complex building history than anticipated.
The results provided in Table 2.3 indicate the problems associated with any attempt to interpret the building history of Cultra House. Firstly, two of the potential Type A timbers, Q8875 and Q8878, contained too few growth rings to provide definitive dates. This reduced the number of samples with definitive end dates to two. One of these, Q8873, was felled in the early summer of 1670. In contrast, the last ring on Q8880 had complete spring and summer wood and was felled in either late 1666 or early 1667. This inconsistency was further compounded by the observation that the timbers with Type B dates were scattered across an extended period, ranging from the late sixteenth to the late seventeenth century. Additional evidence from the timbers indicated mortice holes and other woodworking details that were redundant to the contexts where the timbers were found, something that could be consistent with timbers having been re-used. Indeed, examination of the Ordnance Survey map for the area indicated a complete change in the orientation of the house between the production of the mid-nineteenth century map and the 1930s map. This observation raised the spectre that the roof had been dismantled and re-constructed using material from more than one source or phase.
Table 2.3: Detail of the individual samples from Cultra House, Co. Down. The letters after the number of rings indicate if the dating was of Type A, B or C.
In contrast with Rathfarnham, where a highly consistent set of dates were in good agreement with known historical documentation, the situation at Cultra, despite the fact that eight timbers could be dated by dendrochronology, is disappointing. The best that can be said is that the Cultra roof appears to contain a group of timbers (namely Q8873, Q8877, Q8879 and Q8880) felled in the vicinity of 1670. If this was the original building phase then it may have contained at least some re-used timbers and, in addition, there is a possible hint of later activity in the form of Sample Q8881 (though even here it has to be remembered that this sample could have been felled as early as 1671). The limitation imposed on interpretation by incomplete tree-ring samples is well illustrated by this example. As noted, there is no actual guarantee that the roof, as observed, originally belonged to Cultra House at all.
Early Christian Examples: Nendrum Tidal Mill and Island MacHugh Crannog
The Nendrum tidal mill in Strangford Lough, Co. Down (McErlean and Crothers 2001), offers a further insight into timber dating. Recent excavations at the site have produced a series of timbers relating to the mill structure and an accompanying revetment that was used to capture the water to run the mill at low tide. In total, 27 samples were submitted for dating. This included no less than 18 samples with complete sapwood; unusual by most standards, as detailed above. Archaeologically, the mill complex has three clear, well separated, phases. However, virtually all the timber samples turn out to be from the initial phase of construction. In this case detailed examination of the dates shows that three samples were felled either late in AD 618 or early in 619, and no less than six timbers were felled in the early summer of 619. This is suggestive that either the initial pulse of building activity was undertaken in the spring or early summer of 619, or that this represented the first accumulation of timbers for subsequent building. Felling activity then continued with another three timbers felled either late in 619 or early in 620, and a further three felled in either late 620 or early 621. The question raised by this spread of dates is whether building was spread through time or whether timbers were stockpiled from 618/619 for use in construction in 620/621. This is compounded by the observation that a number of the timbers had mortise holes with no obvious functionality. While the excavator interpreted these mortises as being due to the re-use of timbers, both the dates, and the pristine nature of the sapwood, argue against re-use (mortises towards the ends of heavy timbers may have been simply to assist with transport).
While the issue of stockpiling timbers or extended building may sound esoteric, we appear to see the beginnings of such a pattern in the early seventh century. At Island MacHugh Crannog, Co. Tyrone, palisade oak posts, with complete sapwood, indicated felling in at least seven separate episodes from AD 601 to 619 (specifically 601, 602, 603, 608, 611, 614 and 619)–a period assumed too long to be simply due to stockpiling (Baillie 1993). In fact, the long distribution of dates from the site raises questions about the date of a site such as Island MacHugh–is it AD 601 when activity began, is it 610, the mean date of the activity, or is it 601–619, the identifiable period of activity? This is a type of problem not normally encountered by archaeologists. However, the reason for detailing the issue is that Crone (2000) has observed exactly comparable activity at Buiston Crannog in Scotland. In that case building activity started in AD 550, interestingly close to the date of the earliest Ulster crannogs that were constructed from around AD 550 onwards (Baillie 1982). There then followed major felling phases for palisade construction and refurbishment in 594, 608, 613, 614, 615, 616, 620, 629 etc. Crone recognises the difficulty in interpreting such dates and, as part of a complex set of arguments, notes:
‘Where all the dated sequences were felled in the same year and there is no spread of dates, we may assume that the felling date reflects the construction date. However, where there is only a single sequence with bark edge providing the felling date, we cannot know whether this represents the actual construction date or one of the years in which timber for stockpiling was felled, construction taking place some time afterwards’ (Crone 2000, 54).
Taking Crone’s argument at face value, we could toy with the idea that the Nendrum mill and revetment was actually constructed in the winter of AD 620 or the spring of 621; after all, is there any point in partially constructing a mill? However, the main point is that more thought is necessary on issues relating to tight chronology and how to interpret sets of data such as these. What is important is the observation that complete sapwood offers up a whole new realm of chronological detail for investigation.
Wider Considerations
The story of the construction of the Irish oak master chronology has been detailed elsewhere (Baillie 1982; 1995a). Suffice to say that large numbers of samples (of the order of 10,000) were accumulated from a wide variety of natural and archaeological contexts over some three decades. As a result of that effort we can look at quite a range of factors involving the past that have never previously been available. As indicated in Figure 2.1, we now have some measure of the distribution of archaeological oaks through time. We can parallel that observation with the results of the dating of an extensive, near-random, collection of bog oaks–trees which grew on bog surfaces and were preserved buried in the peat. If we look at a plot of the frequency of these naturally preserved trees it is apparent that there were periods from which trees survive in quite large numbers separated by other periods when bog oaks were relatively scarce. Figure 2.3 shows the overall bog oak distribution for Ireland. We can also speculate why this pattern might exist (Baillie and Brown 1996). The important aspect from an environmental archaeological perspective is that all the information is precisely dated. Thus, for the first time, we can see certain hints of at least some aspects of what was happening in the Irish past in real time. In addition, because of the vast amount of effort put into radiocarbon chronology by archaeologists in Ireland we can make some moves towards integrating the two records. Thus, we know from calibrated radiocarbon assay that the Neolithic started/arrived in the period 4000–3800 cal. BC, while the Bronze Age starts roughly in 2500–2300 cal. BC. We can now view cultural change against a backdrop of both archaeological dates and some environmental information.
Figure 2.3: A frequency distribution of Irish bog oaks through time. The distribution of lake edge pines (P) is shown for comparison.
Table 2.4: Provisional list of notable prehistoric dates from the various strands of tree-ring evidence.
In addition to these sources, there are other dates that have come directly from analysis of the tree-ring chronologies themselves, several of which have given rise to some level of debate. These include the so-called ‘narrowest-ring events’ that appear as episodes of very unpleasant growth conditions for bog oaks in Ireland. These are extreme events in the sense that at least some of them show up in other records outside Ireland hinting at global environmental downturns (Baillie and Munro 1988; Baillie 1995a). The concept that global events could influence populations in Ireland takes a little getting used to. However, once it is realised that any major dust-veil event can affect the environment of most of the planet by placing a veil of material between the Sun and the Earth’s surface (whether it is caused by explosive volcanism or by extraterrestrial loading of the atmosphere as a result of impacting space debris), it is not so surprising that from time to time Irish trees and Irish human populations were adversely affected by global environmental downturns. To elaborate, it is now possible to imagine dust-veil events that would cause several years of extreme climate, with cold being the principal vector.
| DATE BC | ARCHAEOLOGICAL EVENT |
|---|---|
| c. 5500* | Mesolithic microblade to large blade transition. |
| c. 4000* Well-defined | start of Neolithic. |
| c. 3200* | Neolithic settlers arrive in Boyne Valley. |
| c. 2300* | End of Neolithic, start of the Bronze Age. |
| c. 1700–c.1200* | Middle Bronze Age. |
| c. 1200–c.1000* | ‘Bishoplands phase’ of Late Bronze Age. |
| c. 1000–c. 900* | ‘Roscommon phase’ of Late Bronze Age. |
| c. 900–c. 600 | ‘Dowris phase’ of Late Bronze Age. |
| c. 7th–3rd centuries* | One of most enigmatic periods in Irish prehistory. |
Such conditions could have caused the loss of consecutive harvests and could have caused extreme discomfort for agriculture-based populations. Once such possibilities are realised, extreme cases have to be considered, i.e. there could have been circumstances that collapsed agriculture altogether, requiring its re-introduction to the island. There is currently no particular evidence to suggest that this happened in the case of Ireland, but the possibility needs to be borne in mind. There are therefore several ways of deducing aspects of change in the Irish past and it is interesting to draw together a provisional list of dates from the various strands of tree-ring evidence (originally published in Baillie 2001), as shown in Table 2.4.
If we turn to Cooney and Grogan’s (1994) book Irish Prehistory, A Social Perspective it is possible to list the following dates cited by them as change points in Irish prehistory (Table 2.5); the dates would be regarded as generally acceptable by most prehistorians. The similarity between the dates in Table 2.5 (those marked with *) and some of the dates provided in Table 2.4, especially given the difficulties experienced by archaeologists when attempting to date cultural change, implies that environmental factors that affected the trees or their survival, had some direct bearing on the happenings and changes, identified in the Irish archaeological record.
Longer Distance Connections
Shortly after the ‘narrowest ring events’ were discovered in the Irish oak record it became clear that the environmental effects were not always restricted to Irish trees. From early on in the tree-ring project, it was possible to compare tree-ring patterns from different areas within Ireland with those from Scotland, England and Germany. In fact it was these long-distance comparisons that allowed the chronologies to be replicated. So, for example, we can compare long records from Ireland and England back to 5000 BC. Most of the time the growth patterns are very similar, but from time to time notable opposite departures are observed. These ‘differences’ imply changed growth conditions on opposite sides of the Irish Sea and these episodes may have interesting implications for how the environment affected populations in the different regions. Such thinking can be extended by making comparisons between the long Irish, English and German oak records, all of which exist for the whole of the last seven millennia.
Figure 2.4: Comparison of the northern hemisphere pine density record, NHD1 (from Briffa et al. 1998), with a generalised European oak chronology. There is notable similarity in response at AD 1601, contrasted with notable differences at AD 1635 and AD 1641.
It is also now possible for some periods (especially in the last few millennia) to make comparisons globally, i.e. to view Irish and European oak growth in the wider context of tree-growth throughout the temperate world. Such a statement would have been impossible before 1990. To take a recent example, Briffa et al. (1998) have analysed ring widths and ring densities in pine chronologies from around the Northern Hemisphere. They note that maximum late-wood density is a good reflection of summer temperature, so they can produce a summer temperature record for a large swathe of the Northern Hemisphere for high latitude/high altitude sites. They note that the coolest years in the last 600 years consistently coincide with known volcanic events. This makes sense in that explosive volcanoes inject dust and acid into the stratosphere causing cooling at the earth’s surface. However, the trees they are using are relatively rarefied, growing in temperature sensitive areas. We, on the other hand, are dealing with temperate oaks. Because all the tree-ring records are precisely dated, European dendrochronologists can compare their growth records with this published density record (called NHD1). When we do this we find that there are interesting differences in response. Figure 2.4 shows the two records across the early seventeenth century AD. There are some cooling events, such as AD 1601, that show in both records. However, there are other episodes where the oaks show no significant effects, such as at AD 1641 when the pines show a notable cooling, and some, for example around AD 1635, where the oaks are badly affected at times when the pines show no density reduction. Examination of such records allows us to get a feel for local and global effects involving some aspect of temperature. There is no doubt that as more and more regional chronologies become available a whole range of interesting comparisons will become possible between both temperature and moisture-sensitive records and tree-ring data.
Some Point Events–Splitting Up Archaeological Time
The principal effect of producing a dendrochronological time-scale has been to introduce real dates into the previously somewhat vague archaeological time-scale that existed in Ireland. In our view, the main effect of identifying events of various kinds has been to split up ‘time’ into more manageable units. To give an example, if there are independently observed events at around 4000 BC, 3200 BC and 2350 BC in tree-ring records, then, from a discussion point of view, the Irish Neolithic is forced to conform–to fit itself around–those events. Perhaps not surprisingly it does just that, with the generally accepted arrival of the Neolithic being around 4000 cal. BC, the Early Neolithic continuing until around 3200 cal. BC and the Bronze Age beginning (and the Neolithic ending) somewhere in the vicinity of 2500–2300 cal. BC. Moreover, if people wish to explore the possibility of pre-elm decline cereal pollen (e.g. Edwards and Hirons 1984) then this can be assigned to the period before 4000 cal. BC, i.e. before the environmental episode which seems to have taken place centred around 3900 BC on the tree-ring scale. Indeed, we can look at this episode in just a little more detail. Ballynagilly, the Neolithic house site in Co. Tyrone, was located on a low hill beside a peat bog. The palynology of this bog was studied by Jon Pilcher as part of his PhD thesis (Pilcher 1970). A layer of charcoal in the bog, consistent with the Neolithic occupation of the house site dates to around 3900 cal. BC. Pilcher noted that at the time when the Neolithic house was occupied, the bog was a lake and the charcoal represented deposition in lake sediment (lake mud texture as opposed to peat texture). This was interesting because in 1989 a pit was dug into the bog at Ballynagilly in order to obtain bulk samples for research into pre-elm decline cereal occurrence. When the Neolithic charcoal layer was reached it was discovered that a full pine trunk was lying horizontally just below the charcoal. This tree must have grown in relatively dry conditions (certainly not in a lake). So, with 185 growth rings and a radiocarbon determination (on rings 51–70) of 5259 ± 23 BP, i.e. c. 4200–4000 cal. BC, we can surmise that there had been a long, relatively dry, period in the centuries before 4000 cal. BC. This fits well with a separate observation published by Baillie and Brown (1996). Here, date ranges of radiocarbon-dated pines from locations around lake margins (trees that had come to light when lake levels were lowered, or whose location indicated that water levels was relatively low when the trees were alive) were plotted against the overall replication of bog oaks. It was observed that there were several occurrences of lake-edge pines between 5000 and 4000 cal. BC, implying generally drier conditions during that period. Figure 2.5 shows the life spans of dendrochronologically-dated oaks that conform to this same model, i.e. oaks from lake margins. Here, we also see the occurrence of many such oaks between 5000–4000 BC. Putting all this information together it seems that around 4000 cal. BC (in a range suggested elsewhere to be bracketed by 4000–3800 cal. BC) there is a change from relatively dry to relatively wet conditions. So, although the story requires further refinement there is little doubt that a picture is emerging that allows us to see the archaeological Neolithic in a broader environmental context.
Figure 2.5: The frequency of Irish bog oaks (from Fig. 2.3) with the dates of lake-edge oaks plotted for comparison. It is evident that conditions must have been much drier between 5000–4000 BC.
Most of the evidence involving lake-edge oaks and pines relates to trees from small lakes. However, in Ireland, Lough Neagh is of special interest because of two factors; firstly its enormous size, and secondly its single exit to the sea via the Lower Bann. This means that Lough Neagh levels may not be controlled purely by the amount of precipitation. The single exit means that the level could be partly controlled by blockages of various kinds. On at least two occasions, most notably at 2350 BC (Baillie 1995a) and at AD 540 (Baillie 1995b), there are good grounds for suggesting that there were significant changes in the level of the Lough. The former event is suggested by damage/ anomalies in trees from both the north and south of the Lough, the latter by the discovery of an unfinished dug-out in the Lough at around AD 540, coupled with the observation of dramatic growth effects in several bog oaks from the lake edge. In the case of the 2350 BC ‘event’ (starts 2354, lasts until 2345), the strangest twist is the coincidence of the date with a reference in the Anno Mundi section of the Irish Annals that refers to lakes ‘breaking out’ with a date of ‘2341’ BC and mentioning the ‘Plain of Lough Neagh’ (Baillie 2001).
We can obviously look at other events, in some cases very short-term events. For example, the European oak chronologies show an interesting growth downturn at AD 235–7. What serves to make this interesting is that the downturn may have been due to the explosive eruption of the Taupo volcano in New Zealand. For some time, there has been a debate as to whether Taupo erupted around AD 186 or around AD 235. Although the answer is not known with any certainty, workers who have studied the exceptionally explosive nature of the eruption have gone so far as to suggest a global shock wave (Lowe and de Lange 2000). This implies it is possible that people in Ireland at the time may actually have heard, or felt the effects of, a volcano on the opposite side of the world. Ironically, the archaeological record in Ireland is sufficiently thin during the first few centuries AD that there is no way of knowing whether the events around AD 235 had any effect on Irish populations.
Conclusions
Dendrochronological dates and environmental information from the tree-ring chronologies allow the piecing together of new pictures of the Irish past. At a detailed level, the examples given demonstrate the potential of having access to such information. They also serve to reinforce the major step down from absolute dating precision given by complete tree-ring samples to the much less precise ranges deduced from incomplete samples. In this context it is worth repeating the comments relating to the tenth century BC construction episode as an example. Because it is possible to have independent information on past environmental conditions at annual resolution, we want to test the hypothesis that building in wet areas was conditioned by environmental considerations. This demands that the dates of construction be ascertained to equivalent precision i.e. that complete sapwood is sought specifically. However, this exposes a dilemma. Excavators requiring dendrochronological dates are normally asking fairly coarse questions, such as ‘is our site/structure sixteenth or tenth century BC’? Once the excavator has an answer to, say, decadal precision, that is adequate for his or her purpose. The dendrochronologists are, however, asking more refined quasi-annual questions. In an ideal world, once a site/ structure is dated, more samples should be obtained to allow the maximum dating refinement. Unfortunately, this is not in the archaeologist’s brief; neither is it possible for the dendrochronologist to return to excavation sites and arbitrarily acquire more samples with complete sapwood. The only logical answer is that budgets include provision for the acquisition of additional samples with sapwood. Only such an approach will help us out of the impasse where 28 Late Bronze Age sites can be dendrochronologically dated but, of these, only five can be realistically interpreted.
References
Baillie, M. G. L. 1982. Tree-Ring Dating and Archaeology. London: Croom-Helm.
Baillie, M. G. L. 1989. Hekla 3–just how big was it? Endeavour 13, 78–81.
Baillie, M. G. L. 1990. Irish tree-rings and an event in 1628 BC, pp. 160–6 in Hardy, D. A. (ed.), Thera and the Aegean World (Vol. 3). London: The Thera Foundation.
Baillie, M. G. L. 1992. Dendrochronology and past environmental change. Proceedings of the British Academy 77, 5–23.
Baillie, M. G. L. 1993. Archaeological wood in Ulster. Archaeomaterials 7, 139–50.
Baillie, M. G. L. 1995a. Dendrochronology and the chronology of the Irish Bronze Age, pp. 30–7 in Waddell, J. and Shee Twohig, E. (eds.), Ireland in the Bronze Age. Dublin: The Stationery Office.
Baillie, M. G. L. 1995b. A Slice Through Time: Dendrochronology and Precision Dating. London: Routledge.
Baillie, M. G. L. 1999. A view from outside: recognising the big picture. Quaternary Proceedings 7, 1–11.
Baillie, M. G. L. 2001. Some things we will never know: exploring the limits of cause and effect in environmental research, pp. 43–54 in Raftery, B. and Hickey, J. (eds.), Recent Developments in Wetland Research (Monograph Series No. 2). Dublin: Department of Archaeology, University College Dublin.
Baillie, M. G. L. and Brown, D. M. 1991. A dendro-date from Haughey’s Fort? Emania 8, 39–40.
Baillie, M. G. L. and Brown, D. M. 1996. Dendrochronology of Irish bog trackways, pp. 395–402 in Raftery, B., Trackway Excavations in the Mountdillon Bogs, Co. Longford (Irish Archaeological Wetland Unit, Transactions Vol. 3). Dublin: Department of Archaeology, University College Dublin.
Baillie, M. G. L. and Brown, D. M. 1998. The dendro-date from Haughey’s Fort. Emania 17, 45–6.
Baillie, M. G. L. and Brown, D. M. 2002. Oak dendrochronology: some recent archaeological developments from an Irish perspective. Antiquity 76, 497–505.
Baillie, M. G. L. and Munro, M. A. R. 1988. Irish tree-rings, Santorini and volcanic dust veils. Nature 332, 344–6.
Briffa, K. R., Jones, P. D., Schweingruber, F. H. and Osborn, T. J. 1998. Influence of volcanic eruptions on Northern Hemisphere summer temperature over the last 600 years. Nature 393, 450–5.
Brown, D. M. 2002. Dendrochronological dated building from Ireland. Vernacular Architecture 33, 71–3.
Cooney, G. and Grogan, E. 1994. Irish Prehistory: A Social Perspective. Bray: Wordwell.
Crone, A. 2000. The History of a Scottish Lowland Crannog: Excavations at Buiston, Ayrshire 1989–90. Edinburgh: Scottish Trust for Archaeological Research.
Edwards, K. J. and Hirons, K. R. 1984. Cereal pollen grains in pre-elm decline deposits: implications for the earliest agriculture in Britain and Ireland. Journal of Archaeological Science 11, 71–80.
Hillam J., Morgan, R. and Tyers, I. 1987. Sapwood estimates and the dating of short ring sequences, pp. 165–85 in Ward, R. G. W. (ed.), Applications of Tree Ring Studies (BAR International Series 333). Oxford: British Archaeological Reports.
Lowe, D. J., and de Lange, W. P. 2000. Volcano-meteorological tsunamis, the c. AD 200 Taupo eruption (New Zealand) and the possibility of a global tsunami. The Holocene 10, 401–7.
McErlean, T. and Crothers, N. 2001. Tidal power in the seventh and eighth centuries AD. Archaeology Ireland 15 (2), 10–14.
Pearson, G. W., Pilcher, J. R., Baillie, M. G. L., Corbett, D. M. and Qua, F. 1986. High-precision 14-C measurement of Irish oaks to show the natural 14-C variations from AD 1840 to 5210 BC. Radiocarbon 28, 911–34.
Pilcher, J. R. 1970. Palaeoecology and the Radiocarbon Dating of Sites in Co. Tyrone, Northern Ireland. Unpublished Ph.D. Thesis, Queen’s University Belfast.
Pilcher, J. R., Baillie, M. G. L., Schmidt, B. and Becker, B. 1984. A 7272-year tree-ring chronology for Western Europe. Nature 312, 150–2.
Raftery, B. 1990. Trackways Through Time: Archaeological Investigations on Irish Bog Roads 1985–89. Dublin: Headline.