science

1.5 Dendrochronology in Scotland

Dendrochronology is the technique of dating wood through the measurement and analysis of the growth-patterns of the parent tree. It is the most precise dating technique available to building historians and archaeologists because, given a complete sample with bark edge, the calendar year in which the tree was felled can be identified. The statistical correlations on which the dating method is based also enables the source of the wood to be provenanced. Other growth-ring parameters (Maximum density (Polge 1970); Blue Intensity (Wilson et al. 2011) and stable isotopes (McCaroll and Loader 2004)) can also be used but all the dating work reported below is based on ring-width measurement.

In Scotland there is a commercial dendrochronological laboratory at AOC Archaeology Group where Anne Crone specialises in the analysis of standing buildings and archaeological materials. In the School of Geography and Geosciences at the University of St Andrews Rob Wilson is working on the development of a native pine network based on living trees and sub-fossil data. Coralie Mills is a research fellow at St Andrews and provides freelance dendrochronological services for both the cultural and natural heritage sectors.

As in the rest of the UK dendrochronological studies in Scotland have focused primarily on oak (Quercus sp.), because this is the species most commonly used in construction from the Neolithic until the post-medieval period. Chronological and geographical coverage in Scotland is patchy and the existence of robust local chronologies influences the likelihood of successful dating. The bulk of the oak data for Scotland has come from the sampling of standing buildings, with only small amounts coming from archaeological and other sources. Consequently, there is more data for the medieval and post-medieval periods than for earlier periods. From the medieval period one issue dominates cultural dendrochronology in Scotland and that is the distinction between native-grown and imported timber, and the implications this has for woodland history and the timber trade. Until the late 15th century, long-lived native oak was widely available and material with long ring sequences from this period has been found on many urban medieval sites (Crone 2000a) and in buildings such as Glasgow Cathedral, Caerlaverock Castle (Baillie 1997) and Darnaway Castle (Stell & Baillie 1993). Many of the native oak chronologies begin as early as the mid-9th century AD (Crone 2006). Native oak has been identified in only a few later buildings, all late 16th/ early 17th century in date, and this timber tended to be young, poorly-grown oak; consequently the dating of native oak in these later periods remains problematic.

From the late 15th century imported oak was used almost exclusively in building, presumably because native-grown oak was in short supply. There is now an extensive network of oak chronologies throughout Europe and this has facilitated the identification and dating of imported oak in Scotland. Oak boards from the eastern Baltic and beams from Scandinavia have been identified by dendrochronology in many buildings dating from the late 15th century to the late 17th century (ie Crone 2008; Crone & Gallagher 2008). Consequently, a sizeable corpus of Scottish 'import' data now exists which makes it relatively straightforward to date and provenance imported oak.


Sampling underway in the roof of the Mansion House, Drum Castle, Aberdeenshire. Native oak timber, felled over a period of years from AD 1603 to at least AD 1612 was used in the construction of this roof. This is one of a very few buildings of this date which was built with native oak, which probably came from the nearby Forest of Drum. The timber was small in cross-section and knotty, with wavy, irregular grain and compressed growth, probably signalling poor management of the Forest in the previous century, © AOC Archaeology Group.

Considerable progress has been made in the analysis of Scots pine (Pinus sylvestris) in recent years. As with oak it is necessary to be able to distinguish between native-grown and imported pine. By the 17th century European sources of oak were also dwindling and pine imported from Scandinavia and the eastern Baltic replaced oak as the main building timber in Scotland. Although there are still only a relatively small number of dated pine assemblages from Scotland the growing network of master pine chronologies throughout Europe means that dating imported pine in post-medieval and early modern buildings is becoming more routine (Crone & Sproat 2011).

This is not the case with native-grown pine. A programme is underway to develop a continuous reference chronology for the last millennium, primarily for climatic reconstruction, but it will also have applications for the dating of cultural material (Wilson et al. 2011). Commercial exploitation of Scotland's pinewoods began in the 17th century the timber was intended for both shipbuilding and construction. However, at present the tree-ring data comes mainly from the isolated remnant pine woodlands in the Highlands and may not be representative of the timber used in construction. Nonetheless, two vernacular buildings in Upper Deeside have now been successfully dated against local reference chronologies (Mills & Crone 2012).

There is very little chronological coverage earlier than the mid-9th century AD, when many of the medieval chronologies begin. Attempts have been made to date numerous prehistoric and early historic sites which have produced oak timbers but these have been unsuccessful (ie Crone 1998a; 1998b; 2002), largely because the sites are geographically remote from all the reference chronologies available for the these periods but also because they have not produced timbers in sufficient numbers to construct robust site chronologies with strong climatic signals. All the sites that have been successfully dated lie in south-west Scotland, and their dating has relied on their proximity to northern Ireland or northern England, where there are regional reference chronologies constructed from trees which probably grew in similar environmental conditions to those in SW Scotland.

Currently, only two prehistoric sites have been dendro-dated; oak timbers from two crannogs, Cults Loch and Dorman's Island, Whitefield Loch, both in Wigtonshire, have been successfully dated indicating building activity in the 5th centuries and 2nd centuries BC (Cavers et al. 2011). Buiston crannog, Ayrshire produced one of the largest assemblages of oak timbers retrieved from an archaeological site in Scotland and consequently, a robust site chronology was constructed, covering the years AD 250 - 615, with indications of building activity on the crannog in the late 6th century and into the 7th century AD (Crone 2000b). Structural timbers from the Northumbrian monastic settlement at Whithorn have also been successfully dated, producing a site chronology covering the years AD 278 - 752 (Crone 1997).

The only other species which has been used in dendrochronological studies in Scotland is alder (Alnus glutinosa), mainly because it has been used extensively

in the construction of crannogs. There are no master reference chronologies for alder so it cannot be used directly to obtain calendrical dates, its greatest value lying in its potential to provide site-specific chronological relationships. However, it was used successfully at Buiston crannog where the alder chronology could be indirectly linked to the dated oak chronology through stratigraphic connection, thus providing calendrical dates for some of the structures. Work on alder from other crannogs has highlighted problems relating to sequence length, tree structure and growing conditions.

Future directions

Much of the dendro-dating of medieval and early modern Scottish material is now relatively routine and improvements in provenancing is dependent to a large extent on the continuing work of European colleagues. Work is still needed on strengthening the later sections of the native oak chronologies, particularly the 16th and 17th centuries, while the development of the native pine network is critical if we are to date much of the vernacular building stock. Efforts should focus on the development of a continuous tree-ring chronology for the 1st millennium BC, based on crannog timbers, which would help to resolve many chronological issues for this period.

Tags:

6. Statistical modelling in archaeological applications

Throughout the contents of sections 1-5 above there is an implicit use of mathematical or statistical procedures to process, treat and visualise data in many different forms. It is therefore desirable to make a few statements about these procedures that be conveniently under the general term of statistical modelling. This can be regarded as a conceptual description of the processes that generate the data observed. The mathematical formulation of this conceptual description implies setting the corresponding equations, and defining the parameters (unknown quantities) that appear in these equations.  Computationally, the model (estimate parameters) needs to be fitted using data that has been observed or measured.

The first and simplest example of a statistical model comes in 14C calibration where the process starts with calibrating a single measurement.  The process of calibration converts a 14C age (in year BP) and its associated uncertainty to a calendar age with a derived uncertainty.  The 14C age is derived from a measure of the 14C activity in the sample, assumed for most materials (the exception being aquatic or marine samples) to have been in equilibrium with the atmosphere at time of ‘life’.

The statistical model begins here by describing a probability distribution for the 14C date and its uncertainty conditioned on its unknown ‘true’ calendar age (the model parameter).  This distribution is most commonly the Normal or Gaussian distribution.  The 14C distribution is then ‘compared’ to the calibration curve to identify the calendar time window most compatible with the observed 14C activity.  The result is a range of possible (or plausible) calendar ages for the sample identified by the distribution curve on the calibrated date axis (see below for a graph downloaded from the OxCal web site).

The sophistication of calibration modelling has been aided by major developments in the calibration software.  Nowadays, there are a number of calibration programs available for terrestrial samples and for marine samples, the most widely used being Calib, BCal and OxCal and which are easily downloadable from the web or run on-line.  In addition there are a number of special purpose programmes such as as BPeat, BChron, and CaliBomb.   

Chronology building

Figure 5: Downloaded from c14.arch.ox.ac.uk ORAU May 2011

Downloaded from c14.arch.ox.ac.uk/embed.php ORAU May 2011

Dealing with one sample at a time is no longer a common occurrence, as frequently archaeologists will have a set of 14C results, related in some known way. In the simplest chronology construction setting, there would be a series of archaeological samples which are related through stratigraphy (which means that the relative ages of the samples from oldest to youngest can be defined). It may then be possible that some (or all) of the samples can be radiocarbon dated, and then in building the chronology, the 14C dates would need to be calibrated, whilst respecting the stratigraphic relationships. This step requires a more complex model but one which still has at its heart the model for calibrating a single 14C date.

In recent years within Archaeology there has been a major growth in the adoption of Bayesian models (Bayliss and Bronk Ramsay, 2004), characterized by the formal inclusion of the concept of ‘prior’ knowledge or beliefs into the modeling process.

Bayesian modeling and analysis has three key components, formally they are described as the prior, the likelihood and the posterior. Bayes theorem is used to define the posterior in terms of the likelihood (involving the data) and the prior (expressed in terms of the model parameters). The model parameters could for instance be the start and end date of an occupation layer or period, and these quantities will be described mathematically within the model.

First the prior captures expert archaeological knowledge about the parameters, so an archaeologist may be prepared to say that the start date can be no earlier than a particular event or time period, and that similarly the end date can be no later than another. An alternative approach would be to define the parameter of interest as the duration of occupation, so that the prior would provide a statement of the form “duration was greater than 20 years but less than 150 years”. Such statements would be written probabilistically (mathematically), so that in the latter case, one possibility would be to say that duration was equally likely to take any value between 20 and 150 years (or in other words to assume a Uniform distribution for duration). Considerable care is required in constructing the prior (Steir et al. 2000).

The likelihood is the next component and again is presented in probabilistic terms- here the observations are linked to the radiocarbon dates and the unknown parameters. The probability for each radiocarbon date is described given the unknown model parameters, and then these probabilities are generally multipled together to give the likelihood. The likelihood thus depends on the unknown parameters.

Finally, the posterior is then another probabilistic statement about the unknown parameters, but this time (unlike for the prior) the dates have also been used to define this statement, and in this way the archaeological knowledge is combined with the 14C measurements.

The Bayesian approach is probably best attempted in the context of calibration of radiocarbon ages using the readily available OxCal and BCal. For examples of application of the Bayesian approach in Scotland see Hamilton and Haselgrove (2009) and Hall et al. (2010).

General statistical modeling in Archaeology

More traditional statistical modeling, where there is no need for calibration, and where chronology construction is not the primary goal is also commonplace. Examples might include elemental compositions of glass shards, pottery descriptions (pattern, colour….).

Fundamentally, lying behind much of Statistics is the assumption of a population, from which a representative sample of individuals has been drawn. Statistical inference is the process whereby one generalises from the sample of observations to a wider population. The first step in defining a statistical model is to define a probability model that describes the variation in the attribute of interest within the population. A parameter of the probability model describes a population, while a statistic describes a sample. There are arguments concerning statistical models and the validity of the concept of population as applied in an archaeological context. Nonetheless, the idea of population is important, and linked to this is the idea of a probability model, which describes the variation of the variable of interest in the population. Commonly used probability models include the Normal and Poisson.

Hypothesis testing and confidence intervals

Hypothesis or significance testing is a formal method of statistical inference.  The hypotheses are framed in terms of population parameters (such as the population mean or average).

Correlation and regression

When there are several variables (measured on the same artifact), then a common investigation concerns the relationships between the two.  Such a question of interest would be answered using a) either Spearman or Pearson correlation coefficients (depending on distributional assumptions) and b) if it is considered that the value of one variable (response) is in fact determined by the other (explanatory), then one might consider building a regression model.  This latter type of modeling structure can also be extended to the situation where there are potentially many explanatory variables.

Multivariate data

There are a variety of techniques including clustering, principal components analysis and discriminant analysis, commonly described as multivariate techniques, since they are applied to data sets where there are many artefacts, and on each artefact, there are a number of variables measured.  Each specimen whether it be a pottery shard, glass fragment, or coin will have a number of measurements made on it (eg metallurgical analysis, or shape (length, breadth etc) or decoration).  Since the measurements are made on a single specimen, they are likely to be related, so in this sense, we have an extension from the bivariate (two variable- correlation) case to the multivariate (many variable case).

Section 3 outlines the use of principal components and discriminate analysis which have been applied to elemental composition data of pottery.


See also More on Statistical Modelling Applications - just for the ScARF wiki!

Tags: