How to account for and guarantee the quality of dendrochronological results

by Torbjörn Axelson

"No archeologist would accept sending his potsherds to an expert who returned a statement saying they were from this or that period and then let the potsherds disappear without further documentation and the possibility of a follow-up!"

First published as “Redovisning och kvalitetssäkring av dendrokronologiska resultat”, Fornvännen: Journal of Swedish Antiquarian Research, 102: 187–190 (2007) [in Swedish], original published article as a .pdf file.

Dendrochronology is, rightly, a frequently-used tool in archeological and building antiquarian work. For a number of years I have worked with dendrochronology on a hobby basis, mainly with a direction towards local history. I have no experience of my own from archeological work, though gradually I have come to wonder more and more about how dendrochronological investigations are accounted for in archeological contexts.

Therefore I want to draw attention to what I think is deficient in current practice of how these results are documented in e.g. archeological reports and, also, how dendrochronological data sometimes can be found to contain relevant historical information beyond what was initially requested (usually dating and provenance (i.e. origin)).

Finally I will also discuss possible changes in the terms of dendrochronological charges and something on how samples should be stored and how data should be saved.

Sample treatment, measuring, dating, and provenance

Dendrochronology is—as long as we only deal with dating and provenance determination—a simple technology. The tools and knowledge required are quite limited, when compared to many other things. Precision (carefulness) is the primary virtue, though, as with all workmanship, experience is a great asset.

Sample treatment. When a wood sample has been delivered for analysis, it is first prepared to make an even surface. This can be done with a knife or with an even (fine graded) piece of sandpaper. The surface is often treated with oil and/or a pigment to enhance the contrast of the annual rings.

Measuring. When the surface has been prepared, the width of each annual ring is measured, one by one. This can be done directly on the sample with a special measuring instrument, or with a computer program measuring on a high-resolution scanned image of the sample. This gives a series of numerical values, one for each annual ring.

Dating. The series is compared with a reference curve from the same region and of the same sort of wood. Such a curve is built from a great number of series from previously-measured samples. A computer program then searches for the position (on the curve) giving the best statistical match with the sample—typically using ordinary correlation.

Depending mainly on how high a correlation value we can find and how many annual rings could be measured in the sample, we can decide if we have found a reliable dating or not. Unfortunately it is common that the bark surface is missing for the samples being investigated. This means that we can then only give an interval for when the tree ceased to grow and not the exact year when the tree was cut.

Provenance. If we also want to decide on provenance, the computer is made to compare with many different reference curves, created from trees from different places. Hopefully we then find similarities which points to a certain area.

What more can be told by the sample?

When a piece of wood suitable for dendrochronological analysis is found, dating and perhaps provenance are the primary questions we hope to get answered. We should remember, though, that the sample might give more information than that. This information could be relevant for studies within other disciplines, e.g. forest history and climate change. Sometimes it is of such character that it is of interest for historians and archeologists. Let me give some examples.

What is the organizational background to a building? When a farmer built a house of timber, he usually cut the timber within a limited area of the forest. For the walls, he selected blocks of uniform thickness. Therefore, timbers from such a building usually give a quite uniform impression: blocks of about the same age, often with a high mutual correlation between annual rings. On the other hand if we look at a house that has been collectively built, e.g. a church or a parish hall, it was common (at least in the regions which I know best, Dalarna) that each farm or small village got a piece-work contract to deliver a certain number of blocks to the building site. Therefore, the timber in such buildings is often considerably more different in its character, even when the outer dimensions are the same. This is because the trees have been collected from a larger geographical area. Therefore, if we investigate the timber in, say a medieval tithe barn, we might find out not only the dating; at best, we also get a good hint of how the production of the timber was organized.

When everybody is rebuilding. Another example of "extra information" can be noticed in the measurement series created by another hobby dendrochronologist, Tomas Andreasson, who published series from his croft in the Tyresta area outside Stockholm (International Tree Ring Data Bank: swed304 measurement data; see samples DRS13, DRS16, DRS5). The dating was successful and showed that the timber was cut in 1779–1780. In these measurement series it is clearly visible that at least three out of twelve examined trees had a very strong growth increase during one or more summers of 1720–1722; this indicates that these trees had then been freed from other light-hiding neighboring trees. The time corresponds well to when we could expect a great harvest of wood for rebuilding houses and barns after the Russian ravage of 1719.

Information for other sciences. The physical samples may contain more information than that usually produced for a normal dendrochronological analysis. If the sample is a sawn plate it can sometimes contain information on snow and wind conditions or on earth slip caused by erosion or quakes. To produce that type of data, much more extensive measurements are usually necessary. If the question is related to analysis of climate change, it may be necessary to measure the rings of spring and autumn separately or to measure the density of the rings. If that might be done at a later time, then the sample has to be warehoused.

Accounting for results; saving the sample; a national database

If dendrochronological results of today are not carefully accounted for, we have good reason to fear that future scientists will look at them with skepticism. Faced with the possibility that a dating can be questioned from e.g. historical reasons or reasons from history of art, we have to make sure that it can be checked afterwards. Otherwise the money spent on the analysis can later stand out as wasted.

From the point of preservation, wood samples can be divided into two groups: dry and wet. The wet samples demand some form of conservation to be preserved. They might be in such a good condition that they can be carefully dried and then handled the same way as dry samples. Or they might be so easily damaged that they will soon fall into pieces, which demands that dendrochronological measurements have to be done in wet condition and quickly. How wet samples in various degrees of breaking down should be taken care of, to preserve as much information as possible, should be investigated and discussed together by dendrochronologists and museum curators.

In those cases when samples are in such a bad condition that it is not possible to preserve them in a meaningful way, it may be reasonable to consider saving high resolution digital images as documentation. The basic principle should nevertheless be that samples produced from an archeological investigation should as far as possible be given a findings number and be saved and kept together with the rest of the material from the finding place. Consequently, it is the museums and like institutions that should have the responsibility for these samples. There we find working routines for handling such material. The museum collections can also be expected to be much more steady than collections saved at partly-commercial laboratories.

To make saved samples as useful as possible, it would be desirable that all measurement series, datings, specifications on where the physical samples are stored, and information on who has done the dating be registered in e.g. a national (or international) database. Then later inspection and further studies—perhaps based on new methods—could be done more easily. As measurement series and other data are registered in the database, it is also possible for a researcher to find those samples that could be relevant for a certain new investigation. And then order only these samples. As it is reasonably a general and scientific interest that reference material is saved and published, it is important that such a database is accessible over the internet!

A dendrochronological dating done by a commercial laboratory: usually meager account of data and no chance to evaluate the quality

When a dating has been requested from a commercial laboratory—often from a department at a university—it is common to find that the accounting consists of a year for the outermost ring measured and the number of measured rings, with a statement of likely provenance and a discussion on the interval for the cutting year.

Too often, however, there is no account of correlation values found or reference curves used; neither is the measurement series is included. In that way, the dendrochronological analysis evades a meaningful discussion and an independent evaluation of quality.

From a scientific point of view this is completely anomalous. We have to remember that every dendrochronological dating is actually an interpretation done by a human being. I.e. it should be possible to call that interpretation into question in the same way as for other scientific interpretations and results.

No archeologist would accept sending his potsherds to an expert who returned a statement saying they were from this or that period and then let the potsherds disappear without further documentation and the possibility of a follow-up!

Considering the carefulness and profound discussions that are typical for the treatment of material found by archeologists, this meager handling of dendrochronological material and results is puzzling. Are the laboratories unwilling to hand over data or is this only a result of a missing awareness among archeologists about which data to request from those doing the dating?

Who can check the quality of a dating?

Perhaps there is an out-of-date idea among archeologists that dendrochronology is such an alien thing to archeology that a detailed report would be irrelevant?

That may have been a reasonable attitude when dendrochronological analysis was done by hand, with hand-drawn diagrams compared on a light table and measurements done with a microscope. In that time dendrochronological tabular data was in practice almost useless, because of the immense work required to analyze it. Today the computers do the work. Usually we can check computer-stored tabular data faster than it takes to look up the article in the magazine where the table was found. A standard office computer with an office scanner and inexpensive software is sufficient for a lot of dendrochronological work. The quite sophisticated statistics that are behind the scene is usually nothing to care about in normal work.

As time goes by, more data to compare with is published (e.g. at the International Tree Ring Data Bank, www.ncdc.noaa.gov/paleo/treering.html). Though this is perhaps only a rill of data compared to what university institutions have access to.

All in all, a person who wants to learn the dendrochronological craft today can do so without much difficulty, though carefulness and some time are important requirements. The actual crossdating—the matching—is done by the computer instantly when the right reference data is available. Preparing samples—the "woodwork"—and writing reports are consequently found to be the most time-consuming activities.

Perhaps a five-week course in practical dendrochronology would be an appropriate component in university education of archeologists.

The role of laboratories in the deficient reporting

Dendrochronological laboratories finance a great deal of their activities with dating jobs. As a consequence, these laboratories are quite unwilling to publish reference curves, which could be used for future paid assignments.

The hunting for income and the need to keep prices high, together with the unwillingness to publish data for independent scrutiny, threatens to undermine scientific credibility. At the same time, the customer is tempted to limit the number of samples for analysis because of the expense. This makes a risk that data of great scientific value are wasted.

With the current way of financing our university laboratories we have possibly run into the somewhat absurd position that public and taxpayer-funded universities obstruct distribution of knowledge because of commercial considerations.

What to do about this? The responsibility of the customer

Basic data and reference curves have to be made public. The knowledge built up by universities (and also by freelance entrepreneurs) has to be made available also to archeological and historical research in a more productive way than through just plain dates delivered according to an established price list.

It has to be feasible to discuss and make a quality check of datings and of asserted provenance. This can only be done when measurement data is published with local reference curves available to anybody interested. Because of reasons described above, requests for such a change are hardly to be expected from professional dendrochronologists. Such demands have to originate from the customers who order the dating assignments.

With this in mind, it does not seem reasonable to be satisfied with an analysis result in the form of a year number without any measurement values or even a correlation value (or t-value). Publishing of dendrochronological ring width series does not take much space in a report.

The common formats for series (Tucson and Heidelberg) that are readable by computer programs show ten values per line. This makes a typical sample of some 100 annual rings not require more than 10 lines of text and some extra lines for information about place of origin, dating, and like. With a modest text size, quite many samples can be published on each page.

Even when laboratories are willing to hand over data, copyright reasons may prevent this. Many times the laboratory has borrowed reference curves from other labs and is not allowed to hand them over to others.

Therefore, we have reasons to fear that copyright will prevent publishing of the curves used today. In that case it is even more important that public institutions are careful with saving their samples, to make it possible for others to later measure them once more.

As far as possible, old samples should be sent back to their respective museums. The museums may also introduce a requirement that anybody who borrows such a sample should in return hand over a copy of his/her basic measurements to that museum. In this way the amount of freely available data will increase faster and the lack of reference data will gradually be rightened.

Author's web site:
taxelson.se/dendro  [Swedish only]