If you are going to start with dendrochronology dating you will be greatly helped if you find
already published data covering your area of interest and the time span you are interested in.
Lots of data is published in the International Tree Ring Data Bank
(ITRDB)
in Arizona. The ITRDB
works as an exchange center serving us all. If you develop a reference curve for your area
you should really publish it at the ITRDB. Your data can be of great help to other people
interested in dendrochronology.
"The primary purpose for the ITRDB is to provide a permanent location for the storage of well-dated, high-quality dendrochronological data from around the world. This central repository protects data from loss due to: (1) mishandling of tree-ring data, (2) the relocation or termination of laboratories, (3) scientists who move to other projects or retire, or (4) the death of scientists." (ITRDB)
When I started with this in 1995, there was no data published from the area around Stockholm where I live. So I got in contact with a university in Sweden and asked for help with reference data. I was then told that I could indeed use data from them, but only if I signed an agreement saying that I was never to publish any data originating in any way from their data or to use their data in any way for commercial dating jobs. Because as they financed part of their activities with commercial dating jobs, they did not want competing laboratories to learn about their reference curves. As this had effectively prevented me from ever publishing and sharing dendro data with others, I waived that agreement.
From the point of view of an outsider like me, I find such university laboratories to be like any other commercial laboratory - keeping their secrets for the benefit of the company. And - as I see this - by not publishing measurements they can be sure of not having their datings inspected and questioned by others. In all a comfortable position, but not in accordance with my expectations on a university lab.
So from that university and from the ITRDB, where nothing of interest to me was published, I got no reference curves at all. I had to start developing data from the very beginning. I was lucky and found some very old trees which helped me develop my own reference curve from the island of Nämdö covering the years 1995 - 1582. This curve was published in the ITRDB in May 1997.
Gotland. Though Swedish universities keep their secrets, we have recently got help from Swiss people. In August 2002 professor Fritz Schweingruber of the Swiss Federal Institute for Forest, Snow and Landscape Research published their curve from the island of Gotland. This curve covers the time 1987 - 1127. It can be used along the Swedish coast adjacent to Gotland. This curve matches my curve from Nämdö. A lot of thanks go to prof Schweingruber and the Swiss institute for publishing this data!
Today, there is also a 500 year long curve from Dalarna developed by my friend Torbjörn Axelsson living in Björbo. This data is now available at the ITRDB. Even more data is available at Torbjörn's site, see "Other sites" in the left column!
A short curve from Tyresö (Sisshammar) covering 1779 - 1667 has recently been published in the ITRDB by Tomas Andreason.
If you cannot find a reference curve covering the time of interest for your area, you have to develop a
curve yourself. If you end up with curves from dated tree constructions, see that your data is
published at the ITRDB! Even short curves or even undated floating curves may be of interest to others.
But see that you enclose information about your curve and its background with the actual tree ring data.
The people at the ITRDB will publish not only your measuments but also your letter or announcement of the curve.
You will find such files if you use the ftp: interface when you scan contents of the ITRDB.
How to start up with a reference curve?
Using fresh trees
One way to start up with a reference curve is to get samples from trees which are cut by you or by your neighbours. Always be on the alert and ask for a plate when you see people cutting up big trees. Another way of getting samples is asking those guys who are professionally cutting trees in your area. This way you can easily get a reference curve covering about 180 years (we are talking about Scotch pine now). This is also a good way to exercise sample preparation (grinding) and measuring.
When measuring tree rings, you will now and then be fooled by rings which go into each other and look like one ring, especially if you find trees from poor, dry soil or cliffs. But working with newly cut trees means that you have the answer book! You know that all your samples should match together. That situation is a good teacher! And this way you also get accustomed to deviations between trees.
Another thing to remember when doing the measurements is making mean values. Using several radii from the same log makes your data more representative for the log as a whole. This makes your data easier to match against other ring width data series. For a way to build mean values, see the lesson "Fitting curves together" in Using CDendro.
Extending the reference curve with old living trees
When you have that 130-180 year long reference curve, you need to extend it. In my area I was lucky enough to find very old trees (250-400 years old) left in a forest which was owned by a friend. (I.e. it was not too hard to get the permission from the landowner to take samples.) When taking samples from living trees we use an increment borer to take out cores. Increment borers are used by forestry people to measure the tree growth in a forest. And remember, two samples from a log is better than one. Ring width mean values give quality to your measurements!
Old trees are not growing fast. Therefore it is difficult to measure the very narrow rings from the last 100 years. But you don't have to! You already have a lot of measurements from your neighbours' young trees. So start measuring where the rings get wider. Just count the number of young rings so you know where to add the old ones to your reference curve.
Another source is of course old houses which are being demolished or repaired. Local contacts are essential, otherwise the whole house has been burned up before you get there! And remember: Take samples from many logs (20-30) to really get enough reference data. (When in a hurry, this will make a lot of sometimes unhealthy mildewy firewood to take care of.)
Remember, you will often find reused old logs which have been built into the house. See the text about the house Sandviken at this site. In Sandviken we found four different datings of wood! So take many samples when you have the chance!
How many samples make a good reference curve?
My experience says that you get quite a good reference when any year in your curve is covered by 15-20 samples.
Recently I took samples from a demolished house, Kaptensudden (that's the one on the picture above) where the measurements did not match very well to my reference curve. Parts of the timber had very thin rings so it was difficult to measure with my old 600 dpi scanner. With a new 1600 dpi scanner I could later build a reference curve for the house. That curve could be matched towards my Nämdö curve with a correlation coefficient of 0.60, though individual samples had values in the range 0.26 - 0.49. Obviously the logs have been bought from somewhere else.
The final reference curve from Kaptensudden contains 17 samples. With only 8 samples the correlation towards the Nämdö curve became 0.55. With 6 samples it became 0.51. This is an indication of how many samples you need to build a reference curve.
How to validate your reference curve?
Always be on the alert for errors in your ring measurements. Make sure that all curves within a reference curve really fit properly together. Avoid measurements from the very innermost treerings of a log. Sometimes you have rings missing there or it is not obvious what is a ring and what is not.
A missing or an extra ring at the end of a curve may go through at a correlation analysis. Though a visual inspection of the curve compared to a reference would reveal the error.
In the CDendro program you can divide your sample into blocks with a length of e.g. 50 years and check that each block matches properly at its correct place in the reference curve.
Avoid adding samples of only 60 years to your reference curve. This can only be motivated when correlation is high and you know that a sample belongs to a group of contemporary samples from the same place.
Reference curves for different regions.
- I have collected samples from three regions:
- Nämdö in the Stockholm archipelago
- Saltsjöbaden at the boarder of the archipelago 25 km west of Nämdö
- Bettna 115 km WSW of Nämdö and 35 km NW of Oxelösund in Södermanland.
- In addition I have samples collected by Per Zackrisson at "Svensk Byggtradition", Aneby in Småland. Aneby is some 160 km SW of Bettna.
The collections from Saltsjöbaden and from Bettna are not at all as extensive as that from Nämdö. The curves are indeed different but they are so much alike that the Nämdö collection can be used to date samples from Saltsjöbaden (of course - it is very nearby!) but also from Bettna.
The correlation coefficients when comparing between the Saltsjöbaden collection towards the Nämdö reference are 0.63/0.29 (best/next best).
The Bettna collection matches towards Nämdö with 0.49/0.21 (best/next best) and towards Saltsjöbaden with 0.54/0.32 (best/next best).
The first 110 years (which are most representative = many samples) of the Aneby collection matches towards the Nämdö collection with a correlation coefficient of 0.47/0.34 (best/next best). The Aneby collection also matches towards that of Torbjörn Axelson, Björbo, Dalarna (see the "Other sites" section), with 0.50/0.28.
Probably the Nämdö collection can be used as a basis and reference for building new reference curves in quite a large area around Stockholm!