An Irish tree ring chronology
An interpretation of some raw dendrochronology data published by the Queen's University Belfast

- and of English data from the University of Sheffield and English Heritage -

Note: Correlation values mentioned in the text are based on the P2Yrs method, as long as no other method is specified.

The Irish tree ring data

The Belfast QUB data

In April 2010 the Queen's University Belfast published an extensive collection of raw dendrochronology data collected during some 40 years (ref. 20). The collection consists of more than 9.500 measurement series collected from varying places not only in Ireland.

Because of our interest in validating or falsifying the currently used middle European oak chronology, we have put some work on synchronizing parts of the Belfast data (using our program CDendro) so that the data can be used in our validation work.

Though we have only studied a limited amount of the Belfast data, we are greatly impressed by the overall high quality and by the immense work laid down to collect it!

Our work has been directed at crossdating series belonging to the same site and at crossdating mean value chronologies created from such sites. It should then be noted that we could often create more than one series from the same site. It is not only that two series of the same site may come from different times, but also that they may be parallel in time though not matching each other very well. - We have focused on Irish, English and French data series within the Belfast collection.

When a well known dendrochronologist asked us to send him our mean value curves for his analysis, we decided to openly publish the result of our work so that others might possibly benefit of it for some scientific purpose.

We now have three site-wise (one mean value curve for each site) Irish chronologies which are in such a condition that we can publish preliminary versions of them:

  • BelfastAD.fh, a collection of mean value curves covering the time AD 25 - 2006
  • LateBC.fh, a collection - floating in time - but possibly crossdatable towards the German oak chronology to 1155 - 69 BC. It should then be noted that we consider the Celtic and Roman block of the Hollstein reference as floating (see our section on the Hollstein data), i.e. the current dating of the LateBC collection might be wrong!
  • BelfastLong.fh, a bog-oak collection covering 4615 years. We are not able to crossdate it towards the LateBC collection - so it is floating!

You will find the data listed above in the download section below.

We plan to successively analyze and publish more data and hopefully some analysis of how the data possibly relates to the German and French oak chronologies. Updates will be noted on our dendro site at

With best regards
Petra Ossowski Larsson & Lars-Åke Larsson
August 18, 2010 (updated data files Nov 12 2010)

Some more information related to the QUB data:
QUB at Cybis dendro-wiki
On QUB meta data at Cybis dendro-wiki
(These pages are edited mainly by Torbjörn Axelson.)

How to link BelfastAD and LateBC

Our three chronologies out of the QUB material do not cross-correlate, which means that there are two gaps of undefined length in between.

First we try to close the gap between BelfastAD and LateBC, which is 93 years long with the present dating. Mike Baillie had big problems with this gap as he describes in his book "A Slice through Time" (ref. 21). This graphic from his book illustrates the case:

Baillie's young BC-end consists of Garry Bog up to 220 BC, Dorsey/Navan 580 - 100 BC and Corlea 450 - 150 BC. We have verified the links between Garry Bog, Dorsey/Navan and Corlea according to the diagram above - all of them match very well towards each other. Additionally we have got an extension: Annaholty Bog (277 - 69 BC) from a recently excavated causeway in Tipperary. These collections are all contained in our LateBC collection.

Baillie´s old AD-end consists of Teeshan spanning BC 13 - AD 581. We date our QUB-data based Teeshan collection to AD 82 - 581. We cannot find any data to extend it further towards older times. But we have Ballinderry back to AD 25, a collection from 1999 which Baillie did not have when he wrote his book. These collections are contained in our Belfast AD collection.

To close the gap between LateBC and BelfastAD, Baillie turned to England for additional material and got Carlisle and Southwark.

The QUB Carlisle collection has a weak match towards LateBC. There are 69 stems with more than 100 rings from Carlisle in the QUB-material (from 1980/83). Out of these, three collections can be built. One of them (CarlisleQUB3) matches weakly at the younger end of LateBC spanning 264 BC - AD 103 (corr 0.33, TT 4.9, 195 years overlap, compare with the diagram above). (The other Carlisle collections span AD 467-770 and AD 1062-1442.)

No match between Carlisle and BelfastAD. There should be an overlap of 78 years between CarlisleQUB3 and BelfastAD, but we see no acceptable match at all as shown in this table:

Will Southwark bridge the gap? Southwark looks much more promising according to Baillie´s diagram. A fivehundred years long collection which matches Teeshan with TT 6.5 at a 250 years overlap! This is the link Baillie used to finally close the gap. Unfortunately for us, the QUB-material contains only one single stem marked Southwark, and the dating is not sure. This means that Baillie probably had more Southwark data of non-QUB origin at his disposal.

English data. In this situation we established a contact with Cathy Tyers, Sheffield University/English Heritage, who kindly and generously started to provide unsynchronized raw data from English sites.

The QUB Carlisle collection extended with English data has a good match towards LateBC! First we got a lot of samples from Carlisle and other sites in northern England, all with Roman context. We were able to build four additional site collections, which together with CarlisleQUB3 form our 475 years long NorthEnglishRoman chronology. This chronology matches LateBC with corr 0.39, TT 7.3 and an overlap of 303 years at AD 103, thus confirming Baillie's interpretation, i.e. it spans the period 372 BC - AD 103.

This NorthEnglishRoman chronology also matches weakly "as dated" directly against continental Roman-time data - another indication that the preliminary dating of LateBC is consistent with conventional interpretation.

Southwark has a reasonably good match towards LateBC. Recently, we also got data from southern England including the City/Southwark (London) collections. We were able to build a very strong southern England Roman-time chronology (SouthEnglishRoman) consisting of 13 site collections and spanning over 463 years. The SouthEnglishRoman chronology matches directly against LateBC (at AD 207, so it spans 256 BC - AD 207) with corr 0.34, TT 5.0 and 187 years overlap. It also matches NorthEnglishRoman "as dated", corr 0.29, TT 5.8 and a 358 years overlap, and even HollsteinBestRoman, corr 0.35, TT 7.9 and 462 years overlap! This finally confirms the synchronization against the German Roman-time complex and at the same time proves that Hollstein's Roman chronology is consistent upto at least AD 207.

No convincing match between SouthEnglishRoman and Belfast AD:
Though SouthEnglishRoman (256 BC - AD 207) fits towards LateBC, the fit against BelfastAD (AD 25 - 2006), which would close the gap, is completely disappointing and not convincing:

There is a weak match at AD 207 (corr 0.18, TT 2.4 at a 182 years overlap), but nothing even near to corr 0.38 (which would give TT 6.5 at a 250 years overlap according to Baillie´s diagram) which Baillie claims in his book. There are of course a lot of other similarly weak matches as shown above.

Conclusion: LateBC is still floating and more data is needed to clear out the case.

Mean value collection files to download: NorthEnglishRoman and SouthEnglishRoman. See the download section below. (The CarlisleQUB3 mean value curve is a member of the NorthEnglishRoman collection.)

Updated: August 13 2011.

An alternative tail in the old Teeshan end

Mike Baillie wrote us a letter with a comment regarding the length of the Teeshan collection shown in his diagram describing the AD/BC link. Baillie hints that his Teeshan collection also contained material from Allistragh, Balloo and Mill Lough, thus extending the old end back to BC 13!

Allistragh (AD 337 - 39) is already included in our BelfastAD chronology. There is also a properly matching collection of Mill Lough (AD 611 - 81) already included. There are indeed another collection for Mill Lough (AD 209 - 0) and for Balloo (AD 312 - 17), but these match rather weakly towards the current BelfastAD collection so we excluded them.

To allow you to play with the collections, we built an alternative "tail" for BelfastAD containing Allistragh, Balloo and that other Mill Lough, see the download link below for the BelfastAD337to0 collection.

To play with this:

  1. Delete the "Allistragh" member of our BelfastAD collection as "Allistragh" is made a member of the new "tail collection BelfastAD337to0.
  2. Create a mean value curve out of the three members of "BelfastAD337to0".
  3. Add this mean value curve to our BelfastAD collection at AD 337.

Now you can check or uncheck "Ballinderry" and "BelfastAD337to0" alternatively, or use both collections together to study the match between the "new-tailed" BelfastAD collection towards the SouthEnglishRoman collection.

SouthEnglishRoman (as dated) towards BelfastAD with:

  • Ballinderry in the tail (BelfastAD337to0 unchecked): corr 0.19, TT 2.6, 182 years overlap
  • BelfastAD337to0 in the tail (Ballinderry unchecked): corr 0.26, TT 3.8, 207 years overlap
  • Both Ballinderry and BelfastAD337to0: corr 0.26, TT 3.8, 207 years overlap

If we add mean values of SouthEnglishRoman and of NorthEnglishRoman to the LateBC collection, the correlation values between that collection and the BelfastAD collection with alternative tails decrease a little compared to the values above. Still we consider these values unacceptable for chronology building.

Petra Ossowski Larsson, Lars-Åke Larsson. Nov 28 2010.

Mean value files to download: The BelfastAD337to0 collection in the download section below.

Closing the gaps

Status summary

Let us summarize at this point the status of our project on how to build a credible continuous European oak chronology from present time back to Roman time and late BC:
  • Continental western Europe: We have continuous data from present back to AD 814, possibly back to AD 400. The "Carolingian gap", AD 700-800, is still not safely bridged. We have old time data (including Roman time) from AD 336 back to 340 BC (Hollstein Best Roman).
  • Ireland: We have continuous data from present back to AD 25 (Belfast AD). We have old time data from 69 BC back to 1155 BC (Late BC). It is not possible to locate Roman time within the Irish data as the Romans never had activities in Ireland.
  • England: We have a lot of English site collections among the QUB material. These site collections, most from northern England/Scotland, usually crossdate well against the Irish chronologies. The Roman time collections for northern and southern England (North/South English Roman) crossdate well with Irish old time data (Late BC), but rather badly (this is unusual and not expected!) with the continuous Irish AD-curve (Belfast AD).

This is almost exactly what is outlined in a CBA research report from 1994 (ref. 1), with one exception though: the authors claim an extension of the English Roman time data to about AD 300 which would decrease the 200-years gap by 100 years. 2004 this gap has decreased even more: "For the historic period there is now a continuous sequence that runs from the present back to AD 404, and another from Roman contexts covering 434 BC – AD 315. The latter is still reliant on cross-links with Ireland and Germany for its dating since no English tree-ring sequence has been found that spans the fourth century AD." (ref. 2).

New English data helps to finally close the Carolingian gap

Early in 2011, Cathy Tyers sent us a lot of unsynchronized Saxon and Medieval time measurement series from various sites in England. We were now able to build a very strong English Master (see the download section below) spanning the time AD 404 - 1216. This chronology also includes a collection of a number of sites in London (London1) which spans continuously from AD 413 - 1215. As already demonstrated for the Roman time material from London, the correlation to data from the Rhine valley (Hollstein) turned out to be extremely good. This allowed us to validate Hollstein's single site curves for this time (AD 410 - 1242) thus verifying that Hollstein bridged the "Carolingian gap" correctly! Read here!

We were also able to assemble a master for northern England and Scotland (see the download section below), AD 278 - 1588, which crosscorrelates well towards the Irish AD chronology but not so good towards southern England and even less towards the Hollstein master.

New project status

Once the "Carolingian gap" was shown to be properly bridged, we could now concentrate solely on the "Migration gap". In the next section we will describe our efforts to find material from southern England which could extend our Roman time collection (SouthEnglishRoman) through the third century AD.

It should be noted that the closing of the Carolingian gap does not imply a proof against the theory of invented years during Carolingian time! To at all create such a proof we have to be able to present a continous tree ring curve that connects current time tree ring data to tree ring data from wood that is archaeologically anchored in Roman time. With the "Migration gap" with its lack of wood present in England, France, Italy and Germany we do have a problem!

Petra Ossowski Larsson, Lars-Åke Larsson. August 13 2011.

Bits and Pieces - an approach to closing the Migration gap

At present, the situation regarding "the gap" in the English (dendro)chronology is this: we have a reliable Roman chronology up to AD 207, and we have a strong English master down to AD 404. This means a gap of almost 200 years between AD 207 and AD 404.

The Irish data consists of the LateBC curve going from old times up to 69 BC, then the Irish gap of 94 years and then the Belfast AD curve starting at AD 25.
Baillie´s solution was to bridge the Irish gap with English data. As shown in the diagram above, it was easy to crossdate Late BC towards the North and South English Roman curves. Though making these English curves crossdate back towards the Belfast AD curve was much more tricky, they show correlation values of only 0.02 and 0.18 respectively (see the values marked red in the diagram). So we have to use more data to build some kind of "dendrochronological featherbed" that makes the transition from the English curves over to the Belfast AD curve smoother.

There are of course a lot of "leftovers" in the material we got from Sheffield University/English Heritage. The Roman time leftovers from London are especially interesting as English dendrochronologists claim that they reach up into the 4th century AD. Which type of material is available?

First there is an additional collection from Keays Lane in Carlisle (Keays, 319 years) which extends the NorthEnglishRoman master to AD 118 and actually improves its conventional fit against BelfastAD (P2Yrs: corr. 0.18, TT1.7, though Baillie/Pilcher only corr. 0.00, TT 0.0), but on the other hand slightly lowers its fit against LateBC (from corr. 0.39, TT 7.3 to corr. 0.38, TT 7.1).

There are also small collections and single samples with a distinct, good match within the current range of the Roman chronology, but which are not needed to build this master. E.g. Milk Street consisting of one 105 years long sample (BC 90 - 15 AD), or Triangle consisting of two samples which give a 121 years long collection matching with corr. 0.74 at AD 14. Peter's Hill (PeterHill2), two samples giving a 144 years long collection which matches SouthEnglishRoman at AD 25 with corr. 0.46, TT 6.1.

Then there are collections, with good internal correlations, which give a reasonable but not so good match towards the master chronology. These collections might consist of wood which has grown in distant locations or in a different biotope and has been moved to London. For example wood from Yorkshire seems to require an own separate reference chronology as it matches rather weakly against the "mainstream" masters from northern and southern England. For demonstration purposes we include a Yorkshire master (York2, AD 614 - 1183) and two collections: York1 (AD 455 -707) and Castleford (BC 161 - 56 AD).

Adding more London data to make South English Roman match Belfast AD

As indicated in the diagram above, there are a few collections which can be added to the SouthEnglishRoman collection to make it better match the BelfastAD collection. Adding some of these collections are a bit problematic as they somewhat exclude each other as described below.

There is a second collection from New Fresh Wharf (NewFreshWharf2) in the London material. This collection has a weak match against the NewFreshWharf collection already included in SouthEnglishRoman. Our dating suggestion is AD 209.

A collection from Billingsgate (Billingsgate2, AD 178) matches NewFreshWharf2 with really high correlation (corr. 0.63, TT 9.2 at 132 years overlap), but the match towards SouthEnglishRoman is only corr. 0.37, TT 4.9 at 154 years overlap.

By now we have used up almost all samples longer than 100 years and the leftovers are mainly short samples. Crossmatching becomes difficult and ambiguous.

A 129 years long collection from Billingsgate (Billingsgate1) consisting of six short samples with a length between 82 and 54 years matches SouthEnglishRoman with corr. 0.58, TT 7.3 and 107 years overlap at AD 239. This means 30 years extension of the master into the gap, but is it right? Probably yes, because Billingsgate1 matches HollsteinValidatedRomanExtended convincingly "as dated".

Another, 110 years long, collection consisting of eight samples with a length between 100 and 68 years from Baynard's Castle and Chamberlain's Wharf (BaynardChamb) shrinks the gap with additional 10 years. Corr. 0.70, TT 9.6 with 99 years overlap, dated AD 249 towards SouthEnglishRoman extended with Billingsgate1. The match against HollsteinValidatedRomanExtended is still good.

And yes, there is now a better match "as dated" against BelfastAD: corr. 0.25, TT 3.9 with Ballinderry in the tail. We get a somewhat higher correlation, 0.32, TT 5.2, if we replace Ballinderry with BelfastAD337to0, which is Baillie´s tail described above. Though we considered this doubtful - its anchoring in BelfastAD is not good enough!

There is a collection named Tower, consisting of two samples, one 100 years and one 160 years, with high internal crosscorrelation at 84 years overlap. This collection gives a fair match against BelfastAD at AD 328 (corr. 0.35, TT 4.9 at 174 years overlap), but it does not match HollsteinValidatedRomanExtended. This Tower collection matches SouthEnglishRoman extended with Billingsgate1 and BaynardChamb with corr. 0.56, TT 6.5 and 95 years overlap at AD 328! This might be the final proof that the dating of English Roman time wood and Mike Baillie's bridge is correct.
Note: The results are very sensitive to how the BaynardChamb collection is built - removing one out of eight members makes the correlation fall to 0.50/5.4 so we cannot exclude that the high correlation is by mere chance.

A break in the match towards the German masters! Now, a look at the German masters reminds us to be careful! The new extended SouthEnglishRoman (with Billingsgate1, BaynardChamb and Tower), fits nicely towards HollsteinValidatedRomanExtended and also towards A100 up to AD 250. Then something happens!

SouthEnglishRoman with Billingsgate1, BaynardChamb and Tower fits nicely to A100 up to AD 250 where there is a severe break in the match. See the block correlations after and before AD 250 in the yellow boxes. The blockwise correlation is also shown as red figures on the bottom row of the diagram. If the break was caused by a missing or surplus year the correlation would most likely turn negative, but this is not the case here.

The same can be demonstrated with a second collection from Peter's Hill (PeterHill) which matches the new extended SouthEnglishRoman without Tower with corr. 0.55, TT 4.9 and 58 years overlap at AD 294. If the Tower collection is included, the correlation decreases to 0.36, TT 3.8 and the problem starts at AD 245.
This indicates that somewhere around AD 250 there is an error in the European dendro curves.

We are not able to make more progress with the material we have got so far, and we definitely do not think that our results point towards a convincing bridge of the Migration gap. We therefore suggest further investigation with new (more) data which hopefully will lead to an unambiguous solution!

Data to download: In the download section below you will find the collection BitsPieces.fh containing mean values of all small collections mentioned above.

Petra Ossowski Larsson, Lars-Åke Larsson. October 1 2011.

Conclusions so far 2011-12-10

Our project is to verify or falsify the current link of the Christian time era (in common use for about 11 centuries until now) to the Roman time era with dendrochronology. We have reworked great amounts of European dendrochronology to see whether there is any "play" in the time line. This question seems to be regarded as pure nonsense, as the answer already has been given and is commonly accepted. This is an ice cold case! But after having gone more closely into the matter, a number of severe irregularities have emerged, please read the following summary. Therefore we regard it as almost dishonest not to continue this investigation and get to the root of the problem. And we need all help we can get!

Which scientific tools do we have at our disposal?

Dendrochronology of course, as this method usually gives reliable results with one year resolution. Dating with dendrochronology can be great if one has a long local reference curve anchored in recent times and with high internal correlation. However, with increasing geographic distance between the trees of the reference curve and the wood to be dated, the method becomes more ambiguous.

Astronomy usually has an even higher resolution than dendrochronology, but it requires a number of prerequisites to be reliable. First, there have to be observations which can be analysed; as ancient observations are handed down through time, their reliability has to be scrutinised. Second, the models used for recalculation of eclipses and other astronomical phenomena have to be sufficiently precise; the older an observation is, the greater the uncertainty.

Radiocarbon dating, but this method will almost certainly not give the high resolution that is required in our project. Moreover, we have to be very careful with calibrated 14C, as the calibration curves have been calculated using dendrochronology (warning for circular arguments!). However, predating using 14C raw data is an excellent method to sort material for dendrochronological synchronisation.

Archaeology, as remnants like coins, pottery, and wood deposited over time are often recovered in different strata, the oldest at the bottom and the youngest at the top. However, if the strata are disturbed e.g. by digging, little but guessing remains for the archaeologist. Sometimes, a loose find can be related to a certain time or event because of characteristic features.

What are the possible scenarios?

Zero solution, means nothing is wrong with our chronology, and equally with the current dendro-curves. Dionysius Exiguus (ref. 3) got the Roman year count, the birth year of Jesus Christ, and the Coptic year count (ref. 4) (at that time the only Christian year count, starting in AD 284 with the ascension of Emperor Diocletian) exactly right when he wrote his Easter tables AD 525. The new "Christian era" invented by him came into common use about 400 years later.

Chronological error, means that the "historical year 1" related to the time of Emperor Augustus is not the same as year 1 on our current time scale. A mistake made by Dionysius Exiguus could be the cause of a small chronological error. A big chronological error means probably that Dionysius Exiguus' Easter tables have been exploited for some forgery. This could mean more precisely that the Coptic year count (Anno Martyrum, AM) did not start with Diocletian as pretended, but instead with an earlier event important to the early Christians, e.g. Mary's death (Assumptio Mariae).

Dendrochronological error, means that the synchronisation of two dendro-curves is wrong, in this case the curve covering Christian (recent) time and the curve anchored in Roman time. A big dendrochronological error certainly also implies a big chronological error, and these errors do not have to be of exactly the same size. The difference between the two errors will be some sort of offset failure, which could mean that we have to pinpoint the tree ring of "historical year 1" in a different position than defined in the current zero solution. Data from the Roman camp at Oberaden (ref. 5) has been used to synchronise the current European dendro curves against historical Roman time. This data is unpublished and unavailable.

How would our tools indicate different scenarios? And what do we see?

Astronomy. If the time line is correct, an ancient culture's handed-down astronomical observations should fit modern recalculations. This seems to be the case e.g. for the Babylonians, where a large amount of original observations written on clay tablets exist. The so far youngest datable cuneiform clay tablet found in Babylon contains observations from AD 74/75. This means that the Babylonian culture probably is "absolutely dated" on a continuous time line linked to our time.
The Greco-Roman observations however are few and no originals remain. And what about their match towards modern recalculations? As a historian we talked with put it: "You can't trust Roman observations, they are so inaccurate" ...
Now, are there links between the absolutely dated Babylonians and the Romans? Around AD 75, Pliny noted in his "Natural History" (ref. 6), coherent with other historians, that only the temple of Bel is still standing in Babylon, in all other respects the place has gone back to a desert. So, who made astronomical observations and wrote them down in cuneiform among Babylon's ruins at Pliny's time? (ref. 7)
The feeling that something is wrong with the synchronisation of Roman times towards an absolute time line is strengthened when we analyse the few existing Roman observations, among others by Pliny. Read more about this here (ref. 8). If we trust Pliny's observations at all, there is support for two possible scenarios: the zero solution, confirming that Roman observations are quite inaccurate (they do not fit the recalculations too good), or a chronological error of 232 invented years, showing that also Roman observations can sometimes be very precise (they fit the recalculations exactly).

Radiocarbon dating. This dating method assumes that atmospheric carbon dioxide contains over time an almost constant percentage of the radioisotope 14C.
When the carbon dioxide is incorporated in e.g. the wood of a tree, the amount of radioisotope will decrease due to radioactive decay and will be halved over the course of 5730 years. By measuring the remaining radiocarbon activity, the age of the wood can be estimated.
The primary natural source of 14C on Earth is cosmic ray action upon nitrogen in the atmosphere. Due to variation in radiation and to climatic factors the amount of 14C in the atmosphere varies over time. To achieve a better dating precision, calibration curves have been developed based on known age of wood samples which have been absolutely dated against presumedly continuous dendrochronological curves. A minor error in a dendrochronological curve used for calibration would hardly be noticeable, while a large chronological error would show up as some kind of plateau or other anomaly. Though the origin of such an anomaly may as well be e.g. a variation in cosmic ray activity.
As the calibration curves covering the last 3000 years show some prominent anomalies in the BC part (see for example (ref. 9 and 10), we do not consider calibrated radiocarbon dating as a method able to further our investigations.

Archaeology. A large chronological error between two well-developed eras would hardly be conspicuous to the archaeologist. There would be diffuse signs of discontinuity and abandonment which can be explained with war, calamity and people moving somewhere else. There are such "dark" periods during the transition from Roman time to Medieval time all over Europe. However, if there are lots of finds within a well-defined era, a few years discrepancy between different dating methods would rise red alert.
And indeed, there have been objections against dates obtained with dendrochronology since the method was tried first. And the complaints seem to be united: Roman wooden remains are generally dated 30 to 50 years too old by dendrochronology (ref. 11 and 12).
We can see this tendency also in the English Roman dendro data. The results from the excavations at Copthall Avenue in London (ref. 13) are a good example.
Pottery found in different phases at this site generally gives a younger date compared to the dendro-datable timbers found in the same context. In the discussion of the results, this is compensated by a very wide sapwood estimate of 10 to 55 (!) years. The only two timbers showing sapwood are classified as "reused".
Sample OPT679 is of special interest as it most likely is complete (with cambium but without bark). It has been classified as "clearly reused", as it is dated to AD 86 (we can confirm that), but found together with pottery from the early 2nd century. Interestingly, the sapwood range appears to be less than 30 years.

Dendrochronology. This method is self-correcting, because long data series from quite different places over whole Europe replicate (can be synchronised towards each other). Sooner or later a zero-solution would be confirmed or a smaller error would be corrected.
Therefore the fact that an unbridgeable gap in virtually all regional oak chronologies still exists, is a strong warning sign! And the fact that the gap also exists in Ireland, but far earlier than in continental Europe, rules out both a timber shortage because of Roman building activities and a global disaster as the only cause.

Most dendrochronologists accept the Migration gap in their region. Only a few claim that they have bridged it with help of remote material.
We have examined Ernst Hollstein's German, and QUB's Irish oak chronologies. They are said to replicate, and parts of them definitely do so via the English chronologies. The problem is that in Hollstein's material we still find a gap where the bridge is supposed to be, and the Roman time curve can be linked to recent time 207 years later than expected. This "creative" link is not replicated in the Irish/English material.
The Irish oak chronology shows indeed a possible bridge corresponding to the generally accepted "zero solution" using material from southern England. So this could be right, but maybe it is not as signatures are weak and volatile, correlations low, collections thin and overlaps short. At least, this is not a simple dating of a sample, but the synchronisation of two masters in order to show that our chronology is correct! Therefore we request a stronger evidence.
So far we have not seen any replication of Mike Baillie's zero solution. And why is it necessary to bridge the Irish chronologies with so remote material? It is remarkable that the Irish LateBC fits northern England quite well, while the correlation of northern England towards BelfastAD is very low (P2Yrs: corr. 0.18, Baillie/Pilcher: corr. 0.00 !) though the theoretical overlap is 93 years (conventional dating AD 118). Is there a better match somewhere else? Trying LateBC reinforced with NorthEnglishRoman against BelfastAD gives a match at AD 388 (corr. 0.18, TT 3.4 at 363 years overlap) which means a collapse of the gap by 270 years. No extraordinary match at all, but when NorthEnglishRoman is omitted, the correlation increases to 0.22. It turns out that all theoretically possible Irish BC collections show this match towards the Irish AD chronology, with Navan giving corr. 0.28, TT 3.5 at 151 years overlap. This match is weakly also present in southern England. Maybe this is the truth? Nothing really speaks against it, except Dionysius Exiguus.

270 years "play" would certainly mean a large chronology error (why not 232 years as indicated astronomically? (ref. 8). The remaining 38 years up to 270 would please archaeologists who complain about too old dendro-datings of Roman wooden material. And Bede's story about St. Cuthbert visiting Carlisle AD 685 and finding a fortified urban community in existence, with functioning aqueduct and praepositus civitatis (a Roman official title meaning the community's authority), would now become compatible with the archaeological evidence: the Roman city of Carlisle most likely vanished already during the fifth century AD (ref. 14)

What comes next?
There are several German groups claiming to bridge the Migration gap. All are said to replicate Becker's signature more or less. A100 is probably "state of the art". But raw data is unpublished and definitely unavailable (we have asked, our investigation is considered "not meaningful"). A look at this raw data is placed high on our list of desiderata.

In January 2011, Science published an article (ref. 15) where a 2500 years long continuous oak chronology from Germany and France is the basis for climatological studies, proposing reasons for the fall of Rome among others. The fact that the dendro raw data is unpublished does not bother the Science editors (we have asked, we did not even get an answer). A schematic overview of the depth of the data in the Migation gap is shown in the following diagram derived from the article.

The lack of data in the Migration gap is apparent, and we don't know anything about the quality of its synchronisation. Note: with this data climate models are calculated which can have direct influence on our lives via political decisions!

We therefore urge the dendro community to investigate the case and/or to make more raw data available for replication. Further we recommend, for the time being, to avoid using European dendro data of the first millennium AD and the first millennium BC for climatological studies.

Petra Ossowski Larsson, Lars-Åke Larsson. December 12 2011.

Dating and linkage of the long Belfast BC chronology -
an error in one of the key links

A severe error found in the Belfast curve used for 14C calibration!

In August 2010, we published BelfastLong, a preliminary bog-oak chronology covering 4615 years and supposed to (hopefully) be identical with the Belfast Long chronology described in Mike Baillie's "A slice though time". We were not able to crossdate it towards our LateBC collection, which corresponds to Baillie's Garry Bog 2 plus an English collection from Swan Carr, used as the link towards the Long chronology. So our long chronology is still undated and floating!

The diagram is from Baillie's book, and shows how the English Swan Carr collection (S.CARR) connects Baillie's Irish Long Chronology to his Garry Bog 2 (GB2) chronology. Our own corresponding BelfastLong and LateBC collections created from the Irish measurement data are shown in red color in the diagram.

A major portion of the Belfast material can be crossdated to form a nearly 5000 years long chronology covering times before the first millennium BC. It must have been a tremendous job to acquire all this well-matching data! We have now checked and improved our version of this chronology, and there is actually a 75-years gap around BC 2450. However, this gap is securely bridged by English data from Croston moss (1442 years with almost equal overlap on both sides of the gap, corr. 0.33, TT 12.9). The Croston collection has been moved to an English bog oak chronology which is described in (ref. 17).

Our improved Irish long chronology (BelfastLongVersion2 see the download section below) is supposed to overlap LateBC at about BC 900 as described in Mike Baillie's "A slice through time", but we can not find any substantial match at all. Something is wrong.

So let's figure out how our BelfastLong would be dated conventionally. For that we have to do some "reverse engineering" as the published QUB samples are undated. A 761 years long collection from Thorne Moors in Yorkshire, crossmatched by Gretel Boswijk and put on the ITRDB as brit036, gives the answer. It is dated -3016 to -3776 and matches BelfastLong with corr. 0.27, TT 7.7, resulting in a date of -836 for the youngest ring of BelfastLong.

With this dating, BelfastLong overlaps LateBC by 316 years but with a very unsatisfactory correlation (corr. 0.17, TT 3.0). This means that some deeper analysis is required.

The oldest part of LateBC consists of an English collection from Swan Carr, which shows a good match towards the rest of the collection at -311 (corr. 0.30, TT 8.0). Lifting off this collection reveals that the all-Irish parts of BelfastLong and LateBC overlap by 108 years but do not match! An even closer look shows that the overlapping collections are from the same site, Ballymacombs More (county Antrim).

This was not expected! Mike Baillie reported that the Long chronology and LateBC only overlap via SwanCarr and that there is a one-year gap between the two meeting Garry Bog/Ballymacombs More collections, the collection in the Long chronology ending at -948 and the collection in LateBC starting at -946. We find -946 as the oldest ring in our Ballymacombs4 (of LateBC), but -836 instead of -948 as the youngest ring of our Ballymacombs3 (of BelfastLong).

A look into Ballymacombs3 makes clear what has happened. There is one huge, 380 years long oak curve (Q10705) extending the collection by 112 years. And while the rest of the collection was measured early (Q2203 to Q2272, youngest ring at -948 which is the same as Baillie's dating), the huge oak was measured in August 2009 and therefore was unknown when the long Belfast chronology was linked together in 1983.

Q10705 must have drawn some attention as David Brown measured it several times with very similar results. It fits the rest of Ballymacombs3 with a convincing corr. 0.53, TT 10.0, but there is no match towards Ballymacombs4. Even if we suspect a badly matching tail and truncate the sample by 40 years at its younger end, there is no match. (There is no reason to truncate Ballymacombs4 as it matches in its full length towards Swancarr.)

We believe that, if Mike Baillie had known about this stem, he never would have linked the parts of the Belfast BC chronology the way he actually did.

If, with the new data, this link would be considered to be right anyway, it would not be because of dendrochronology. It would be because of other methods' prerogative to decide what is right in chronology. From a dendrochronological point of view this link is most probably wrong. (See also ref. 18)

How wrong? Well, there is no match within sight. And if we add SwanCarr to LateBC again and look for matches which are better than "as dated", the first alternative would mean inserting 179 years into the gap. Better alternatives emerge more than 200 years off. But there is no chance to definitely point out a matchingpoint with the material we have at our disposal, the overlaps become too short and the collections become too thin.

Inserting a great number of years would either push BelfastLong back in time, or LateBC further towards recent time, or maybe both. And as LateBC has a fixed position in relation to BelfastAD, pushing LateBC towards recent time could consequently mean a chronology error if the number of years is sufficiently high.

An error of this kind necessarily also means that there is an error in Pearson's 14C calibration curve (ref. 9)(ref. 10) as it is derived from exactly the collections in which we have found the error. A recent paper by R E Taylor in Radiocarbon (ref. 16) might give a hint in which direction we have to look for the solution of this problem.

The ancient city of Nineveh (near Mosul in Iraq) was sacked by Babylonians and Medes in BC 612, an event which is absolutely dated on an astronomical timeline by Babylonian clay tablets. Some of the city's last defenders died in the gateway they tried to hold and were instantly buried by the collapsing roof of the gateway. Their remains were found in situ in 1990.

14C analysis of a couple of bone collagen samples from these remains gave an average calibrated date which was almost 200 years too old (though the raw values happened to appear in the right range!). The paper includes a thorough discussion where a number of factors which could contribute to this large offset are ruled out because the sample preparation was carefully conducted. However, a heavy fish diet remains as a possible explanation for the anomaly.

As we see it, an alternative to extensive feasting on imported boned, dried and salted codfish (Bacalhau), which could explain the lack of fish bones in the kitchen middens, would be an error in the 14C calibration curve. And this means an error of the latter kind, the one that pushes LateBC towards recent times, probably implicating that we have to face a chronology error. Pushing BelfastLong towards older times would only spoil 14C dating and does not solve the Nineveh-case.

We think that it is high time that the dendro community starts investigating these problems. We volunteer, if nobody has the time or the energy! Just make the raw data available.

Petra Ossowski Larsson, Lars-Åke Larsson. February 5 2012.

A comment about the confidence in the current 14C calibration curve.

We have got a few comments on this investigation. One opinion rests on a solid confidence in the current 14C calibration curve as it is said to be based on American dendro data of the long lived bristlecone pine, and that dendrochronology is known to be correct.

As "calibration curve" we normally refer to the IntCal09 calibration curve. The tree-ring section of it uses in principle the data sets also present in IntCal04. These are derived from Pacific Northwest Douglas fir, Californian sequoia, Alaskan sitka spruce, German oak and pine (Hohenheim) and Irish oak (Belfast).

The marine extension towards older times is based on corals and foraminifera varved sediments.

A list of all 14C wood samples used is available here:

There is no Bristlecone pine within this list!
American data comes from recent trees and older back to BP 2089 with California Sequoia in the oldest portion.

The older section of Intcal09 (before BP 2089) contains ONLY European tree ring data - back to BP 12549! From BP 10059 to BP 2089 there is a mix of German and Irish oaks.

The Hohenheim data behind this is unpublished, and so was the Belfast data until april 2010.

Dendrochronologists who consider the calibration curve as based on American data, might be thinking of the Suess curve which was based on bristlecone pine from Ferguson's investigation (ref. 19) The Suess calibration curve is regarded as historical (see e.g. ) and not contained in the IntCal09 data sets. However, RE Taylor (ref. 16) wants to change that ...

But the bristlecone pine chronology may anyway play an indirect role.

Mike Baillie claims that his long BC chronology was coarsly dated against the Suess curve, and then fine adjusted by dendrochronology. He further claims that the position of certain growth dips in his BC chronology are conform with LaMarche's frost rings in the bristlecone pine chronology (ref. 22).

So, if Baillie managed to synchronise correctly against the American bristlecone pine chronology, and if the bristlecone pine chronology is absolute (note: two big "ifs"), then we have that painful scenario we drew up in our paper above.

In that case IntCal09 would be correct except for the part corresponding to the first millennium BC (a "minor" problem, but see RE Taylor), and the error we found would almost be a proof that we have a larger chronology error between our time and Roman time. The number of years we insert in the gap at BC 950 would relentlessly push LateBC and with it the Roman time curves towards our time with exactly the same number of years!

Of course, nobody will believe this! This is very near to Heribert Illig's phantom time theory which is regarded as ridiculous among dendrochronologists. And this is our dilemma.

Petra Ossowski Larsson, Lars-Åke Larsson. March 10 2012.

More comments about the confidence in the current 14C calibration curve.

We have now digitalized the Suess calibration curve published in 1978 (ref. 23). We also have checked the consistency of the dendro material forming the Irish oak part of IntCal09.

We find it most probable that:

  • the bodies of all three Irish chronology parts are sound and that the 4100 years long sequence of BelfastLong (5150 to 1050 BC) used for radiocarbon measurement is continuous.
  • the 4100 years long sequence of BelfastLong is correctly synchronized against the Suess calibration curve.
  • the Suess calibration curve is absolute.
For details please see our new Excel-workbook, see ref. 24

This means that the conventional dating of BelfastAD is absolutely correct, while the conventional dating of BelfastLong is nearly correct (± 20 years). LateBC, covering the first millennium BC and linking to the European Roman complex, acts as a "slider" between these blocks. Though its wiggle-match against the Suess calibration curve supports its conventional (historical) dating, this is ruled out by the dendro error we have described above. Further dendrochronological investigation is needed to determine the link between LateBC and BelfastLong, and finally the link between LateBC and BelfastAD which defines our chronology.

Petra Ossowski Larsson, Lars-Åke Larsson. September 9 2012.

How to check our work. Access to original data.

The Belfast series are available from QUB's home page (ref. 20), a total of some 9.500 unsynchronized and undated series, all available at separate links. The Belfast data available for download from is only mean value curves created from the Belfast measurement series. We have been somewhat uncertain whether we may republish the original QUB data in synchronized form, therefore we now have generated "Dating reports" for our published mean value curves (see below). Of course we can make the complete data available to researchers who want to analyze what we have done.

Petra Ossowski Larsson, Lars-Åke Larsson. Updated August 10 2012.


Ref. 1: Tyers, Hillam and Groves, "Trees and woodland in the Saxon period: the dendrochronological evidence", CBA research report 89 (1994).

Ref. 2: English Heritage, "Dendrochronology - Guidelines on producing and interpreting dendrochronological dates" (2004).

Ref. 3: (opens in new window)

Ref. 4: (opens in new window)

Ref. 5: Wooden remains from the Roman camp at Oberaden have been measured and dated with dendrochronology by Ernst Hollstein. He listed the results in his green book of 1980 and promised to publish the raw data later, but this never happened. Oberaden could be connected to the campaigns of Drusus in the Lippe valley BC 12 to 9 by coins among other things. The camp was in use only a few years.

Initially, Hollstein synchronised his Roman time curve using the Kölner Rheinbrücke (KORB) which he dated to AD 310 in accordance with historical documents. It turned out that this dating would place Oberaden at BC 38, which was unacceptable for the archaeologists. Therefore Oberaden was used for synchronisation instead and the camp's building year was set to BC 11, which meant that the whole Roman time curve had to be moved 27 years towards recent time. That this was possible dendrochronologically gives a hint about the quality of the link between the two curves. Hollstein was quite unhappy about the new dating of KORB.

Ref. 6: Pliny, Natural History, Book 6, chapter 30. (opens in new window)

Ref. 7: Stephenson F.R.: Historical Eclipses and Earth's Rotation, Cambridge 1997, section 4.4.

Ref. 8: Sections on astronomy under

Ref. 9: Pearson et. al., 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.

Ref. 10: Pearson et. al., 1993. High-Precision 14-C Measurement of Irish Oaks to Show the Natural 14-C Variations from AD 1840 to 5000 BC: A Correction, Radiocarbon, 35, 105-123.

Ref. 11: Baatz, D. (1977): "Bemerkungen zur Jahrringchronologie der römischen Zeit"; in: GERMANIA 55 173-179

Ref. 12: J.Seigne - Dendrochronologie et datations archéologiques pour la période antique. Compte-rendu de la table-rode du 23/01/06 - Les petits cahiers d'Anatole, no.20, 2007. (opens in new window)

Ref. 13: Catharine Maloney with Dominique de Moulins, "The archaeology of Roman London Volume 1: The upper Walbrook in the Roman period", CBA Research Report No 69 (1990) (opens in new window)

Ref. 14: Burnham and Wacher, The small Towns of Roman Britain, University of California Press 1990. See Google books (opens in new window)

Ref. 15: Büntgen et al., 2500 Years of European Climate Variability and Human Susceptibility, Science Express, 13 January 2011. doi:10.1126/science.1197175

Ref. 16: R E Taylor, Radiocarbon, Vol 52, Nr 2-3, 2010, p 372-382.

Ref. 17: English bog oak chronology in the QUB material We removed a 1442 years long collection from Croston moss from our BelfastLong chronology in order to get a more Irish signature. Croston now is the backbone in the 2105 years long EnglishBog chronology which is conventionally dated -1672 to -3776 and matches BelfastLongVersion2 with corr. 0.28, TT 12.9. The oldest part of the chronology is made up by a collection from Thorne Moors (ITRDB brit036). See the download section below

Ref. 18: It might be argued that our findings are based solely on two stems from the same bog which do not match as they should and that this single mismatch can not be a PROOF that the match is wrong. When comparing 110 years long blocks from Ballymacombs3 towards Q10705, the standard deviation interval for corr coeff is 0.33-0.43 with a mean value of 0.38. That should be compared to the current correlation coefficient of only 0.13 between Ballymacombs4 and Q10705.

Ref. 19: Ferguson, C.W. 1969, A 7104-year annual tree-ring chronology for bristlecone pine, Pinus aristata, from the White Mountains, California: Tree-Ring Bulletin, v 29, no. 3-4, p 3-29.

Ref. 20: QUB Dendrochronology raw data

Ref. 21: Baillie, M.G.L. 1995, A slice through time - dendrochronology and precision dating. ISBN 0713476540

Ref. 22: LaMarche, V.C. Jr and Hirschboeck, K.K. 1984, Frost rings in trees as records of major volcanic eruptions, Nature 307, 126-6.

Ref. 23: Suess, H.E., La Jolla measurements of radiocarbon in tree-ring dated wood, Radiocarbon, Vol. 20, No. 1, 1978, P. 1-18.

Ref. 24: Our new Excel-workbook, BelfastRadiocarbon.xlsx. (direct link to .xlsx file)
If that .xlsx file does not open as an Excel .xlsx file but saves as a .zip file, try to download and unpack this ( real zip file containing only the Excel document. The direct link to the .xlsx file works for us in Firefox but not in Internet Explorer.

Files to download:

These collections contain mean value files created out of site collections which we have synchronized out of raw measurement data.
  1. Synchronized Irish QUB data: BelfastAD, LateBC and BelfastLong as described in the section The Belfast QUB data. Available in Heidelberg format (, Tucson .rwl format (, and Besancon .txt format (
  2. The NorthEnglishRoman and SouthEnglishRoman collections as described in the section How to link BelfastAD and LateBC. Available in Tucson .rwl, Heidelberg and Besancon .txt formats.
  3. The BelfastAD337to0 collection as described in the section An alternative tail in the old Teeshan end. Available in Heidelberg and Tucson .rwl formats.
  4. English masters: A NorthEnglishMaster spanning AD 278-1588 and an EnglishMaster spanning AD 404-1216 as described in New English data helps to finally close the Carolingian gap. Only available in Heidelberg format.
  5. BitsPieces.fh containing mean values of all small collections mentioned in the section Bits and Pieces - an approach to closing the Migration gap.
  6. A 2105 years long EnglishBog chronology.
See also ref. 24

Dating reports for published mean value collections:

The purpose of these reports is to enable check of our work. Each link will open within a new tab.
Note: These tables and diagrams show how a large number of site collections are dated and how their individual members (samples) are crossdated towards each other. Some members of these collections do not match well towards the rest of their collection. In these cases their dating have been set by good matching towards the master or towards other site collections.
  1. BelfastAD
  2. BelfastAD337to0
  3. LateBC
  4. BelfastLongVersion2
  5. EnglishBog
  6. EnglishMaster
  7. NorthEnglishMaster
  8. NorthEnglishRoman
  9. SouthEnglishRoman
  10. BitsPieces