Dendrochronological Dating of Roman Time

Abstract

Based on published and otherwise available tree-ring data, we have analyzed the dendrochronological support for the current dating of Roman activities in western Europe. Manuscript rejected by Tree-Ring Research, details of peer review see: www.cybis.se/dendro

Full text

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page1of37

Dendrochronological dating of Roman time
1
2
Petra Ossowski Larsson* and Lars-Åke Larsson, Sweden 3
* Corresponding author: petra@cybis.se 4
5
Abstract
6
Based on published and otherwise available tree-ring data, we have analyzed the dendrochronological support for the current
7
dating of Roman activities in western Europe. Among other things, we have reconstructed the Belfast oak chronology with its gaps
8
and depletions, and found the current links to be too weak to make a definitive statement about the continuity into Roman time of
9
the reviewed curves.
10
11
We have then found a distinct correlation between recent Danish and Norwegian oak curves, and supra-long pine curves from
12
northern Scandinavia. Moreover we have found a distinct correlation between a long north-west European oak curve anchored
13
archaeologically in Roman time, and the Scandinavian pine curves, but 218 years later than expected. There is no correlation at or
14
near to the expected point of match.
15

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page2of37

16
To resolve this problem, more data – still not available – would need to be analyzed, but from a dendrochronological point of view
17
we can not exclude the possibility that Roman time is conventionally dated too old by more than 200 years.
18
19
Introduction
20
Chronology criticism (in German "Chronologiekritik", in English "Revised Chronology") is an umbrella term for various hypotheses
21
doubting the correctness of the historical course widely accepted by historians (i.e. the established historical model). These
22
hypotheses claim that historical events have to be redated and/or that historical eras never took place, which would imply invented
23
years in our history. The academic community has definitely rejected chronology criticism in this form.
24
When the first long continuous tree-ring chronologies were announced in Europe about 30 years ago, they were claimed to confirm
25
the conventionally assumed length of the Christian era, i.e. about 2000 years between our time and Western Roman Emperor
26
Augustus. Moreover, these tree-ring chronologies were used to build calibration curves for radiocarbon dating which allowed
27
floating chronologies from e.g. ancient Egypt to be placed on the time line. However, this prompted the chronology critics to reject
28
dendrochronology and also radiocarbon calibration (e.g., Newgrosh 1992, Blöss & Niemitz 2000) as the dates generated with the
29
new methods did not fit their hypotheses. But there were critics also among historians and dendrochronologists, who complained
30

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page3of37

about inconsistencies and lack of transparency concerning methods and data (Baatz 1977, Cüppers 1979, Baatz 1983, Lambert &
31
Lavier 1991, Seigne 2007).
32
The main subject of our work has been to prove with dendrochronology alone, if possible, where tree-ring sequences measured
33
from wood with archaeologically Roman origin fit on an absolute time line, in order to reject or confirm the existence of invented
34
years in the Christian era.
35
36
The long and supra-long tree-ring chronologies of Europe
37
The first prerequisite for our project was to find demonstrably continuous tree-ring chronologies with an absolute anchor in living
38
trees and with at least 2500 years length. As the Romans seem to have preferred oak as construction wood, there are considerable
39
amounts of internally well replicated tree-ring sequences with archaeologically Roman origin from Germany, France and England.
40
Therefore we first had a look at the more than 2000 years long European oak chronologies.
41
In 1984, a continuous oak tree-ring chronology for western Europe, which spanned more than 7000 years, was announced as
42
completed (Pilcher et al. 1984). It was a joint venture by the dendro-labs in Belfast, Köln and Stuttgart Hohenheim, involving oak
43
chronologies from the north of Ireland, northern Germany and southern Germany. This joint venture was necessary because neither
44
the Irish nor the German chronologies were at that time stand-alone, with the weakest link in the Irish chronology then thought to be
45

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page4of37

at 250 BC, and in the German chronology at 550 BC where there was actually an error and the article added in 71 extra years at
46
that point.
47
Furthermore, in order to connect the BC chronology to the absolute AD chronology, additional material had been necessary. In the
48
Belfast case, the English chronologies from Carlisle and Southwark were used as a bridge. In Germany, the link between late
49
BC/Roman time and early medieval time was reinforced by the West German chronology of Ernst Hollstein.
50
51
The German oak chronologies
52
Köln, Hohenheim, Bernd Becker
53
The raw measurement series from both Köln and Hohenheim are still unpublished and unavailable to us. Though it is claimed that
54
the Hohenheim chronology now confidently bridges the so called "Hallstatt gap" at 550 BC (Friedrich et al. 2004), this is nothing we
55
can verify.
56
For a suitable reference curve over the past 2350 years, the Hohenheim lab (M. Friedrich, personal communication) refers to Bernd
57
Becker's oak chronology from southern Germany (Becker 1981).
58
59

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page5of37

Ernst Hollstein
60
The only so far available German detail-data (not the raw data though) is published as "Jahrringtafeln" by Ernst Hollstein (Hollstein
61
1980). We retrieved 174 single-site mean value curves from the hand drawn tables (details available at:
62
http://www.cybis.se/forfun/dendro/hollstein ). These mean value curves form a 2698 years long chronology (724 BC to AD 1974)
63
which matches the Becker chronology with corr. 0.62, t=37.8 at 2317 years overlap (Hollstein normalization used if not otherwise
64
specified).
65
We then analyzed the internal consistency of the Hollstein chronology with CDendro (see ref.), using the Hollstein normalization and
66
the following criteria: we excluded all sequences shorter than 100 years, and required a correlation coefficient of at least 0.40, a t-
67
value of at least 6.0 and a minimum overlap of 70 years.
68
This resulted in an absolute, continuous chronology (HollsteinRecent, AD 383 to1974) and a floating continuous chronology
69
containing a lot of archaeologically Roman sequences (HollsteinRoman, 546 BC to AD 315). Dating reports and downloads are
70
available at: http://www.cybis.se/papers/data1 .
71
We excluded Hollstein's youngest Roman mean value sequence, the Rhine bridge in Köln (AD 149 to 336), from our
72
HollsteinRoman chronology even though it fits convincingly in its older part. This is because we suspect an error at about AD 250,
73
and because the sequence does not match too well against the Becker chronology.
74

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page6of37

Hollstein used, except for the Köln Rhine bridge, three site collections to bridge the gap between the two chronologies:
75
Broichweiden (AD 129 to 365), Köln Knabengrab (AD 273 to 511) and Berlegem (AD 284 to 550). But these site collections do not
76
by a long way fulfil our criteria for cross correlation and can therefore neither be incorporated in the chronologies, nor be regarded
77
as providing a credible bridge. Figure 1 and Table 1 describe the relation in time of the sequences involved, and their crossdating
78
quality with the dating assumed by Hollstein. The Becker master chronology is present as a reference.
79
80
81
Figure 1: Time line diagram for the conventionally assumed dating; Becker master chronology and Hollstein sequences bridging the period known as the 82
"Migration gap". 83
84

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page7of37

85
Table 1: Crossdating quality at the conventionally assumed position, Becker master chronology and Hollstein sequences bridging the "Migration gap". 86
(An'*'indicatesaverylowcorrelationvaluethoughtheoverlapisnottoolow(corrCoeff<0.2andoverlap>40).Sortkeyisusedtoletthememberwiththebestmeanvaluematchtowardsanythree
87
othermembersappearintheleftuppercornerofthetable.)
88
Table 1 shows the excellent compliance of the two Hollstein main chronologies with the Becker master. The correlation of the
89
HollsteinRecent chronology is better than that of the HollsteinRoman chronology probably because Becker and Hollstein partly used
90
the same samples for this time, e.g. the tree coffins from Oberflacht, Hüfingen and Zöbingen.
91
However, the three site collections forming the bridge do not correlate properly with the Hollstein or Becker chronologies. Although
92
both Berlegem and Köln Knabengrab probably are synchronized and dated correctly, this can not be demonstrated with the
93
available reference curves.
94

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page8of37

Therefore, the overall conclusion concerning Hollstein's West German oak chronology is that it fails to bridge the gap in the fourth
95
century AD and, consequently, that its Roman block is floating. The bridge in the Becker chronology can not be analyzed as the raw
96
data is not available.
97
98
The Belfast oak chronology
99
The Irish oak chronology announced in 1984 (Pilcher et al. 1984) spanned more than 7000 years, but with two important gaps at
100
950 BC and the BC/AD transition which had to be bridged with English sequences. In 1995, M.G.L. Baillie (Queen's University
101
Belfast, QUB) described the history behind this chronology from an Irish point of view in his book A Slice Through Time. A recent
102
paper (Brown & Baillie 2012) confirms that the original gaps and depletions in the Irish oak tree-ring record are still evident.
103
In 2010, the complete QUB raw data was published on the Internet, unsynchronized and undated but with site information, as 9500
104
single files. This publication allowed us to check the consistency of the Irish oak chronology according to the references mentioned
105
above. Again we used CDendro with the Hollstein normalization and the following criteria when building site collections: we
106
excluded all sequences shorter than 100 years, and required a correlation coefficient of at least 0.35, a t-value of at least 6.0 and a
107
minimum overlap of 70 years. Mean value curves of our site collections including dating reports are available at
108
http://www.cybis.se/papers/data1 .
109

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page9of37

110
Figure 2: Time line diagram for the conventionally assumed dating of the Belfast oak chronology. 111
112
Summing up, we agree with the description of the Belfast oak chronology according to Brown & Baillie (2012), with the following
113
comments and exceptions:
114
115
The gap at 2400 BC
116
This gap is not mentioned in the recent paper by Brown and Baillie, but we actually find a gap between 2479 and 2405 BC which we
117
are not able to close with Irish sequences. However, the gap is confidently bridged by an English collection from Croston moss
118
(3123 to 1682 BC, 1442 years long with almost equal overlaps on both sides of the gap, corr. 0.30, t=8.4 and corr. 0.39, t=10.6,
119
respectively).
120

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page10of37

This gap is discussed in Baillie et al. (1983), under the heading "The 1546 link". In that paper the gap is bridged by the sample
121
Q1546 and the link is called "weakest link in the Belfast Long chronology". We cannot reproduce the link described in that paper.
122
123
The 9th century AD depletion
124
We found two samples labelled "Ballinderry" which stretch over the weak period AD 839 to 895: Q9850, dated AD 804 to 1041, and
125
Q9846, dated AD 782 to 1021. This is worth mentioning as there is no other Irish timber bridging the gap, except the ship timber
126
with almost the same range found at Skuldelev,Denmark, but originating from Ireland.
127
128
The Swan Carr collection
129
Our interpretation of the QUB samples labelled "Swan Carr" is an 844 years long collection, conventionally dated 1155 to 312 BC.
130
QUB always refers to a length of 775 years, dated 1155 to 381 BC (e.g. Baillie 2009). We do not know when this dating appears for
131
the first time, but in a paper (Hillam et al. 1990) we find Baillie et al. (1983) as the reference for Swan Carr, though the dating is not
132
mentioned there.
133

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page11of37

A look into our SwanCarr collection reveals that there is only one sample reaching to the fourth century BC, Q4415. This sample is
134
327 years long and matches well (corr. 0.48, t=8.1, 218 years overlap) towards the rest of the collection. The end date is 312 BC.
135
To get an end date of 381 BC, this sample has to be truncated (there is no other sample labelled "Swan Carr" which produces an
136
end date of 381 BC). But there is no reason to truncate Q4415; though it is only one stem it matches in its full length towards Garry
137
Bog 2 (GB2; corr. 0.35, t=6.6). Therefore we suspect that QUB's end year 381 BC is just an early writing error which never has
138
been corrected.
139
140
The gap at 948 BC
141
We also agree with the description of this gap in the recent paper by Brown and Baillie, except for two points.
142
1) The crossdating quality of the match of Swan Carr towards the Belfast Long chronology is given as a t-value of 7.6 at 206 years
143
overlap in figure 4 of Brown & Baillie 2012. This high value is due to a drawing error; the best match is still claimed to be t=4.7
144
(Baillie et al. 1983) (D. Brown, personal communication).
145
2) We found a 380 years long oak curve from Ballymacombs More (Q10705M, measured in 2009, dated 1210 to 837 BC) extending
146
the Belfast Long chronology by 112 years with a convincing corr. 0.49, t=9.2. But there is no match towards Garry Bog 2 (GB2; corr.
147

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page12of37

0.13, t=1.4 at 110 years overlap, see table 2) though Garry Bog 2 also contains samples from that same Ballymacombs More,
148
especially in the oldest overlapping part.
149
Figure 3 and table 2 show the relation in time of the collections and samples involved in the 948 BC gap, and the quality of their
150
matches with the proposed dating.
151
152
153
Figure 3: Time line diagram for the conventionally assumed dating, bridging the 948 BC gap in the Belfast chronology.
154
155

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page13of37

156
Table 2: Crossdating quality at the conventionally assumed position, bridging the 948 BC gap in the Belfast chronology.
157
According to QUB, the bad match of Q10705M towards GB2 is due to the sample's last 40 years being very narrow and difficult to
158
measure, and periods with problematic rings and eccentric growth patterns with included sapwood. Therefore the tree-ring pattern
159
could be truncated at 950 BC, and only the inner section back to 1210 BC could be used. (D. Brown, personal communication).
160
In our opinion, this is a very drastic unparalleled measure which would mean that the whole overlapping part has to be cut away just
161
at the critical key link between the two chronologies. If we instead truncate Q10705M at 877 BC thus removing the 40 youngest
162
narrow rings, the correlation towards Garry Bog 2 becomes only slightly better (corr. 0.18, t=1.5 at 70 years overlap). Therefore we
163
suspect that the original link between the Belfast Long chronology and Garry Bog 2 is wrong and that the gap has to be widened
164
with a considerable amount of years as we could not find a consistent match in the existing material.
165
166

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page14of37

Conclusions about the Belfast chronology
167
It is evident that the Belfast oak chronology has two remaining weak links in the first and tenth centuries BC which means that it
168
actually consists of three parts:
169
• The absolute AD-part, corresponding to our BelfastAD, anchored in living trees and covering the last two millennia.
170
• The floating part between the weak links (GB2, BelfastArchaeol and SwanCarr), roughly covering the first millennium BC and
171
clearly connecting to the Carlisle and Southwark chronologies containing English series of archaeologically Roman origin
172
(Pilcher et al. 1984).
173
• The floating Long chronology, corresponding to our BelfastLong948, roughly covering the second to sixth millennia BC,
174
coarsely put in place by radiocarbon wiggle-matching against the Suess calibration curve (Suess 1978, Baillie et al. 1983).
175
Therefore when analyzed this chronology (as well as the Hollstein chronology) fails to provide an unambiguous bridge between
176
recent time and Roman time and does not solve our problem.
177
178
The English oak chronologies
179

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page15of37

A lot of archaeological oak timbers have been retrieved in England, both from Roman and later times. It has been possible to build
180
robust Roman masters up to about AD 300, and absolute recent masters back to about AD 400. But a dendrochronological bridge
181
between these blocks has not yet been established. This is described in detail in a Council for British Archaeology (CBA) Research
182
Report (Tyers et al. 1994), and also in the leaflet Dendrochronology (English Heritage 2004).
183
This means that the situation in England is very similar to what we find in the Hollstein chronology, a fact which is also highlighted
184
and discussed in the CBA Research Report.
185
186
The French oak chronologies
187
The gap between Roman and recent time chronologies is also evident in France (Lambert 2008). The recent chronology for
188
northern and central France is called "Historic Oaks" and spans about AD 500 to 2000, while the Roman chronology is called
189
"Classic Oaks" and spans 449 BC to AD 193 (Durost 2005).
190
A new master compiled by us from French data included in the QUB material, and a lot of series from north-eastern France put on
191
the Digital Collaboratory for Cultural Dendrochronology (DCCD) by Willy Tegel, provides additional years reaching into the gap from
192
both sides but still does not bridge the gap. Mean value curves of site collections including dating reports are available at
193

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page16of37

http://www.cybis.se/papers/data1 . As these samples are of archaeological origin, they are generally shorter and we therefore
194
included samples down to 75 rings length in our site collections.
195
FranceAbsoluteAD, AD 343 to 2008, is discontinuous but well replicated by the Becker chronology, HollsteinRecent and Historic
196
Oaks (see Figure 4 and Table 3). All sample identifiers with prefix "Q" or other letters denominate QUB measurement series; all
197
other samples are retrieved from Willy Tegel's projects placed on the DCCD in 2012.
198
TegelRoman, 511 BC to AD 217, is a strong Roman master which is well replicated by the Becker chronology, HollsteinRoman and
199
Classic Oaks (see Figure 4 and Table 3). The samples were retrieved from Willy Tegel's projects placed on the DCCD in 2012.
200
Figure 4 and Table 3 show the relation in time of the collections mentioned, and the quality of their matches with the conventional
201
dating.
202
203
Figure 4: Time line diagram for the conventionally assumed dating of recent and Roman blocks of some European oak chronologies.
204

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page17of37

205
206
Table 3: Crossdating quality at the conventionally assumed position for recent and Roman blocks of some European oak chronologies. 207
208
The overall conclusion is again: excellent replication within the recent block and the Roman block respectively, but the blocks fail to
209
confirm the bridge defined by the Becker master.
210
211
The Scandinavian pine chronologies
212
There are also supra-long absolute masters of other species. One of these is a pine curve from northern Finland (Eronen et al.
213
2002), (Helama et al. 2008) which was kindly made available to us by Mauri Timonen for research at single sample level. Therefore
214

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page18of37

we were able to check its synchronisation and found that this master curve is indeed continuous over more than 7600 years. A pine
215
curve from Torneträsk in northern Sweden (Grudd et al. 2002), given to us as a mean value curve by Håkan Grudd, matches the
216
Finnish master clearly in its full length and therefore is absolute as well.
217
However, both these pine masters do not contain archaeological material which can be linked to e.g. Roman activities; they are
218
mostly built from "anonymous" stems preserved for centuries in the lakes from which they were retrieved.
219
220
Crossdating European oak with Scandinavian pine
221
After having checked some available important long and supra-long tree-ring chronologies of Europe, we apparently have a
222
dilemma. The oak chronologies containing material of archaeologically Roman origin appear to have gaps or at least somewhat
223
uncertain continuity between recent time and Roman time, and on the other hand, the supra-long continuous pine curves do not
224
contain Roman material.
225
Our challenge now is to find a match between European oak and Scandinavian pine masters, i.e. both interregional and
226
interspecies. Is this possible, and under what conditions? As a first step we experimented with absolutely dated, recent material.
227
Interspecies correlation over short distances
228

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page19of37

The QUB raw data includes a lot of sub-fossil pine samples which can be crossdated to form quite long site collections. These are
229
the same sites which also yielded long oak collections, and there is a clear cross correlation between oak and pine (corr. coeff.
230
about 0.20). This has been investigated and described in detail at QUB (Pilcher et al. 1995), and gives an indication that an
231
interspecies match is possible, but that the correlation is so low that very long and strong chronologies are necessary to reach
232
significance.
233
The supra-long masters from Finland and Sweden are both from the north far above the tree line for oak, so we will not find recent
234
oak chronologies from the same region. The nearest absolute oak chronology we could find is from southern Norway (Christensen
235
& Havemann 1992, on ITRDB). We assembled a mean value curve of 26 site members, 281 years long and dated AD 1709 to
236
1989. This mean value curve shows excellent cross correlation "as dated" towards both Torneträsk and Finland, according to Table
237
4.
238
We also compared a long oak master from West Denmark (WestDK, National Museum Copenhagen) with the Scandinavian pine
239
masters.
240
end
date
overlap
Corr. coeff. and TTest in position
"as dated" towards
FinlandPine TorneTräskGrudd
SNorwayOakRecent 1989 280 0,25 4,2 0.33 5,8
WestDKOak 1986 1786 0,14 6,0 0.16 6,7
241

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page20of37

Table 4: Crossdating quality at the conventionally assumed position, for recent Scandinavian oak towards Scandinavian pine.
242
The Danish "as dated" match is evidently significant, which can be shown with figure 5 displaying the correlation coefficients and t-
243
values for all possible positions of WestDKOak towards Torneträsk pine. The true synchronous position (1986) has a well
244
discriminated t=6.7, compared to all false matches with the "next best" reaching t=4.3.
245
246
Figure 5: Correlation coefficients and t-values for all possible positions of WestDKOak towards TorneträskGrudd pine. Hollstein normalization used, diagrams 247
with other normalization methods are available at http://www.cybis.se/papers/data1 . 248
249
Interspecies correlation over long distances
250

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page21of37

As expected, the correlation becomes lower with increasing distance. Our long and well replicated absolute oak masters from
251
Ireland, Germany and France show quite low but still positive correlation towards Scandinavian pine.
252
However, combined as a mean value curve (AbsoluteAll) these three oak masters show "as dated" as the best match towards
253
Torneträsk pine for an average of the normalization methods, but of course this is not significant.
254
end
date
overlap
Corr. coeff. and TTest in position
"as dated" towards
FinlandPine TorneTräskGrudd
BelfastAD 2006 1979 0,03 1,4 0,07 3,0
HollsteinRecent 1974 1591 0,03 1,3 0,06 2,3
FranceAbsoluteAD 2008 1436 0,04 1,4 0,08 3,0
AbsoluteAll 2008 1979 0,04 1,9 0,08 3,7
255
Table 5: Crossdating quality at the conventionally assumed position for recent European oak towards Scandinavian pine. 256
The Becker master chronology, from southern Germany, shows very low correlation (0.02) against Scandinavian pine.
257
258
An attempt to date European oak of Roman time towards Scandinavian pine
259

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page22of37

In a similar approach to that described above, we combined our Roman time oak masters from Ireland, Germany and France and
260
cross correlated the mean value curve (RomanAll) towards the pine masters. This mean value curve is 1470 years long and dated
261
1155 BC to AD 315 (fig. 6). We expected the true match to appear among the 20 best matches.
262
263
Figure 6: Time line diagram for the conventionally assumed dating of members of RomanAll.
264
We found that "as dated" (AD 315) did not appear at all, but instead there is a best match 218 years later (end date AD 533). This
265
match is well discriminated and significant towards Finland pine (corr. 0.16, t=6.2). Fig. 7 shows correlation coefficients and t-values
266
for all possible positions of RomanAll towards Finland pine.
267

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page23of37

268
Figure 7: Correlation coefficients and t-values for all possible positions of RomanAll oak towards FinlandPine. Hollstein normalization used, diagrams with 269
other normalization methods are available at http://www.cybis.se/papers/data1 .
270
To exclude the possibility of a spurious match, we tested the independent French Roman time dataset called "Classic oaks" (Durost
271
2005). This curve, 642 years long and dated 449 BC to AD 193, also did not match at the expected position (corr. -0.01, t=-0.3), but
272
218 years later (corr. 0.15, t=3.9, end date AD 411) towards Finland. If this curve is combined with our other Roman time masters,
273
the match towards Finland 218 years later than expected increases slightly to corr. 0.16, t=6.3.
274
We also made a block analysis of this match between European oak and Finnish pine (block length 350 years, block distance 30
275
years), which demonstrated that the match is consistent over its whole length and that the assumed end year AD 315 is not
276
suggested at all by CDendro.
277

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page24of37

Table 6 shows correlation and t-value for each single member of the mean value curve towards the Scandinavian pine masters,
278
both for the expected "as dated" position and for the match 218 years later. It is evident that all members show much better values
279
at the unexpected new offset.
280
281
end
date
overlap
Corr. coeff. and TTest in position
"as dated" towards Corr. coeff. and TTest in position
"as dated" plus 218 years towards
FinlandPine TorneTräskGrudd FinlandPine TorneTräskGrudd
RomanAll 315 1469 0,01 0,3 0,01 0,3 0,16 6,2 0,11 4,3
TegelRoman 217 727 -0,01 -0,3 -0,01 -0,4 0,13 3,6 0,09 2,3
HollsteinRoman 315 860 -0,06 -1,7 -0,12 -3,5 0,15 4,3 0,08 2,3
GB2
-220 726 0,06 1,5 0,04 1,0 0,12 3,4 0,08 2,1
SwanCarr -311 843 0,04 1,0 0,04 1,3 0,09 2,6 0,09 2,6
ClassicOaks 193 641 -0,01 -0,3 -0,06 -1,5 0,15 3,9 0,10 2,5
282
Table 6: Crossdating quality at both the conventionally assumed position and with 218 years offset for Roman time European oak towards Scandinavian 283
pine.
284
A compilation of all the collections and chronologies mentioned above, their conventional dates and their new dates suggested by
285
this study is given in Table 7.
286
Chronology Start
year End
year Reference Suggested
start year Suggested
end year Comments
AbsoluteAll 25 2008 --- 25 2008 compiled from BelfastAD,FranceAbsolutAD,
HollsteinRecent
Becker -369 1950 Becker 1981 --- 1950 a)

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page25of37

Chronology Start
year End
year Reference Suggested
start year Suggested
end year Comments
BelfastAD 25 2006 Brown&Baillie 2012 25 2006 b)
BelfastArchaeol -569 -68 Brown&Baillie 2012 -351 150 b)
BelfastLong948 -5451 -948 Brown&Baillie 2012 --- --- b)
Berlegem 284 550 Hollstein 1980 284 (?) 550 (?) a)
Broichweiden 129 365 Hollstein 1980 --- --- a)
ClassicOaks -448 193 Durost 2005 -230 411 a)
Croston -3197 -1681 Baillie 1995 --- --- b)
FinlandPine -5633 2004 Eronen et al. 2002,
Helama et al. 2008 -5633 2004 e)
FranceAbsolutAD 343 2008 Tegel 2012 343 2008 d) Tegel, b) QUB
GB2 -946 -220 Brown&Baillie 2012 -728 -2 b)
HistoricOaks 672 2004 Lambert 2008 672 2004 a)
HollsteinRecent 383 1974 Hollstein 1980 383 1974 c)
HollsteinRoman -545 315 Hollstein 1980 -327 533 c)
KölnKnab 273 511 Hollstein 1980 --- --- a)
KölnRhineBr 149 336 Hollstein 1980 --- --- a)
Q10705M -1209 -836 Brown pers. comm. --- --- b)
RomanAll -1154 315 --- -936 533 compiled from GB2, HollsteinRoman,
SwanCarr, TegelRoman
SNorwayOakRecent 1709 1989 Christensen &
Havemann 1992 1709 1989 c)
SwanCarr -1154 -311 Brown&Baillie 2012 -936 -93 b)
TegelRoman -510 217 Tegel 2012 -292 435 d)
TorneTräskGrudd -5319 1997 Grudd et al. 2002 -5319 1997 a)
WestDKOak 200 1986 Nationalmuseum
Copenhagen 200 1986 a)
287
Table 7: The collections and chronologies mentioned above with their conventional dates and their new dates suggested by this study. 288
a) dated mean value curve available from originator 289
b) undated raw measurement data available from originator, our compilation 290
c) dated mean value site collections available from originator, our compilation 291

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page26of37

d) dated raw measurement data available from originator, our compilation 292
e) dated raw measurement data available from originator 293
294
Discussion
295
With the data available to us, it is not possible to demonstrate a significant bridge between Roman time chronology and recent
296
chronology for European oak. Above all, this seems to be due to a severe timber depletion between AD 200 and 400 in most places
297
in Europe.
298
So we changed strategy when we understood that the situation is different for Scandinavian pine and that there are indeed
299
continuous and absolute supra-long tree-ring chronologies readily available. With these pine chronologies we are able to
300
demonstrate correlations towards long recent, absolute oak chronologies from north-western Europe which confirm the dating of the
301
oak curves. However, the match of the European Roman oak complex extended with Irish late BC collections against the absolute
302
Scandinavian pine masters does not confirm the conventional dating. Instead there is a significant match 218 years later than
303
expected.
304
We can not exclude the possibility that the significant match 218 years later than expected indicates that the Roman oak complex
305
has a wrong conventional dendro-date. This would mean that the floating middle part of the Belfast chronology (Garry Bog 2) has to
306

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page27of37

be slid 218 years towards recent time, opening a gap of probably more than 200 years towards the prehistoric Belfast Long
307
chronology. An error of this kind has been predicted after our analysis of Q10705M (see above).
308
A direct consequence of a misdating of Roman time would also be that the Becker chronology is in error. In the publication of that
309
chronology (Becker 1981), figure 1 gives the distribution of site chronologies and shows a marked timber depletion between AD 200
310
to 400. The only site chronology linking Roman time and early medieval time is D5 made up from Danube (Donau) river valley oaks.
311
Figure 8 is a redrawn version of Becker’s figure 1.
312

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page28of37

313
Figure 8: Time line diagram for TegelRoman compared to elements of the Becker chronology, and internal replication (number of trees) of the Becker 314
chronology across the Migration period gap, with conventional dating.
315
316

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page29of37

As discussed earlier, the TegelRoman chronology matches the Becker chronology at AD 217 (as conventionally dated) with corr.
317
0.43, t=11.7 at 586 years overlap. Therefore it is remarkable that the youngest 110 years of TegelRoman also show a second
318
match exactly 218 years later at AD 435 (Table 8).
319
320
Table 8: Matching positions and crossdating quality for the youngest 110 years of TegelRoman towards the Becker chronology with various normalization 321
methods shown.
322
The second match is in fact better for all normalization methods (P2Yrs, Baillie-Pilcher, Cross84, Besançon Index E and Hollstein),
323
except Gleichläufigkeit and Skeleton Chi2.
324
As the old Becker chronology has been revised (Friedrich et al. 2004), we tried to get more recent data to see if this double match
325
with 218 years interval was still present in the Hohenheim data. We found a 215 years long section of the Hohenheim chronology
326
covering AD 200 to 414 in an article by David Holt (Holt 2011). The youngest 110 years of TegelRoman show the 218 years offset
327
against a site collection with data from the Danube valley with even better correlation:
328

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page30of37

329
Table 9: Matching positions and crossdating quality for the youngest 110 years of TegelRoman towards a part of the Danube chronology (Hohenheim), 330
various normalization methods shown. There is of course insufficient overlap for the “as dated” case.
331
332
Another direct consequence of shifting HollsteinRoman 218 years towards recent time would be that this collection comes to fit
333
exactly between two narrow ring events visible in the FinlandPine curve, at 330 BC and AD 536 to 542 (Helama et al. 2013), see
334
Figure 9. In the Finnish pine master, the event at 330 BC caused a servere timber depletion which made it difficult to bridge this
335
period (Eronen et al. 2002). The 330 BC event is clearly evident in the Torneträsk pine master as well (Grudd et al. 2002).
336
The oldest tree-ring of HollsteinRoman (conventionally 546 BC) directly succeeds a severe and still unexplained timber depletion in
337
the German oak record known as the "Hallstatt gap" (Becker 1993, Friedrich et al. 2004), which with 218 years offset would
338
coincide with the Scandinavian pine depletion at 330 BC.
339

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page31of37

340
Figure 9: Part of FinlandPine showing the narrow ring events at 330 BC and AD 536 to 542 resp., with HollsteinRoman fitting exactly in between if redated 341
218 years. HollsteinRoman (red normalised curve and green ring width curve) matched against FinlandPine (black normalised curve and blue ring width 342
curve) using the Hollstein normalization, offset 218 years, corr.coeff. 0.14, t=4.3 at 860 years overlap. 343
344
Conclusions
345
We would like to see our findings and the questions arising from them as the start of a scientific discussion, and a call for further
346
investigations. If the dating of Roman time is wrong, this would also have consequences for the radiocarbon calibration curve
347

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page32of37

(IntCal13), at least for the part corresponding to the first millennium BC, as both the Belfast chronology and the Becker chronology
348
have been fundamental for its construction. However, the overall course of the calibration curve has been confirmed by
349
measurements of wood used to build the Torneträsk pine master (Grudd et al. 2002). Regrettably, no Torneträsk samples have
350
been carbon dated between 2170 and 2770 BP which corresponds to Garry Bog 2.
351
Finally, an error of the size mentioned would have consequences for our calendar as it seems to indicate the existence of invented
352
years in the Christian era.
353
354
Acknowledgements
355
We would like to thank Georges (Joel) Lambert, Samuli Helama, Robert Porter and Olivier Bouriaud for insightful comments and
356
constructive feedbacks on an earlier version of the manuscript. Special thanks to Robert Porter who has done the proof reading and
357
style revision.
358
359
References
360
* Baatz, D., 1977. Bemerkungen zur Jahrringchronologie der römischen Zeit. Germania 55, 173-179. 361
362

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page33of37

* Baatz, D., 1983. Besprechung: Ernst Hollstein, Mitteleuropäische Eichenchronologie. Bonner Jahrbücher des Rheinischen Landesmuseums in Bonn 183, 363
718-720. 364
365
* Baillie, M.G.L, Pilcher, J.R. and Pearson, G.W., 1983. Dendrochronology at Belfast as a background to high-precision calibration. Radiocarbon 25 (2), 171-366
178. 367
368
* Baillie, M.G.L. 1995. A slice through time - dendrochronology and precision dating. Routledge, London 1995. 369
370
* Baillie, M.G.L. 2009. The radiocarbon calibration from an Irish oak perspective. Radiocarbon 51 (1), 361-371. 371
372
* Blöss, C. and Niemitz, H.-U., 2000. C14-Crash. Das Ende der Illusion, mit Radiokarbonmethode und Dendrochronologie datieren zu können. Verlag IT&W, 373
Berlin 2nd ed. 2000. 374
375
* Brown, D.M. and Baillie, M.G.L., 2012. Confirming the existence of gaps and depletions in the Irish oak tree-ring record. Dendrochronologia 30 (2), 85-91. 376
377
* Becker, B. 1981. Fällungsdaten Römischer Bauhölzer anhand einer 2350-jährigen Süddeutschen Eichen-Jahrringchronologie. Fundberichte aus Baden 378
Württemberg No.6, 369-386. 379
380
* Becker B. 1993. An 11,000-year German oak and pine dendrochronology for radiocarbon calibration. Radiocarbon 35 (1), 201–13. 381
382

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page34of37

* CDendro available at: http://www.cybis.se/forfun/dendro 383
384
* Christensen, K. and Havemann, K. 1992. ftp://ftp.ncdc.noaa.gov/pub/data/paleo/treering/updates/contributions/christensen_denmark_oak/norge/ 385
386
* Cüppers, H. 1979. Preface in: Hollstein 1980. 387
388
* Dendrochronology. English Heritage, June 2004. 389
390
* Durost, S. - Dendrochronologie et dendroclimatologie du 2è âge du Fer et de l'époque romaine dans le Nord et l'Est de la France. Datations, Système de 391
références et modélisations. A. Daubigney Dir., Thèse de doctorat, Université de Franche-Comté, Besançon, 2005, 175 pages, 100 annexes. 392
393
* Durost, S. and Lambert, G., 2005. Revision dendrochronologique du nord de la France a l’age du Fer et au debut de l’epoque romaine. 394
In : Barral, Ph., Daubigney, A., Dunning C., Kaenel G., Rouliere-Lambert M.-J. (eds.) — L’age du Fer dans l’arc jurassien et ses marges - Depots, lieux sacres 395
et territorialite a l’age du Fer. Actes du 29e colloque international de l’AFEAF, Bienne, 5-8 mai 2005. Besancon : Presses Universitaires de Franche-Comte, 396
2007, p. 13-30 (Annales Litteraires de l’universite de Franche-Comte ; 826 / Serie Environnement, societe et archeologie ; 11). 397
398
* Eronen, M., Zetterberg, P., Briffa, K.R., Lindholm, M., Merilainen, J., Timonen, M., 2002. The supra-long Scots pine tree-ring record for Finnish Lapland: Part 399
1, chronology construction and initial inferences. The Holocene 12(6): 673-680. 400
401

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page35of37

* Friedrich, M., Remmele, S., Kromer, B., Hofmann, J., Spurk, M., Kaiser, K.F., Orcel, C. and Küppers, M., (2004). The 12,460-year Hohenheim oak and pine 402
tree-ring chronology from central Europe. Radiocarbon 46 (3), 1111-1122. 403
404
* Grudd, H., Briffa, K.R., Karlen, W., Bartholin, T.S., Jones, P.D., Kromer, B. 2002. A 7400-year tree-ring chronology in northern Swedish Lapland: natural 405
climatic variability expressed on annual to millennial timescales. The Holocene 12(6): 657-665. 406
407
* Helama, S., Holopainen, J., Macias-Fauria, M., Timonen, M. & Mielikäinen, K. 2013. A chronology of climatic downturns through the mid- and late- Holocene: 408
tracing the distant effects of explosive eruptions from palaeoclimatic and historical evidence in northern Europe. Polar Research 32, 15866. 409
410
* Helama S., Mielikäinen K., Timonen M. & Eronen M. (2008): Finnish supra-long tree-ring chronology extended to 5634 BC. Norsk Geografisk Tidsskrift-411
Norwegian Journal of Geography 62 (4): 271-277. 412
The curve can be downloaded here: http://lustiag.pp.fi/ClimateFromTreeRings_gb.htm 413
414
* Hillam, J., Groves, C.M., Brown, D.M., Baillie, M.G.L., Coles, J.M. and Coles, B.J. 1990. Dendrochronology of the English Neolithic. Antiquity 64 (243), 210-415
220. 416
417
* Hollstein, E. Mitteleuropäische Eichenchronologie. Verlag Philipp von Zabern, Mainz am Rhein 1980. 418
419
* Holt, D.H. 2011. Germania and climate variability in 3rd and 4th centuries AD: A methodological approach to dendroclimatology and human migration. 420
Physical Geography, 32(3), pp. 241–268. 421

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page36of37

422
* Lambert, G., 2008. A century for dendrochronology and archaeology. Quiet activities. 423
In: LEHOËRFF A. (ed.). — Construire le temps. Histoire et méthodes des chronologies et calendriers des derniers millénaires avant notre ère en Europe 424
occidentale. Actes du XXXe colloque international de Halma-Ipel, UMR 8164 (CNRS, Lille 3, MCC), 7-9 décembre 2006, Lille. Glux-en-Glenne : Bibracte, 425
2008, p. 113-122 (Bibracte; 16). 426
427
* Lambert G.-N. & Lavier C. (1991) A new historical master chronology for dendrochronology 428
of the oak in the East of France. Questions about the dating in a large geographical area. Dendrochronologia, 165-179. 429
430
* Newgrosh, B., 1992. Living with Radiocarbon Dates - a response to Mike Baillie. Journal of the Ancient Chronology Forum 5, 59-67. 431
432
* Pilcher, J.R., Baillie, M.G.L., Brown, D.M., McCormac, F.G., McSweeney, B.and A. McLaurence, 1995. Dendrochronology of sub-fossil pine in the north of 433
Ireland, J. Ecology, 83(4), 665-671. 434
435
* Pilcher, J.R., Baillie, M.G.L., Schmidt, B. and Becker, B., 1984. A 7,272-year tree-ring chronology for western Europe. Nature 312 (5990), 150-152. 436
437
* QUB measurement series available at: from 2010 until 2013 http://chrono.qub.ac.uk/Resources/dendro_data/dendro.html; again available 2014 at 438
http://www.chrono.qub.ac.uk/bennett/dendro_data/dendro.html , but 539 of the old samples missing (including Q10705M), and 50 new samples added. 439
440

DendrochronologicaldatingofRomantime,draft,2014‐04‐18,Page37of37

* Seigne, J. - Dendrochronologie et datations archéologiques pour la période antique. Compte-rendu de la table-rode du 23/01/06 - Les petits cahiers 441
d'Anatole, no.20, 2007. http://hal.archives-ouvertes.fr/docs/00/16/91/28/PDF/pecada_20.pdf (opens in new window) 442
443
* Suess, H.E. 1978. La Jolla measurements of radiocarbon in tree-ring dated wood. Radiocarbon 20 (1), 1-18. 444
445
* Tegel, Willy: 122 objects placed on 2nd June 2012 in the DCCD data repository at http://dendro.dans.knaw.nl 446
447
* Tyers, I., Hillam, J., and Groves, C., 1994. Tree and woodland in the Saxon period: the dendrochronological evidence, in Environment and Economy in 448
Anglo-Saxon England (ed J Rackham). CBA Research Report 89, 12-22. 449
450
* WestDK, Chronology for Jylland and Fyn 200-1986, by Nationalmuseum in Copenhagen. Was available at their web site at least until 2009, but now seems 451
to have disappeared. See also http://www.cybis.se/wiki/index.php?title=WestDK 452
453
454