A Simple Crossdating Program for Tree-Ring Research

Item type Article

Authors Baillie, M. G. L.; Pilcher, J. R.

Citation Baillie, M.G.L., Pilcher, J.R. 1973. A simple cross-dating

program for tree-ring research. Tree-Ring Bulletin 33:7-14.

Publisher Tree-Ring Society

Journal Tree-Ring Bulletin

Rights Copyright © Tree-Ring Society. All rights reserved.

Downloaded 16-Feb-2017 23:03:04

Link to item http://hdl.handle.net/10150/260029

TREE -RING BULLETIN, Vol. 33, 1973

A SIMPLE CROSSDATING PROGRAM FOR TREE -RING RESEARCH

M. G. L. BAILLIE and J. R. PILCHER

Palaeoecology Laboratory,
Queen's University, Belfast

ABSTRACT

A crossdating program for tree -ring research has been written to compare
ring patterns of individual trees and composites. The program written in
FORTRAN calculates the t value for correlation at every point of overlap of the
two chronologies. The program is small enough to be used on a routine basis with
a large number of trees. As the chronologies must be free from errors, the
program is more suited to the study of oaks than coniferous trees.

INTRODUCTION
Recent developments in dendrochronology in Europe, especially the construction
of long standard chronologies from oak timbers, have created a demand for a
straightforward, powerful correlation method.
The computer program described in this paper was devised to compare the ring
patterns of individual timber samples with those of established chronologies whether
floating or standard. It is also a powerful tool for establishing the position of highest
correlation between the ring patterns of individual trees.
The program (Figure 1) is designed to be used for comparing the ring patterns of
chronologies without missing or double rings. This is normally the case in oak
chronologies from Europe (Huber and Giertz 1970; Baillie 1973). Where an error is
suspected this must be checked before the program is used.
A standard program exists for the calculation of the percentage agreement
coefficients at all positions of overlap between two chronologies (Eckstein and Bauch
1969). The drawback of the percentage agreement method is that it is non - parametric;
while calculating a figure of merit for the similarity of the two ring patterns, it takes no
account of the magnitudes of the year -to -year changes in ring width. Significant
crossdating is only demonstrated by the percentage agreement method when there is a
long overlap between two chronologies and the agreement figure is better than 60
percent. When comparisons of ring patterns of around 100 years length are made, the
normal distribution of agreement coefficients resulting from random mis- matches tends
to obscure the coefficient of true crossdating. Since with European oak the ring patterns
are frequently of 100 to 200 years in length, a program was written to calculate the
product moment correlation coefficient r at each position of overlap. This coefficient is
parametric and thus takes account of the magnitudes of the yearly width variations. The
value of r takes no account of the length of overlap, so the value of Student's t is
calculated from r to introduce a measure of significance in relation to length of overlap.
The form presented here is the simplest version for the comparison of one
chronology against a single longer chronology. It is written in FORTRAN IV and uses a
card reader and line printer. Versions have been written for use on a teletype computer
terminal and to handle more than one data set per run.
8 BAILLIE AND PILCHER

MASTER CROS

C STATISTICAL CROSSDATING PROGRAM FOR TREE RING DATA WRITTEN BY

C M.G.L.BAILLIE AND J.R.PILCHER
C QUEEN'S UNIVERSITY BELFAST
C
INTEGER NAMA(8).NAMB(8)
DIMENSION A(500),B(250),C(750)
C
C DIMENSIONS OF A AND B DETERMINE MAKTMUM LENGTHS OF SERIES THAT
C CAN BE HANDLED
C
COMMON A,B,C,AMEAN,MMEAN

C INPUT AND STANDARDIZATION

C a sasaaasa
C
C READ SITE NAME
C
READ(5,1)(NAMA(I),I51,6)
1 FORMAT(6A4)
C
C READ NAME OR NUMBER OF FIRST TREE

READ(5,101)NAMA(7),NAMA(8)
101 FORMAT(2A4)

C READ NUMBER OF YEARS IN FIRST TREE, AND DATE OF FIRST RING IF

C KNOWN, IF DATE IS NOT KNOWN LEAVE BLANK
C
READ(5,102)K,MDATE
102 FORMAT(13,14)
IF(MDATE)0,0,4
IDATE15
00 TO 27
4 !DATE a MDATE 14

C READ WIDTHS OF FIRST TREE

77 READ(S,3)(A(1),11,K)
3 FORMAT(25F3.0)

C STANDARDIZATION OF FIRST TREE

KK.4

Figure 1. Listing of program CROS.

 A Simple Crossdating Program 9

AMERNa0.0
DO 7 la1,K
A( I)4ALOG( S00a A( 1+ 2)/( A( I ) +A(I +1)+A(1+2)+A(1+3)+A(I +4)))
7 AMEANaAMEAN +A(1)
AMEANaAMEAN /K

C READ DATA FOR SECOND TREE, SITE, NAME, TEARS, WIDTHS.

PEAD(1,1)(NAMR(I),Ia1,6)
READ(1,101)NANB(7),NAM1(8)
READ(1,2)L
2 FORMAT(I3)
READ(1,3)(1(I),Ia1,L)
C STANDARDIZATION OF SECOND TREE

LaL.4
RMEANa0.0
DO A 141.1
R(1)ALOG(SOOat(1+2)/(B(1)+PO +1)+R(I+2)+R(I+3)aR(1+4)))
8 RMEAN4BMEAN+e(1)
RMEANaBMEAN/1

C COMPARISON SECTION
C saaaaaasasass

C SLInE TREES PAST EACH OTHER AND CALL SUBROUTINE CALK AT EACH POINT
C OF DVERLAP.THIS IS DONE IN THREE STAGES.

L9I 9
111.111
00 12 NF1,L11
12 CALI CALR (NF ,1,NF+10,L9.NF,Nia1 ,NF)
L111
KLaKL+1
DO 13 NG41,K1
13 CALL CALR(NG,NG,NG+L1r1,L,L11+NC)
K1OK10
DO 14 NN41,1.11
14 CALI CALR(NH,KL+NH,K,1,LaNH,K104NH)
C
C OUTPUT SECTION
C 2222222 711.27122Z
C
WRITE(2,19)

Figure 1, continued
10 BAILLIE AND PILCHER

19 PORMAT(1H1, / /40X,34H CROSS DATING BY PROGRAM + +CROSS + +)

WRITE(2,15)NAMA,NAMB
15 FORMAT (1)10,20X,20H COMPARISON OF TREE ,8A4,5H AND ,8A4)
IF(IDATE15)0,0,20
URITE(2,21)
21 FORMAT(/ 17H YEARS OF OVERLAP, 20X,27H 'T' VALUES FOR COMPARISONS)
GO TO 23
20 WRTTE(2,22)
72 FORMAT( /19H DATE OF OUTER YEAR,20X,27H 'T' VALUES FOR COMPARISONS)
;3 IEND ((K *L 22)/10)+10
DO 18 ED 1,IEND,10
K10 KD49
WRITE (2,16)IDATE,(C(I),IKD,K10)
16 FORMAT(2X,I4,6)(0OF10.2)
18 IDATE IDATE 10

C VALUES OF 'T' ARE SCANNED FOR ANY OVER A PRESET VALUE OF 3.5
C THERE VALUES ARE OUTPUT WITH THE OVERLAPS AS A SUMMARY

DO 24 L1,IEND
IF(C(L)- 3.5)24,24,0
tF(MDATE)0,0,26
LAP 1414
WRITE(2,25) 0(1),LAP
25 FORMAT(10X,6H 'T' 4F,F4.2,4M AT ,I4,17N YEARS OF OVERLAP)
GO TO 24
26 LAPL13 +MDATE
WRITE(2,28)C(1),LAP
28 FORHAT(10X,6H 'T' ,F4.2,23H ,DATE OF OUTER YEAR .14)
24 CONTINUE
WRITE(2,29)
79 FORMAT(1H1)
PAUSE
END

SUBROUTINE CAIR (N,NG,NS,NT,NV.NW)

C CALK CALCULATES THE CORRELATION COEFICIENT NNO 'T' VALUE

C FOR EACH POSITION OF OVERLAP AS SET IN THE SUBROUTINE CALLS

DIMENSION A(500),11(250),C(750)
COMMON A.B,C,AMEAN.BMEAN
S10.0
0.0.0
53.0.0
JJ NT -1
DO III NG.NS
XXA(II) -AMEAN
JJJJ +1
VY A(JJ) -BMEAN
s1 0+xx+xx
s2 ç2 +VYYY
s3s3 +xx +YY
1 CONTINUE

C NEGATIVE CORRELATIONS ARE SET 0.0

IF(s3.LE.0.0)GO TO 2
SIGX SORT(S1 /NV)
SIGV SORT(S2 /NV)
CAL 1S3 /(NV +SIGXSIGY)
CAL 2SORT(NV -2.0)
CAL 3SORT(1.CAL10,CAL1)
C(N4)CAL1CAL2 /CAL3
GO TO 3
2 C(NW)0.0
3 RETURN
END
FINISH

Figure 1, continued
 A Simple Crossdating Program 11

STANDARDIZATION AND CALCULATION OF t

The matching of two tree -ring patterns involves the cross -correlation of two sets of
numerical ring widths. When mis- matched, the correlation between these sets approxi-
mates that between two sets of random numbers. When the ring patterns represent the
same span of years the correlation should be high, assuming the trees have been growing
under similar conditions.
If any set of values (xy) shows a trend when plotted as co- ordinates, then x andy
are said to be correlated. If the values of x increase and decrease when the values of y
increase and decrease, then the correlation is positive. This is the expected condition for
the ring widths of two trees growing over the same period of years. Thus the basic
assumptions of the dendrochronological method argue for the use of a direct parametric
correlation method. The degree of correlation between x and y is measured by r, the
product moment correlation coefficient, defined thus
Ei xiyi - N zÿ
r
1/(Ei xi2 - Nx 2) (Ei Yi2 - N y-2)
where x and ÿ are the means of all the x and y values respectively. The calculation of
Student's t provides a measure of the probability of the observed value of r having arisen
by chance. The value of t is defined as
rVN- 2
t -
N/1 -r2
where N is the number of degrees of freedom, in this case the number of years overlap
between the ring patterns.
For the values of t obtained by this calculation to provide a valid measure of
probability, the sets of values x and y must be bivariate -normal (Parker 1973). In order to
satisfy this condition, any trends have to be removed from the basic data. The program
carries out a simple standardization where each ring width is converted to a percentage of
the mean of the five ring widths of which it is the centre value. In this form the data
varies about a mean of 100 but is not normally distributed. Normalization is achieved by
taking log to base e of the percentage figures. Sample sets of data treated in this way were
tested for skewness and excess and were found to satisfy the conditions of normality.
Probability levels can be obtained from tables of Student's t.

APPLICATIONS AND LIMITATIONS

As presented here program CROS uses relatively little computer store and running
times are fast. It is designed for routine use in the crossdating of large numbers of timbers
of unknown age. It is used by us for crossdating individuals and for crossdating against
standard and floating chronologies. Its advantage over the percentage agreement
calculation is that the distinction between true crossdating and random similarity is
greatly enhanced and a much greater statistical confidence can be placed on the results.
Table 1 shows a series of comparisons carried out by the two methods with the
confidence limits indicated by each method. It can be seen that the confidence limits are
in general several orders of magnitude higher in the case of the correlation coefficients.
Even when the computer indicates a high degree of confidence in the crossdating, this
must still be checked visually.
 12 BAILLIE AND PILCHER

CROSS DATING 8Y PROGRAM CROSS

COMPARISON OF TREE MILLSSROUGM 0 538 AND $1115180068 Q 536
YEARS OF OVERLAP 'T' VALUES FOR CuMPAR150A5
15 0.00 0.00 0.00 0.59 0.77 0,41 0.00 0.10 0,66 0.00
25 0.00 0.75 0.12 0.00 0.94 0,00 0.94 0.68 0.00 0.00
35 2.22 0.00 0,00 0.00 0,00 t.eó 0.45 0.00 0.61 0.62
4S 0,00 0.00 2.11 0.00 0.00 1.13 1.41
SS 0.65 0.00 0,00
1,32 2,n3 0,00 0.00 1.59 0.14 0.00 1,27
65 0,55 0.00
0.00 1,61 0.00 0.00 ',75 0,00 0.53 0.00
75 0.00 0.00
0,25 0.40 0.00 1.47 0,00 0,00 2.28 0.99 0.00
85 0,57 0,53 0.00
G.89 0.00 0,00 0.00 1,16 1.89 0.00
99 0,28 1.46 0.00 1,22
0.00
0.00 0.36 0.00 1.19 0.00 0.00
103 1,76 0.80 0.00 0.00 1,31 0.31 0.17 1.57
115 0.00 0.00
1.63 0,61 0.00 1.52 0,00 0,un 0.64 2.93
12S 0.00 0,00
0.95 0.00 1.24 1.09 0.00 0.00 0.41 0.03 0.00
13S 1.76 1.52
0,00 0.00 0.00 0,05 1.62 0.06 0.00 0.00
145 0.36 2.16 0.00 0.00
0.00 0,OU 1,52 1.12 0.00 0.00
155 1,32 1.06 0.00 0.75
0.00 '.10 0,00 1,50 1.36 0.00
165 0.00 0,00 0.00 0.07
0.44 c,21 0.36 0.77 0.00 0,00
17S 1.47 0.00 2.07
0.00 0.00 (.00 5.29 1,95 0.00 0,00
18S 1,26 0.61
0.72 0.52 0.00 0.00 1.63 0.91 0.00 0.00
19S 1,14 0,00 0.00
0.26 0.57 0,00 0.42 0.00 0.00 0.00
205 0.75 0.00 1.77
0,00 1,66 1,05 0.00 0,00 0.00 0.17
215 0,47 0.00 2.60
0.00 1.31 0,94 0.00 0.51 0.00 0,00 1.89
225 0,11 0.12 0.84 0.00 0.00 0.48 0.00 0.19
233
0.28 0.50
0.00 0.00 2.55 0.38 2,83 0,00 0.00 0,00 2.61
245 0.68 0.00
0.00 1.21 0.78 1,52 0,11 0.00 0,00 0.45 0,40
ASS 0.37 0.00 0.00 0.13 0,03 0,19 0.00 0,57 0.00 0.00
'T' 45.29 AY 180 YEARS OF OVERLAP

CROSS 04718G nv PROGR41' CROSS+

COMPARISON OF TRFE 81LLS8ROU68 Q 538 AND MILLSSROUGM 0 536
DATE OP 00,88 YEAR 'T' VALUES FOR COMPAPISONS
1467 0.00 0.00 0.00 0.59 0.27 0.41 0.00 0.10 0.66
1477 0.00
0.00 0.75 n.12 0.30 0,94 0.00 0.94 0,68 0.00 0.00
1487 2.22 0.90 0.00 0.00 0.00 1.66 0,45 0.00 0.61 0.62
1497 0.00 0.00 7.11 0.00 ;.00 1.13 1.47 0.65 0.00 0.00
1507 1.32 1.03 0.00 0.00 2.59 0.14 3.00 1.27 0.55
1517 0.00
0.00 1.61 0.00 0.00 0.75 0.00 0.53 0.00 0.00 0.00
1527 0.25 0.40 0.00 1.47 0.00 000 2.28 0.99 0.00 0.00
1537 0.57 0.53 ,i.89 0.00 0.00 0.60 1.16 1.89 0.00 0.00
1547 0.28 1.'.6 0.00 0.00 1.22 0.36 0.00 1.19 0.00
1557 0.00
1.76 0.80 0.00 0.00 1.31 0.31 0.17 1.57 0.00 0.00
1567 1.63 0.61 0.00 1.52 0.00 0.10 0.64 2.93 0.00 0,00
1577 0.95 0.00 0,24 1.09 l,.,(0 0.00 0.41 0.03 0.00
1587 1.52
1.76 0.00 6.00 0,00 .05 1.62 0,06 0.00 0.00 0.00
1597 0.36 2.46 0.00 0,00 0.00 1.52 1.12 0.00 0.00
1607 0.75
1.32 1.06 1.00 0,00 0.00 0.00 1,50 1.36 0.00 0.07
1617 0.00 0.n0 0,00 0.44 6.21 0.36 0.77 0.00 0.00 2.07
1627 1.47 0.00 íi.00 0.00 0.00 5.29 1.95 0,00 0.00 0.61
1637 1.26 0.72 rß,52 0.00 0.00 1.63 0.91 0.00 0.00
1647 0.00
1.14 0.26 0.00 0.57 -
0.00 0.42 0.00 0.00 0.00 1,77
1657 ((,75 0.00 0,00 1,66 1,05 0,00 0.00 0,00 0.17
1667 0,47 0,00 2,60
0.00 1.31 0.94 0.00 0.51 0.00 0.00 1,69
1677 0.11 0.32 1,84 0.00 0,00 0,48 0.00 0.19 0,28 0,50
1687 0.00 0,00 2.55 0.38 2.63 0,b0 0,00 0.00 2.61
1697 0.00
0.88 0.00 1.21 0.78 1,52 0.11 0.00 0.00 0.45 0.40
1707 0,57 0.00
0.00 0.13 u,03 0,19 0.00 n.57
.

0.00 0.00
'T' 5.29 .DATE OF OUTER YEAR 1632

Figure 2. Output from program CROS showing result when chronologies are floating
(above) and when the date of the longer one is known (below).
 A Simple Crossdating Program 13

Table 1. Comparison of the results of testing crossdating by the percentage agreement

method and by the calculation of t values.

Laboratory Years Percent Limits1 Students Limits1

Numbers for of Agreement t
Tree A Tree B Overlap

537 538 160 65 10' 3.87 10'

535 542 156 65 10' 5.46 5 x 10 °
536 538 106 65 10' 529 10'
547 535 116 63 10' 5.19 5 x 10r'
536 537 114 62 10' 3.59 5 x 10'
542 537 157 58 10-' 3.09 3 x 10-5
553 542 119 56 10' 3.89 10'
536 544 122 55 10- 3.39 10'
'Figures in these columns indicate the probability of the percentage agreement or t value occurring
randomly

INPUT AND OUTPUT

As both the nature of the computer peripherals and the data format at different
institutions vary, the program is presented in the form used in the Palaeoecology
Laboratory. The data is prepared on cards, one set per chronology. The first card of the
set gives the title or site name of the set (6 A 4 Format)and the second card an identity
number of sample name (2 A 4 Format). The third card gives the number of years in the
series (up to 999) followed by the calendar year of the first measurement if known (13,14
Format). If the year columns are left blank the comparison is assumed to be between
undated individuals and the results are given in terms of the number of years overlap. The
remaining cards of the set contain the ring width measurements (25 F3.0 Format or
25 I 3 Format). The format cards can easily be changed to suit different data formats.
The two styles of output are shown in Figure 2. The upper one being a comparison
of undated specimens, the lower one the same data with the year of the first ring
measurement given. In the first output `years of overlap' indicates the total number of
years that the shorter chronology has been moved relative to the longer one. At the end
of the table of t values there is a summary of all values greater than an arbitrary preset
value of 3.5.

ACKNOWLEDGEMENTS

The authors would like to thank Dr. D. Chambers of the Applied Mathematics
Department, Queen's University for advice on the application of correlation coefficients.
14 BAILLIE AND PILCHER

REFERENCES
Baillie, Michael G. L.
1973 A dendrochronological study in Ireland with reference to the dating of Medieval and
post -Medieval timbers. MS. Doctoral dissertation, Queen's University, Belfast.
Eckstein, D. and J. Bauch
1969 Bietrag zur Rationalisie -rung eines Dendrochronologischen Verfahrens und zur Analyse
Seiner Aussagesicherheit. Fortwiss 88:230 -50.
Huber, B. and V. Giertz
1970 Central European dendrochronology for the Middle Ages. In Scientific Methods in Medieval
Archaeology, edited by Rainer Berger, pp. 201 -12. University of California Press, Berkeley.
Parker, R. E.
1973 Introductory statistics for biologists. The Institute of Biology, Studies in Biology No. 43,
London.