Jökull - 01.01.2009, Page 59

Holocene sediment- and paleo-magnetic characteristics from the Iceland and E-Greenland margins
DISCUSSION
Our major question is whether there is any clear dif-
ference in magnetic properties between the Holocene
sediments on the East Greenland versus the Iceland
shelves? In an analysis of surface sediment magnetic
properties (Watkins andMaher, 2003), sites from East
Greenland and Iceland were assigned to their clus-
ters 1 and 3, which indicated a strong association
with low-coercivity minerals and median ferromag-
netic grain-size. We synthesize the main components
of our data through Principal Component Analysis
(PCA) (Davis, 1986). The data were standardized so
that equal weight was given to each variable. Fifty-
three percent of the variability is associated with the
1st PC axis and a further 28% are associated with
the 2nd axis (Table 2). The 1st PCA is associated
with the strong inverse correlations (Figure 3D) be-
tween the MDF and ARM(J0)/ARM(J20) (Appendix
2) with the NRM(J0)/NRM(J60) and kARM also cor-
relating with those two variables respectively. This
axis largely reflects magnetic grain-size. The 2nd PC
has strong positive associations with the two magnetic
concentration variables and a weaker negative corre-
lation with MAD. A plot of the PC scores on the 1st
two axes shows a tight clustering of sites from the East
Greenland margin whereas the data from Iceland are
significantly more scattered on this plot (Figure 4).
A hierarchical cluster diagram based on these six
parameters shows that five of the East Greenland sites
are similar to each other with Euclidean distances of
<10 (0 = identical), but the Iceland sites group into
two distinct subsets (Figure 5). In group B, all four
sites lie in Vestfirðir (NW Iceland) (Figure 1), but
#324 lies on the N Iceland shelf. Group C consists
of sites from the SW, NW, and N Iceland (Figure
1). Analysis of the mineral content of many of these
cores indicated that the Vestfirðir sites had the lowest
weight% of volcanic glass, but also the highest car-
bonate content (Appendix 2). Either of these factors
could cause the Vestfirðir sites to cluster together.
An important question in magnetic measurements
is the amount of redundancy in stepwise demagneti-
zation measurements. What, if any, do the succes-
sive demagnetization steps tell the investigator? This
will probably vary from region to region. Redundancy
is the squared correlation (r2) between two parame-
ters, and the information that is not held in common
is 1-r2. At most sites the shared variance between
NRM(J0) and NRM(J60) is >80%, but are more vari-
able for the ARM demagnetization steps. In the case
of ARM(J0)/ARM(J20 or J30) the unexplained vari-
ance is generally <10%whereas for the kARM/k ratio
it is between 20 and 100% for half the cores. Some
of the poor correlations between susceptibility and
kARM are associated with the measurement of suscep-
tibility in the u-channel, but it might also be because
they respond oppositely to grain-size changes with in-
creasing grain size corresponding to an increase in k
but a decrease in kARM.
Table 2: Results of Principal Component Analysis (PCA) on data (see Appendix 2). Data were standardized.
– Niðurstöður höfuðþáttagreiningar (PCA) á gögnunum (sjá viðauka 2). Gögnin hafa verið stöðluð.
1) Variables = massMS, kARM, ARM(J0)/ARM(J20), NRM(J0)/NRM(J60), NI(30)
Variance explained: 1st PCA explained 57% of the variance
2nd PCA explained 35% of the variance
Loadings (as listed above): 1st PCA: 0.36, 0.59, 0.03, 0.48, 0.54
2nd PCA: 0.42, 0.52, -0.22, -0.16, -0.69
2) CV% data for the five variables listed under #1 above
Variance explained: 1st PCA explained 45% of the variance
2nd PCA explained 23% of the variance
Loadings (as listed above #1): 1st PCA: 0.51, 0.58, -0.28, 0.48, 0.32
2nd PCA: 0.25, 0.05, -0.67, -0.31, -0.62
JÖKULL No. 59 59