Paleoceanography
Supporting Information for
Geochemical multi-element signatures of glacial and interglacial facies of the
Okhotsk Sea deepwater sediments during the past 350 kyr: A response to global
climate changes at the orbital and millennial scales
Eugene P. Chebykin1,2, Sergey A. Gorbarenko3, Ol’ga G. Stepanova1, Vsevolod S.
Panov4, Evgeny L. Goldberg 1,4
1Limnological Institute of the Siberian Branch of the Russian Academy of Sciences, 664033, UlanBatorskaya st. 3, P.O. Box 278, Irkutsk, Russia
2Institute of the Earth Crust of the Siberian Branch of the Russian Academy of Sciences, 664033,
Lermontov st. 128, Irkutsk, Russia
3V.I. Il’ichev Pacific Oceanology Institute of the Far East Branch of the Russian Academy of
Sciences, 690041, Baltiyskaya st. 43, Vladivostok, Russia
4Institute of Archeology and Ethnography of the Siberian Branch of the Russian Academy of
Sciences, 630090, Academician Lavrentyev ave. 17, Novosibirsk-90, Russia
Contents of this file
Text S1 (Description and testing the mathematical model for the calculation of
weight fractions of glacial and interglacial facies in Okhotsk Sea sediments)
Figures S1 to S3
Introduction
The supplemental materials include description and testing of the mathematical
model for the calculation of weight fractions of glacial and interglacial facies in
Okhotsk Sea sediments.
1
Text S1. Description and testing of the mathematical model for the calculation of
weight fractions of glacial and interglacial facies in Okhotsk Sea sediments.
1. Mathematical model for the calculation of weight fractions of glacial and
interglacial terrigenous facies in Okhotsk Sea sediments
For Okhotsk Sea sediments, we have tested a mathematical model of mixing of
two main types of geochemical facies (which differ by their dominating elemental
composition), delivered to the sea during considerably diverse climatic periods, e.g.,
glacial and interglacial, with a little admixture of the third facies (not more than 10%). For
modeling, 40 typically terrigenous elements were taken (Li, Be, Na, Mg, Al, K, Ti, V, Cr,
Fe, Co, Ni, Ga, Ge, Rb, Y, Zr, Nb, Cs, REE, Hf, Ta, W, Tl, Pb, Bi and Th). The elemental
composition characterizing a typical interglacial facies was set equal to the average
composition of the modern OS sediments (the first 5 cm of the core PC-7R – age 1-2.2
kyr), and the composition of a typically glacial facies was set equal to the average
composition of the sediments formed during the Last Glacial Maximum (64-99 cm, age
18-25 kyr). Both types of facies (glacial and interglacial) are specific mixtures of main
terrigenous sources: riverine (Amur River mainly), coastal and ice abrasion and do not
include any influence of volcanogenic matter. The relations between main terrigenous
sources must be different during glacial and interglacial periods. Let us assume typical
glacial and interglacial facies as “end members” in our model.
Bulk element contents in the sediments (С_tot) were preliminarily recalculated to
the terrigenous part (С_ter), considering biogenic dilution:
C_ter = C_tot / (1-D)
(1),
where D (“diluent”) is the weight fraction of biogenic components (CaCO 3, SiO2_bio and
2* TOC) in bulk sediments.
The number of modeled horizons was set to be identical to that of the core
investigated, i.e., 1723. The simulated weight fraction of the interglacial facies of
terrigenous matter (Q1), which was to be found by performing an inversion, was set in
the range from 0 to 1, scaling the shape of the global stack of oxygen isotopic
composition of marine benthic foraminifera [Lisiecki and Raymo, 2005] within the last
350 kyr (MIS 1- MIS 9). We choose this proxy only as an example for modeling, i.e. it is
possible to take any climate-change like profile or even profile with any shape. The third
terrigenous facies, which differs from the first two facies can be tephra because the OS
is in the direct vicinity of the zone of active volcanism (the Kurile-Kamchatka region).
Tephra presence is discretely found along the core as cryptotephras [Gorbarenko et al.,
2014] and as pure tephra layers that are up to several cm thick [Derkachev et al., 2012].
Concerning the elemental composition of the third (non-essential) facies, we have
assumed an averaged composition of pure layers of tephra at the horizons, with the
highest values of volcanogenic indicators [Gorbarenko et al., 2014]: 120-124, 10171021, 1515-1519 and 1533-1536 cm.
To test the model for a general case, the optional elemental composition of the
sources was randomly set using a random number generator.
The weight fraction of the third facies (Q3) was set as a value not more than 10%
using a random number generator, and the weight fraction of the second main facies
(Q2) (glacial terrigenous matter) was calculated as follows using the leftover principle:
Q2 = 1 - Q1 - Q3 (at Q2 < 0, the value of Q3 was corrected as Q3* = Q3+Q2). We
optionally modeled tephra layers by increasing the weigh fractions of the third facies Q3
up to 80% at the same horizons as those horizons found in the core, including adjacent
horizons with elevated values of volcanogenic indicators (115-125 cm, 1004-1029 cm
and a double layer in the interval of 1513-1537 cm). In the model, we also optionally
2
included the process of dilution of terrigenous sediment with “element-transparent”
admixture (i.e., biogenic compounds), which did not contain test terrigenous elements
(40 elements). The amount of diluent (“element-transparent” admixture) at each horizon
was set as a sum of the main biogenic components (CaCO3, SiO2_bio and 2* TOC) that
were determined in the samples of the core investigated. The fraction of the diluent D in
bulk samples of studied core reached 60% during optimuma of the warmest interglacial
stages MIS 1, MIS 5.5 and MIS 9.3 and varied by ca. 10% on average during other
periods (Figure S3а, S3e). Bulk contents of the elements in the modeled horizons were
calculated as follows:
C = (C1*Q1+ C2*Q2 + C3*Q3)*(1-D)
(2),
where C1, C2, C3 are the element contents in interglacial, glacial and volcanic
facies, respectively; and D is the diluent weight fraction in the bulk sediments.
We have also tested the composition of sediments expressed as ratios of
terrigenous elements, which, evidently, do not depend on the dilution of the terrigenous
part of sediments with an “element-transparent” admixture. We have used Al as a
reference element for normalization. Ratios of the contents of each element (El) to the
content of Al (R = El/Al) in each of the tested terrigenous matter sources were obtained.
R values in model horizons were calculated using the following equation:
R = (R1*Q1+ R2*Q2 + R3*Q3)
(3),
where R1, R2 and R3 are the ratios of El/Al in interglacial, glacial and volcanic
facies, respectively.
Optionally, using a random number generator, analytical errors of element (or
their ratios) measurements were generated, with values not more than ± 5 rel.% (percent
relative error).
The distribution of element contents or of their R values constructed according to
(2) or (3) along the model core depth were nondimensionalized (to independently
equalize the contributions of all elements based on their absolute contents/relations in
the sediments) in two ways. The first method is normalizing to the arithmetic mean (each
value of a data series is divided by the arithmetic mean of the data series). The second
method is linear conversion in the range {0-1} as follows:
C* = [C – min(С)] / [max(С) – min(С)]
(4),
where min(С) and max(С) are minimal and maximal values, respectively, of
variable С (C is the content of an element (El) in the sediments or its ratio R = El/Al) in
the data series (along the core depth).
Thus, each horizon represents a vector in an n-dimensional space (n is the
number of elements, i.e., 40 elements), the coordinates of which are dimensionless
contents (or ratios) of the elements. The complete coincidence of compared horizonvectors denotes the complete identity of their elemental compositions. To measure the
degree of similarity among the horizon-vectors compared with a standard one (typically
glacial or interglacial), we tested coefficients of correlation (COR) and of covariation
(COV). The sets of tested coefficients (COR or COV) of the horizon-vectors against the
standard vector (along the core depth) reflect by their physical sense` the inputs of
glacial (or interglacial) facies into OS sediments. The re-normalization of obtained COR
or COV profiles (sets of ones along the core depth) according to (4) must yield a
quantitative assessment of the weight fractions of required facies (Q1, Q2) in sediments
[Goldberg et al., 2001; Goldberg et al., 2001a; Goldberg, 2008; Goldberg and Grachev,
2008]. Let us call such calculations (COR or COV profiles renormalized according to (4))
“correlator” or “covariator”.
2. Testing of mathematical model for the calculation of weight fractions of glacial
and interglacial terrigenous facies in Okhotsk Sea sediments
3
A method for the quantitative assessment of weight fraction of two main facies in
bottom sediments based on their multi-elemental composition was first proposed by E.L.
Goldberg and applied for Lake Baikal sediments [Goldberg et al., 2001; Goldberg et al.,
2001a; Goldberg, 2008; Goldberg and Grachev, 2008]. The essence of this method is
the multi-elemental comparison of each horizon of a sediment core with a chosen
“standard”, i.e., sediments whose elemental composition formed during one of the
extreme climatic periods (interglacial optimuma, glacial maximuma). For such a
comparison, the most “conservative” clastogene elements are used (REE, Ti, Th, Nb,
Be, Li, etc.), which characterize terrigenous components of matter and are weakly
involved in diagenetic processes. The obtained characteristics (set of correlation or
covariation coefficients between the elemental composition of a chosen sample and that
of a “standard” reference interval) at reasonable suppositions and at a specific
nondimensionalization yield a quantitative assessment of the weight fraction of either of
the facies sought in the sediments at a given time slice.
Verifying this approach for OS sediments using the proposed mathematical
model with random number generation has shown that the rigorous calculation is only
possible when using a more “sensitive” statistical function, i.e., the covariation coefficient
(COV), and under the condition that the effect of the dilution of terrigenous matter with
an “element-transparent” admixture is considered, i.e., calculations are performed using
element contents per terrigenous part of the sediments or using terrigenous element
ratios, which evidently do not depend on the presence of a diluent in sediments (Figure
S2). It has been stated [Goldberg et al., 2001a; Goldberg, 2008; Goldberg and Grachev,
2008] that the dilution of terrigenous sediments with a matter that does not contain
terrigenous elements (“element-transparent” admixture) does not change the values of
the correlation coefficient (COR) of the horizons compared; therefore, bulk element
contents in the sediments can be used. This statement is true only if COR is calculated
by absolute
element
contents in the sediments.
Data normalization
(nondimensionalization of element content distributions along the core depth) allows the
equalization of contributions of all elements independent of their absolute contents (i.e.,
the contributions of elements with contents from 1 to 10 ppm will be comparable with
those elements whose contents vary from 1 to 10%); however, data normalization alters
the proportions of the elemental composition, depending on the degree of sediment
dilution with an “element-transparent” admixture and on methods of data
“nondimensionalization”, finally influencing the shape of calculated “correlator” profiles
(compare COR in Figure S2 and in Figure S3). In the case of data normalization for
arithmetic means, sediment dilution with an “element-transparent” admixture slightly
influences the COR profile shape (compare Figures S2а-S3a, S2b-S3b and S2c-S3c). If
data normalization were performed using a linear operator in the range {0-1} according
to (4), then the shape of COR profiles considerably differs (compare Figure S2а-S3e and
S2e-S3f), and the profile similarity increases only in the presence of large episodic
inputs of the third facies (volcanogenic one in this case) (compare Figures S2f-S3g). A
principal problem of COR application is the inability to obtain rigorous calculation of
weight fractions of main facies in the sediments under any conditions.
The usage of the covariation coefficient (COV) with other factors being equal
yields a more rigorous calculation compared with using COR (Figure S2 and S3),
although due to its properties, COV will depend on the degree of sediment dilution in any
case.
Next, we will use definitions and signs. Parameters at which “covariators” (COV)
or “correlators” (COR) are obtained will be marked with the following indices and
prefixes:
tot – bulk element contents in the sediments (according to (2));
4
ter – element contents per terrigenous part of the sediments (according to (2) at
D = 0);
rat – element ratios (R = El/Al);
volc – influence of the third terrigenous matter source is considered (volcanic, Q3
<10%);
tephr – the presence of tephra layers is considered, including pure tephra layers
and horizons with elevated values of volcanogenic indicators, the same as for the PC-7R
core (115-125 cm, 1004-1029 cm and a double layer within 1513-1537 cm; for these
intervals, Q3 is given within 20-80%);
error – analytical errors of element (or of their ratios) measurements are
generated, with not more than ± 5 ref. %;
{aver} – element contents (or their ratios) along the core depth are normalized for
arithmetic means;
{0-1} – element contents (or their ratios) along the core depth are normalized
using a linear operator in the range {0-1} according to (4);
prefix W – the reconstruction of interglacial (“warm”) terrigenous matter source
weight fractions (Q1) obtained by a comparison of the horizon-vectors (normalized
element composition) of the core using a typical interglacial matter vector;
prefix С – the reconstruction of the inverted profile of glacial (“cold”) terrigenous
matter source weight fractions (Q2) obtained by a comparison of the horizon-vectors
(normalized element composition) of the core with a typical glacial matter vector and
taking the opposite sign (i.e., values of COV or COR coefficients multiplied by minus 1
before re-normalization according to (4)).
For example, the descriptor W_COV(ter, volc, tephr, error, {aver}) will denote that
the reconstructed profile of the weight fraction of the interglacial (W) source of
terrigenous matter was obtained using a statistical function COV (“covariator”) with the
following parameters: element contents per terrigenous part of the sediments (ter) were
used, with the influence of volcanic source (volc), tephra layers (tephr) and analytical
errors of element measurement (error) considered, and the resulting model element
contents (considering the combined influence of volc+tephr+error) along the core depth
were then normalized for arithmetic means ({aver}).
2.1. Calculations using element contents per terrigenous part of sediment and
element ratios
The usage of “correlators” W_COR(ter) and W_COR(rat) for calculation of weight
fractions of main facies in the sediments is unacceptable in the case of a mathematically
rigorous model of two facies mixing because W_COR takes either minimal or maximal
values (Figure S2a, S2h). The addition of “measurement noise” as analytical errors of
element measurements (error) (<± 5 ref.%) and/or “natural noise” as the random
contribution of the volcanic facies (volc) Q3 (<10%) to a rigorous model makes the
W_COR profile more “noisy” but, simultaneously, more similar (in general) to the Q1
profile, the sought preassigned weight fraction of the interglacial facies (Figure S2b, S2i).
Under identical conditions as for W_COR, W_COV profiles (“covariators”) practically
coincide with profile Q1. If we exclude all analytical errors and the impact of the third
facies (a rigorous mathematical model with two end members (facies)), then W_COV
profiles exactly coincide with the sought profile of Q1 (Figure S2a, S2h), independent of
the data normalization method.
Similar to the case with W_COR, in principle, the addition of perturbations, such
as analytical errors of element measurements and of “natural noise”, into the rigorous
model does not change the shape and numerical values of reconstructed W_COV
5
profiles but only makes these profiles more “noisy” (compare Figure S2а with S2b and
S2e; and S2h with S2i and S2l).
Calculations of W_COV(ter…) and W_COV(rat…) under the influence of (error)
and/or (volc) parameters are sensitive to the data nondimensionalization method. When
normalizing {0-1} data, W_COV profiles are more “noisy” (Figure S2e, S2l) than in the
case of {aver} normalization (Figure S2b, S2i).
The addition of tephra layers (tephr) into the model results in the generation of
spikes on the profiles in the sites where these layers are situated (Figure S2c, S2f, S2j,
S2m). At {0-1} normalization, these spikes are considerably larger, resulting in a change
in the scaling (Figure S2f, S2m) due to the re-normalization of COV profiles according to
(4) at the final stage of obtaining “covariators” (see item 1).
A discrepancy in W_COV and С_COV profiles can serve as a test for the
presence of tephra layers in natural sediments or for considerable inputs of other “third”
facies (Figure S2d, S2k, S2g, S2n). While testing a general model (the random element
compositions of facies and the random Q1 (Q2) profile shape and location of tephra
layers along the core depth), the discrepancy at spiked locations can have both one-way
and opposite directions, as well as mixed ones (one-way directions at some horizons,
opposite directions at other horizons or some extremely weak spikes). In the special
case for the OS (i.e., the specific element composition of glacial and interglacial facies
and of regional tephra, Figure S1), opposite characteristics of spike discrepancies on the
profiles of W_COV and С_COV will be always observed only if {0-1} data normalization
has been performed. This property can be used for the identification of tephra layers in
OS sediments, and in the case of calculations using the element ratios W_COV(rat, {01}) and C_COV(rat, {0-1}), results in greater discrepancies for the identification of trace
contents of tephra (cryptotephras), the visual detection of which is difficult or impossible.
If the sediments were formed by only two main facies (no admixture of the third
one), then W_COV and C_COV profiles will be identical. The presence of slight
admixtures from the third facies and of typical analytical errors of element determination
practically do not affect the similarity between W_COV and C_COV profiles. This
property allows the applicability of the model proposed for the sedimentation objects
studied to be tested.
Analyzing all variants of W_COV(ter, …) and W_COV(rat, …) calculations, one
can conclude that the best approximations of the sought profile Q1 are obtained if {aver}
data normalization is performed. Let us call such calculation Q1-type.
2.2. Calculations using bulk element contents in the sediments
If we use bulk element contents in the sediments, then the shape of
reconstructed W_COV profiles during scaling will be greatly similar to that of profile Q1
in bulk sediments (Q1_dil), i.e., “covariators” will imitate the shape of the Q1 profile
corrected for biogenic dilution (Q1_dil = Q1*(1-D)) (Figure S3). For the specific elemental
composition of glacial and interglacial facies of OS sediments (Figure S1), the exact
calculation of the shape of the Q1_dil profile results from {0-1} data normalization (Figure
S3e); however, this situation is a special case. When solving a general problem (random
elemental composition of “end members”, i.e. of typical glacial and interglacial facies),
we generally fail to obtain exact reconstructions of the Q1_dil profile shape due to the
diluent impact. The most divergences appear at horizons with high biogenic diluent
content (more than 40%).
In the sediments of the studied core, such horizons correspond to the warmest
substages of interglacial periods (MIS 1, MIS 5.5, MIS 9.3), when Okhotsk Sea
productivity was maximal. If bulk element contents are normalized to arithmetic means
({aver}), then the W_COV profile at those horizons will always have elevated values
6
compared with the Q1_dil profile (Figure S3а), as well as when performing {0-1}
normalization, then the W_COV profile at those horizons may have any values (including
lowered) compared with the Q1_dil profile, but less by its divergence amplitudes in most
cases (random element compositions of facies). Let us call the latter ones
(W_COV(tot,…{0-1})) calculations of Q1_dil-type.
In principle, the addition of perturbations, such as analytical errors of elements
measurements (< 0.5% ref., error) and “natural noise” (Q3 < 10%, volc), into the rigorous
model does not change the shape and numerical values of calculated W_COV profiles
but only makes the profiles more “noisy” (compare Figure S3a with S3b; and S3e with
S3f).
Notably, even if the shape of the W_COV profile is identical to Q1_dil, it is
impossible to exactly calculate the weight fractions sought because the diluent (D)
fraction is unknown a priori. Moreover, if D is known, then it is much more reasonable to
use Q1-type calculation (item 2.1) and to make a necessary re-calculation using D for
obtaining the sought Q1_dil profile. Therefore, all Q1_dil type calculations will have a
semiquantitative character and only more or less correctly reflect the shape of the profile
Q1_dil.
The addition of tephra layers into the model, such as while calculating using
element contents per terrigenous part of the sediments or using elements ratios (item
2.1), results in the appearance of spikes on COV(tot, …, tephr) profiles in the locations
where these layers are added (Figure S3c, S3d, S3g, S3h), and these spikes also have
greater magnitudes in the case of {0-1} data normalization (Figures S3g, S3h). In the
latter case, the W_COV(tot, …, tephr, {0-1}) profile shape can be considerably distorted
in the segments with a high content of diluent D and, when comparing with the
C_COV(tot, …, tephr, {0-1}) profile, show oppositely directed differences not only in the
tephra layer locations but also in the segments with a large fraction of diluent D (Figure
S3h). This fact does not allow the use of the W_COV – C_COV comparison method for
a reliable identification of tephra layers or of large episodic inputs of other third facies.
While adding tephra layers, the shape of the W_COV(tot,…, tephr, {aver}) profile,
contrary to W_COV(tot,…, tephr, {0-1}) profile, one does not change the segments with
a large fraction of D (compare Figures S3b-S3c and S3f-S3g), and when comparing with
the “cold covariator” С_COV(tot,…, tephr, {aver}), the differences appear only in tephra
layer locations (Figure S3d). However, the use of this profile pair generally does not
always allow the reliable identification of tephra layers because corresponding spikes on
W_COV(tot, …, tephr, {aver}) and C_COV(tot, …, tephr, {aver}) profiles may not be
manifested or are weakly expressed (depending on the distribution of random
measurement errors and “natural noise”).
The usage of the COR statistical function does not allow exact calculations of any
sought profile shapes (Q1, Q1_dill) to be obtained, as shown in Figure S3. Shapes of
“correlators” W_COR(tot, …) are similar to those shapes obtained when using elements
contents per terrigenous part of the sediments W_COR(ter, …) and ratios of elements
W_COR(rat, …), as shown in Figure S2. Thus, we can conclude that “correlators” reveal
only general peculiarities of variations of two main facies at the orbital-time scale and
can be used for tentative reconstructions, as was performed for Lake Baikal sedimentary
records [Goldberg et al., 2001; Goldberg et al., 2001a; Goldberg, 2008; Goldberg and
Grachev, 2008].
Analyzing the variants of W_COV(tot, …) calculation, one can conclude that the
best approximations to the shape of the Q1_dil profile are obtained if {0-1} data
normalization is performed. However, large episodic inputs from third terrigenous
sources can considerably deform the shape of calculated profiles not only at their
locations but also in the segments with a large content of “element-transparent”
7
admixture diluent. In this case, {aver} data normalization provides more predictable
results with moderate systematic elevation in the segments with high dilution. Thus, the
use of bulk element contents in the sediments allows a semiquanitative assessment of
weight fractions of two main facies in bulk sediments to be obtained; in this case,
systematic errors will be higher compared with calculations of Q1-type.
References
Goldberg, E.L., M.A. Grachev, M.A. Phedorin, L.A. Kalugin, O.M. Khlystov, S.N.
Mezentsev, I.N. Azarova, S.S. Vorobyeva, Т.О. Zheleznyakova, G.N. Kulipanov, V.I.
Kondratyev, E.G. Miginsky, V.M. Tsukanov, K.V. Zolotarev, V.A. Trunova, Yu.P.
Kolmogorov, V.A. Bobrov (2001), Application of synchrotron X-ray fluorescent analysis
to studies of the records of paleoclimates of Eurasia stored in the sediments of Lake
Baikal and Lake Teletskoye, Nuclear Instruments and Methods in Physics Research
Section A, 470(1-2), 388-395, doi:10.1016/S0168-9002(01)01084-1.
Goldberg, E.L., M.A. Fedorin, M.A. Grachev, K.V. Zolotarev, O.M. Khlystov (2001a),
Geochemical indicators of paleoclimate changes in sediments of Lake Baikal, Russian
Geology and Geophysics, 42(1), 66-76.
Goldberg, E.L. (2008), Tracer Elements and Uranium-Series Isotopes in the Bottom
Sediments of Lake Baikal as Geochemical Climate Proxies for High-Resolution
Reconstructions of Regional Change. Chapter 1, in Climate Change Research Progress,
edited by L.N. Peretz, pp. 13-85, Nova Science Publishers, Inc., New York,
Goldberg, E.L. and M.A. Grachev (2008), High-resolution lake paleoclimate records of Lake
Baikal. Chapter 2, in Integrative projects of RAS SB “Late Cenozoic global and regional
environmental and climate changes in Siberia”, iss. 16, edited by A.P. Derevyanko, pp.
109-172, RAS SB Publishing House, Novosibirsk (in Russian).
Gorbarenko, S.A., N. Harada, M.I. Malakhov, T.A. Velivetskaya, Yu. P. Vasilenko, A.A.
Bosin, A.N. Derkachev, E.L. Goldberg, A.V. Ignatiev (2012), Responses of the Okhotsk
Sea environment and sedimentology to global climate changes at the orbital and
millennial scale during the last 350 kyr, Deep-Sea Research II, 61-64, 73–84,
doi:10.1016/j.dsr2.2011.05.016.
Gorbarenko, S.A., E.P. Chebykin, E.L. Goldberg, O.G. Stepanova (2014), Chronicle of
regional volcanic eruptions recorded in Okhotsk Sea sediments over the last 350 ka,
Quaternary Geochronology, 20, 29-38, doi: 10.1016/j.quageo.2013.10.006.
Lisiecki, L.E. and M.E. Raymo (2005), A Pliocene-Pleistocene stack of 57 globally
distributed benthic δO18 records, Paleoceanography, 20(1), PA1003,
doi:10.1029/2004PA001071.
8
Al Na Li
Be Mg K
Ti Cr Co Ga Rb Zr Cs Ce Nd Eu Tb Ho Tm Lu Ta Tl Bi
V
Fe Ni Ge Y Nb La Pr Sm Gd Dy Er Yb Hf W Pb Th
1.4
1.2
a
Cold / Warm
1
0.8
0.6
2.5
0.4
0.2
b
1.5
1
0.5
3.5
3
Tephr / Cold
2.5
Tephr / Warm
0
2
0
c
2
1.5
1
0.5
0
Be Mg K
Al Na Li
V
Fe Ni Ge Y Nb La Pr Sm Gd Dy Er Yb Hf W Pb Th
Ti Cr Co Ga Rb Zr Cs Ce Nd Eu Tb Ho Tm Lu Ta Tl Bi
Figure S1. Ratios between interglacial (Warm), glacial (Cold) and volcanogenic (Tephr)
facies spectra of elements normalized to the Al content in Okhotsk Sea sediments (core
MR06-04 PC-7R at 5116.87’N, 14912.57’E).
9
The calculations using elements contents
per terrigenous part of the sediments
1
3
Fraction
5
7
The calculations using elements
ratios (R = El / Al) in the sediments
MIS
1
9
Normalizing to arithmetic mean {aver}
1
3
Fraction
5
7
9
Normalizing to arithmetic mean {aver}
1
a
0.8
0.6
0.4
0.2
0.2
0
1
0
1
b
0.6
0.4
0.4
0.2
0.2
0
1
c
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
0
1
i
0.8
0.6
0
1
{aver}
or
{0-1}
0.6
0.4
0.8
h
0.8
{aver}
or
{0-1}
j
0
1
d
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
k
0
0
Normalizing to range { 0-1 }
1
Normalizing to range { 0-1 }
1
e
0.8
0.6
0.6
0.4
0.4
0.2
0.2
0
1
0
1
f
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
0
1
m
0
1
g
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
0
0
0
l
0.8
200
400
600
800 1000 1200 1400 1600
n
0
200
Model core depth, cm
Q1 ;
400
600
800 1000 1200 1400 1600
Model core depth, cm
W_COV ;
W_COR ;
C_COV
Figure S2. Modeling of the calculation of weight fractions of glacial and interglacial
facies in Okhotsk Sea sediments using element contents per terrigenous part of the
sediments and element ratios (R = El / Al) in different methods of data
10
nondimensionalization (normalization to arithmetic means {aver} – (a, b, c, d, h, i, j, k)
and linear conversion in the range {0-1} – (a*, h*, e, f, g, l, m, n)).
Q1 – sought preassigned weight fraction of the interglacial matter facies;
W_COV – calculations of Q1 obtained using the covariation coefficient (“warm
covariator”, see text S1);
W_COR – calculations of Q1 obtained using the correlation coefficient (“warm
correlator”, see text S1);
C_COV – calculations of the reversal profile of the weight fraction of the glacial facies
(Q2) obtained using the covariation coefficient (“cold covariator” see text S1).
(a), (h) – rigorous model with two facies; (b), (e), (i), (l) – model with two facies impacted
by (volc + error); (c), (f), (j), (m) – model with two facies impacted by (volc + error +
tephr); (d), (g), (k), (n) – comparison of W_COV with C_COV calculated with (volc +error
+ tephr) parameters.
(volc) – random impact of third (volcanic) facies (Q3 < 10%);
(error) – random errors of elements (or their ratios) determination (< 5 ref. %);
(tephr) – impact of tephra layers (with fraction of volcanic source Q3 =20-80%) at the
selected intervals.
*Calculations with the parameter {0-1} for a rigorous model are identical to
reconstructions with the parameter {aver}.
11
The calculations using elements
contents in bulk sediments
1
3
Fraction
5
7
1
9
W_COV,
fraction
1
0.8
0.4
Q1_dil
D
0
1
b
0.8
7
0.4
0.2
c
9
W_COV,
fraction
1
e
Q1
1
0.8
0.6
0.4
0.2
0
Q1_dil
D
f
0.8
1
0.8 0.6
0.6 0.4
0.4
0.2 0.2
0
0
1
0.6
0
1
5
Normalizing to range { 0-1 }
0.8
1
0.8 0.6
0.6 0.4
0.4
0.2 0.2
0
0
1
Q1
0.6
0.2
3
Fraction
Normalizing to arithmetic mean {aver}
a
The calculations using elements
contents in bulk sediments
MIS
1
0.8
0.6
0.4
0.2
0
1
g
0.8
0.8
1
0.8 0.6
0.6 0.4
0.4
0.2 0.2
0
0
1
0.8
0.6
0.4
0.2
0
d
1
0.8
0.6
0.4
0.2
0
0.8
1
0.8 0.6
0.6 0.4
0.4
0.2 0.2
0
0
0.6
0.4
0.2
0
h
1
0.8
0.6
0.4
0.2
0
200
400
600
800 1000 1200 1400 1600
0
Model core depth, cm
Q1 ;
Q1_dil ;
200
400
600
800 1000 1200 1400 1600
Model core depth, cm
D;
W_COV ;
W_COR ;
C_COV
Figure S3. Modeling of the calculation of weight fractions of glacial and interglacial
facies in Okhotsk Sea sediments using element contents in bulk sediments in different
methods of data nondimensionalization (normalization to arithmetic means {aver} – (a, b,
c, d) and linear conversion in the range {0-1} – (e, f, g, h)).
Q1 – sought preassigned weight fraction of the interglacial facies;
Q1_dil – Q1 per bulk sediments (Q1_dil = Q1*(1-D));
D – weight fraction of the “diluent” (“element-transparent” admixture, which does not
include terrigenous elements) in bulk sediments;
W_COV – calculations of Q1 obtained using the covariation coefficient (“warm
covariator”, see text S1);
W_COR – calculations of Q1 obtained using the correlation coefficient (“warm
correlator”, see text S1);
C_COV – calculations of the reversal profile of the weight fraction of the glacial facies
(Q2) obtained using the covariation coefficient (“cold covariator” see text S1).
(a), (e) - rigorous model with two facies; (b), (f) - model with two facies impacted by (volc
+ error); (c), (g) - model with two facies impacted by (volc + error + tephr); (d), (h) –
comparison of W_COV with C_COV calculated with (volc +error + tephr) parameters.
12
0