Professional Documents
Culture Documents
Malinverno Et Al 2012
Malinverno Et Al 2012
1029/2012JB009260, 2012
B06104 1 of 17
B06104 MALINVERNO ET AL.: MHTC12 GEOMAGNETIC POLARITY TIME SCALE B06104
Figure 1. Red boxes indicate ocean floor areas with the M-sequence magnetic lineations used in this
study. The shaded background shows 120–160 Ma oceanic crust where M-sequence magnetic anomalies
have been observed from the age grid of Müller et al. [2008], with Cretaceous (K) ages in green and Juras-
sic (J) in blue.
[4] Here, we provide an updated M-sequence GPTS, NW of Australia (Figure 1); available studies are summarized
named MHTC12, that addresses some of the shortcomings by Tominaga and Sager [2010]. Several M-sequence GPTSs
of previous attempts. Following the “least favoritism” prin- have been constructed by assuming constant spreading rates
ciple advocated by Huestis and Acton [1997], the GPTS between radiometric dates in the Hawaiian magnetic linea-
obtained here minimizes the global variability of spreading tions [Larson and Pitman, 1972; Larson and Hilde, 1975;
rates in all the available marine magnetic anomaly sequen- Channell et al., 1995]. The most recent of these time scales
ces, thereby replacing the necessity for a constant spreading [Channell et al., 1995], hereafter referred to as CENT94, is
rate assumption. Moreover, MHTC12 includes information based on a single magnetic anomaly profile in the Hawaiian
on polarity chron durations estimated from astrochronology. lineations and three dates for CM0, CM16, and CM25.
This integration is accomplished by propagating uncertain- [7] In the most recent Geological Time Scale compilation
ties from the input data, which weigh the extent to which [Gradstein et al., 2004], Ogg and Smith [2004] constructed a
each piece of information constrains the final GPTS. We M-sequence GPTS with a composite magnetic anomaly
argue that such an integration of diverse data sources is an record based on the Hawaiian lineations in the M0-M25n
advance from the usual “winner-take-all” strategy where the interval, the Japanese lineations in the M25r-M27n interval
GPTS is based on the set of data deemed most accurate (e.g., [Handschumacher et al., 1988], and a deep-tow survey of the
a selected set of magnetic lineations). Japanese lineations in the M27r-M41 interval [Sager et al.,
1998]. In this GPTS, hereafter named GTS2004, Ogg and
Smith [2004] used a “minimalist” spreading rate model
2. Background (four intervals of constant spreading rate separated by small
2.1. Notation jumps or linear changes), three radiometric age constraints
(for CM0, CM26, and CM41), and a few polarity chron dura-
[5] Following general nomenclature [Opdyke and Channell, tions from astrochronology (see below). Although GTS2004
1996; Gee and Kent, 2007], M-sequence magnetic anomalies goes back to CM41, acceptance of the pre-CM29 reversal
are named M0r, M1n, etc., and the corresponding polarity record has been debated because these anomalies have low
chrons are CM0r, CM1n, etc. Magnetic polarity is explicitly amplitudes and the record is only from one spreading system,
noted by appending “n” for normal and “r” for reversed. We the Japanese lineation set [Sager et al., 1998; Tominaga
denote years of age as “a” and years of duration as “yr,” with et al., 2008]. To date, CM29 is the oldest chron where
the usual prefixes (e.g., 1 Ma = 1 million years ago). higher amplitude magnetic anomaly lineations are available
2.2. Previous M-Sequence GPTSs from multiple locations, and we focus on the CM0-CM29
interval in this paper.
[6] The M-sequence magnetic anomalies have been [8] Most recently, Tominaga and Sager [2010] con-
mapped in the oldest parts (about 120–160 Ma) of the world’s structed a GPTS for the CM0-CM29 interval based on an
oceans, and the best continuous lineation sets can be found in updated set of 52 magnetic anomaly profiles from the Pacific
the Western Pacific, Northern Atlantic, and Indian Ocean
2 of 17
B06104 MALINVERNO ET AL.: MHTC12 GEOMAGNETIC POLARITY TIME SCALE B06104
Figure 2. Block model distances used in this study, after TS10 except for the interval M6n-M15r in the
Keathley block model, which was not included in the TS10 compilation. Data in this interval were inserted
using the distances published in CENT94. Note the different distance scales.
(Figure 2). Their GPTS, hereafter named TS10, is not built Japanese and Phoenix lineations, and is constrained by two
assuming a constant spreading rate on a single set of linea- age calibration points (CM0 and CM26). This is equivalent
tions. Instead, TS10 is based on a composite profile obtained to assigning a constant spreading rate to the composite pro-
by averaging rescaled distances measured on the Hawaiian, file and assuming a constant ratio of spreading rates between
3 of 17
B06104 MALINVERNO ET AL.: MHTC12 GEOMAGNETIC POLARITY TIME SCALE B06104
the three lineation sets. The up-to-date magnetic block model to have remained rather constant throughout the Phanerozoic
data published by TS10 will be the basis for the GPTS pre- [Berger et al., 1992; Laskar et al., 2004, 2011].
sented here. [14] The duration of Mesozoic intervals have been esti-
[9] The GPTSs described above relied on the assumption mated by counting bedding couplets and bundles interpreted
of constant (or smoothly varying) spreading rates in a limited to correspond to precession and eccentricity cycles [Fischer
set of magnetic lineations. It was also recognized, however, and Schwarzacher, 1984; Herbert and Fischer, 1986;
that a valid GPTS should result in small spreading rate fluc- Herbert, 1992; Herbert et al., 1995; Fiet and Gorin, 2000] or
tuations on other ridges [Cande and Kent, 1992; Channell by some form of spectral analysis [Park and Herbert, 1987;
et al., 1995]. Huestis and Acton [1997] pointed out that Hinnov and Park, 1998; Meyers and Sageman, 2007;
there is no a priori reason to expect that spreading rates Meyers, 2008; Malinverno et al., 2010]. The durations of
vary smoothly at one particular location while they fluctuate CM0 through CM2 estimated by Herbert [1992] were used to
more erratically at all other ridges. They argued for a global constrain GTS2004. In this study, we directly incorporate,
approach where a GPTS is constructed by minimizing the into the GPTS construction process, the durations of CM0
spreading rate variation in all the magnetic anomaly linea- through CM3 and their uncertainties estimated by astro-
tions examined and ran a test inversion on a demonstration chronology [Fiet and Gorin, 2000; Malinverno et al., 2010].
data set. We adopt the same philosophy and add uncertainty
quantification in the GPTS construction process. 3. Data
2.3. Astrochronology in the Mesozoic 3.1. Magnetic Anomaly Block Model Distances
[10] Over the last few decades, time series analyses of [15] The first step in GPTS construction is to generate
sediment properties have conclusively shown periodicities block models consisting of alternating magnetic polarity
linked to Milankovitch orbital cycles [e.g., Hays et al., 1976; intervals. The positions of polarity changes in these block
Martinson et al., 1987]. This discovery sparked considerable models are adjusted to match measured magnetic anomaly
interest in using orbital cycles to measure time in the geo- profiles projected on a line perpendicular to the local strike of
logical record, and astronomical cycle stratigraphy started its magnetic lineations and deskewed so that the anomalies are
development [Fischer et al., 1990; Shackleton et al., 1990; symmetrical. Two recent GPTS studies list M-sequence
Hilgen, 1991; Schwarzacher, 1993]. block model distances derived in this fashion, CENT94
[11] In principle, an astronomically based time scale can [Channell et al., 1995] and TS10 [Tominaga and Sager,
be constructed by correlating records of sediment properties 2010]. CENT94 constructed block models for the Western
controlled by climate (e.g., carbonate content) with orbital Pacific Hawaii (based on 1 profile), Japanese (6), and Phoe-
solutions (for eccentricity, obliquity, and precession) or with nix lineations (4), supplemented by a North Atlantic Keath-
inferred insolation for a particular season and location. The ley block model computed from the spreading rates of
most recent version of the Neogene geological time scale Klitgord and Schouten [1986]. TS10 examined a larger set
[Lourens et al., 2004] represents a compilation of numerous of magnetic anomaly profiles and built block models for the
such individual astrochronological studies. Western Pacific Hawaii (based on 11 profiles), Japanese (28),
[12] Neogene astrochronology has significant implications and Phoenix lineations (13), the Indian Ocean NW of Aus-
for global variations in oceanic spreading rates. Wilson tralia (7), and the North Atlantic Keathley (4) and Canary (7)
[1993] and Krijgsman et al. [1999] found that the spread- lineations (Figure 1). The Western Pacific data in TS10
ing rates implied by astronomical dating in the last 10 Ma included 9 out of the 11 profiles used by CENT94. The block
were less variable than rates computed from the Cande and models of TS10 were built by averaging polarity boundaries
Kent [1992] GPTS; astronomical age control tends to determined for several profiles, thus reducing the noise due to
steady spreading rates [see also Gordon, 1993; Baksi, 1994; irregularities in the crustal recording process, and are the
Langereis et al., 1994; Krijgsman et al., 1999]. This obser- most comprehensive up-to-date compilation. We use for our
vation supports the notion that minimizing global spreading GPTS the TS10 block model distances, adding the distances
rate fluctuations is a useful strategy to construct a GPTS given by CENT94 for the interval M6n-M15r in the Keathley
[Huestis and Acton, 1997]. block model (Figure 2). Our magnetic anomaly nomenclature
[13] Astronomical dating in the geological past beyond a follows TS10, except that M25Ar, M25n5, and M25r5 in
few tens of Ma is more challenging than in the Cenozoic. The TS10 have been changed to M25Ar.1r, M25Ar.1n, and
solar system is chaotic, and because of sensitivity to initial M25Ar.2r, respectively, to follow general nomenclature rules
conditions the phases of orbital cycle solutions are accurate [e.g., Gee and Kent, 2007].
only to about 50 Ma [Laskar, 1999; Laskar et al., 2004,
2011]. Astrochronology in the Mesozoic cannot, therefore, 3.2. Radiometric Dates
be based on direct temporal correlation, but instead relies on [16] Radiometric dates relevant to the GPTS in the CM0-
detecting periodicities in sedimentary records that correspond CM29 time interval are listed in Table 1. Most of these dates
to supposed periods of orbital cycles. The resulting “floating” do not have a direct or precise tie to the reversal sequence,
time scales constrain the duration of intervals but need and previous M-sequence GPTSs have been constructed on
independent information to obtain absolute ages [Hinnov, the basis of two (TS10) or three dates (CENT94, GTS2004).
2004]. An additional difficulty is that tidal dissipation pro- Our GPTS is constrained by two radiometric dates, and the
gressively decreases the rotation rate of the Earth and GPTS ages we obtain will be later compared to all dates in
increases the Earth-Moon distance, so that the obliquity and Table 1.
precession periods were shorter in the geologic past; on the [17] The first date used here is 121.2 0.5 Ma for the
other hand, the long eccentricity cycle (405 kyr) is expected middle of CM0r. This date is from the results of He et al.
4 of 17
B06104 MALINVERNO ET AL.: MHTC12 GEOMAGNETIC POLARITY TIME SCALE B06104
[2008] with an increased uncertainty to account for the Ludden [1992] on a celadonite vein recovered from Ocean
unknown position of the dated material within CM0r, which Drilling Program (ODP) Site 765 in the Argo Abyssal Plain.
lasts about 0.5 Myr. All uncertainties reported here are one This age was assigned to the middle of CM26n by
standard deviation; radiometric age uncertainties are based GTS2004, but the end of CM26n better matches the location
on analytical error and do not therefore include uncertainties of Site 765 in the polarity block model of Sager et al.
related to causes such as standard age uncertainty or sample [1992]. This is a minor difference given the large uncer-
alteration. He et al. [2008] dated reversely magnetized lavas tainty of this date.
in northeast China and argue that the ages of these volcanics
correspond to CM0r. Their conclusion is consistent with 3.3. Chron Durations From Astrochronology
the original interpretation of Pringle and Duncan [1995], [21] The polarity chron durations from astrochronology
who assigned to the end of CM1r a 124.6 Ma date measured used in this study are listed in Table 2. These chron dura-
on reversely magnetized basalt sampled at MIT Guyot tions, estimated by Malinverno et al. [2010] and by Fiet and
(Table 1). An age of 121 Ma for the onset of CM0r was used Gorin [2000] on Cretaceous pelagic sequences in northern
in the CENT94 GPTS. and central Italy, have been chosen because they are closely
[18] The age of CM0r, however, is debated. The critical tied to a magnetic stratigraphy and include a quantified
question is which reversed polarity chron is being dated by uncertainty. The durations of CM0r, CM1n, CM1r, and
the work of Pringle and Duncan [1995]: CM0r or CM1r? Ogg CM3n in Table 2 are the same (within uncertainties) as those
et al. [2004] assigned the 124.6 Ma age of Pringle and Duncan estimated for CM0r by Huang et al. [2010] and for the other
[1995] to CM0r and concluded that the onset of CM0r (the chrons by Herbert [1992].
Barremian-Aptian boundary) should be at 125 Ma. This age
assignment is supported by a recent astrochronology from the 4. Monte Carlo GPTS Construction
Piobbico Core (Central Italy) that estimates an Aptian duration
of 13.42 Myr and an age of 125.45 Ma for the Barremian- [22] The problem of GPTS construction can be stated as
Aptian boundary [Huang et al., 2010]. On the other hand, an inverse problem: find ages for polarity chron boundaries
cyclostratigraphy of the same Piobbico Core and neighboring in a model parameter vector m that simultaneously agree
outcrops by Herbert et al. [1995] gave a duration of 10.1 Myr with radiometric dates (vector d), minimize the variability of
for the interval between the end of the Aptian and the end of spreading rates evaluated from block model distances (vec-
CM0r, implying a Barremian-Aptian boundary at 122 Ma. tor b), and fit polarity chron durations from astrochronology
[19] The alternative age assignments for CM0r are dis- (vector c). We apply a Bayesian formulation of the inverse
cussed at length in the papers cited above. Our preference is problem [e.g., Tarantola and Valette, 1982; Jackson and
for the 121 Ma date used in the CENT94 GPTS because it is Matsu’ura, 1985; Duijndam, 1988; Malinverno, 2002],
difficult to reconcile a 125 Ma age for CM0r, which is the where the solution is a posterior distribution of GPTS vec-
last reversed-polarity chron preceding the Cretaceous Long tors that can be written as
Normal superchron, with the radiometric date of 121.2 Ma pðm j d; b; cÞ ¼ const: pðmÞpðd j mÞ pðb j mÞ pðc j mÞ; ð1Þ
measured in reversed-polarity volcanics by He et al. [2008].
The MHTC12 GPTS obtained here consequently assumes a where p(m) is a prior distribution and p(d ∣ m), p(b ∣ m), and
date of 121.2 0.5 Ma for the middle of CM0r. We rec- p(c ∣ m) are likelihood functions (the normalizing constant
ognize, however, that the age of CM0r is not settled, and we
include in the auxiliary material an alternative GPTS, Table 2. Chron Durations From Astrochronology Studiesa
MHTC12–125, constructed in the same way as MHTC12
Polarity Chron Estimated Duration (Myr) Reference
except for an age of 125 0.5 Ma assigned to the onset of
CM0r.1 CM0r 0.49 0.05 Malinverno et al. [2010]
[20] The second date used here is 155.3 3.4 Ma CM1n 2.28 0.24 Fiet and Gorin [2000]
assigned to the end of CM26n. This age was obtained by CM1r 0.43 0.21 Fiet and Gorin [2000]
CM3n 0.7 0.19 Fiet and Gorin [2000]
1
Auxiliary materials are available in the HTML. doi:10.1029/ CM3r 1.75 0.24 Fiet and Gorin [2000]
2012JB009260. a
Uncertainties are one standard deviation.
5 of 17
B06104 MALINVERNO ET AL.: MHTC12 GEOMAGNETIC POLARITY TIME SCALE B06104
can be ignored). The prior distribution is based on the radiometric dates within their uncertainty and have a broad
expected statistical characteristics of the GPTS, and the distribution of polarity chron durations.
likelihood functions measure how closely a given GPTS in [26] Figure 3b shows the result of accepting or rejecting
the vector m agrees with the data in the vectors d, b, and c. candidate GPTSs generated by the previous random walk
[23] A way to solve the inverse problem is to find an algorithm depending on how much they minimize the global
optimal GPTS as the value of m that maximizes the value of variation of spreading rates. The probability of accepting a
the posterior distribution in equation (1), for example using a candidate depends on the likelihood function of spreading
nonlinear optimization method as done by Huestis and Acton rates p(b ∣ m), defined in Appendix A. An important parameter
[1997]. Optimization, however, does not quantify the uncer- in this likelihood is a measure of how much the spreading rates
tainty that is inherent to the GPTS, e.g., because of spreading in each block are expected to fluctuate. As the lineation sets
rate fluctuations or imprecise dating constraints [Agrinier et al., considered here have markedly different spreading rates, we
1999]. Instead of striving to find a best value for the GPTS, quantify the spreading rate variation with a global coefficient
we apply a Monte Carlo sampling method to obtain a large of variation (CV), the standard deviation of spreading rates
ensemble of possible GPTSs that agree with the data. The mean divided by their average. In practice, it is impossible to obtain
value of this ensemble will give the reference GPTS, and its a GPTS that makes spreading rates constant in all the block
variability will measure the uncertainty associated with the models, and we determined a target value for the CV by run-
GPTS (e.g., quantified by the standard deviation of the sampled ning sampling experiments and observing how much the var-
GPTS values). A similar strategy to obtain a GPTS using a iation of spreading rates can be actually minimized. We found
genetic algorithm has been proposed by Acton and Huestis sampling could reduce the global CV to about 50% at best, and
[1994]. we use this value for the likelihood function in our final sim-
[24] The sampling can be carried out efficiently by Markov ulation. This value is confirmed by an analysis of the CV
chain Monte Carlo (MCMC) methods [Gilks et al., 1996], expected for estimated errors in the block model distances,
which have been used extensively in geophysical inverse asymmetric spreading, and long-term variations in spreading
problems [e.g., Mosegaard and Tarantola, 1995; Sen and rate (Appendix B).
Stoffa, 1995; Malinverno, 2002; Sambridge and Mosegaard, [27] The sampling results in Figure 3b are clearly different
2002; Sambridge et al., 2006]. In these methods, the vector from those in Figure 3a. Rather than fluctuating erratically
m is iteratively modified to perform a random walk in the throughout, the sampled GPTSs in Figure 3b progressively
space of possible model parameters and each step in the ran- adjust to minimize spreading rate variations during a “burn-in”
dom walk only depends on the previous value of m (the phase that lasts a couple of thousand iterations and then display
Markov property). We follow the practical strategy outlined changes much smaller than those in Figure 3a. After the “burn-
by Mosegaard and Tarantola [1995], which can be broken in” phase, the ensemble of GPTSs in Figure 3b will have a
in two parts. First, a random walk process generates a candi- distribution defined by the product p(m) p(d ∣ m) p(b ∣ m) of
date GPTS that agrees with prior information on the distribu- equation (1), the prior times the likelihoods of the radiometric
tion of polarity chron durations and with radiometric dates. dates d and of the spreading rates given by the block model
The candidate GPTS is then accepted or rejected following distances in b.
the Metropolis algorithm [Metropolis et al., 1953; Chib and [28] Finally, Figure 3c shows the result of including the
Greenberg, 1995] depending on the extent it minimizes the likelihood function p(c ∣ m) of polarity chron durations from
total variability of spreading rates and how well it fits polarity astrochronology, defined in Appendix A. Astrochronologies
chron durations from astrochronology. When repeated many constrain the youngest five polarity chrons in the M-sequence
times, this simple recipe samples an ensemble of GPTSs that GPTS (Table 2), and Figure 3c shows that the variation in
will be asymptotically distributed as in the posterior distri- the duration of these chrons is reduced compared to
bution of equation (1) [Mosegaard and Tarantola, 1995]. Figure 3b. After the “burn-in” phase, the ensemble of GPTSs
Details on the implementation of the MCMC algorithm are in Figure 3c will be distributed as in the full posterior dis-
in Appendix A, and in the rest of this section we illustrate tribution of equation (1) and will simultaneously agree with
the Monte Carlo sampling procedure with a simple example. radiometric dates, minimize the global variation of spreading
[25] Figure 3a shows the results of the random walk process rates, and fit polarity chron duration constraints.
that generates candidate GPTSs. The sampling starts from an
arbitrary GPTS constructed by concatenating polarity chron
durations sampled from the prior distribution. We use here a
5. Application to the CENT94 Block Models
lognormal prior distribution for the polarity chron durations [29] To demonstrate how the Monte Carlo sampling method
with a mean and variance taken from the log-chron durations minimizes global spreading rate variations, we applied it to
in previous M-sequence GPTSs (CENT94, GTS2004, and the Hawaii, Japanese, Phoenix, and Keathley block model
TS10). Ages in the GPTS are then rescaled to match dates distances in CENT94. The CENT94 GPTS was built assuming
sampled from the normal distribution that defines the uncer- constant spreading rates in the Hawaiian lineations, and it
tainty of the two radiometric dates used here (for clarity of results in variable spreading rates in the other block models
illustration, these dates are fixed to their mean value in the (CV of 45–56%; Figure 4a). We computed an alternative
example of Figure 3). The random walk modifies iteratively GPTS by running the Monte Carlo sampling method with the
this initial GPTS by choosing one chron at random, sampling polarity block models and absolute age constraints used in
a new duration for this chron from the prior distribution, CENT94. The resulting GPTS allows for some variability of
and rescaling the GPTS to match new values for radiometric spreading rates in the Hawaii block model (CV = 24%;
dates sampled from their distribution. Repeating this random Figure 4b), but the rate fluctuations in the other three block
walk process generates an ensemble of GPTSs that match models are smaller compared to CENT94. In the Japanese and
6 of 17
B06104 MALINVERNO ET AL.: MHTC12 GEOMAGNETIC POLARITY TIME SCALE B06104
Figure 3. Illustrative Monte Carlo sampling of M-sequence GPTSs for different constraints. Each sampled
GPTS is plotted in the usual convention (black is normal polarity, white is reversed). (a) The algorithm
generates candidate GPTSs whose chron durations are distributed as in the prior distribution and that
match two radiometric dates. The two dates (121.2 Ma and 155.3 Ma) are fixed for clarity in this illus-
tration. (b) Candidate GPTSs that minimize spreading rate variations are preferentially accepted. (c) In
addition to minimizing spreading rate variations, the GPTSs are also constrained by the durations
estimated by astrochronology of the five youngest polarity chrons.
7 of 17
B06104 MALINVERNO ET AL.: MHTC12 GEOMAGNETIC POLARITY TIME SCALE B06104
Figure 4. (a) Spreading rates implied by the CENT94 GPTS in the block models of Channell et al.
[1995]. CENT94 was built assuming constant spreading rates in the Hawaiian block model. Percentages
in parentheses indicate the coefficient of variation (CV = standard deviation/average) of spreading rate
in each block model. (b) A time scale obtained by applying the Monte Carlo method described here mini-
mizes spreading rate fluctuations in all the block models.
Keathley block models, the CV of spreading rates decreases (CV = 47%). The alternative GPTS also attenuates some
to 28–30%, a value comparable to that in the Hawaii block prominent spreading rate spikes (e.g., in CM24 of the Japanese
model. As in the results of Channell et al. [1995], the and Keathley block models). The total CV of spreading rates,
Phoenix lineations display the most variable spreading rates which was 43% in CENT94, is reduced to 33%. Age
8 of 17
B06104 MALINVERNO ET AL.: MHTC12 GEOMAGNETIC POLARITY TIME SCALE B06104
Figure 5. (a) MHTC12 is the average M-sequence GPTS in the Monte Carlo simulation. (b) The asso-
ciated uncertainties in the ages of chron boundaries and (c) the chron duration uncertainties. Uncertainties
are one standard deviation.
differences between CENT94 and the alternative GPTS are not generated by Monte Carlo sampling, and the uncertainties in
insignificant, reaching 400 kyr around CM1 and CM10 and ages of chron boundaries and chron durations are the standard
800 kyr around CM22. deviations of the ensemble. The age uncertainties we calculate
[30] This alternative GPTS demonstrates that the Monte are close to those reported by Hinnov and Ogg [2007], who
Carlo method can reduce the variability of spreading rates on estimated a 95% confidence limit (two standard deviations) of
multiple block models from several spreading centers. The rest 4 Myr in the Upper Jurassic interval of the M-sequence
of this paper describes the new MHTC12 GPTS we obtain by (approximately CM29-CM19), decreasing in the Cretaceous
minimizing spreading rate fluctuations for the TS10 polarity to 1 Myr around CM0r. The ages and uncertainties in Table 3
block models. are also included in the auxiliary material, together with results
for an alternative MHTC12–125 obtained in the same way
except for an age of 125 0.5 Ma assigned to the onset of
6. The MHTC12 M-Sequence GPTS
CM0r (see the “Radiometric dates” section). Details on the
and Its Uncertainty sampling procedure are in Appendix A.
[31] MHTC12 is shown in Figure 5 and listed in Table 3. [32] Figure 6 compares the variability in spreading rates
The MHTC12 ages are the average of the ensemble of GPTSs for MHTC12 to that of TS10. TS10 was constructed by
9 of 17
B06104 MALINVERNO ET AL.: MHTC12 GEOMAGNETIC POLARITY TIME SCALE B06104
10 of 17
B06104 MALINVERNO ET AL.: MHTC12 GEOMAGNETIC POLARITY TIME SCALE B06104
Figure 6. Spreading rates implied by the (a) TS10 GPTS and (b) MHTC12. Percentages in parentheses
indicate the coefficient of variation (CV = standard deviation/average) of spreading rate in each block
model. The TS10 GPTS results in lower CVs for the block models that were used to construct the time scale
(Hawaii, Japanese and Phoenix) compared to those of block models from other spreading centers. The
GPTS obtained here, which uses all block models, decreases the total CV of spreading rates (44% versus
51% for TS10) and the differences between the CVs of different block models.
comparison underscores an important point: when different [34] This point is further illustrated in Figure 8, where the
kinds of data are included in the GPTS construction, the ages in the MHTC12 ensemble for the middle of CM0r and
final GPTS will not necessarily match exactly each piece the end of CM26n are compared with the means and standard
of information. deviations of the corresponding radiometric dates (see the
11 of 17
B06104 MALINVERNO ET AL.: MHTC12 GEOMAGNETIC POLARITY TIME SCALE B06104
Figure 7. Chron durations in MHTC12 (gray histograms, average and standard deviation in black lettering)
compared to chron duration estimates from astrochronology used in constructing the time scale (mean and
standard deviation in red symbols and letters). The sampled chron durations do not match exactly the
uncertainties of the astrochronology estimates because the sampled GPTSs also need to minimize the total
variability of spreading rates.
“Radiometric dates” section). While the distributions are the ensemble that is slightly younger and has a smaller uncer-
same for CM0r, they differ somewhat for CM26n, whose tainty than the radiometric date. This explanation has been
radiometric age has a large uncertainty of 3.4 Myr. This tested by running the Monte Carlo simulation without the
difference is slight: more than 95% of the ages sampled in the chron durations from astrochronology; in this case, the
MHTC ensemble for CM26n are within the 95% bounds of sampled age distribution of CM26n matches that of the
the radiometric age (two standard deviations). The reason corresponding radiometric age (actual results not shown).
for this difference is that the polarity chron durations from [35] MHTC12 is constrained only by two radiometric
astrochronology place a weak constraint on the average dates, and Figure 9 compares the sampling results to the other
spreading rates. This constraint propagates to the older end dates listed in Table 1. The comparison in Figure 9 excludes
of the GPTS, resulting in an average age in the sampled the 125 0.5 Ma age assigned to the onset of CM0r by
Figure 8. Ages in MHTC12 (gray histograms, average and standard deviation in black lettering) com-
pared to radiometric dates used in constructing the time scale (mean and standard deviation in red symbols
and letters). The age of CM26n does not match exactly the uncertainty of the corresponding radiometric
date because the chron durations from astrochronology constrain the average spreading rates.
12 of 17
B06104 MALINVERNO ET AL.: MHTC12 GEOMAGNETIC POLARITY TIME SCALE B06104
7. Conclusions
[38] Geomagnetic polarity time scales (GPTSs) are cur-
rently constructed from magnetic anomaly lineations by
assuming nearly uniform spreading rates over time at one or a
few selected spreading centers. A drawback of this procedure
is that spreading rates at other mid-ocean ridges will vary
more erratically, whereas an optimal GPTS should minimize
the variation of spreading rates at as many spreading centers
as possible. This need is underscored by astrochronology-
based GPTSs obtained for the last few Ma, which result in
Figure 9. Comparison between ages in the sampled GPTSs smaller spreading rate fluctuations compared to those implied
and radiometric dates from Table 1. The dashed line marks by GPTSs constructed from marine magnetic anomalies.
equal ages, the gray region shows the uncertainty of MHTC12, [39] We describe a Monte Carlo method that improves
and the error bars denote the uncertainties of the measured GPTS construction by minimizing spreading rate variations
dates (uncertainties are one standard deviation). MHTC12 in a global data set and by also incorporating information on
was constrained only by the two dates shown in black. Radio- chron durations estimated from astrochronology. We apply
metric dates that were not used in constructing the GPTS, the method to construct a new version of the Late Jurassic-
shown in red, generally agree with MHTC12 given the respec- Early Cretaceous M-sequence GPTS based on magnetic
tive uncertainties. lineations from the Western Pacific, North Atlantic, and
Indian Ocean NW of Australia. This MHTC12 GPTS also
accounts for the duration of five polarity chrons (CM0r
GTS2004, which is an alternative interpretation (see the through CM3r) established from astrochronology studies of
“Radiometric dates” section). The sampled GPTS ages are Cretaceous sedimentary sequences. Absolute ages are con-
consistent with all the radiometric dates, given the respective strained by two radiometric dates, 121.2 0.5 Ma for the
uncertainties. We note that the uncertainties in Figure 9 for middle of CM0r and 155.3 3.4 Ma for the end of CM26n.
the dates not used in constraining the GPTS are minimum While MHTC12 is our preferred time scale, the CM0r date is
values, because they represent analytical errors only and do debated, and we include in the auxiliary material an alter-
not include uncertainties due to approximate correlations to native GPTS (MHTC12–125) constructed with an onset of
the polarity reversal record. For example, the 137 Ma date of CM0r at 125 0.5 Ma.
Bralower et al. [1990] has been assigned to the middle of the [40] The Monte Carlo sampling procedure generates a
CM16n-CM16r interval in Figure 9; this date, however, large ensemble of GPTSs that simultaneously agree with
could correspond to any point in this interval, which spans radiometric age constraints, minimize the total variation of
almost 2 Myr. spreading rates, and fit polarity chron durations estimated by
[36] Figure 10 compares MHTC12 ages to previous astrochronology. The average and standard deviation of this
M-sequence GPTSs. MHTC12 is our preferred time scale, ensemble provide a reference time scale and its uncertainty,
but to make a fair comparison Figure 10 shows differences respectively. A key feature of this approach is that it
with MHTC12–125 for GTS2004 and TS10, which assumed accounts for and propagates the uncertainties of the input
an age of 125 Ma for the onset of CM0r. Some of the differ- data, which weigh how much each piece of information
ences with CENT94 and GTS2004 (e.g., in the interval constrains the time scale.
CM10N-CM14) are likely due to the different polarity block [41] MHTC12 reduces the global variability of spreading
models used to build these GPTSs. Age differences with TS10 rates compared to a previous GPTS based on the Western
are smoother, as TS10 was based on the same Pacific block Pacific lineations only, fits radiometric date and astro-
models, but they still can exceed 1 Myr (e.g., around CM20). chronology data within their uncertainties, and is in broad
While MHTC12 is an important update of the M-sequence agreement with other radiometric dates that were not used
GPTS, it is sobering to note that age differences with previous to constrain the solution. An added benefit of this proce-
time scales are within two standard deviations of the uncer- dure is that it is fully automated, so that the time scale can
tainty inherent to MHTC12 (Figure 10). Improving the accu- be easily updated by repeating the Monte Carlo sampling
racy of the M-sequence time scale will require more absolute with future data additions (magnetic block model distances,
age constraints and chron duration estimates from astro- radiometric dates, or polarity chron duration estimates from
chronology. As new data become available, the time scale can astrochronology).
13 of 17
B06104 MALINVERNO ET AL.: MHTC12 GEOMAGNETIC POLARITY TIME SCALE B06104
Figure 10. Comparison between ages in MHTC12 and ages in (a) CENT94, (b) GTS2004, and (c) TS10.
Age differences on the vertical axes are ages in MHTC12 minus ages in the other time scale. The dashed
line marks zero age difference and the gray region shows the uncertainty in MHTC12 (one standard
deviation).
Appendix A: Markov Chain Monte Carlo rates implied by the GPTS in vector m is smaller. The
Algorithm likelihood is
A1. Likelihood of Block Model Distance Data pðb∣mÞ ¼ const:
(P " #)
[42] This likelihood function minimizes the total variation B
Ni XB
1 X
Ni
2
of spreading rates: for block model distances in a vector b, exp Pi¼1
B uij mi Dtij ;
i¼1 Ti i¼1 2ðvmi Þ2 j¼1
the likelihood is higher if the total variation of spreading
ðA1Þ
14 of 17
B06104 MALINVERNO ET AL.: MHTC12 GEOMAGNETIC POLARITY TIME SCALE B06104
where B is total number of block models, Ni is number of [46] 2. Accept the candidate GPTS with a probability
blocks in i-th block model, Ti is total duration of i-th block equal to the ratio of the likelihoods as in
model, as in
pðb∣mcand Þ pðc∣mcand Þ
a ¼ min 1; : ðA6Þ
X
Ni pðb∣mÞ pðc∣mÞ
Ti ¼ Dtij; ðA2Þ
j¼1
In practice, generate a random number r from a uniform
distribution between 0 and 1. If r < a, set m = mcand;
where Dtij is duration of j-th chron of the i-th block model otherwise, leave m unchanged. Note that a depends only
(from GPTS in m), uij is spreading rate of j-th block in i-th
on the ratio of the likelihoods, so that the constants in
block model, as in equations (A1) and (A5) are irrelevant.
Dxij [47] 3. Every 100 iterations, save the current value of m.
uij ¼ ; ðA3Þ [48] 4. Repeat from step 1.
Dtij
[49] The ensemble used to obtain the final GPTS (Figure 5,
where Dxij is width of j-th block in i-th block model (from Table 3, and supplementary material) was obtained by com-
distances in b), v is coefficient of variation (CV) of bining 500 independent chains started from different random
spreading rates (see the text and Appendix B), and mi is initial GPTSs. This strategy ensures that the starting point does
average spreading rate in i-th block model, weighted by not affect the solution. The final variability of the sampled
chron duration as in GPTSs is much less than the differences between the starting
GPTSs, indicating that the chains were not trapped in sec-
ondary modes and that they converged to sample the global
1X Ni
1X Ni
mode of the posterior distribution. The histograms of sampled
mi ¼ uij Dtij ¼ Dxij : ðA4Þ
Ti j¼1 Ti j¼1 chron ages and durations (Figures 7 and 8) do not show any
sign of multimodality. Each chain was run for 50,000 itera-
The likelihood in (A1) is weighted by polarity chron dura- tions, and to safely discard the burn-in interval, the first 10,000
tion, so that a given difference from the average spreading iterations in each chain were ignored. The sampling algorithm
rate in a longer chron decreases the likelihood more than the saves only one GPTS every 100 iterations (see step 4 above).
same difference in a shorter chron. This is done to reduce the redundancy of the samples: con-
secutive samples are very similar because only one chron is
changed for each iteration. The interval between saved sam-
A2. Likelihood of Chron Duration Data ples is close to the number of chrons in the M-sequence (102).
[43] This likelihood function is higher if the polarity chron The final ensemble consisted of 200,000 GPTSs. The con-
durations implied by the GPTS in vector m are closer to the vergence of the simulation was assessed by comparing the
chron duration estimates in a vector c. The likelihood is statistics of the two halves of the ensemble. The average ages
of the polarity chron boundaries in each half were very close,
" 2 # with differences that were at most 1% of the standard devi-
XC
cj Dtj
pðc j mÞ ¼ const: exp ; ðA5Þ ation. Again, this close comparison confirms that the inde-
j¼1
2s2j pendent chains were sampling the global mode of the posterior
distribution.
where C is total number of chron durations from astro-
chronology, cj is mean of j-th chron duration (from astro- Appendix B: CV of Spreading Rates
chronology data in c), sj is standard deviation of j-th chron
duration, and Dtj is duration of j-th chron (from GPTS in m). [50] The likelihood of block model distance data, which
quantifies the total variation of spreading rates (equation (A1)),
needs a value for v, the coefficient of variation (CV) of
A3. MCMC Sampling Algorithm
spreading rates. To estimate an expected CV in the set of
[44] The MCMC algorithm used to generate samples from block model distances used here, we consider three reasons
the posterior distribution of GPTS vectors m in equation (1) for measured spreading rates to depart from a constant value.
starts from a random initial GPTS constructed in two steps. [51] Errors in measured model block widths. The spreading
First, a random GPTS vector m is constructed by concatenat- rate calculated in a given polarity block equals the block width
ing the desired number of chron durations sampled from their divided by the duration of the polarity chron. For a given chron
prior distribution (see Section 4). The chron ages in m are then duration, an error in the measured block width directly trans-
rescaled to match values for the absolute dates that were lates into an error in the calculated spreading rate, so that the
sampled from the distributions that define the uncertainty of CV of the spreading rates equals the CV of the block widths.
radiometric dates. Sampling then proceeds as follows: We use here the polarity block distance data published by
[45] 1. Generate a candidate GPTS in a vector mcand by (a) TS10, who estimated block model uncertainties by comparing
picking at random a chron in m and perturbing its duration multiple magnetic anomaly profiles. TS10 report standard
(these perturbations are designed to sample the prior distri- deviations of the polarity block widths in the Western Pacific
bution of chron durations) and (b) rescaling the ages in mcand as 2–36 km (average 11 km) for Hawaii, 1–36 km (14 km) for
to match dates sampled from the distribution of radiometric the Japanese, and 2–33 km (9 km) for the Phoenix model. The
dates. average standard deviations are large compared to the average
15 of 17
B06104 MALINVERNO ET AL.: MHTC12 GEOMAGNETIC POLARITY TIME SCALE B06104
block widths (19.4, 19.7, and 11.2 km in the Hawaii, Japanese, Baksi, A. K. (1994), Concordant sea-floor spreading rates obtained from
and Phoenix block models, respectively), and these averages geochronology, astrochronology and space geodesy, Geophys. Res. Lett.,
21, 133–136, doi:10.1029/93GL03534.
are likely biased by a few large standard deviations. We Berger, A., M. F. Loutre, and J. Laskar (1992), Stability of the astronomical
obtained an alternative estimate of the standard deviations of frequencies over the Earth’s history for paleoclimate studies, Science,
the block widths from the standard deviations of the block 255, 560–566, doi:10.1126/science.255.5044.560.
Billups, K., H. Pälike, J. E. T. Channell, J. C. Zachos, and N. J. Shackleton
model distances listed by TS10. To deduce the errors in block (2004), Astronomic calibration of the Late Oligocene through Early
widths from those of the distances, consider a simple case Miocene geomagnetic polarity time scales, Earth Planet. Sci. Lett.,
where the blocks are referred to a common origin so that the 224, 33–44, doi:10.1016/j.epsl.2004.05.004.
distance to a polarity block boundary is the sum of all the Bralower, T. J., K. R. Ludwig, J. D. Obradovich, and D. L. Jones (1990),
Berriasian (Early Cretaceous) radiometric ages from the Grindstone
preceding block widths. Assuming that errors in block widths Creek Section, Sacramento Valley, California, Earth Planet. Sci. Lett.,
are uncorrelated, the variances of the distances are expected to 98, 62–73, doi:10.1016/0012-821X(90)90088-F.
grow from the origin, and the average increase in variance per Cande, S. C., and D. V. Kent (1992), A new geomagnetic polarity time
scale for the Late Cretaceous and Cenozoic, J. Geophys. Res., 97(B10),
block is the average variance of block widths. In the TS10 13,917–13,951, doi:10.1029/92JB01202.
data, the expected increase of variance is not observed in the Cande, S. C., and D. V. Kent (1995), Revised calibration of the geomagnetic
Western Pacific because several intervals with different ori- polarity time scale for the Late Cretaceous and Cenozoic, J. Geophys.
gins were used to construct composite block models. In the Res., 100(B4), 6093–6095, doi:10.1029/94JB03098.
Channell, J. E. T., E. Erba, M. Nakanishi, and K. Tamaki (1995), Late
Atlantic Keathley lineations, however, there is a clear linear Jurassic-Early Cretaceous time scales and oceanic magnetic anomaly
increase in the variance of the distances that averages 31.6 km2 block models, in Geochronology, Timescales, and Stratigraphic Correla-
per block, giving a standard deviation for the block widths of tion, edited by W. A. Berggren et al., pp. 51–63, Soc. for Sediment.
Geol., Tulsa, Okla.
5.6 km. When divided by the average block width of 12 km, Channell, J. E. T., C. E. Casellato, G. Muttoni, and E. Erba (2010), Magne-
this gives a CV of 0.47 for the block widths and for the tostratigraphy, nannofossil stratigraphy and apparent polar wander for
spreading rates estimated in this block model. Using the same Adria-Africa in the Jurassic-Cretaceous boundary interval, Palaeogeogr.
Palaeoclimatol. Palaeoecol., 293, 51–75, doi:10.1016/j.palaeo.2010.04.
standard deviation for the block widths in the Western Pacific 030.
Hawaii lineations (average block width 19.4 km) gives a CV Chib, S., and E. Greenberg (1995), Understanding the Metropolis-Hastings
of 0.29 for the spreading rates. The average of the Atlantic and algorithm, Am. Stat., 49, 327–335.
Pacific CVs is 0.38. Duijndam, A. J. W. (1988), Bayesian estimation in seismic inversion, part I:
Principles, Geophys. Prospect., 36, 878–898, doi:10.1111/j.1365-2478.
[52] Asymmetric spreading. In block models constructed 1988.tb02198.x.
from a single ridge flank, asymmetric spreading during a Fiet, N., and G. Gorin (2000), Lithological expression of Milankovitch
polarity chron will result in a variation of measured spread- cyclicity in carbonate-dominated, pelagic, Barremian deposits in central
Italy, Cretaceous Res., 21, 457–467, doi:10.1006/cres.2000.0220.
ing rate. The Keathley and Canary block models are from Fischer, A. G., and W. Schwarzacher (1984), Cretaceous bedding rhythms
conjugate ridge flanks, and averaging their distances would under orbital control?, in Milankovitch and Climate: Understanding the
remove the effect of asymmetric spreading. As all other Response to Astronomical Forcing, edited by A. Berger et al., pp. 163–175,
block models in our data set are from a single ridge flank, D. Reidel, Dordrecht, Netherlands.
Fischer, A. G., P. L. de Boer, and I. Premoli Silva (1990), Cyclostratigraphy,
however, we could not eliminate this cause of spreading rate in retaceous Resources, Events and Rhythms, edited by R. N. Ginsburg
variability and opted not to average the Keathley and Canary and B. Beaudoin, pp. 139–172, Kluwer Acad., Dordrecht, Netherlands.
block model distances. A histogram of spreading asymme- Gee, J. S., and D. V. Kent (2007), Source of oceanic magnetic anomalies and
the geomagnetic polarity timescale, in Treatise on Geophysics, vol. 5,
tries from the global survey of Müller et al. [2008] gives a Geomagnetism, edited by G. Schubert, pp. 455–507, Elsevier, Amsterdam,
CV of 10/50 = 0.2 for spreading rate fluctuations due to doi:10.1016/B978-044452748-6.00097-3.
asymmetries. Gilks, W. R., S. Richardson, and D. J. Spiegelhalter (Eds.) (1996), Markov
Chain Monte Carlo in Practice, Chapman and Hall, London.
[53] Long-term variations in spreading rate. The best- Gordon, R. G. (1993), Orbital dates and steady rates, Nature, 364, 760–761,
studied C-sequence magnetic anomaly corridor in the South doi:10.1038/364760a0.
Atlantic shows long-term variations in spreading rate [Cande Gradstein, F. M., J. G. Ogg, and A. G. Smith (Eds.) (2004), A Geologic Time
and Kent, 1995]. This long-term variation, as summarized in Scale 2004, Cambridge Univ. Press, Cambridge, U. K., doi:10.4095/
215638.
Ogg and Smith [2004, Figure 5.3], is consistent with a CV of Handschumacher, D. W., W. W. Sager, T. W. C. Hilde, and D. R. Bracey
5/22.5 = 0.22. (1988), Pre-Cretaceous tectonic evolution of the Pacific plate and exten-
[54] The total CV of spreading rates due to these three sion of the geomagnetic polarity reversal timescale with implications for
the origin of the Jurassic “quiet zone”, Tectonophysics, 155, 365–380,
causes is the square root of the sum of the variances and doi:10.1016/0040-1951(88)90275-2.
comes to 0.48, or about 50%. Hays, J. D., J. Imbrie, and N. J. Shackleton (1976), Variations in the Earth’s
orbit: Pacemaker of the Ice Ages, Science, 194, 1121–1132, doi:10.1126/
[55] Acknowledgments. A.M. and J.H. were supported by NSF grant science.194.4270.1121.
OCE 09–26306, M.T. was supported by a Woods Hole Oceanographic He, H., Y. Pan, L. Tauxe, H. Qin, and R. Zhu (2008), Toward age determi-
Institution postdoctoral scholarship, and J.E.T.C. was supported by NSF nation of the M0r (Barremian-Aptian boundary) of the Early Cretaceous,
grant OCE 09–60999. Thanks to two anonymous reviewers for comments Phys. Earth Planet. Inter., 169, 41–48, doi:10.1016/j.pepi.2008.07.014.
that improved the manuscript. Figure 1 was constructed with the GMT soft- Heirtzler, J. R., G. O. Dickson, E. M. Herron, W. C. Pitman III, and X.
ware of Wessel and Smith [1991]. This is LDEO contribution number 7551. LePichon (1968), Marine magnetic anomalies, geomagnetic field rever-
sals, and motions of the ocean floor and continents, J. Geophys. Res.,
73, 2119–2136, doi:10.1029/JB073i006p02119.
Herbert, T. D. (1992), Paleomagnetic calibration of Milankovitch cyclicity
References in Lower Cretaceous sediments, Earth Planet. Sci. Lett., 112, 15–28,
doi:10.1016/0012-821X(92)90003-E.
Acton, G. D., and S. P. Huestis (1994), Genetic algorithm optimization for Herbert, T. D., and A. G. Fischer (1986), Milankovitch climatic origin
seafloor spreading rates and geomagnetic polarity time scales, Eos Trans. of mid-Cretaceous black shale rhythms in central Italy, Nature, 321,
AGU, 75(44), Fall Meet. Suppl., Abstract GP42B-14. 739–743, doi:10.1038/321739a0.
Agrinier, P., Y. Gallet, and E. Lewin (1999), On the age calibration of the geo- Herbert, T. D., I. Premoli Silva, E. Erba, and A. G. Fischer (1995), Orbital
magnetic polarity timescale, Geophys. J. Int., 137, 81–90, doi:10.1046/ chronology of Cretaceous-Paleocene marine sediments, in Geochronology,
j.1365-246x.1999.00755.x.
16 of 17
B06104 MALINVERNO ET AL.: MHTC12 GEOMAGNETIC POLARITY TIME SCALE B06104
Timescales, and Stratigraphic Correlation, edited by W. A. Berggren et al., Meyers, S. R., and B. B. Sageman (2007), Quantification of deep-time
pp. 81–93, Soc. for Sediment. Geol., Tulsa, Okla. orbital forcing by average spectral misfit, Am. J. Sci., 307, 773–792,
Hilgen, F. J. (1991), Extension of the astronomically calibrated (polarity) doi:10.2475/05.2007.01.
time scale to the Miocene/Pliocene boundary, Earth Planet. Sci. Lett., Mosegaard, K., and A. Tarantola (1995), Monte Carlo sampling of solutions
107, 349–368, doi:10.1016/0012-821X(91)90082-S. to inverse problems, J. Geophys. Res., 100, 12,431–12,447, doi:10.1029/
Hinnov, L. A. (2004), Earth’s orbital parameters and cycle stratigraphy, in A 94JB03097.
Geologic Time Scale 2004, edited by F. M. Gradstein, J. G. Ogg, and Müller, R. D., M. Sdrolias, C. Gaina, and W. R. Roest (2008), Age, spread-
A. G. Smith, pp. 55–62, Cambridge Univ. Press, Cambridge, U. K. ing rates, and spreading asymmetry of the world’s ocean crust, Geochem.
Hinnov, L. A., and J. G. Ogg (2007), Cyclostratigraphy and the astronom- Geophys. Geosyst., 9, Q04006, doi:10.1029/2007GC001743.
ical time scale, Stratigraphy, 4, 239–251. Ogg, J. G., and A. G. Smith (2004), The geomagnetic polarity time scale, in
Hinnov, L. A., and J. Park (1998), Detection of astronomical cycles in the A Geologic Time Scale 2004, edited by F. M. Gradstein, J. G. Ogg, and
stratigraphic record by frequency modulation (FM) analysis, J. Sediment. A. G. Smith, pp. 63–86, Cambridge Univ. Press, Cambridge, U. K.
Res., 68, 524–539. Ogg, J. G., F. P. Agterberg, and F. M. Gradstein (2004), The Cretaceous
Huang, C., L. A. Hinnov, A. G. Fischer, A. Grippo, and T. D. Herbert period, in A Geologic Time Scale 2004, edited by F. M. Gradstein, J. G.
(2010), Astronomical tuning of the Aptian Stage from Italian reference Ogg, and A. G. Smith, pp. 409–440, Cambridge Univ. Press, Cambridge,
sections, Geology, 38, 899–902, doi:10.1130/G31177.1. U. K.
Huestis, S. P., and G. D. Acton (1997), On the construction of geomagnetic Opdyke, N. D., and J. E. T. Channell (1996), Magnetic Stratigraphy,
timescales from non-prejudicial treatment of magnetic anomaly data from Academic, San Diego, Calif.
multiple ridges, Geophys. J. Int., 129, 176–182, doi:10.1111/j.1365- Pälike, H., R. D. Norris, J. O. Herrle, P. A. Wilson, H. K. Coxall, C. H.
246X.1997.tb00947.x. Lear, N. J. Shackleton, A. K. Tripati, and B. S. Wade (2006), The heart-
Jackson, D. D., and M. Matsu’ura (1985), A Bayesian approach to nonlinear beat of the Oligocene climate system, Science, 314, 1894–1898,
inversion, J. Geophys. Res., 90, 581–591, doi:10.1029/JB090iB01p00581. doi:10.1126/science.1133822.
Klitgord, K. D., and H. Schouten (1986), Plate kinematics of the central Park, J., and T. D. Herbert (1987), Hunting for paleoclimatic periodicities
Atlantic, in The Geology of North America, vol. M, The Western North in a geologic time series with an uncertain time scale, J. Geophys.
Atlantic Region, edited by P. R. Vogt and B. E. Tucholke, pp. 351–378, Res., 92(B13), 14,027–14,040, doi:10.1029/JB092iB13p14027.
Geol. Soc. of Am., Boulder, Colo. Pringle, M. S., and R. A. Duncan (1995), Radiometric ages of basement
Krijgsman, W., F. J. Hilgen, I. Raffi, F. J. Sierro, and D. S. Wilson (1999), lavas recovered at Loen, Wodejebato, MIT, and Takuto-Daisan guyots,
Chronology, causes and progression of the Messinian salinity crisis, northwestern Pacific Ocean, Proc. Ocean Drill. Program, Sci. Results,
Nature, 400, 652–655, doi:10.1038/23231. 144, 3–26.
Langereis, C. G., A. A. M. van Hoof, and F. J. Hilgen (1994), Steadying the Renne, P. R., C. C. Swisher, A. L. Deino, D. B. Karner, T. L. Owens, and
rates, Nature, 369, 615, doi:10.1038/369615a0. D. J. DePaolo (1998), Intercalibration of standards, absolute ages and
Larson, R. L., and T. W. C. Hilde (1975), A revised time scale of magnetic uncertainties in 40Ar/39Ar dating, Chem. Geol., 145, 117–152,
reversals for the Early Cretaceous and Late Jurassic, J. Geophys. Res., 80, doi:10.1016/S0009-2541(97)00159-9.
2586–2594, doi:10.1029/JB080i017p02586. Sager, W. W., L. G. Fullerton, R. T. Buffler, and D. W. Handschumacher
Larson, R. L., and W. C. Pitman III (1972), World-wide correlation of (1992), Argo Abyssal Plain magnetic lneations revisited: Implications
Mesozoic magnetic anomalies, and its implications, Geol. Soc. Am. Bull., for the onset of seafloor spreading and tectonic evolution of the eastern
83, 3645–3662, doi:10.1130/0016-7606(1972)83[3645:WCOMMA]2.0. Indian Ocean, Proc. Ocean Drill Program, Sci. Results, 123, 659–669.
CO;2. Sager, W. W., M. A. Weiss, M. A. Tivey, and H. P. Johnson (1998), Geo-
Laskar, J. (1999), The limits of Earth orbital calculations for geological magnetic polarity reversal model of deep-tow profiles from the Pacific
time-scale use, Philos. Trans. R. Soc. A, 357, 1735–1759, doi:10.1098/ Jurassic “Quiet Zone”, J. Geophys. Res., 103, 5269–5286, doi:10.1029/
rsta.1999.0399. 97JB03404.
Laskar, J., P. Robutel, F. Joutel, M. Gastineau, A. C. M. Correia, and Sambridge, M., and K. Mosegaard (2002), Monte Carlo methods in geo-
B. Levrard (2004), A long-term numerical solution for the insolation physical inverse problems, Rev. Geophys., 40(3), 1009, doi:10.1029/
quantities of the Earth, Astron. Astrophys., 428, 261–285, doi:10.1051/ 2000RG000089.
0004-6361:20041335. Sambridge, M., K. Gallagher, A. Jackson, and P. Rickwood (2006), Trans-
Laskar, J., A. Fienga, M. Gastineau, and H. Manche (2011), La2010: A new dimensional inverse problems, model comparison and the evidence,
orbital solution for the long-term motion of the Earth, Astron. Astrophys., Geophys. J. Int., 167, 528–542, doi:10.1111/j.1365-246X.2006.03155.x.
532, A89, doi:10.1051/0004-6361/201116836. Schwarzacher, W. (1993), Cyclostratigraphy and the Milankovitch Theory,
Lourens, L. J., F. J. Hilgen, N. J. Shackleton, J. Laskar, and D. S. Wilson Elsevier, Amsterdam.
(2004), The Neogene period, in A Geologic Time Scale 2004, edited by Sen, M. K., and P. L. Stoffa (1995), Global Optimization Methods in
F. M. Gradstein, J. G. Ogg, and A. G. Smith, pp. 409–440, Cambridge Geophysical Inversion, Elsevier, Amsterdam.
Univ. Press, Cambridge, U. K. Shackleton, N. J., A. Berger, and W. A. Peltier (1990), An alternative astro-
Ludden, J. N. (1992), Radiometric age determination for basement from nomical calibration of the lower Pleistocene time scale based on ODP Site
Sites 765 and 766, Argo Abyssal Plain and northwestern Australian mar- 677, Trans. R. Soc. Edinburgh Earth Sci., 81, 251–261, doi:10.1017/
gin, Proc. Ocean Drill. Program, Sci. Results, 123, 3–26. S0263593300020782.
Mahoney, J. J., R. A. Duncan, M. L. G. Tejada, W. W. Sager, and T. J. Bralower Tarantola, A., and B. Valette (1982), Inverse problems = quest for informa-
(2005), Jurassic-Cretaceous boundary age and mid-ocean-ridge-type mantle tion, J. Geophys., 50, 159–170.
source for Shatsky Rise, Geology, 33, 185–188, doi:10.1130/G21378.1. Tominaga, M., and W. W. Sager (2010), Revised Pacific M-anomaly geomag-
Malinverno, A. (2002), Parsimonious Bayesian Markov chain Monte Carlo netic polarity time scale, Geophys. J. Int., 182, 203–232, doi:10.1111/
inversion in a nonlinear geophysical problem, Geophys. J. Int., 151, 675– j.1365-246X.2010.04619.x.
688, doi:10.1046/j.1365-246X.2002.01847.x. Tominaga, M., W. W. Sager, M. A. Tivey, and S.-M. Lee (2008), Deep-tow
Malinverno, A., E. Erba, and T. D. Herbert (2010), Orbital tuning as an magnetic anomaly study of the Pacific Jurassic Quiet Zone and implications
inverse problem: Chronology of the early Aptian oceanic anoxic event for the geomagnetic polarity reversal time scale and geomagnetic field
1a (Selli Level) in the Cismon APTICORE, Paleoceanography, 25, behavior, J. Geophys. Res., 113, B07110, doi:10.1029/2007JB005527.
PA2203, doi:10.1029/2009PA001769. Wan, X., R. Scott, W. Chen, L. Gao, and Y. Zhang (2011), Early Cretaceous
Martinson, D. G., N. G. Pisias, J. D. Hays, J. Imbrie, T. C. Moore Jr., and stratigraphy and SHRIMP U-Pb age constrain the Valanginian-Hauterivian
N. J. Shackleton (1987), Age dating and the orbital theory of the Ice boundary in southern Tibet, Lethaia, 44, 231–244, doi:10.1111/j.1502-
Ages: Development of a high-resolution 0 to 300,000-year chronostrati- 3931.2010.00238.x.
graphy, Quat. Res., 27, 1–29, doi:10.1016/0033-5894(87)90046-9. Wessel, P., and W. H. F. Smith (1991), Free software helps map and display
Metropolis, N., A. W. Rosenbluth, M. N. Rosenbluth, A. H. Teller, and data, Eos Trans. AGU, 72, 441, doi:10.1029/90EO00319.
E. Teller (1953), Equation of state calculations by fast computing Wilson, D. S. (1993), Confirmation of the astronomical calibration of the
machines, J. Chem. Phys., 21, 1087–1092, doi:10.1063/1.1699114. magnetic polarity timescale from sea-floor spreading rates, Nature, 364,
Meyers, S. R. (2008), Resolving Milankovitchian controversies: The Trias- 788–790, doi:10.1038/364788a0.
sic Latemar Limestone and the Eocene Green River Formation, Geology,
36, 319–322, doi:10.1130/G24423A.1.
17 of 17