Isotope stratigraphy has become the method of choice for investigating both
past ocean temperatures and global ice volume. Lisiecki and Raymo (2005)
published a stacked record of 57 globally distributed benthic
Sedimentary archives retrieved by ocean drilling since 1968 by the Deep Sea Drilling Program (DSDP, 1968–1983), the Ocean Drilling Program (ODP, 1983–2003), the Integrated Ocean Drilling Program (IODP, 2003–2013) and the International Ocean Discovery Program (IODP, since 2013) provide key records needed to better understand processes and interactions of the Earth system. Over almost 5 decades of coring, ocean drilling samples and data have contributed significantly to major breakthroughs in our understanding of Earth history – including such basic tenets as seafloor spreading, a detailed history of reversals of the Earth's magnetic field and evolution/extinction of marine species. Included in this list is the advancement of stable isotope stratigraphy and the recognition of the critical part played by variations in the Earth's orbital parameters in climate history. Sites drilled during ODP Leg 154 on the Ceara Rise have played a significant role in creating age models for the Neogene based on astrochronology.
Stable isotope stratigraphy has become the method of choice for investigating
both past ocean temperatures and global ice volume. When global ice volumes
are large, such as times of vast continental ice sheets, the world oceans
become enriched in
The LR04 stack is a significant contribution for having demonstrated the
global semi-synchrony of the overall behavior of the
There are 21 records included in LR04 that extend to ages older than 3 Ma included in LR04, and only 14 that have data older than 4 Ma. As the numbers in the stack shrink, the importance of having well-spliced records grows. A number of records used in LR04 contain problematic succession with respect to their composite record. Site 982, for example, is one of the high-resolution sites that extends beyond 3 Ma (Venz et al., 1999; Venz and Hodell, 2002), and has been used subsequently to transfer age models to other isotope records (Drury et al., 2017). However, there is controversy over the composite record of 982 as well as the age model (Lawrence et al., 2013; Khélifi et al., 2012; Bickert et al., 2004).
For the interval 1.7–5.3 Ma the LR04 stack depends heavily on the spliced
records from Leg 138 – the S95 benthic composite stack (Shackleton et al.,
1995). It was noted in Lisiecki and Raymo (2005) that for marine isotope
stages (MIS) M2 and MG2 at 3.35 Ma there is a mismatch of data and a
potential coring or splicing problem in Site 846. Even so, Site 846 was used
for the initial alignment in LR04 from 2.7 to 5.3 Ma along with Site 849
(1.7–3.6 Ma) and Site 999 (3.3–5.3 Ma). Any problem in the spliced
records of the sites used for initial alignment will propagate through the
stack if not compensated for by a large number of additional sites. Thus we
might expect a greater possibility of erroneous correlation in older less
repeated parts of the stack, particularly where the amplitude of the
In order to provide a precise age model the LR04 stack was tuned to a
nonlinear ice volume model forced by insolation (65
Here we revisit data collected during, and subsequent to, ODP Leg 154
(Fig. 1). The LR04 stack includes benthic stable isotope data from ODP Leg
154 sites 925, 927, 928, and 929. Site 927 was used for initial alignment
from 0 to 1.4 Ma in LR04. Site 926 is also considered a primary site for
timescale constructions for 0–15 Ma and is independent of LR04. In this
study, we use newly developed software to check and improve the composite
records of Leg 154. We then stretch and squeeze data outside the splice, use
core images to correlate all sites to the Site 926 depth scale, orbitally
tune the core images, and compare the age model with the LR04 stack for the
past 5 Myr. The new software
system greatly facilitates the construction of benthic
The location of ODP Leg 154 sites.
The proliferation and diversity of the data collected both during and after
ocean drilling cruises can at times be somewhat overwhelming for the
individual scientist. Data are now freely available through online databases
maintained by the ocean drilling infrastructure for cruise results (e.g.,
LIMS, JANUS), by national efforts (e.g., NGDC) or
community efforts (e.g., PANGAEA). However, a unified and consistent
system for integrating disparate data streams such as micropaleontology,
physical properties, core images, geochemistry, and borehole logging has not
been widely available. In this section we describe a system that we have
developed over several years to work with ocean drilling data and images
(CODD – Code for Ocean Drilling Data). CODD takes advantage of the versatile
graphical user interface and analytical functions contained in the
IGOR™ graphing and analysis program
commercially available from Wavemetrics, Inc. One of the great advantages of
a modern analysis program like IGOR™ paired
with new computers and fast processors is the ability to use images as data.
Rather than a static picture of a core or section, images are scaled and
plotted along with traditional data versus depth or age. Core images may be
squeezed, stretched, subsampled, and concatenated, allowing for great
versatility. The CODD set of ocean drilling macros for
IGOR™ and a user guide are freely available
at
The heart of the CODD data structure is the coring matrix – a 3 layered array in which the top layer contains the original depth to the top of each section (m b.s.f. – meters below seafloor) sorted by core (rows) and sections (columns). The middle layer contains the length of the sections and the third layer the composite depth (m c.d. – meters composite depth). Sample depths are calculated by referencing the proper layer and coordinate by core and section and then adding the sample interval. The reverse process of returning the core, section, and interval designation of a given sample depth is accommodated by comparing it to the section top depth plus the section length to find where the sample originated.
Creating a composite core image from a core table image.
A standardized naming convention is essential to efficient processing of multiple and diverse data streams. In CODD data are assigned three-part names: hole, technique and information, separated by underscores. Thus, gamma-ray attenuation depths are U925A_GRA_MBSF and U925A_GRA_MCD with data as U925A_GRA_GRA. Core, section, interval and age are similarly named. Isotope data might be U925A_Iso_d18O and U925A_Iso_d13C. While the hole and technique designations must be identical for a single data set, the information may be anything the user desires, including new data like ratios created from existing information. IGOR™ records data processing steps and the use of a standard naming convention allows users to repeat processing for different data by simply replacing one hole or technique with another in the recorded steps. It also simplifies the development of automation macros. This is essential for processing large amounts of data from multiple drill-holes and drill sites – especially when changes to composite records (splices) are needed.
Ever since IODP Leg 200, core section images have been captured by line scanners as discrete files which are easily loaded into analysis programs with little or no preparation. However, the only access to core images from the first approximately 200 ocean drilling cruises are through digitized photographs of entire cores laid out on a table in parallel sections (Fig. 2a). CODD includes a module for cutting core section images from core table photo images, correcting them for uneven lighting, scaling them to m b.s.f. (meters below seafloor) and combining them into a single core image (Fig. 2b) through a series of simple steps. In general, the outer 5 % of each section image is excluded to minimize friction effects of coring that tend to bend horizontal layers. In practice it takes between 1 and 2 min to go from loading a core table photo to producing a scaled composite core image. The visualization and impact of the scaled composite are very much different from the core table photo and of much greater value during data analysis. The use of scaled composite core images has proven to be particularly effective in creating site splices or for the checking of existing splices.
Lighting correction is a necessary step when using images cut from core table photos because the light source used for the original photos was colocated with the camera above the center of the core table, resulting in the center of the picture being brighter than the edges (Schaaf and Thurow, 1994; Nederbragt and Thurow, 2001, 2005). This effect is illustrated by profiles of lightness from HSL (hue, saturation, lightness) representations of section images plotted together (Fig. 2a inset). For these sections the variability in the intensity of lightness, excepting some spikes representing darker layers, is around 50 units of lightness (out of a full scale of 0–255). The difference from the center to the ends of the best-fit line to the profiles is approximately 25 lightness units, so uneven lighting has a significant effect on the section images. When the core table photos are viewed, the observer's eyes and mind make a correction and the uneven lighting seems subtle, but we have found that when stringing section images together to make a composite core image the 1.5 m long lighter/darker cycles are readily apparent. As many ocean drilling sediment cores vary in lightness as a function of carbonate and/or biogenic silica content (e.g., Balsam et al., 1999), lighting cycles in core images degrade the usefulness of core color or lightness profiles as proxies for other properties of interest or for spectral analysis. Thus, CODD processing of core table photos includes a step which fits a line to the lightness profiles and then applies a “flattening” filter which brightens the section images away from the center according to the fit. While not perfect, the process removes most of the 1.5 m color cyclicity (Fig. 2b). There is also lighting variation across the core box images that can produce a 9/10 m cycle in the spliced composite images. It appears to be somewhat more diffuse than the along-core section variation and has not hindered the present work. We are developing a process to correct for lighting variation in the entire core box image prior to cutting the individual section images. This may also allow us to remove the color cast present in many of the older core box images, such as the purplish hue seen in Fig. 2a.
In the same manner that sections may be strung together to make a composite
core image, extracted splice sections of core images from different holes can
be merged into scaled spliced site image (Fig. 3a). Splicing is a two-step process, the
first of which involves offsetting the m b.s.f. depth for individual cores
to a composite depth by aligning features in data collected from multiple
holes. It is worth noting here that it is rare that all features in
individual cores from different holes align – coring disturbance (e.g.,
extension or compression at the top and bottom of piston cores; see Ruddiman
et al., 1987, for an in depth discussion) or natural variability means that
while one feature may align, another is offset (e.g., Lisiecki and Herbert,
2007). The individual setting the splice (the correlator) makes a decision as
to which feature to align based on overall considerations of the splicing
process. Once the core offsets are set, the correlator chooses tie points
between holes to produce as complete a sedimentary record as possible while
avoiding any possible duplication. In the past this has been done using data
profiles of properties measured on whole round core sections – primarily
density from gamma-ray attenuation (GRA) and magnetic susceptibility (MS) as
well as reflectance spectrophotometer intensity (RSC) on split sections. This
can prove to be tricky when using data that are replete with similar cycles.
Cycle skipping or doubling is a constant source of potential error and the
inclusion of images in the process helps greatly. While checking splices or
splicing cores and choosing tie points we used the same criteria as typically
used by the shipboard stratigraphic correlator for (I)ODP expeditions. The
splice should contain no coring gaps and disturbed sections are avoided.
Where possible we avoided using the top and bottom
Core 925B-2H was not used for the Site 925 splice, and while there is good alignment between the core image and data and the spliced image and data at 13–14 m c.d., shallower portions of the core are not well aligned with the splice. Yellow numbers indicate tie points used to stretch the image and data so that they are in better agreement with the splice. Choice of tie points is cursor driven and stretching can be recalculated in real time.
An example from Ceara Rise Site 927 demonstrates image utility while examining an existing splice. A 10 m long section of images and data is presented in Fig. 3. Poor agreement between offset data from all three holes of Site 927 occurs around 50 m c.d., immediately below a splice tie in the published splice for the site (Fig. 3a). The images show poor agreement between the light and dark bands in cores 927C-05H and 927B-06H. A better solution is obtained by reducing the offset of 927B-06H by 1.6 m to align the peak in RSC seen around 50.2 m c.d. in 927C-05H with a similar peak at 51.8 m c.d. in 927B-06H (Fig. 3b, c). Fortunately, because the core images are depth scaled, CODD allows us to shift and re-splice both core images and all other data sets using a simple algorithm. The resultant shift shows better agreement between images and data from both holes. Significantly, the shift illustrated removes one 40 kyr obliquity cycle from the isotope record (Bickert et al., 1997) and will alter a tuned age model accordingly.
Spliced images and MS data from ODP sites 926 and 927. There have been small adjustments to the published splices for each site. Site 927 data and image are plotted versus the Site 927 depth scale on the bottom of each graph and versus the Site 926 depth scale at the top. Green numbers indicate tie points between the sites used to stretch the Site 927 image and data.
Traditionally, once the splice has been set, subsequent samples are taken and measurements made only from the core material included in the splice. While three or more holes are often cored at sites devoted to paleoceanographic studies, the volume of samples available within a splice is equivalent to a single hole. Moreover, since archival halves of each core are reserved for later sampling, it is often difficult to obtain new samples along a heavily sampled section of the splice. More material is available from sections of cores not included in the splice, but, as mentioned above, the process of aligning and offsetting cores from adjacent holes by matching features is imperfect due to coring effects and natural variability (e.g., Lisiecki and Herbert, 2007; Wilkens et al., 2009). Misalignment of off-splice features may add significant noise when in-splice and out-of-splice data are combined. In order to align features from sections of core not included in the splice, it is necessary to stretch or squeeze images and data outside the splice. Magnetic susceptibility data have been stretched from the off-splice data to the splice in Fig. 4. Using CODD, sets of tie points between off-splice data and the splice for each hole (yellow numbers in Fig. 4) are selected using cursors. Stretched data and images are updated in real time. The tie points allow investigators to interpolate out-of-splice m c.d. depths to their equivalent levels in the splice.
The ability to squeeze and stretch data and images has a second useful application. Sites drilled in the same general area of the ocean, such as those on the Ceara Rise, often share many physical features in data such as density, magnetic susceptibility, or color in their sediment columns. In a manner similar to the process of stretching and squeezing off-splice data to the splice, CODD employs a cursor-driven routine to stretch data and images from different sites to a single common depth scale using similar features. The segment of the stretch of Site 927 to Site 926 between tie points 60 and 80 is illustrated in Fig. 5. In total, 428 pairs of tie points were identified while matching the upper 304 m c.d. of Site 927 to the upper 285 m c.d. of Site 926. Additional constraints such as paleomagnetic reversals and biostratigraphic events may be included, helping to guide the correlation. In practice a user views multiple data types and images simultaneously and tie points selected from one data set are mapped to all others at the same time.
Once data and images from the individual sites have been tied to a common depth scale, the final CODD processing step is to set everything to a single age model. We used the age models of Bickert et al. (1997) and Tiedemann and Franz (1997), adjusted for our splice corrections and updated to Laskar et al. (2004), to compare age-scaled images and data from the various Ceara Rise sites. An example comparing sites 926 and 927 is presented in Fig. 6. Comparison of the composite images is remarkable for the fact that individual sedimentary layers that represent sometimes less than 10 kyr are readily identifiable between sites. This suggests that in areas where the sediment has enough color variation highly targeted samples may be collected that represent precisely the same event at multiple sites.
Laskar et al. (2004) orbital calculation compared to the Site 926
composite image and MS data.
MS data and the composite image of Site 926 are compared with orbital
calculations using Laskar et al. (2004) in Fig. 7. The orbital curve was
calculated using 100 % of the eccentricity (
Benthic oxygen isotope data from all Ceara Rise sites compared with
one another and a smoothed composite of all data compared to LR04.
We checked the entire splices of sites 925, 926, 927, 928 and 929 for the
last 5 Myr. Most of the
changes in the published splice tables were minor, although several, such as
the one illustrated in Fig. 3, were large enough to affect age models based
on orbital tuning. Data from samples outside of the revised splices were
aligned with the splice based on stretching and squeezing of the
out-of-splice data. Mapping pairs to convert depths outside of the splice to
the composite depth are provided in the Supplement. For the interval spanning
0 to 5 Ma we compiled 5533 benthic
Agreement amongst the different Ceara Rise sites is good in terms of the shapes of the curves, while there is a spread in absolute values. This is likely due to the water depths at the different sites, which ranged from 3040 m at Site 925 to 4355 m at Site 929. Offsets in benthic oxygen isotope data between Site 925 and Site 929 in some intervals (e.g., 3.6 to 4.5 Ma) have been suggested to indicate a relatively warmer and saltier North Atlantic Deep Water than today (Billups et al., 1997).
Detail from Fig. 8 comparing individual holes to one another and a smoothed composite to LR04 for the intervals 1.5 to 2.0 and 4.0 to 5.0 Ma. For better illustration we plotted the initial alignment target records of the LR04 stack. For the 1.5 to 2.0 Ma interval these are the records from ODP sites 677 and 849; for the interval 4.0 to 5.0 Ma these are the records from ODP sites 846 and 999. Some records have been shifted as indicated in the figure for better comparison of the data with each other. Note the differences between LR04 and the Ceara Rise average at 1.80–1.85 Ma, although the initial alignment targets are more similar to the Ceara Rise smooth. Also note the difference between 4.0 and 4.5 Ma. The Site 999 record is from a single hole and the splice of the Site 846 record might be erroneous. The age model for the Ceara Rise is very robust in this interval (see Fig. 10) pointing to potential inconsistencies in the age model construction of the Site 846 and Site 999 records.
The overall agreement between the Ceara Rise smoothed composite oxygen isotope curve and the LR04 global compilation is generally quite good, although there is a definite difference in absolute values with the Ceara Rise data exhibiting consistently lower values of about 0.2 ‰ than LR04 (Fig. S1 in the Supplement). The 0.2 ‰ offset is well within the potential regional differences of up to 0.3 ‰ cited by Lisiecki and Raymo (2005). The consistency of the difference over the entire 5 Myr scope of this study is remarkable given the regional mix of data used for LR04.
Detail from CODD tuning of Site 926 magnetic susceptibility and core
images to insolation. Panel
While the agreement between Ceara Rise and LR04 oxygen isotope data is good,
there are discrepancies in some intervals. The two curves are out of sync
between 1.80 and 1.90 Ma with LR04 exhibiting two maxima, whereas Ceara Rise
contains only one. As this is close to a point where the LR04 stack switched
from Site 677 (0–2 Ma) and Site 927 (0–1.7 Ma) to Site 849
(1.7–3.6 Ma), misalignments in the stack between single sites with the
original spliced records could have led to a mismatch here. Tuning for Site
926 in this interval is robust and does not allow a shift that could
accommodate the mismatch. Hence, the interval from 1.80 and 1.90 Ma in the
LR04 stack has to be revised. Even larger differences are seen between 4.0
and 4.5 Ma (Fig. 9). Data from Site 929 have been shifted
Accessing uncertainty in the age model is difficult and cannot be discussed in this paper as it would require extensive testing. However, in Zeeden et al. (2013, 2014) this is already done with regards to the uncertainty in the target curve. The outstanding match of sedimentary pattern and insolation calculations, keeping in mind that the Laskar et al. (2004) model is based on a relatively short time of observational data, gives confidence that the error for the Miocene is less than a single precession cycle. Due to the excellent match in patterns we think the main error lies in the accuracy of the target (precession and obliquity). The error in precession maxima and minima positions will be only relevant for times older than 5 Ma (see Lourens et al., 2004), as already discussed in the Zeeden et al. (2013, 2014) papers.
Independent tuning of Site 926 images and physical property data to the Laskar 2004 orbital solution and integration of available benthic stable isotope data from the Ceara Rise provides a new regional reference section for the equatorial Atlantic covering the last 5 million years. Comparing the CODD-based new stack from the Ceara Rise to the LR04 stack reveals overall very good agreement, suggesting that most of the LR04 stack is robust for the interval from 0 to 4 Ma. Disagreement in the interval from 1.8 to 1.9 Ma (Fig. 9) points to uncertainties in the records of sites 677 and 849. The record of Site 677 (Shackleton et al., 1990) has a gap in the composite around this time interval at 85 m c.d. Our unpublished re-examination of the Mix et al. (1995) Site 849 age model suggests that it might be affected by issues in the composite record revolving around core 849C 5H at around 52 m c.d. Construction of an equatorial Pacific stack, presently underway, should resolve the issue.
A comparison of LR04 (Red) to Ceara Rise (grey and black (smooth)) to obliquity and insolation from Laskar et al. (2004). Note that the interval 4.0 and 4.5 Ma exhibits poorly defined obliquity cycles leaving insolation dominated by precession.
A comparison of LR04 and Ceara Rise (smooth) to Site 925 and Site 929 benthic isotope data. LR04 assignment of variability in the interval from 4.0 to 4.5 Ma to precession peaks may have resulted in the mismatch with the Ceara Rise stack.
The differences between LR04 and the Ceara Rise average between 4 and 4.5 Ma
reveals a more complex matter that questions assumptions made in LR04. The
tuning in Site 926 (Fig. 10) in this interval is robust and cannot be
changed. The match between the precession-dominated insolation curve and the
dark/light pattern shown in the composite site image is excellent. To match
the LR04 and the Ceara Rise isotope stacks, the Ceara Rise stack needs to be
shifted by 21 kyr to older ages between 4.1 and 4.3 Ma – which is not
possible without changing the phase relation between insolation and the
dark/light pattern of the Ceara Rise sediments. The LR04 stack is basically
tuned to obliquity in this interval with lighter
Further study of splices and age models used in the data contributing to LR04 will be needed before these discrepancies can be fully resolved. Such clarification is a necessary step in the ongoing effort to create a global correlation of isotope and other data that can be resolved at the isotopic stage level. Such examination of other areas of the oceans will also aid in the development of regional isotope curves to compare with our findings for the Ceara Rise. The CODD approach is a useful tool for extending oxygen isotope reference records into the Miocene and beyond. Combining multiple records from several sites drilled in an oceanic region is greatly facilitated by CODD and helps to form a regional stratigraphic framework. Stacked records from different regions, such as the equatorial Pacific, are urgently needed to test and verify the completeness of each record as gaps can occur on a regional scale. Establishing high-resolution age models on a regional scale is key to understanding paleoceanographic changes on orbital timescales for the entire Cenozoic.
We have demonstrated a new system for capturing core images as data using newly developed CODD software. The ability to transform core table photos and line scans of core sections into data as depth- or age-scaled core images has helped greatly in the task of revising published splices for Ceara Rise sediments cored during ODP Leg 154. Comparison of the revised data with the LR04 global oxygen isotope stack reveals that there are sections of the stack that are not well resolved. Further study of data contributing to LR04 will lead to a clarification of the misfits we have found as well as establishing other regional isotope offsets from a global stack. The CODD software package thus can play a key role in the construction of a new generation of the benthic isotope stack and surely will be very helpful in extending the stack into the Miocene. The next important step will be to form a more robust and accurately tuned initial signal used to form the benthic isotope stack.
Data are available at
The authors declare that they have no conflict of interest.
Development of CODD was partially supported by post-cruise funds from U.S. Science Support for Roy H. Wilkens. Financial support for this research was also provided by the Deutsche Forschungsgemeinschaft (DFG) to Thomas Westerhold and Anna J. Drury. Edited by: Arne Winguth Reviewed by: Torsten Bickert and Christian Zeeden