The CIP initiative

The study of astronomical climate forcing and the application of cyclostratigraphy have experienced a spectacular growth over the last decades.  In the field of cyclostratigraphy a broad range in methodological approaches exist. However, comparative studies between the different approaches are lacking. Different cases demand different approaches, but with the growing importance of the field questions arise about reproducibility, uncertainties and standardization of results. The radioisotopic dating community, in particular, has done far-reaching efforts to improve reproducibility and intercomparison of radioisotopic dates and their errors. To satisfy this need in cyclostratigraphy, we initiated a comparable framework for the cyclostratigraphic community. The aims were to investigate and quantify reproducibility of, and uncertainties related to, cyclostratigraphic studies and to provide a platform to discuss the merits and pitfalls of different methodologies, and their applicability.

The Cyclostratigraphy Intercomparison Project (CIP) assessed the robustness of cyclostratigraphic methods using an experimental design of three artificial cyclostratigraphic case studies with known input parameters. Each case study is designed to address specific challenges that are relevant to cyclostratigraphy. Case 1 represents an offshore research vessel environment, as only a drill-core photo and the approximate position of a late Miocene stage boundary are available for analysis. In Case 2, the Pleistocene proxy record displays clear nonlinear cyclical patterns and the interpretation is complicated by the presence of a hiatus. Case 3 represents a Late Devonian proxy record with a low signal-to-noise ratio with no specific theoretical astronomical solution available for this age. Each case was analyzed by a test group of 17-20 participants, with varying experience levels, methodological preferences and dedicated analysis time. During the CIP 2018 meeting in Brussels, Belgium, the ensuing analyses and discussion demonstrated that most participants did not arrive at a perfect solution, which may be partly explained by the limited amount of time spent on the exercises (~4.5 hours per case). However, in all three cases, the median solution of all submitted analyses accurately approached the correct result and several participants obtained the exact correct answers. Interestingly, systematically better performances were obtained for cases that represented the data type and stratigraphic age that were closest to the individual participants’ experience. This experiment demonstrates that cyclostratigraphy is a powerful tool for deciphering time in sedimentary successions and, importantly, that it is a trainable skill. Finally, we emphasize the importance of an integrated stratigraphic approach and provide flexible guidelines on what good practices in cyclostratigraphy should include. Our case studies provide valuable insight into current common practices in cyclostratigraphy, their potential merits and pitfalls. Our work does not provide a quantitative measure of reliability and uncertainty of cyclostratigraphy, but rather constitutes a starting point for further discussions on how to move the maturing field of cyclostratigraphy forward.

A paper was published and is available under this link.

The Task:

During the summer of 2027 IODP Expedition 666 “Prelude on the Messian salinity crisis” successfully recovered a complete and continuous core. The core has a total length of 26.38 m and exists of pelagic carbonate-rich sediments which look cyclic. Correlations based on seismic profiles and biostratigraphy suggest that the core contains the Tortonian-Messinian boundary around 15-20 m core depth and does not contain any Pliocene material.

As the only shipboard cyclostratigrapher you are asked to look at the color record of the full core and provide following output to the expedition stratigraphic correlator:

Q1: A best estimation on the total duration (in kyr) of the recovered core based on cyclostratigraphy. What is the uncertainty on this estimation? Do you suspect the presence of any hiatus(es)?

Q2a: A floating age model for following core depths: 4.0, 7.5, 15.0, 22.5 and 25.0 m. [0m = 0 kyr]. What are the uncertainties on these?

Q2b: Optional: an absolute age model for the same core depths (Tuning). Uncertainty?

Q3: Stratigraphic positions (in m) of potential 2.4 Myr eccentricity cycle extreme(s)? Uncertainty?

The design of Case 1 allows for a comparison of scenarios that are more naturally prone to cyclostratigraphic methodologies that are based on approaches that are purely visual, purely numerical or a combination of both. The following key papers provide some more background:

• Sapropel cyclostratigraphy in the Mediterranean (Hilgen et al., 1991)
• Cyclostratigraphy Messian salinity crisis (Krijgsman et al., 1999)
• Numerical solution insolation and astronomical parameters (Laskar et al., 2004)

SPOILER ALERT: If you want to analyse Case 1 without apriori knowledge on how the signal was created, stop reading and watching now. We strongly recommend that you try to solve the case (see the task above) before.

This video explains step-by-step how we converted an insolation signal into an artificial color record reflecting a lithological succession. We used the MATLAB coding language for this case, the code used in this case study is available for download here. If you are not familiar with MATLAB, there are many potential sources of information available, like for example this video.

This video discusses a few possible ways of analyzing Case 1, based on the approaches of some of the CIP participants.

During the IODP Expedition 999 “Quaternary High Latitudes”, a core showing quasi-cyclic variability in proxy data was recovered. The topmost (and thus most recent) sediment is missing for unclear reasons. It is also not clear how much sediment and time is missing. The core has a total length of 10.00 m, and exhibits pattern which seem cyclic. You have a quickly measured record of the magnetic susceptibility (signal origin unclear, but somehow related to paleoclimate/paleoenvironment). Investigation of the biostratigraphy suggests the core to represent sediments with a maximal age of 2 Ma and a minimal age of 0.5 Ma, spanning a maximum time of 1.5 Myr. As cyclostratigrapher you are asked to look at the proxy record of the core, and provide following information:

Question/Q1: A best estimation on the total duration (in kyr) of the recovered core based on cyclostratigraphy. What is the uncertainty on this estimation? Do you suspect the presence of any hiatus(es)?

Q2a: A floating age model containing for following core positions: 2.0, 4.0, 6.0, 8.0 m. What are the uncertainties on these?

Q2b: Optional: an absolute age model for the same core positions (Tuning). Uncertainty?

Q3: Stratigraphic positions (in m) of potential 405-kyr eccentricity cycle extreme(s)? Uncertainty?

The information, image above and this data file were provided.

The design of Case 2 was tailored to include several challenges for cyclostratigraphy, namely: Nonlinear reaction to insolation, changing sedimentation rate, a hiatus and noise. The rationale is also outlined in the video below.

This video explains the creation of the signal in detail in the video below. It uses the programming language R, and the Rstudio software. Creating the signal includes using an insolation curve (from Laskar et al. 2004), and applying the Imbrie & Imbrie (1980) ice model. The R script of how this signal was created is available as Supplement to the CIP paper, and a version creating nicer plots is provided here.

This video discusses several possibilities of tackling this case, partly based on the approaches of some of the CIP participants. The case was best resolved through tuning to a target, as often done in the Quaternary and Neogene.

This video presents how the participant performed in Case 2.

The Task:

A team of mo­ti­va­ted mas­ter stu­dents ge­ne­ra­ted a high-re­so­lu­ti­on (15-cm spaced) pro­xy re­cord of a 394.5 m thick Late De­vo­ni­an sec­tion in Aus­tra­lia, ent­i­re­ly Fa­men­ni­an in age. The sec­tion was de­po­si­ted in an ex­ter­nal car­bo­na­te ramp set­ting. The co­nodont bio­stra­ti­gra­phy of this sec­tion is known from the li­te­ra­tu­re, and was con­struc­ted ba­sed on 40 co­nodont sam­ples at 10-me­ter in­ter­vals throughout the ent­i­re sec­tion.

Your mas­ter stu­dents took off to other ad­ven­tures, and you are left with this ex­cep­tio­nal­ly high- re­so­lu­ti­on pro­xy re­cord for the Fa­men­ni­an. Can you di­still a cy­clo­s­tra­ti­gra­phic sto­ry for your next pa­per?

Q1: A best esti­ma­ti­on on the to­tal du­ra­ti­on (in kyr) of the re­co­ve­r­ed core ba­sed on cy­clo­s­tra­ti­gra­phy. What is the un­cer­tain­ty on this esti­ma­ti­on? Do you sus­pect the pre­sence of any hia­tus(es)?

Q2: A floa­ting age mo­del for fol­lo­wing stra­ti­gra­phic le­vels: 145.0, 155.0, 175.0, 185.0, 350.0 m. Out­put is as­ked in age, youn­ger than the base of the sec­tion. [0 m = 0 kyr]. What are the un­cer­tain­ties on the­se?

Q3: Stra­ti­gra­phic po­si­ti­ons (in m) of po­ten­ti­al 2.4 Myr ec­centri­ci­ty cy­cle ex­tre­me(s)? Un­cer­tain­ty?

SPOILER ALERT: If you want to ana­ly­se Case 3 wi­thout a-prio­ri know­ledge on how the si­gnal was dis­tor­ted. Strop rea­ding and watching now, and start ana­ly­sing the si­gnal.

Case-3-Signal.csv (60,6 KB)

The vi­deo be­low ex­plains how the cli­ma­te-si­mu­la­tor out­put was trans­fer­red from the time-do­main into the depth-do­main, ap­p­ly­ing an oscil­la­ting se­di­men­ta­ti­on rate evo­lu­ti­on, and how the si­gnal was sub­se­quent­ly bu­ried in red and whi­te noi­se.

This vi­deo is using the R lan­gua­ge wi­t­hin Rstu­dio: If you are an ab­so­lu­te be­gin­ner, you can learn how to in­stall R and RStu­dio as well as the ba­sics of sta­tis­ti­cal com­pu­ting in R by watching this YouTube vi­deo.

This vi­deo ex­plains how one could use re­la­tive­ly sim­ple and "tra­di­tio­nal" time-se­ries ana­ly­sis tech­ni­ques to come clo­se to the cor­rect so­lu­ti­on of Case 3. This ana­ly­sis so­le­ly re­li­es on spec­tral ana­ly­sis, band­pass fil­te­ring, li­ne­ar in­ter­po­la­ti­on and am­pli­tu­de de­mo­du­la­ti­on using the Hil­bert trans­form. 

The ori­gi­nal Case 3 file that the CIP par­ti­ci­pants re­cei­ved can be down­loa­ded here.

The do­cu­ment be­low con­ta­ins all code that is show­ca­sed in the vi­deo.

This vi­deo ex­plains how one could use the ti­me­Opt and etime­Opt func­tions in as­tro­chron to come clo­se to the cor­rect so­lu­ti­on of Case 3.