Marine Sediment Age–Depth Models and Hiatuses

The age of a sediment layer is a model, not a label

A marine sediment core is a physical sequence of material deposited through time. Researchers can measure dates at a limited number of levels, but a climate proxy, fossil sample or chemical observation often occurs between those dated horizons. An age–depth model translates the scattered chronological evidence into plausible calendar ages throughout the core. This guide explains how to read that translation without confusing a measured age, an interpolated estimate, a depositional assumption or a probability interval. It also shows why erosion, mixing and changing accumulation rates can invalidate a deceptively smooth curve.

Did you know that two age–depth curves can pass through the same dated horizons and still give materially different ages between them? One curve may assume constant deposition within intervals. Another may allow accumulation rates to change gradually according to prior information and observed sedimentology. A third may recognize an interval of missing time. None of these differences is a computational failure by itself. They arise because a finite number of dates does not uniquely determine everything that happened between samples. The model must tell the reader what it assumes.

An interpretable chronology has at least four layers: the observed sample and depth, the independent dating method, the conversion to a consistent calendar timescale, and the model of how those dates relate to stratigraphic order. Each layer contributes uncertainty. If an article reports that an event happened at 9,500 years BP, ask which measured samples surround it, how their ages were calibrated, how the depth was established, and whether the quoted figure is a median, an estimate from a linear interpolation or a posterior distribution from a more elaborate model.

What the core depth coordinate really means

Core depth may be measured as centimeters below seafloor, meters below seafloor, composite depth after multiple holes were spliced or another expedition-specific scale. Core expansion, loss, disturbance and breaks can make those numbers differ. A sample described as occurring at 70 cm may refer to the center of a two-centimeter sampling interval rather than a mathematical point. An age–depth model that expects one coordinate per observation should retain the sampling interval or its relevant uncertainty instead of pretending every value has zero thickness.

Suppose two adjacent sections overlap in a composite splice. Moving from an original section depth to composite depth is an explicit transformation. If the scientific article hides that transformation, later users may accidentally apply ages from one depth scale to measurements recorded on another. This can create apparent leads and lags between climate indicators even though all numerical values are internally consistent. State the depth datum in figure labels, datasets and narrative prose. Do not merge ‘mbsf’, ‘mcd’ and laboratory sample positions into an unlabeled depth column.

Physical compaction is another issue. Centimeter thickness is not a constant mass of sediment. A core may contain water-rich mud near the surface, more compact layers below and discrete coarse event beds. Depth-based accumulation rates express net thickness per unit time, whereas mass accumulation rates require dry mass per area. A model can use depth as the independent coordinate while additional density data are needed to translate its chronology into a flux. This distinction becomes especially important when comparing carbon burial rates among sites.

Observed ages: several clocks with different meanings

Marine cores may be constrained by radiocarbon determinations, volcanic ash correlations, paleomagnetic events, biostratigraphic datums, radioisotope profiles and other stratigraphic information. Each is a different kind of evidence. A shell radiocarbon determination may require marine calibration and an appropriate local reservoir correction before it becomes a calendar-age distribution. A tephra tie point relies on eruption identification, independent eruption age and the assessment of whether the ash was deposited promptly. A fossil datum may refer to a biological event whose occurrence time varies across oceans.

One can be tempted to put every number in a table and fit a line. That is risky if a conventional radiocarbon age of 8,000 years is mixed with a calibrated tephra age of 8,000 cal BP, because identical-looking numbers occupy different time scales. Likewise, a lead-210 age of 40 years before core collection does not share the 1950 BP origin by default. A defensible data table records time unit, calendar origin, dating method, reference curve, event definition and uncertainty alongside each numeric estimate.

Dating the material is not always identical to dating its final position in sediment. An older shell may be eroded, carried elsewhere and redeposited into a younger bed. A turbidite can deliver particles of many ages in a single event. A volcanic glass shard may be chemically associated with an eruption long before its final burial. An age–depth model should distinguish the age of a measured object from the depositional age targeted by the chronology. An old transported specimen is not direct evidence that the enclosing bed is equally old.

Time conventions and why 1950 matters

Calibrated years BP, often abbreviated cal BP, conventionally use AD 1950 as the zero reference. A geological age of 6,000 cal BP is not a count of 6,000 years backward from the calendar year when a core was recovered. Conventional radiocarbon years BP also use the 1950 reference but are derived from a radiocarbon measurement and are not themselves calendar ages. A report that drops the prefixes can confuse two numbers that represent different transformations from evidence to time.

Suppose a recent radionuclide profile suggests a deposit is about 30 years older than collection in 2026. Its approximate calendar placement is 1996. Converting that result into a cal BP-style year requires a clearly stated calendar convention, and for such a recent record an explicit CE date may be easier to read. Conversely, for a late-Pleistocene marine core, calibrated BP may be the appropriate plotting standard. The most useful rule is to disclose the origin rather than treating the word ‘present’ as self-explanatory.

Age signs also matter on graphs. Sediment generally becomes older downward in an undisturbed core, so age values typically increase with depth when expressed in BP. But plotting software may reverse a y axis, or put depth on the vertical and age on the horizontal. A steep visual line may represent rapid accumulation under one axis orientation and slow accumulation under another. Always read axis labels and compute rate in physical units before interpreting a slope from appearance alone.

The simplest construction: piecewise linear ages

An elementary age–depth model joins successive age controls with straight lines. Within each interval it assumes a constant mean depth accumulation rate, though that assumption need not be credible over very long periods. If a dated point at 40 cm is 5,000 cal BP and another at 100 cm is 14,000 cal BP, the depth difference is 60 cm and the age difference is 9,000 years. A linear interpolation gives t(z)=5,000+(z−40)×9,000/60 for depths z in centimeters between those two points.

At 70 cm, exactly halfway between 40 and 100 cm, the interpolated age is 5,000+30×150=9,500 cal BP. The interval-average depth accumulation rate is 60/9,000≈0.00667 cm per year, or 0.0667 mm per year. The age gradient is its reciprocal: 150 years per centimeter. These numbers are pedagogical calculations; they do not establish that the marine core actually accumulated at a constant rate through every century. The model can predict an age between tie points without observing the true depositional history there.

If a shallow anchor at zero centimeters is set to age zero and the 40-centimeter point is 5,000 cal BP, its average accumulation rate is 40/5,000=0.008 cm per year, or 0.08 mm per year. The two piecewise sections have different average rates. This creates a corner in the age–depth line at 40 cm. Such a corner is mathematically allowed but may not represent an actual abrupt oceanographic transition. The analyst should decide whether the data justify that structure or whether a smoother rate model is more appropriate.

A basic uncertainty exercise

Imagine the 40-centimeter age is 5,000 ± 200 cal years and the 100-centimeter age is 14,000 ± 500 cal years, with approximately independent symmetric uncertainties for this arithmetic demonstration. The midpoint interpolation at 70 centimeters gives central age 9,500. With equal weights of 0.5 on the two anchors, a simple propagated standard uncertainty is sqrt((0.5×200)²+(0.5×500)²), about 269 years. This number summarizes only the stated input uncertainty under the linear interpolation assumption.

Real age–depth uncertainty is usually wider or more structured because the accumulation model itself is uncertain. We may have chosen the wrong shape between dated horizons, ignored a hiatus, assumed the dates are independent when they share a calibration correction, or lost depth precision in core recovery. Adding an illustrative separate standard model uncertainty of 200 years in quadrature would make the toy combined uncertainty sqrt(269²+200²), about 335 years, but such an adjustment is defensible only when the extra component has a justified probabilistic meaning.

The main lesson is not to memorize 269 or 335. It is to identify what an uncertainty range includes. A plot that shades a 95-percent chronology interval may represent a posterior distribution across many model realizations, not a simple twice-standard-error band. A narrow ribbon generated by a rigid model can be misleading if alternative depositional histories were never considered. Reader confidence should increase when uncertainty sources are explained, not merely when the curve looks smooth.

Why a smooth line may be a strong assumption

A long marine sequence may contain intervals of slow background deposition, high productivity, rapid terrigenous input, carbonate dissolution and glacially controlled sediment redistribution. There is no universal rule that sedimentation rate stays constant between dated horizons. Choosing a smooth curve expresses a preference for gradual changes; choosing straight segments expresses local constant averages; choosing flexible accumulation processes permits more complex histories. The right model is the one whose assumptions suit the stratigraphy and available evidence.

A spline can pass precisely through known points but behave implausibly between them, including producing negative accumulation or oscillatory rates if unconstrained. Some interpolation methods respect monotonic older-with-depth behavior; others do not without additional restrictions. A curve that locally makes deeper material younger may be mathematically possible but geologically inconsistent with an undisturbed continuous sequence. The problem must be resolved through model constraints or investigation of the anomalous data, not hidden by formatting.

Strong assumptions do have legitimate uses when transparently chosen. If a teaching task asks for the average rate between two well-dated horizons, a linear calculation may be the simplest correct answer. If a research claim concerns millennial shifts in ocean circulation, more flexible chronological modeling and sensitivity testing may be necessary. Model complexity should match the precision of available evidence and the question asked. More parameters do not automatically make an age curve truer.

Bayesian chronology and prior knowledge

Bayesian age–depth models combine evidence from dated samples with assumptions about plausible accumulation histories. These prior assumptions may concern the typical rate and how strongly rates tend to persist between adjacent depth intervals. The resulting posterior represents modelled plausible ages given the dating evidence and the stated priors. A posterior credible interval is therefore conditional on a complete statistical model. It is not an unconditional guarantee about the true seabed’s history.

One influential approach, often called Bacon, was described by Blaauw and Christen in 2011 using a gamma autoregressive model for accumulation rates. The concept is to break the core into sections and give neighboring rates a structured relationship instead of drawing an arbitrary smooth curve through all dates. The priors matter: a model that expects slow, persistent deposition can interpret sparse ages differently from one allowing abrupt rate shifts. Reading a model figure responsibly means looking for the calibration data, section design and accumulation priors.

Bayesian methods can also include explicit outlier treatments for inconsistent dates, but that does not eliminate the need to diagnose why a sample is anomalous. An outlier probability is not geological evidence of reworking by itself. The model may downweight a date that is perfectly measured but incompatible with a faulty assumption elsewhere. Good authors examine the physical sample and sediment context, report sensitivity to priors and present plausible alternatives when the depositional interpretation is unsettled.

Monte Carlo ensembles and the credible ribbon

Age–depth software may simulate thousands of plausible curves, each representing one sample from a statistical distribution consistent with the inputs and priors. At a given depth, the set of modeled ages can be summarized by a median and percentile intervals. Connecting these summaries creates the familiar age–depth line with a shaded envelope. The shading does not mean every curve inside it has equal probability, nor does the outer boundary necessarily represent a unique physical history.

Uncertainty may change sharply with depth. Near several reliable tie points, plausible curves can cluster. In a long interval with no independent dates, they may fan out, especially when accumulation rate is poorly constrained. But overconfident prior assumptions can artificially suppress that widening. The reader should look at the distribution of acceptable rate histories and conduct sensitivity analyses. A flat narrow band in an underdated section deserves scrutiny rather than immediate trust.

Model outputs are also dependent across depth. If a curve is shifted older by an uncertain reservoir correction, neighboring depths may shift together. Treating every sampled depth’s median age as an independent observation exaggerates information. When estimating the lag between two proxy events, uncertainty in their relative timing may be smaller or larger than the individual absolute-age uncertainties, depending on shared errors. Ideally compute such lags from the same ensemble of chronological realizations, retaining their dependence.

The role of radiocarbon calibration

A conventional radiocarbon determination enters a marine sediment chronology with its own calibration problem. Marine organisms often require a marine calibration curve such as Marine20, with appropriate local reservoir consideration for the setting and time period. An atmospheric plant fragment may instead need an atmospheric curve. The calibrated result can be asymmetric or multimodal. Reducing it to a single normal date before modeling discards information that may matter to the posterior age–depth relationship.

Age models differ in how they accommodate calibration likelihoods. Some software directly integrates radiocarbon determinations with their curve and corrections; other workflows calibrate first and pass numerical summaries into a later modeling stage. Both can be used carefully, but the second risks losing probability structure or shared uncertainty if compressed too aggressively. The author should state which operation was performed and how a multimodal calibration distribution was represented.

A particularly important shared assumption is local marine ΔR. If multiple dates from one core use the same regional reservoir estimate, its uncertainty is not independent for each specimen. Treating each as though it had a fresh unrelated ΔR error can distort chronology confidence. Likewise, temporal variations in reservoir conditions may make a single offset inappropriate across glacial and interglacial intervals. Calendar curves are only as credible as the carbon reservoir assumptions passed into them.

Tephra, paleomagnetism and fossil markers

A chemically matched volcanic ash horizon can provide an independent age tie point when its eruption date and depositional equivalence are secure. A magnetic polarity reversal or excursion may supply a stratigraphic marker when identification and timing are well established. Fossil assemblages can suggest biozones or ecological changes, but some fossil first and last appearances are diachronous between basins. These controls should not all be assigned identical uncertainty structures merely because they appear in the same age table.

Suppose a tephra at 142 cm has an independently dated eruption age, while a radiocarbon sample at 120 cm has a calibrated distribution. The two observations may be nearly independent in their underlying dating methods, but the interpretation of tephra as primary fallout still matters. If the ash was reworked, its eruption age is an upper bound on the possible depositional timing in the sense that the bed cannot be older than the eruption material it contains, subject to age conventions. A fixed age at 142 cm would overstate the evidence.

Chronostratigraphic matching by proxy shape can create circularity if the same proxy alignment is later used to claim the two environmental records changed at the same time. When a curve is tuned to an orbital target, a separate age-model inference should not present the resulting periodicity as wholly independent confirmation. Document which observations are independent clocks, which are cross-core correlations and which are tuning constraints. A chronology can be useful while still having correlated or model-dependent features.

Hiatuses: time that is not represented by sediment

A hiatus is a gap in depositional record caused by nondeposition, erosion or other processes that leave time unrepresented in the recovered stratigraphy. On an age–depth graph, a hiatus may produce a jump in age across a very small depth interval. For example, sediment immediately above an erosional surface might be 5,000 cal BP while sediment immediately below is 8,000 cal BP. The three thousand years between them are not preserved as a continuous three-thousand-year stack of mud at that horizon.

Trying to fit a perfectly smooth, continuous accumulation curve across the gap spreads the missing time into sediment layers that did not form during it. That can falsely slow the inferred deposition rate, shift proxy changes and create spurious event durations. Core photographs, erosion surfaces, changes in fossil assemblages or missing magnetostratigraphic intervals may indicate the problem. Age anomalies alone can suggest a hiatus but do not uniquely diagnose one. The geological evidence should be considered alongside the dates.

Some chronology tools can model explicit hiatuses, while others require segments to be fitted separately or event horizons to be specified. The interpretation must identify which depths lie above and below the gap and how its duration is constrained. If duration is uncertain, report a range or scenarios. A visible step in a model is not a defect when it reflects missing stratigraphy. It is often the most honest way to represent what the core does not contain.

Turbidites and rapidly deposited event beds

Submarine landslides, density currents and storm-driven transport can deliver thick sediment packages over a comparatively short time. A turbidite may contain graded beds and an erosional base. If its thickness is interpreted as thousands of years of steady accumulation, an age–depth model can misdate the sediments around it and exaggerate the duration of environmental fluctuations. Rapid deposition should be represented as an event or high-rate interval when supported by sedimentology.

An event bed can also contain older reworked fossils, making its apparent age measurements older than its final deposition. A date from a shell embedded in the turbidite might be a maximum constraint on deposition, not a direct timestamp of the event. The model should not resolve this discrepancy by quietly moving the shell or inventing a tiny analytical error. Evidence from grain size, radiography, microfossil preservation and multiple independent samples can clarify the physical history.

For a numerical illustration, imagine 20 cm of sediment emplaced in one storm-driven event rather than over 2,000 years. A naive constant-rate assumption would imply 0.1 mm/year for that unit. But the true event-scale accumulation could have been vastly faster, with almost no elapsed calendar time represented by the bed. The average relation between layer thickness and total age span becomes a poor description of the process. Models that separate event beds can protect the chronology of slower background deposition.

Age reversals and why they need investigation

An age reversal occurs when a deeper dated sample appears younger than one above it, according to the ages being compared. It can reflect reworked old material in the shallower layer, laboratory contamination, sample mislabeling, core disturbance or uncertainty ranges that overlap enough to make apparent ordering inconclusive. It may also indicate that the assumed carbon reservoir correction is wrong. Reversal is a clue to investigate, not automatic proof that the entire core is upside down.

Check whether the two dates are truly different at the stated confidence level, preferably after calibration and with shared errors considered. Then examine each sample’s material, taphonomy, lab report, stratigraphy and depth scale. A well-preserved in-situ shell might be stronger evidence than an abraded mixed shell. If one date is rejected, document why with the original observation retained. If the evidence remains ambiguous, use model sensitivity analyses rather than presenting the exclusion as an unquestioned fact.

Chronological software can sometimes accommodate anomalous dates through outlier models. Such tools are valuable because they expose disagreement instead of simply deleting it. But the posterior probability of a date being an outlier is conditional on the rest of the model. A genuinely unexpected depositional event may make a correct measurement look incompatible with an oversimplified smooth accumulation prior. A responsible interpretation checks both the model and the material.

Sediment accumulation rates and their units

An age–depth model gives elapsed time as a function of depth. Its local slope can be used to describe net linear accumulation, but the units depend on which axis is considered. If 60 cm accumulates over 9,000 years, the mean linear accumulation rate is 0.00667 cm/year or 0.0667 mm/year. The inverse gradient is 150 years/cm. Calling 150 years/cm a sedimentation rate in mm/year without inversion is a unit error. The safest practice is to write the calculation and dimensions explicitly.

Mass accumulation requires a density field. If a layer has dry bulk density 600 kg/m³ and a local linear rate of 0.0000667 m/year, the approximate dry mass accumulation is 600×0.0000667≈0.040 kg/m²/year, conditional on relevant compaction interpretation. If a constituent has concentration 2 grams per kilogram of dry sediment, a simple estimated burial flux would be 0.040×2=0.080 grams/m²/year. Both calculations are illustrative, not measurements of an actual marine location.

The local derivative of a flexible age model can be especially uncertain. Small changes in model curvature can produce large changes in instantaneous rate even when the ages at measured horizons are well constrained. Authors should specify whether they report a smoothed local rate, interval-averaged rate or mass accumulation rate. If an environmental proxy appears to change abruptly, a high or low apparent concentration rate may partly reflect changing sediment accumulation rather than changing source intensity.

Worked comparison of two acceptable model shapes

Consider anchors at 0 cm and 0 cal BP, 40 cm and 5,000 cal BP, and 100 cm and 14,000 cal BP. A piecewise linear model gives age 9,500 cal BP at 70 cm. A different model might allow slower accumulation from 40 to 70 cm and faster accumulation from 70 to 100 cm, while preserving both anchor ages. For instance, a hypothetical curve could place 70 cm at 10,500 cal BP and still meet the end constraints. The central difference of 1,000 years represents a model choice that the two anchors alone cannot settle.

Such flexibility must not be used arbitrarily. Sedimentological changes, additional dated horizons and statistical priors should determine which shapes are plausible. If an independent ash marker at 70 cm securely dates to 9,600 ± 100 cal BP, the 10,500-year option becomes less plausible. If a thick disturbed layer occurs there, both simple curves may be inadequate. The exercise demonstrates why an additional genuinely independent tie point can be more valuable than reporting extra decimal places for the original endpoints.

A sensitivity report might present multiple plausible chronologies and ask whether the scientific conclusion survives all of them. If a claimed lead-lag relationship between two ocean proxies changes sign depending on the allowed accumulation shape, the evidence does not support a precise causal sequence. If all plausible curves place an event within a broadly similar millennium, the more modest conclusion may be robust. Statistical uncertainty becomes useful when it guides the strength of an interpretation.

Dating density, sample quality and independence

Adding more dates does not always improve a chronology as much as one might hope. Ten dates from reworked material can reinforce the same bias. Multiple dates sharing an uncertain reservoir correction may contribute less independent information than they appear to. A carefully sampled primary ash horizon or other genuinely independent clock can improve constraints differently from another similar shell determination. Evaluate sample quality, dating-method independence and depth distribution rather than counting age points alone.

Date placement matters. Three strong age controls clustered near the core top may leave an enormous gap in the middle. A model may look tight at those shallow depths but rely heavily on priors farther down. Conversely, one isolated old date may anchor the bottom while failing to resolve intervening rate shifts. A useful sampling plan places independent age controls where the scientific question requires temporal resolution and where sedimentological complexity is greatest.

Chronology inference benefits from quality flags. Record whether a specimen was intact or reworked, whether its calibration is multimodal, whether a tephra correlation is tentative and whether a depth interval intersects a disturbance. Such flags guide sensitivity analyses and prevent future analysts from treating every number as equivalent. An explicit ‘not suitable for direct depositional age’ note can be more valuable than an additional apparently precise radiocarbon date.

How to read an age–depth figure

First find the depth axis and its datum. Then find the age axis, its reference and whether values are calendar BP, conventional radiocarbon years or CE dates. Identify the symbols representing independently dated horizons; differentiate them from the smooth model line. Look for uncertainty ranges on input dates and the age-model envelope. Check whether the plot includes identified hiatuses, rejected dates, event beds or extrapolated sections. A model extending far beyond the oldest or youngest anchor deserves explicit caution.

Next inspect where the uncertainty band widens. These are often the intervals where chronology is least constrained, though a restrictive prior can sometimes hide the expected widening. Compare any sharp changes in line slope with sedimentary changes and whether the inferred rate is physically plausible. If the figure includes an inset accumulation-rate panel, check the units and whether uncertainty is shown. A large rate peak may be a real depositional event or an artifact of sparse tie points.

Finally ask what the paper wants the figure to prove. If the claim concerns the order of two events separated by only a century, a thousand-year-wide chronology envelope cannot securely establish which came first. If the claim concerns a broad glacial-versus-interglacial transition, that same age uncertainty may be acceptable. A chronology is neither good nor bad in isolation; it is adequate or inadequate for a stated temporal question.

Uncertainty in synchronizing different cores

Comparing two marine sediment records introduces chronology uncertainty from each core. An event at 9,500 cal BP in Core A and another at 9,700 cal BP in Core B do not necessarily differ by a securely known 200 years. The individual model distributions may be hundreds of years wide, and shared calibration errors may make the difference behave differently from a naive square-root-of-sums calculation. A robust lead-lag claim ideally samples age–depth realizations from each model and evaluates the distribution of event-time differences.

If the two cores share the same volcanic ash marker, that shared age uncertainty may shift both chronologies together. Their relative timing near the shared horizon can be more precise than their absolute dates. By contrast, if each uses a different reservoir model, local ΔR histories can introduce differential error. Distinguish common-mode errors from independent uncertainty. Summing all reported ± values without considering dependence can either exaggerate or understate the reliability of synchrony.

Tuning one core to resemble another may help establish a relative framework, but it weakens the independence of subsequent claims that their proxy changes are simultaneous. Readers should know whether synchrony was inferred from independent dates, a shared physical event marker or aligned climate signals. Those are different levels of evidential independence. The best explanation states which component of the time relationship was observed and which was imposed by model design.

A reproducible age-model workflow

Begin with a master depth table preserving collection position, composite adjustments and sample interval bounds. Record all original laboratory determinations without replacing them by preferred dates. For radiocarbon, store the conventional age, uncertainty, dated material, curve, reservoir assumptions and resulting calibration information. For independent event markers, retain the source of the event age and the depositional context. Flag reworking, mixed layers and possible hiatuses before modeling. This makes it possible to audit later decisions.

Choose a modeling approach matched to the stratigraphy and temporal goal. Fit a simple piecewise model as a transparent benchmark, then consider flexible or Bayesian approaches when data and scientific questions justify them. Test alternate reasonable accumulation priors, reservoir corrections and treatments of anomalous dates. Check posterior age order, rates and consistency with independent markers. Save software version, input files, configuration and random seeds where relevant so the result can be reproduced rather than merely redrawn.

Report median or central ages with credible or confidence intervals whose meaning is stated. Show chronological uncertainty at the depths of substantive proxy interpretations, not only at dated points. Do not imply finer resolution than the model warrants. If a hiatus or event bed is plausible but uncertain, make that possibility explicit in the figure and prose. A carefully limited age model is much more useful than a neat curve that silently erases geological complexity.

Independent transfer problems and answers

Problem A: a core is dated 5,000 cal BP at 40 cm and 14,000 cal BP at 100 cm. Assuming linear interpolation, what is the age at 60 cm? The fraction is 20/60, so age=5,000+(1/3)×9,000=8,000 cal BP. Problem B: what is the average depth accumulation rate between those controls? It is 60 cm over 9,000 years, or about 0.00667 cm/year. Problem C: what does 150 years/cm represent? It is the age gradient, the reciprocal of the interval-average linear accumulation rate.

Problem D: if a depth interval has 20 cm of sediment emplaced nearly instantaneously, is thickness divided by an assumed long timespan a good event rate? No; it ignores rapid deposition. Problem E: a 12,000 radiocarbon-year date and a 12,000 cal BP ash marker are entered as identical values. What is the category error? Uncalibrated radiocarbon years and calibrated calendar years are different timescales. Problem F: a piecewise model gives the same center age as a Bayesian model but a much narrower uncertainty band. Are they equivalent? No; their probabilistic assumptions and uncertainty statements may differ.

Problem G: two dates appear reversed, but their calibrated distributions overlap heavily. Is overturning established? No; consider uncertainty and other explanations. Problem H: a secure tephra occurs at 142 cm, but the layer is demonstrably reworked. Is its eruption age a fixed depositional date? Not without more evidence. Problem I: two proxy changes differ by 200 modeled years, while the inferred lag varies from negative to positive under plausible chronologies. May the paper assert which event led? Not securely. The correct conclusion reflects the model sensitivity.

Mistakes and repairs that protect the timeline

Error one is treating the age model as observed truth. Repair by showing measured dates separately from modeled interpolation. Error two is mixing calendar and radiocarbon ages. Repair by calibrating with appropriate curves and documenting conventions. Error three is extrapolating far beyond data without caution. Repair by presenting large uncertainty or limiting the modeled domain. Error four is silently forcing a smooth curve across an erosional hiatus. Repair by recognizing missing time through the stratigraphy.

Error five is deleting every date reversal until the line becomes pretty. Repair by examining samples and reporting the reason for any rejection. Error six is converting a slope visually without checking which axis is which. Repair by deriving depth per time with explicit units. Error seven is treating independently calibrated dates as independent even when they share a common regional ΔR. Repair by considering correlated uncertainty. Error eight is claiming that alignment of tuned proxies proves independent synchronous causation.

The recurring conceptual mistake is to hide a model assumption inside a number that looks like a measurement. A scientific chronology is a set of inferences from dated samples and sedimentary context. Its value comes from making those inferences inspectable. Translators and science writers should preserve ‘modeled’, ‘estimated’, ‘conditional’, ‘calibrated’, ‘event layer’ and ‘hiatus’ rather than replacing them with everyday terms that suggest unwarranted precision.

Frequently asked questions

Can an age–depth model return an exact date for every centimeter? It can output a central estimate at every centimeter, but that does not make each date independently observed or exact. Why can age uncertainty grow between dated horizons? Because changing accumulation histories remain plausible in underconstrained sections. Why might a Bayesian model differ from a straight-line calculation? Priors and accumulation-rate structure encode additional assumptions. Is the model closest to the most dates always best? Not necessarily, because date quality and shared biases matter.

Does a hiatus mean no time passed? No, it means a span of time is missing from the preserved deposit. Can a thick turbidite represent a very short period? Yes. What is a tie point? An event or dated sample that constrains age at a specified depth under an explicit interpretation. Why keep calibrated probability distributions rather than medians alone? They may be asymmetric or multimodal and carry information relevant to chronological inference. Can a perfectly precise laboratory measurement date a reworked shell’s final deposition? Not by itself.

How should two cores be compared? Preserve the uncertainty and dependence in each chronology and evaluate the distribution of relative event times. When should one report a rate? When the modeled age interval and depth or mass definition are clear, and the rate’s uncertainty and averaging window are meaningful. What is the most practical question for a reader? Which parts of this curve came directly from independent evidence, and which parts were created by the chosen model of deposition?

Scientific language, translation and useful judgment

Words such as ‘age’, ‘date’, ‘calibrated age’, ‘age model’, ‘sedimentation rate’ and ‘accumulation rate’ are not interchangeable labels. A measurement may date carbon fixation, a volcanic eruption or an isotope signal; a model estimates when a sediment depth was deposited. An age-depth plot uses both. Preserving this distinction in English or another language is essential because one small change from ‘consistent with’ to ‘proves’ can transform a conditional interpretation into a false claim.

Keep the calendar scale visible: cal BP, conventional 14C years BP, CE years or years before collection. Keep core depth units and composite-depth transformations explicit. State whether a ± interval is analytical, calibrated, modeled or a sensitivity range. Translate ‘hiatus’ as missing preserved time, not as ‘no change’. Translate ‘age reversal’ as an observed or modeled ordering conflict, not an automatic overturning of strata. These careful terms perform real scientific work.

To learn the method rather than memorize jargon, take one proxy observation at an undated depth. Identify the nearest reliable controls, calculate a simple baseline interpolation, examine realistic deposition alternatives, inspect the uncertainty band and decide what temporal statement survives. That procedure is transferable to groundwater histories, archaeological layers, ecological archives and long-form science explanation. A good age–depth model does not eliminate ambiguity. It tells us what time-related claims the evidence can support.