Marine sediment XRF core scanner results are easy to plot and surprisingly easy to misread. A bright Fe peak, a low Ca count or a dramatic Fe/Ti ratio is an observation about the instrument signal at a position in a sediment core. It is not, by itself, a measured chemical concentration, proof of a change in river input, or a dated environmental event. Learning to translate these signals into defensible statements is the purpose of this guide.
To read an XRF core scanner profile responsibly, we need to follow the whole chain: a split core and its physical condition, an X-ray measurement, a fitted spectral peak, a count or count rate, a ratio or calibration, a depth alignment, and finally a geological interpretation. If one link is missing, a neat graph may tell a story stronger than its evidence permits. We will walk through the chain with labelled hypothetical numbers and checks that a reader can repeat.
Here “translate easily” does not mean replacing a technical term with a friendly but inaccurate synonym. It means translating raw instrument evidence into ordinary language while retaining units, uncertainty, sample condition and the distinction between interpretation and measurement. A good translation can be simple without becoming certain about things the scanner cannot establish. This is useful in marine geology, paleoclimate, lake and estuary sediments, laboratory reports and interdisciplinary conversations.
First decide what question the core can answer
A marine sediment core records material accumulated at a location, but accumulation is not a single process. Land-derived mineral grains, shells and biogenic carbonate, organisms, organic matter, dissolved constituents and transported particles can all contribute. Mixing by animals, erosion, resuspension, diagenesis and gaps in recovery can disturb a simple chronological picture. Element signals are therefore clues within a sedimentary system, not a direct recording of one environmental variable.
Begin with a specific question. Are we trying to recognise lithological boundaries, identify a carbonate-rich interval, compare terrigenous delivery with biogenic material, locate unusual metal enrichments, or align two nearby cores? Those are different tasks. A profile that is excellent for finding a boundary may still be unsuitable for estimating annual iron flux. The necessary evidence depends on the claim. A ratio can highlight a contrast without measuring a flux, whereas a flux requires concentrations, accumulation rates and density information.
Write the intended claim before choosing a proxy. For example, “the interval has relatively more Fe signal than Ti signal” is different from “more iron arrived from the continent.” The first refers to observed fluorescence and a denominator. The second implies a source mechanism and needs independent sedimentological or geochemical support. A useful reading protocol makes this difference visible from the beginning rather than repairing it after publication.
How X-ray fluorescence creates an elemental signal
An XRF instrument illuminates a sample with primary X-rays. Some photons eject inner-shell electrons from atoms in the sample. When electrons from higher energy levels fill those vacancies, characteristic X-ray photons can be emitted. Their energies help identify elements, while their observed intensities contain information about abundance as well as geometry, attenuation and instrument configuration. The scanner does not label a sediment particle “river-derived” or “marine-derived”; it detects radiation associated with atoms.
The detector accumulates a spectrum with photon energy on one axis and recorded events on the other. Element peaks may overlap, sit above a background or be affected by escape peaks and detector response. A peak-fitting procedure assigns counts or areas to named elements. That fitted number should be read alongside acquisition voltage, filter settings, dwell time, detector type, calibration and the software’s processing assumptions. A data table containing only a coloured Fe curve hides important measurement context.
Not every element is equally well measured under one setting. Lower-energy fluorescence is attenuated more strongly by water, air paths and material above the emitting grains. Different excitation settings favour different parts of the elemental range. An apparent absence can reflect insufficient sensitivity rather than true absence. Before interpreting a near-zero point, ask whether the configured measurement could reliably detect that element in that matrix.
Counts are not automatically concentrations
A raw Fe value of 12,000 counts is a tally associated with a fitted signal under particular measurement conditions. A value of 6,000 counts may mean less Fe in comparable material, but it can also reflect a shorter dwell time, wetter sample, rough surface, changing geometry, spectrum-fitting differences or another matrix effect. Counting events and calculating a weight percentage are different operations. The units should make the distinction obvious to the reader.
A count rate is often written in counts per second, abbreviated cps. If a fitted peak contains 9,000 counts over a 30-second acquisition, a simple rate is 300 cps. That calculation is useful for comparing runs with different dwell times when all other relevant settings are comparable. It does not correct attenuation or matrix chemistry. Dividing by time removes one source of unfair comparison, not every source. Always report whether the original instrument file stores counts, cps, peak areas or transformed intensity.
An elemental concentration such as 4.2 wt% Fe states a fraction of sample mass, using a defined analytical convention. Some laboratories report iron as Fe2O3 equivalent rather than elemental Fe; those numbers are not interchangeable. A defensible conversion from scanner intensities to concentration needs suitable calibration samples or validated methodology, preferably spanning the sediment compositions in question. A label saying “Fe (%)” on a raw-count plot is a substantive error, not a harmless formatting choice.
Keep four units separate on every plot
The horizontal or vertical coordinate of a core profile might be centimetres below a section top, metres below seafloor, a composite depth scale, or an age estimate in years. These are not equivalents. The plotted elemental variable might be raw counts, cps, a dimensionless ratio, log ratio, weight percent, or an enrichment factor calculated against a background. A reader should never have to guess which transformation was used.
When two elements are measured over different live times or excitation configurations, raw count ratios may not mean what a familiar interpretation assumes. Check the instrument protocol and whether intensities are directly comparable; if not, adjust only according to a documented analytical method. Do not silently invent a universal normalisation that makes every setting equivalent. The same caution applies to data stitched from different laboratories or generations of scanner.
A good axis label might read “Fe/Ti fitted intensity ratio (dimensionless, same acquisition)” or “Fe2O3 (wt%, calibrated to discrete samples).” A poor label would read simply “iron abundance” when the column is an uncalibrated detector output. Precise labels seem like an editorial detail, but they prevent the central error in many downstream interpretations.
Why the wet surface matters
Soft marine sediments contain porewater. Water contributes to absorption and scattering and alters the mass and geometry presented to the instrument. When porosity or water content changes downcore, element intensities can change even when the dry solid’s chemical composition changes less. This effect is particularly important for relatively low-energy fluorescence. It is therefore risky to use a wet-core intensity curve as if every millimetre of the core were a dry, homogeneous reference powder.
Surface preparation matters as well. A freshly split surface may be flat and smooth, or it may contain cracks, shell fragments, air pockets, smeared clay, grooves and sections disturbed by cutting. A protective film can introduce an additional attenuation path. An edge may sit farther from the detector than the middle, and an exposed void can produce an anomalously weak signal. These problems are physical, not statistical; smoothing a curve does not make the missing material reappear.
The IODP XRF facility specifically describes its strongest results on smooth, fine-grained material and warns that coarse, desiccated, crumbly or fractured cores are difficult. This is a practical reading rule: inspect section photographs and instrument notes near every extreme excursion. A low total signal coincident with a visible crack is a reason to flag that interval before proposing a rapid environmental shift.
Translate water effects without overselling a correction
Suppose several elements fall together in a water-rich interval. That pattern might be consistent with stronger attenuation or reduced exposed solid mass. It might also reflect a real compositional change or a mixture of both. A careful caption says that water content is a potential common-mode influence and checks available moisture-density data. It does not assert that every synchronized decline is an artefact or that one numerical correction works at all energies.
If dry, compositionally analysed discrete samples are available, use them to assess whether scanner intensities track actual concentrations across varying water content. Residual plots can then reveal systematic bias with water content or density after calibration. Where no such samples exist, treat the scanner series as relative signals and make the physical limitation prominent. Reporting an uncertainty is more truthful than fabricating a concentration correction from an unrelated site.
Do not remove all low-signal intervals automatically. Low counts can carry genuine sedimentological information, and cleaning them may erase useful boundaries. The right choice is to annotate, examine, cross-check and then decide which analytical inference can be supported. Maintain both the untouched measurements and a clearly identified quality-screened derivative so a future reader can reconstruct the decision.
Grain size and mineralogy influence elemental response
Two sediments with identical bulk Fe concentration can produce different spectra if grain size, mineral distribution, surface texture or density differs. Fine particles may form an even surface, while coarse grains leave gaps between mineral phases. A shell-rich patch can dominate a small measurement footprint. Since the beam sees a finite area and depth, the result may not represent the entire core’s dry bulk composition in the way a homogenised powder measurement does.
Mineralogy also changes absorption and enhancement. A heavy-element-rich matrix can attenuate other fluorescence lines, and overlapping spectral features can complicate peak fitting. Calibration built on a narrow range of mud may perform poorly on carbonate-rich sand or an ash-rich layer. A sudden change in the relation between peak intensity and chemistry is sometimes the geological signal, and sometimes a failure of the chosen measurement model.
A practical defence is to pair XRF with core photographs, grain-size description, density and a subset of independent chemical analyses. When those records disagree, the disagreement is valuable: it directs attention to a matrix effect, a lithological change or a depth-matching problem. The purpose of an independent method is not merely to make the scanner look correct; it is to test whether the inference survives another measurement route.
Read Fe/Ti as a comparison with a denominator
An Fe/Ti intensity ratio compares two fitted signals. If Fe is 8,000 counts and Ti is 4,000 under the same appropriate conditions, the ratio is 2.0. The statement this supports is limited: the measured Fe signal is twice the measured Ti signal. It does not establish that the sediment contains exactly twice as much elemental iron as titanium by mass, because fluorescence efficiency and matrix effects differ between elements.
Ratios are attractive because some shared fluctuations can cancel. If both signals are multiplied by the same factor, their quotient stays unchanged. But common-mode cancellation is imperfect when elements have different fluorescence energies, response functions or peak-fitting biases. A ratio can also rise because its numerator increases, its denominator falls, or both happen. Plot both constituent series beside the quotient whenever the mechanism matters.
A near-zero denominator creates a special hazard. Ti = 50 counts and Fe = 5,000 counts yields a mathematical ratio of 100, but a small Ti count may be close to a detection or fitting limit, making the ratio unstable. Impressive peaks caused by denominator collapse should not be treated as strong events. Mark low-denominator positions, inspect the raw spectrum and consider whether a qualified ratio should be reported at all.
A useful proxy is not a universal fingerprint
In some sedimentary contexts Fe/Ti, Ti/Ca or other element ratios are interpreted as indicators of changing relative terrigenous and carbonate components. That can be useful when endmember compositions and depositional processes support it. Yet Fe may respond to redox alteration, sulfide formation and diagenesis; Ti may be associated with relatively resistant detrital phases; Ca may occur in biogenic carbonate as well as detrital minerals. A proxy requires local justification.
For instance, an increase in Ti/Ca might reflect higher detrital input, lower carbonate production, carbonate dissolution, a difference in grain size, or a combination. The quotient alone cannot determine which process moved. A strong interpretation checks carbonate percentages, shell observations, particle size, magnetic or mineralogical records, and the sedimentary setting. Different nearby locations may produce the same ratio pattern for different reasons.
The discipline here resembles translating a sentence whose noun has multiple meanings. We identify the context before choosing the sense. For an XRF proxy, that context is the core’s lithology, depositional environment, calibration and complementary measurements. A glossary helps, but it cannot replace the evidence. Precision becomes easier after we distinguish what was measured from what the site history might explain.
Worked example one: counts, cps and a ratio
All numbers in this example are hypothetical and are included to teach calculation, not to report a field observation. At depth 42 cm, the fitted Fe peak has 9,000 counts and the Ti peak 4,500 counts in a 30-second acquisition. Their count rates are 300 cps and 150 cps. The corresponding fitted-intensity ratio Fe/Ti is 9,000 divided by 4,500, or 2.00. If both peaks were fitted under a directly comparable configuration, the same quotient is obtained from the count rates.
At 44 cm another scan reports Fe = 8,400 counts and Ti = 2,800 counts over the same dwell time. Rates are 280 and 93.33 cps, with Fe/Ti = 3.00. The ratio has risen by 50% from 2 to 3 even though the Fe count fell from 9,000 to 8,400. The stronger effect is the decline of the Ti denominator. This is why “iron increased sharply” would be the wrong summary of this fictional comparison.
A readable report would say: “Fe/Ti fitted intensity ratio increased from 2.00 to 3.00, driven primarily by lower Ti intensity; the dry elemental concentrations and source mechanism are not established by these two scans.” Then examine total counts, water content and lithology. The arithmetic is trivial; the discipline lies in the words around the result.
Counting uncertainty and what it leaves out
When detections approximately follow independent Poisson counting statistics, the standard deviation of a raw count N is about the square root of N. For 9,000 Fe counts and 4,500 Ti counts, the simple fractional counting uncertainties are about 1.05% and 1.49%. For their ratio, first-order propagation gives a fractional standard uncertainty of the square root of 1/9000 plus 1/4500, approximately 1.83%. A ratio of 2.00 therefore has a counting-only standard uncertainty of roughly 0.037.
That calculation is an educational approximation, not a full instrument uncertainty budget. Fitted spectral peak areas may be correlated, background subtraction changes the error model, and detector drift, water content, beam geometry and calibration may dominate. Count-based intervals also behave badly near zero, and negative net peak areas after subtraction cannot be handled as ordinary positive Poisson counts. Report the model being used instead of presenting a small counting error as overall accuracy.
A meaningful uncertainty statement separates repeatability, calibration error and representativeness. A scanner might repeatedly measure the same smooth spot with high precision while still disagreeing with a dry bulk assay because of matrix effects. A single narrow uncertainty band should not conceal this distinction. When environmental interpretation depends on a small difference, evaluate whether that difference exceeds both technical and geological variability.
Worked example two: the illusion of an iron peak
Imagine three fictional depths with Fe/Ti ratios of 1.5, 3.0 and 1.5. A quick glance suggests a central iron excursion. But suppose the Fe counts are 6,000, 6,000 and 6,000 while the Ti counts are 4,000, 2,000 and 4,000. Fe did not change at all in the raw series. The ratio peak comes entirely from Ti being lower at the middle position. It may still be a real compositional signal, but the explanation must start with the observed components.
Now imagine instead that both counts halve at the middle depth: Fe falls to 3,000 and Ti to 2,000. The ratio remains 1.5, concealing a pronounced shared intensity decrease. The quotient has hidden a potential water-content, geometry or total-solid-mass change. Ratios can reveal contrasts and suppress common changes at the same time. The best graphic frequently includes Fe, Ti, Fe/Ti and a simple quality-control track.
Finally suppose Ti is reported as zero at one position. Dividing by zero is not a dramatic environmental anomaly; it is an undefined ratio. Investigate whether the element was not detected, the fitted net peak was zero, the record is missing, or a parsing step replaced a blank with zero. Leave the ratio missing until its analytical status is resolved. A software plot that draws an enormous spike is not a substitute for this decision.
Transformations: natural logs are not magic repairs
A log ratio such as ln(Fe/Ti) is often useful because it expresses relative composition on a scale where multiplicative contrasts become additive. For ratios 1, 2 and 4, natural-log values are 0, approximately 0.693 and approximately 1.386. Doubling the quotient moves the log ratio by the same amount each time. This helps when variations span orders of magnitude and when calibration models are formulated in log-ratio space.
The transformation does not repair invalid input. Neither a zero nor a negative fitted net ratio has a real natural logarithm in the usual statistical treatment. Adding a small arbitrary constant to every count changes low-concentration behaviour and should never be concealed. If censoring or nondetection is important, adopt a defensible statistical treatment and state it. If the goal is simply a descriptive core profile, retaining marked missing values may be more transparent.
Do not confuse log ratios of counts with log ratios of concentrations. The scanner response functions differ, so a regression calibrated with paired chemical observations is still needed before a count log ratio becomes an estimate of a chemical log ratio. A mathematical transformation can improve the behaviour of data under a model; it cannot create independent chemical evidence.
Calibration begins with representative paired samples
To estimate concentrations, collect or select discrete material whose location corresponds to the scanned interval, then measure that material with an appropriate quantitative laboratory method and documented reference standards. A good calibration set spans the major lithologies, water contents and compositional range of the core rather than clustering only in easy, uniform mud. Aligning sample depths and scanner positions is part of the calibration problem; a misplaced shell-rich layer can spoil an otherwise sound regression.
The 2020 IODP Site U1474 data report provides a concrete methodological example: investigators selected 25 discrete calibration samples and used a univariate log-ratio approach for major elements. Their published results show the logic of converting relative XRF signals with independent composition data rather than merely relabelling scanner counts as percentages. Their exact coefficients belong to their samples and instrument conditions; they should not be copied onto an unrelated core.
A practical calibration notebook records core and section identifiers, scan settings, fitted peak definitions, discrete sample mass, preparation, chemical reference method, uncertainty, matching depth and exclusion rationale. It also preserves raw and processed measurements. That may seem administrative, yet it allows another analyst to distinguish genuine model performance from a result obtained by repeatedly retuning choices until the graph looks convincing.
Worked example three: a log-ratio calibration
Consider an explicitly illustrative calibration equation, ln(C_Fe/C_Ti) = -0.4 + 0.9 × ln(I_Fe/I_Ti). Here C denotes concentrations in the same mass-based units and I denotes fitted intensities collected consistently. The constants are made up for the arithmetic lesson; no real sediment site is being calibrated. For an intensity ratio of 2, ln(2) is about 0.693, so the right-hand side is -0.4 + 0.624 = 0.224.
Exponentiating gives an estimated concentration ratio of exp(0.224), about 1.25, not 2.00. This illustrates why counts and chemistry cannot simply share a numerical axis. The result is a calibrated ratio conditional on the chosen model and its evidence; it does not yet tell us the absolute mass percentage of either element. Estimating Fe wt% separately would require enough independent compositional constraints and a suitable closure or full calibration approach.
For a scanner ratio of 4, the same fictional equation yields ln(C_Fe/C_Ti) = -0.4 + 0.9 × 1.386 = 0.848, so C_Fe/C_Ti is approximately 2.34. Doubling the intensity ratio from 2 to 4 does not double the calibrated chemical ratio. The exercise teaches why transformations, coefficients and uncertainty must travel together when an analytical result is translated for a non-specialist reader.
Validation must use data the fit did not simply memorise
A calibration can fit its own training pairs well and still predict badly outside them. If samples are few and close together in the same homogeneous layer, neighbouring observations are not truly independent tests. With enough material, reserve calibration checks from distinct lithologies or sections, fit without those samples and compare predictions with withheld chemistry. Report prediction bias and scatter in the actual units needed by the intended interpretation.
Residuals should be inspected against concentration, water content, sediment type, depth, scan session and detector settings. A systematic pattern may reveal an omitted matrix effect, an inappropriate common denominator or depth misalignment. Merely reporting a correlation coefficient misses this structure. A high correlation can coexist with a slope error, and a highly significant regression may still be unfit for estimating small changes in concentration.
When a model performs poorly for a particular facies, the honest answer is often to report relative intensities there rather than forcing an implausible concentration number. Alternatively, collect more paired measurements across that facies and rebuild the calibration with appropriate validation. Keeping limits visible is an advantage: it stops a specialist result from being turned into false certainty by an attractive plot.
Worked example four: a measurement-time trap
Imagine one scan gives Fe = 10,000 fitted counts in 20 seconds, and another gives Fe = 15,000 counts in 40 seconds. The raw counts increased by 50%, but the respective count rates are 500 and 375 cps. On a time-normalised basis the second signal is lower by 25%. A conclusion that the second sample is more iron-rich from raw counts alone reverses the direction of the observed rate difference.
Even after using cps, the result remains an instrument intensity. If the second scan used a different voltage or filter, or if its sample is wetter, an apples-to-apples chemical comparison has not been established. Store setting changes as metadata, not as a footnote forgotten during plotting. A rate correction answers “How many fitted events per second?” It does not answer “How many milligrams of iron per gram of dry sediment?”
The correct two-sentence translation is: “Fitted Fe intensity per unit live time is lower in the second scan, although its total count is larger because it ran longer. A compositional interpretation would require matched acquisition conditions and relevant matrix or calibration checks.” This is an example of a simple explanation that actually protects the scientific meaning.
Spatial resolution is not the same as independent information
A scanner can collect points every 2 mm, but a beam has a footprint and the sediment may be disturbed over a larger length. Adjacent measurements can share material, and instrument response can smear a sharp boundary. Sampling spacing and true independent geological resolution are therefore different. A series with thousands of closely spaced numbers may carry much less independent information than the row count suggests.
At IODP Site U1474, investigators documented a densely sampled 2 mm profile that was noisy at the short counting time and improved interpretation by integrating over a coarser effective interval. That approach trades spatial detail for counting stability. It is reasonable for broad changes, but it may blur a thin ash layer or sharp contact. The analyst should select smoothing or integration according to the question and report the effective scale.
A moving average can also smear spikes into neighbouring intervals and can generate misleading edge behaviour. Keep the unfiltered series, state the window width and avoid presenting processed peaks as if they were independent original observations. Where a feature matters, compare raw spectra, section photographs and repeat measurements. Signal processing is an analytical decision with consequences, not a cosmetic graph-setting.
Depth, section joins and offsets
A long core may contain many sections. Each has its own top, bottom, gaps, disturbed ends and sometimes repeated or missing material. The scanner measures coordinates relative to a section; geological interpretation commonly needs a larger depth system. Converting between section centimetres, depth below seafloor and a composite splice requires documented offsets and rules. A curve shifted by a few centimetres can be falsely aligned with a boundary or another proxy.
Never treat a depth correction as an age conversion. Depth represents position within recovered material, while age requires a separate chronological model and its uncertainty. An XRF event at 75 cm cannot be called a specific calendar year simply because another layer elsewhere has a known age. Sediment accumulation can vary, mixing can smear marker horizons and cores can contain hiatuses.
When comparing XRF with the adjacent guide on sediment lead-210 dating, retain the chronology as an independent model. The lead-210 ages depend on radionuclide inventories and assumptions, while the XRF intensities depend on their own analytical chain. They may be brought together only after depth ties are checked. For a practical dating primer, see https://edukatesg.com/2026/10/28/how-to-translate-easily-sediment-lead210-supported-excess-crs-inventory-age/ .
A quality-control track should travel with the chemistry
A useful profile includes more than colourful element curves. Include section boundaries, missing positions, scan settings, total or selected peak intensity, signal-to-noise markers, calibration support and known surface defects. If an interval was excluded, name the criterion. If a detector malfunction or low-signal episode is suspected, describe its evidence rather than silently deleting the rows. A reader can judge a profile only when quality flags are legible.
At IODP Site U1474, the authors discussed anomalously low counts in some sections and explored whether ratios were less disturbed than individual intensities. Their conclusion was cautious, not a permission to ignore every instrument fault. A ratio may remain coherent if both signals are affected similarly, but an unknown fault can also change them differently. Corroboration from independent sections, spectra or discrete measurements is still essential.
Version the transformations in the data workflow. The raw table should remain intact; subsequent files can be identified as time-normalised, quality-screened, ratio-derived, log-ratio-derived or calibrated. Store sufficient instructions to regenerate each one. For a scientific article, a simple method statement connecting these stages is often more valuable than a dense wall of unexplained coefficients.
Detection limits and zeros in elemental columns
A zero can mean no fitted net peak, a concentration below a reporting threshold, a missing scan, a failed fit or a software default. These are analytically distinct. Never replace all with the same value and then compute ratios. Where a lab provides a detection limit, recognise that the exact concentration of a censored value is not known; it lies in a range defined by the method’s assumptions.
Near a detection limit, random variation and background estimation may dominate the signal. Reporting a ratio to four decimal places can be particularly misleading. The sensible response may be “not estimable at this depth” rather than a precise quotient. A plotting script should retain missing-value codes and quality flags until the evidence supports a numerical treatment.
Detection limits also vary across elements, settings and matrices. A table of instrumental capabilities under specified settings is not a universal guarantee for every sediment section. For low-level analytes, ask how the instrument operator defines detection, what standards were used and whether the reported peak was consistently identified. Reliability requires method context as much as a number.
Core photographs are part of the evidence
A scanner graph alone cannot show that a peak crosses a shell fragment, a fracture, a section boundary or a compacted layer. Photographs and visual descriptions help identify these conditions. They also distinguish a broad colour or texture change from a narrow localized anomaly. A hand specimen or smear-slide observation may reveal whether a high-Ca point is dominated by a visible shell rather than a region-wide change in carbonate production.
Where the beam footprint is small relative to heterogeneity, repeat a measurement away from an unusual clast or use a homogenised discrete sample for comparison. If the scientific question is the composition of the clast itself, then targeting it may be exactly right; if the question is bulk mud chemistry, the same target could mislead. Measurement design and geological question must remain aligned.
A good figure caption tells readers whether suspicious positions were rescanned, whether fractured intervals were flagged, and how section joins were handled. Those details make the graph readable even to people who never visit the laboratory. The aim is not to overwhelm readers but to supply the few physical facts necessary to trust the interpretation.
Relative enrichment is not a deposition rate
A higher Fe/Ti ratio describes relative instrumental or calibrated chemical composition, depending on processing. It is not a mass accumulation rate. To calculate deposition of an element over time, one typically needs a defensible dry mass accumulation rate and an independently calibrated elemental mass fraction, together with chronological support and corrections appropriate to the setting.
For a hypothetical layer with a dry mass accumulation rate of 0.20 g cm^-2 yr^-1 and measured Fe mass fraction 0.040, the implied Fe mass accumulation rate is 0.0080 g Fe cm^-2 yr^-1, equivalent to 80 g Fe m^-2 yr^-1. The conversion uses 10,000 cm² per m² and is valid only under the stated premises. An uncalibrated scanner count of 4,000 cannot be inserted for 0.040 in that equation.
Even a calculated flux may represent redistribution rather than changing source supply. Sediment focusing, lateral transport and changes in accumulation can alter local deposition. The model should distinguish an elemental mass flux at the core location from a watershed export rate. Both can be scientifically interesting; they are not interchangeable just because their units contain a time denominator.
Distinguish lithological boundaries from climate events
A change in Ti/Ca, Fe/K or another XRF ratio may locate a distinct sediment interval. That is a valuable observation. Attributing the boundary to drought, stronger monsoon rainfall, currents, glacial dust or biological productivity requires a causal argument appropriate to the site. These mechanisms can have overlapping fingerprints, and several may act together.
Start by checking whether the excursion corresponds to a grain-size shift, a colour change, altered microfossils, magnetic susceptibility, a dated event marker or another independent proxy. Consider alternative causes deliberately, including dilution by another component and post-depositional redistribution. The best-supported explanation should account for the full pattern, not only the one ratio that originally attracted attention.
Where alternatives remain plausible, make that uncertainty explicit: “The rise is consistent with a relative increase in detrital contribution, but carbonate dilution and grain-size effects have not been separated.” Such language is informative rather than evasive. It names the leading interpretation, its evidential basis and the work still needed.
Worked example five: a misleading carbonate story
Take two hypothetical intervals. In A, calibrated CaCO3 mass fraction is 40% and a terrigenous mineral fraction is 60%. In B, CaCO3 drops to 20% while the terrigenous fraction rises to 80%. A detrital-to-carbonate indicator would rise strongly even if the absolute delivery of detrital material per year stayed constant. Composition is a proportion of a whole; increasing one fraction can arise because another shrinks.
Suppose both intervals accumulate 100 g of dry sediment per square metre per year, giving 60 and 80 g of terrigenous material per square metre per year. In that particular scenario the terrigenous accumulation does rise. But if B accumulates only 75 g of total dry sediment per square metre per year, its 80% terrigenous fraction gives the same 60 g as A. The concentration contrast alone cannot tell the rate story.
This exercise is why a sediment proxy should not be marketed as a direct delivery meter without accumulation evidence. A careful caption says “higher relative terrigenous fraction” where that is what is supported. A broader source-flux interpretation needs sedimentation, dry density, chronology and depositional context. The denominator question is central to both XRF ratios and broader marine mass balances.
Cross-core comparison and scanner differences
Two cores may show similar Fe/Ti shapes yet have different absolute intensities because of their instruments, water content, preparation and settings. A common ratio axis is not automatically a common geochemical scale. Likewise, a change in software peak-fitting algorithms can alter fitted intensities even when the physical core has not changed. Cross-core interpretations should compare acquisition methods before correlating peaks.
Use discrete standards, interlaboratory tests or mutually calibrated reference materials when quantitative comparability is important. If those are unavailable, state that correspondence is qualitative and verify major features with independent stratigraphic markers. The 2020 multi-laboratory intercomparison of XRF core scanning demonstrates why practical calibration and instrumentation differences deserve attention, rather than assuming every scanner produces numerically exchangeable profiles.
For a visual correlation exercise, mark distinctive stratigraphic features first and align cores only through justified tie points. Do not stretch the depth axis repeatedly until every crest matches. Genuine records can differ because local deposition, erosion and mixing differ. Differences between cores are not always mistakes to be removed.
How to diagnose a strange profile systematically
When a profile looks suspicious, first ask which column is strange: a raw count, a count rate, a single ratio, multiple elements or only the calibrated product. Then identify whether the anomaly occurs at an end of section, a known disturbance, a low-intensity interval, a change of instrument setting or a distinct visible lithology. The order matters because these checks are quick and prevent confident interpretation of technical artefacts.
Next inspect the spectrum or available peak-fitting diagnostics, then compare the numerator and denominator separately. Check whether the anomaly survives a reasonable quality screen and whether other independent proxies agree. A feature that disappears only after a large smoothing window may be real but thin; a feature that vanishes when a bad denominator is marked missing may be a mathematical artefact.
Finally state what is unresolved. If a cracked interval prevents reliable measurement, say so. If the scanner signal is trustworthy but its environmental meaning is ambiguous, report the stable observation and competing hypotheses. Diagnostic clarity is usually more useful than a dramatic explanation that outruns the available data.
A field-to-figure workflow that readers can repeat
Begin with an inventory of cores, sections, photographs, sampling dates, acquisition conditions and available independent analyses. Preserve the original scanner export and establish a consistent depth identifier for every row. Identify missing scans, duplicated intervals and visibly disturbed positions. These records are the raw evidence from which the reader-facing interpretation will be built.
Next verify units and live time, inspect peak quality, and plot individual element intensities before computing a ratio. Choose quality exclusions by a documented rule. Where concentrations are required, pair relevant discrete measurements, fit and validate an appropriate calibration and report errors in the prediction units. Keep raw and transformed tracks distinguishable. Interpret environmental mechanisms only after these analytical stages are intelligible.
End with a figure caption that can stand alone: site and core, depth scale, measured quantity, units, relevant scanner setting, ratio definition, quality flags, calibration status and interpretation limits. This approach produces useful explanation without converting every technical choice into an unexplained acronym. It is also a reproducible way to audit an older profile before citing it in a new article.
Transfer problem one: count rate versus total counts
Try this fictional problem without looking at the answer. Sample A has a fitted Mn signal of 6,000 counts in 20 seconds; sample B has 9,000 counts in 45 seconds. Which has the higher count rate, and what can you claim about Mn concentration? Write the units and the analytical caveat before choosing an interpretation.
Answer: A has 300 cps and B has 200 cps. Thus A has a higher fitted Mn count rate even though B has more total events. This does not prove A has a higher Mn mass fraction because comparable scanner settings, sample physical conditions and calibration have not been established. The answer is complete only when it separates the time-normalised instrument observation from the desired chemical statement.
If someone writes “B is more Mn-rich because 9,000 exceeds 6,000,” the repair is to expose the time denominator. If someone writes “A is exactly 50% richer in Mn,” the repair is to acknowledge that a 50% difference in count rate is not automatically a 50% difference in concentration. The numerical work is a route to correct language, not an excuse for stronger claims.
Transfer problem two: what made the ratio change?
At position P the fictional Fe and Ti fitted counts are 7,200 and 3,600. At Q they are 7,200 and 2,400 under the same acquisition settings. Calculate both Fe/Ti ratios, identify the component that changed, and propose two distinct reasons that would require additional testing.
Answer: Fe/Ti is 2.0 at P and 3.0 at Q. Fe counts stayed fixed while Ti counts fell by one-third, so the ratio increased because of its denominator. The change could represent a shift in mineral composition, but it could also be influenced by spectral fitting or matrix effects that affect Ti more strongly. Compare spectra, water content, visual lithology and discrete chemistry before settling on a source mechanism.
This problem illustrates why ratios are not free-standing environmental variables. A scientifically useful statement identifies the mathematical mechanism of an observed ratio change first. Mechanistic environmental interpretation comes afterward, with a different evidence requirement. The same habit transfers to isotope ratios, biodiversity indices and any measurement in which the denominator can move.
Transfer problem three: a log-ratio result
A fictional calibration uses ln(C_X/C_Y) = 0.2 + 0.8 × ln(I_X/I_Y). If the scanner intensity ratio is 1, estimate the calibrated concentration ratio, then explain why the answer is not a weight percent. Treat the equation as an illustrative teaching device; do not apply it to an actual core.
Answer: ln(1) = 0, so ln(C_X/C_Y) = 0.2. Exponentiation gives C_X/C_Y about 1.22. The result expresses one concentration relative to another. Without an absolute measurement or valid compositional closure, it does not establish the individual fraction of X or Y in the sample. It is also subject to uncertainty in the coefficients and paired calibration data.
The common mistake is to say that a ratio of 1 in the scanner means equal chemical concentrations. Response factors and a calibration intercept can break that equivalence. The correct translation states the measured ratio, the applied model and the nature of the resulting estimate. It preserves both the transformation and its assumptions.
Transfer problem four: concentration is not flux
Imagine two layers with dry sediment accumulation of 50 and 100 g m^-2 yr^-1, and independently measured Fe mass fractions of 10% and 5%. Calculate the Fe mass accumulation rate in each layer. Then decide whether the layer with higher Fe percentage had more Fe deposited per year.
Answer: the first gives 50 × 0.10 = 5 g Fe m^-2 yr^-1; the second gives 100 × 0.05 = 5 g Fe m^-2 yr^-1. The fluxes are equal although the Fe fractions differ by a factor of two. These are hypothetical teaching figures and would depend on a valid chronology, dry mass accumulation estimate and quantitative concentration measurements in any real application.
The repair to a mistaken answer is to identify the missing mass-per-time denominator. Chemistry alone gives composition; composition multiplied by a suitable total mass rate gives an element mass rate. Neither should be confused with a catchment-wide source flux without further transport evidence. This same lesson applies to the adjacent ocean particle-flux and sedimentation topics.
Frequently asked questions: concentrations and ratios
Can I call an Fe/Ti scanner curve an “iron concentration record”? Not without a defensible concentration calibration. Call it an Fe/Ti fitted-intensity ratio, specifying whether count or rate conventions apply. Can a ratio reduce wet-core effects? Sometimes shared fluctuations partially cancel, but unequal attenuation and matrix responses mean it cannot be assumed. Does a high correlation with laboratory chemistry guarantee good predictions? No; look at biases, residuals and independent validation.
Can I compare ratios collected on different scanners? Only with careful attention to settings, response differences and the calibration or standardisation method. May I plot a log ratio? Yes, when the contributing values are valid positive intensities and the transformation is labelled. Should missing and zero signals be replaced with small arbitrary numbers? Not silently. Those substitutions can dominate the low-signal part of a ratio record.
Is a large change in Ti/Ca proof of stronger river discharge? No. It may be consistent with changing relative detrital and carbonate contributions, but carbonate production or dissolution, dilution, grain size and local reworking can also matter. What should a non-specialist reader remember? XRF is strong at locating relative chemical contrasts along a core; turning those contrasts into absolute chemistry, dates or environmental causes requires additional evidence.
Frequently asked questions: precision, resolution and interpretation
Does collecting data every millimetre guarantee millimetre-scale events? No. Beam footprint, sample disturbance, smoothing and natural mixing limit effective resolution. Is a very small counting uncertainty proof of accuracy? No. Systematic matrix effects and calibration error may exceed the count-based component. Can I delete low-signal rows because they look ugly? No; examine spectra and sample condition, document any exclusion and retain raw evidence.
What if a core contains sand, shell fragments and mud? Expect analytical representativeness and matrix response to vary. Consider stratified calibration, discrete samples and clear quality flags rather than applying a mud-based model without validation. What if two environmental interpretations remain plausible? State both and identify the measurement that would discriminate them. A useful scientific explanation need not pretend that an unresolved mechanism has been solved.
What is the quickest quality question to ask of any XRF figure? “What exactly is on this axis?” If the answer is unclear, request the raw units, transformation, calibration and depth system before inferring a climate or sedimentation process. Those four items often resolve more confusion than another paragraph of interpretation.
How this connects to other sediment measurements
XRF core scanning describes relative elemental signals and, with support, calibrated chemistry. Laser-diffraction grain-size analysis describes the distribution of particle-size equivalents under a different physical measurement model. Comparing them can help distinguish a compositional shift from a change in sediment texture, but neither is a substitute for the other. The adjoining guide on marine sediment laser diffraction is at https://edukatesg.com/2026/10/28/how-to-translate-easily-marine-sediment-laser-diffraction-grain-size-d10-d50-d90/ .
Benthic chamber oxygen flux addresses an entirely different quantity: oxygen exchange across a sediment-water interface under chamber conditions. Pairing its results with sediment chemistry may help form ecological hypotheses, but chemical intensities cannot be translated directly into chamber oxygen flux. For those units, chamber assumptions and examples, read https://edukatesg.com/2026/10/28/how-to-translate-easily-benthic-oxygen-flux-chambers-incubation-slopes-area-leakage/ .
The transferable skill is vocabulary precision. “Counts,” “counts per second,” “ratio,” “concentration,” “mass accumulation,” “depth,” and “age” are different concepts. A learner who can use each term correctly has gained more than a new graph-reading trick; they can carry the same reasoning into science reports, evidence discussions and unfamiliar quantitative writing. To practise this skill, rewrite every technical claim with its quantity, unit, denominator and uncertainty intact, then check that the simpler wording still makes exactly the same evidential claim.
The complete reader’s decision rule
Before trusting a sediment XRF interpretation, ask five questions in order. What physical material and surface were actually measured? What instrument signal and units appear in the file? Which corrections, ratios and calibrations created the displayed series? What independent evidence connects that analytical series to a geological process? Which alternative explanations and uncertainties remain? A strong analysis can answer each question or explicitly identify the unresolved step.
Do not turn the five questions into a rigid promise that every core can produce absolute concentrations or a climate reconstruction. Sometimes the defensible conclusion is a high-resolution relative profile that guides where to sample next. Sometimes carefully validated chemistry supports a stronger compositional estimate. The usefulness of the technique lies in matching the claim to what the experiment can truly measure.
The simplest accurate translation is often the best final sentence: “This interval shows an elevated Fe/Ti XRF intensity ratio relative to nearby sediment; the compositional and environmental cause requires comparison with core condition, calibrated chemistry and independent stratigraphic evidence.” It tells readers what changed without hiding uncertainty. That is how a technical graph becomes useful knowledge.
Sources and method notes
International Ocean Discovery Program JRSO, XRF Core Scanning Facility, https://www.iodp.tamu.edu/labs/xrf/ — instrument scope, sample preparation and operational limitations.
Babin and colleagues (2020), Data report: X-ray fluorescence core scanning of IODP Site U1474 sediments, Proceedings of the International Ocean Discovery Program 361, https://publications.iodp.org/proceedings/361/201/361_201.html — acquisition, sample calibration and observed quality issues.
Tjallingii and colleagues (2007), Influence of the water content on X-ray fluorescence core-scanning measurements in soft marine sediments, https://doi.org/10.1029/2006GC001393 — water-content influence on wet-core fluorescence intensities.
Dunlea and colleagues (2020), Intercomparison of XRF Core Scanning Results From Seven Labs and Approaches to Practical Calibration, https://doi.org/10.1029/2020GC009248 — instrument comparisons and calibration considerations.
Google Search Central, Creating helpful, reliable, people-first content, https://developers.google.com/search/docs/fundamentals/creating-helpful-content — general guidance on useful, reliable explanations, not a scientific measurement manual. All numerical examples above are explicitly illustrative and are not observations or claimed laboratory calibrations.