A grain-size curve is not a photograph of every sediment grain
Marine sediment laser diffraction uses the way suspended particles scatter light to estimate a distribution of equivalent particle diameters. A report may display a smooth curve with numbers such as D10, D50 and D90 and imply that the seabed is mostly sand, silt or clay. The interpretation is more delicate: these values depend on optical assumptions, sample dispersion, analysis range, reporting convention and the relationship between irregular mineral shapes and the spherical model used to turn scattering into sizes.
This guide explains the measurement before the classification. We will trace sediment from the corer to the laboratory suspension, see how a distribution is reconstructed from scattering, translate percentiles correctly, work through calculations and diagnose contradictory results. We will also distinguish a physical particle-size distribution from naturally occurring flocs and aggregates. All numerical examples are illustrative classroom data, not measurements from a real coastline or expedition.
A particle-size result becomes useful when the reader knows the measured basis. The difference between the smallest ten percent by volume and the smallest ten percent by number is enormous when a sample contains both clay-sized particles and sand grains. A sample with thousands of fine grains can still have most of its particle volume in comparatively few large grains. Precision begins by asking which distribution the instrument actually reports.
From suspended particles to a diffraction pattern
When a laser beam crosses a sample suspension, particles scatter light into different directions. The angular distribution and intensity contain information about particle size, refractive index and shape. Instruments collect scattered light using multiple detectors and use an inverse mathematical model to estimate the particle-size distribution that would explain those signals. The model is not watching each grain pass through a camera and measuring its longest diameter individually.
ISO 13320:2020 provides a framework for laser diffraction particle-size analysis and instrument qualification in many two-phase systems. Its applicability spans approximately 0.1 micrometres to 3 millimetres under standard conditions, with possible extensions through suitable instruments. The standard’s reported distributions are based on scattering-equivalent spheres, including when actual particles are nonspherical. This limitation is essential for platy clays, fibres, needle-shaped minerals and particles with irregular outlines.
For large particles relative to light wavelength, a simplified Fraunhofer diffraction treatment may sometimes be invoked. More general Mie scattering models account for refractive index contrasts and absorption and can be important for finer material. The chosen optical model and its input parameters should be documented. A researcher comparing two studies cannot assume their D50 values are interchangeable if one used a different refractive index, dispersant, instrument range or analysis model.
The inversion from scattering to a size distribution is not uniquely determined by a single detector signal. Software uses an instrument-specific mathematical forward model and optimisation to retrieve a plausible distribution. This is why an extremely smooth graph or many decimal places do not prove equally precise grain dimensions. Calibration, measurement stability and the appropriateness of optical assumptions matter more than how polished the curve looks.
Why preparation changes the thing being measured
Marine sediment contains mineral grains, organic matter, shell fragments, microbial films, salts and aggregates that may break apart under handling. A subsample might be wet-sieved to separate coarse material, chemically treated to remove organic matter, dispersed with a chemical agent, sonicated to break aggregates and circulated through an instrument. Each step has a purpose, but each can also change the physical particle population whose size is ultimately reported.
Suppose a muddy estuary has abundant fluffy flocs formed from clay particles and organic polymers. If a laboratory thoroughly disperses these flocs, the instrument may report the sizes of their constituent mineral particles rather than the sizes of the aggregates that settle through water in nature. Both measurements are legitimate for different questions. Neither should be described merely as the real particle size without specifying whether the method preserved or destroyed floc structure.
Salt removal and chemical dispersants deserve explicit mention. Electrolyte conditions influence how clay particles aggregate, so adding freshwater, salt solution or dispersant can change the suspension. Removing organic matter may clarify mineral grain-size distribution, yet also eliminate natural binding and reduce the size of organic-rich aggregates. The reader must match the pretreatment to the intended use, such as geological provenance, sediment transport, ecological turbidity or engineering classification.
Subsampling is another source of error. Sediment settles during storage and larger grains can segregate from fine material. Scooping a convenient spoonful from the top of a jar may not preserve the original distribution. A representative, well-homogenised subsample or validated splitting method is needed. Repeating a single biased aliquot three times can give beautiful analytical repeatability while still misrepresenting the original seabed sample.
In the laboratory, obscuration indicates the degree to which the particle suspension attenuates the light beam. Too concentrated a suspension may produce multiple scattering; too dilute a suspension can give poor signal-to-noise. The valid working range depends on the instrument and particle sizes. The analyst should adjust concentration within the method’s accepted region and record it, rather than assuming more solids always yield better data.
Ultrasonic dispersion is not automatically neutral. It may break loosely attached sediment flocs as intended, but can also alter fragile particles or aggregates relevant to the study question. A dispersion-energy sensitivity test can reveal whether the reported D50 stabilises as treatment increases. If results continue to change substantially, the protocol should explain how its chosen endpoint corresponds to a meaningful physical definition.
The three percentiles everybody sees
D10 denotes the diameter below which ten percent of the measured distribution lies on the specified cumulative basis. D50 is the median diameter, and D90 is the diameter below which ninety percent lies. In many laser diffraction outputs, these refer to a volume-weighted distribution, often written Dv10, Dv50 and Dv90. They are not automatically percentiles of particle count, mass, projected area or sediment layer depth.
If Dv50 is 45 micrometres, half the particle volume represented by the model is in particles smaller than approximately 45 micrometres and half in particles larger. It does not mean that half the individual grains are smaller. Nor does it say the average diameter is 45 micrometres. The numerical median and arithmetic mean are different statistics, especially for a broad or bimodal distribution.
Dv90 of 300 micrometres does not imply that ninety percent of particles are exactly 300 micrometres. It is a cumulative threshold: ninety percent of modelled particle volume is in the distribution below that diameter. A common mistranslation is to write ninety percent of grains are 300 micrometres, which transforms a threshold into a singular particle size and ignores the weighted basis.
The spread can be summarised using a span, often calculated as (Dv90 − Dv10)/Dv50. This ratio is dimensionless when all diameters use the same unit. It offers a simple measure of distribution breadth but loses information about separate modes or long tails. Two highly different distributions can share the same Dv10, Dv50 and Dv90. Always inspect the full curve before identifying a sediment mixture from three summary numbers alone.
Worked example one: D10, D50, D90 and span
Imagine a fictional laser diffraction report with Dv10 = 4 micrometres, Dv50 = 48 micrometres and Dv90 = 320 micrometres. The span is (320 − 4)/48 = 316/48, approximately 6.58. The numerator and denominator are both lengths, so the result carries no length unit. A span this broad signals substantial dispersion, but it does not identify which proportion is organic aggregate, mineral silt or sand without more evidence.
The report can correctly be described as follows: ten percent of the modelled particle volume is associated with diameters below 4 micrometres; half is below 48; and ninety percent is below 320. It cannot be paraphrased as ten percent of counted grains are smaller than 4 micrometres unless the instrument actually produced a number-based distribution. A very large grain may contribute volume comparable to numerous smaller grains, so a number-based translation could be radically wrong.
Now convert the same figures to millimetres. Four micrometres equals 0.004 millimetres; 48 equals 0.048; and 320 equals 0.320. Recomputing span gives (0.320 − 0.004)/0.048 = 6.58. If one value is accidentally left in micrometres while the others are in millimetres, the nonsensical result will expose a unit mismatch. A useful checking habit is to calculate all ratios after first moving every quantity into one unit system.
A report rounded to whole micrometres should not be rewritten with six decimal places of certainty. Instrument qualification, sample preparation and model assumptions remain broader sources of error. A single precise-looking percentile is a conditional estimate from a defined method. The correct presentation preserves that condition rather than implying an exact physical ruler measurement of every grain.
Sand, silt, clay and the danger of classification labels
Geologists often use conventional size boundaries such as 63 micrometres between sand and silt, and 2 micrometres between silt and clay in a widely used grain-size framework. Other standards and disciplines may use different class boundaries. More importantly, clay-sized material is a size category, while clay minerals are a mineralogical category. A particle smaller than 2 micrometres is not automatically a clay mineral; a fragment of a clay mineral may also appear in a larger aggregate.
A laser diffraction distribution can estimate the modelled particle volume below these size thresholds when the curve and chosen boundaries are known. However, it is risky to equate a laser diffraction clay fraction uncritically with a clay fraction measured by gravitational sedimentation or pipette methods. Different physical principles and nonspherical particle effects can produce systematic differences. Method-specific classification should accompany any summary of sand, silt and clay percentages.
For engineering purposes, grain-size categories are only one part of material behaviour. Plasticity, mineralogy, porewater chemistry and grain arrangement influence strength and permeability. For environmental purposes, grain size may correlate with adsorption surface area and contaminant distribution, but it does not alone identify contamination. Rewriting a sediment report as fine grains therefore means toxic would be a causal leap unsupported by particle sizing.
A chart can look bimodal when the sediment contains a mixture of relatively fine mud and coarser sand. The two peaks may correspond to different transport modes or sources, but the chart alone cannot prove their provenance. Instrumental artefacts, incomplete dispersion and model inversion can also influence shapes. Sedimentary environment, replicate distributions and independent mineral or elemental observations help decide whether a second peak is a physical population.
Worked example two: grain number and grain volume
For intuition, imagine one spherical grain 100 micrometres in diameter and another 10 micrometres in diameter of the same density. The larger sphere has a diameter ten times greater and therefore a volume one thousand times greater, because volume scales with diameter cubed. One thousand small spheres have approximately the same total volume as one large sphere. A volume-weighted instrument will treat their combined groups differently from a particle-counting method.
If a sample contains a thousand 10-micrometre grains and one 100-micrometre grain, the small grains constitute 1000/1001, or about 99.9 percent of the counted grains. Yet they collectively represent about half of the idealised spherical particle volume because each has one-thousandth of the large grain’s volume. A numerical description that omits the denominator can turn the same sample into both overwhelmingly fine and roughly half coarse without either calculation being wrong.
These perfect spheres are illustrative only. Real sediment grains have irregular shapes and variable density; an optical instrument estimates equivalent diameters from scattering rather than physically weighing each sphere. The example still exposes the conceptual issue: a percentage has meaning only after its basis has been named. The reader should distinguish volume fraction, mass fraction, count fraction and fractional area before comparing results from instruments that use different weights.
When distributions must be merged
A laboratory may use sieving for coarse sediment and laser diffraction for finer material. A full-sample distribution then requires measured mass fractions of the separated portions and compatible bases. If a sample was 30 percent coarse material by dry mass and 70 percent fine material by dry mass, simply averaging the two separately normalised percentage curves would misrepresent the original composition. The coarse and fine contributions must be weighted according to the actual sample fractions and the chosen reporting basis.
Mixing distributions is also different from averaging their percentiles. Suppose one sample has Dv50 = 20 micrometres and another Dv50 = 200 micrometres. The median of a 50:50 mixture cannot generally be found by calculating (20 + 200)/2 = 110 micrometres. The answer depends on the shapes of the full cumulative distributions and on whether 50:50 refers to dry mass or modelled particle volume. Percentiles are nonlinear summaries of distributions.
A full sediment archive should preserve enough information to reconstruct the percentage finer than each stated size boundary. Retaining only Dv50 discards information about tails that may be environmentally or engineeringly significant. Fine-grained contaminant association and coarse-grained seabed mobility can both matter in the same sample. Reporting the cumulative curve or binned data with its measurement method gives future readers more than a single median can provide.
Data quality has layers
Instrument repeatability can be assessed by rerunning the same well-dispersed suspension, but this tells little about field sampling representativeness. Replicate subsamples from one container test splitting and homogenisation. Replicate core samples test spatial heterogeneity. Calibration standards assess instrument performance. These are different sources of evidence, and a published uncertainty statement should not collapse them into one percentage without explanation.
The USGS Bascom laser diffraction laboratory description explains modern measurements using liquid suspensions and scattering patterns. The USGS Woods Hole sediment laboratory describes complementary grain-size and compositional equipment. These first-party examples show why a complete sediment description may require several distinct measurements rather than treating laser diffraction as a complete mineralogical and geochemical assay.
A traceable reference material is valuable, but it usually tests the instrument against a defined particle population under controlled conditions. Agreement with a size standard does not prove that saline organic-rich estuarine flocs were dispersed appropriately or that the subsample was representative. Quality assurance should match the scale of the claim: detector qualification, sample preparation, field sampling and interpretation each have their own failure modes.
A field investigator translating a particle-size report into a map should ask whether samples represent similar depth intervals, sedimentary environments and preparation methods. Comparing a surface grab from one station with a deep core section from another can confound habitat differences with sampling depth. Maps of median size may be informative, but interpolation across unmeasured habitats must not imply that seabed composition was directly observed between stations.
Worked example three: read a cumulative distribution
Imagine a fictional cumulative volume distribution whose fraction finer than 2 micrometres is 12 percent, finer than 20 is 35 percent, finer than 63 is 70 percent and finer than 500 is 96 percent. Using a 2-micrometre clay-size boundary and a 63-micrometre sand-size boundary, the estimated clay-size fraction is 12 percent, the silt-size fraction is 70 − 12 = 58 percent, and the sand-size fraction greater than or equal to 63 micrometres is 100 − 70 = 30 percent.
The three fractions sum to 12 + 58 + 30 = 100 percent. This basic closure check catches several reporting errors. Do not calculate sand as 96 − 70 = 26 percent merely because the distribution was given to 500 micrometres; that would omit the remaining 4 percent above 500. If the instrument’s reliable measurement range stops below the largest particles, an external coarse-fraction measurement may be needed instead of pretending the unmeasured tail is zero.
Now consider the median. Since 35 percent is finer than 20 micrometres and 70 percent is finer than 63, Dv50 lies between 20 and 63 micrometres. Its precise interpolated value cannot be recovered without knowing the distribution within that interval and the interpolation convention. A simplistic linear interpolation in diameter would yield 20 + (50 − 35)/(70 − 35) × (63 − 20), approximately 38.4 micrometres. That is only an illustrative assumption, not the original instrument’s recovered median.
Many particle-size graphs use logarithmic diameter axes. Interpolating a cumulative distribution linearly in logarithmic diameter instead gives a different midpoint estimate: exp[ln(20) + (15/35) × (ln(63) − ln(20))], approximately 32.7 micrometres. Neither interpolation is uniquely justified by the four cumulative points alone. A sound published Dv50 comes from the complete recorded or modelled distribution, not from choosing whichever interpolation produces a desired classification.
The word cumulative is essential. A cumulative curve tells how much of the distribution is finer than a threshold. A frequency curve tells how much lies within a bin or range. If a graph is labelled frequency percent against diameter, reading 70 percent at 63 micrometres as the proportion below 63 may be completely wrong. Begin by identifying the curve type, the axis scale, and whether the plotted percentage is weighted by count, area, mass or volume.
Worked example four: geometric diameter conversion and phi units
Sedimentologists sometimes use the phi scale, defined as φ = −log₂(d in millimetres). A grain with d = 1 millimetre has φ = 0; a grain of 0.5 millimetre has φ = 1; a grain of 0.25 millimetre has φ = 2. The larger the phi value, the smaller the grain. This inversion can confuse readers who expect numbers to increase with physical size. The phi axis also encodes multiplicative size differences using additive steps.
For a diameter of 0.0625 millimetres, or 62.5 micrometres, φ = −log₂(0.0625) = 4. A particle of 0.004 millimetres has φ approximately 7.97. These conversions are useful when comparing a laser diffraction report in micrometres with a traditional sedimentology plot in phi classes. It is dangerous to insert a diameter already expressed in micrometres into the formula as if it were millimetres; that introduces a factor of one thousand inside the logarithm.
A distribution percentile on the phi scale requires attention to direction. Dv90 is the large-diameter cutoff below which 90 percent of volume lies, but in phi terms the large diameters are at lower phi numbers. Replacing Dv90 with a same-numbered high percentile of phi values may reverse the tail. A careful conversion maps each physical diameter to phi without reassigning cumulative fractions blindly. In sediment archives, preserve raw diameter thresholds alongside transformed axes.
If a paper reports a mean phi size, it is not the same as taking a mean diameter in millimetres and converting that one number. Because the logarithm is nonlinear, transforming before averaging differs from transforming after averaging. Readers must know how the mean was defined and weighted. Percentiles, means and modes respond differently to distributions; none becomes interchangeable simply because they all have the word size in the column header.
Distinguish grain size from settling behaviour
A particle’s terminal settling velocity depends on more than its optical equivalent diameter. Density contrast, shape, fluid viscosity, turbulence and aggregation all matter. Two particles with the same reported laser-diffraction diameter can sink differently if one is dense quartz and another a porous organic aggregate. A median sediment size is therefore not a directly measured settling velocity or a guaranteed indicator of the transport energy required to move the material.
Stokes’ law can estimate slow settling of sufficiently small spherical particles in a low-Reynolds-number regime, but its assumptions should be checked before application to irregular sand grains or complex marine flocs. Introducing a Stokes settling speed as though it were measured by a diffraction analyser would mix two models and imply unjustified certainty. The laser instrument measures a scattering pattern; a separate hydrodynamic model estimates motion under specified conditions.
If sediment is sampled after a storm, coarse particles may have been winnowed or deposited through a process influenced by waves and currents. A grain-size distribution can support a sediment-transport hypothesis when combined with environmental observations, but it does not by itself reveal exact current velocity or wave height. Nearby records of bottom currents, wave exposure and seabed morphology help distinguish transport explanations. The ocean ADCP guide
A change from mud to sand down-core might represent a change in sediment supply, habitat conditions, current sorting, erosion or an event bed. A chronological explanation additionally requires an age model. The sediment lead-210 age guide
Optical assumptions in fine sediment
For mineral grains with high optical contrast relative to water, a chosen refractive index may work acceptably within a validated method. However, clay plates, shell fragments and organic aggregates have differing shapes and optical properties. The inverse model expresses these as equivalent scattering sphere diameters. The reported value should not be described as the grain’s largest dimension, geometric thickness or hydrodynamic diameter unless a separate technique establishes that relation.
Particles near the lower end of an instrument’s measurable range can be especially sensitive to optical model parameters, signal strength and background. An analyst should investigate whether a small apparent clay tail is robust to reasonable changes in input assumptions and sample handling. The presence of a tail in a visually smooth software plot does not prove its absolute abundance with high precision. A model may distribute inferred volume across bins as part of its optimisation process.
The ISO 13320 discussion of qualification matters because laser diffraction is a method family rather than a single immutable instrument. Detector layout, light source, inversion method, disperser configuration and size range differ between instruments. A result from one lab may show a different fine fraction from another even when both are competently operated. Interlaboratory comparisons and transparent methodological metadata can identify systematic differences. Scientific comparison should begin with methodological compatibility, not merely matching the same unit labels.
If an experiment measures undispersed aggregates to study natural settling, then chemical dispersion in the analysis would answer a different question. If the purpose is mineralogical provenance, disaggregating flocs may be appropriate. Thus the correct preparation is conditional on the research objective. A reader evaluating a dataset should ask: are we measuring primary grains, loosely bound aggregates or operationally defined particles after a particular treatment? The method’s answer belongs in the interpretation.
How a quality-control record helps
A useful data package states the instrument model, optical configuration, sample medium, refractive index assumptions, dispersion protocol, sonication settings, obscuration range, replicate design and size bins. It also identifies how large grains were handled and whether organic matter or carbonate was removed. Reporting Dv50 to a decimal place without any of this supporting information is insufficient for a reproducible environmental comparison.
Control samples should be chosen intelligently. A traceable particle-size standard can reveal instrument drift and gross calibration errors. Repeated suspensions reveal analytical repeatability. Independent splits from a homogenised sample reveal preparation variability. Replicate field samples expose spatial variation. These are separate layers of uncertainty. A small instrument repeatability figure should not be used as the error bar for an ecological map spanning heterogeneous tidal habitats.
The USGS grain-size data release
A strong uncertainty discussion also distinguishes between a systematic shift in the entire size distribution and variability among field replicates. Different preparation protocols may shift all Dv50 values in one direction; repeated sampling may reveal scattered values because actual sediment varies. These are not the same problem and they require different responses. Method harmonisation addresses systematic comparability; spatial sampling addresses heterogeneity. Combining them indiscriminately into one standard deviation hides the mechanism.
What a scatter plot cannot prove
A scatter plot of Dv50 against organic carbon concentration may show an association because fine sediment retains organic matter more effectively or because depositional settings influence both. The association alone cannot establish that small grains caused carbon accumulation. Changes in biological production, mineral surface chemistry, hydrodynamics and sampling depth can influence both variables. A sound interpretation proposes a mechanism and evaluates alternate explanations before writing a causal headline.
Similarly, grain-size classes cannot reveal precise mineral compositions. Quartz sand, carbonate fragments and volcanic glass can occupy overlapping diameter ranges but differ chemically and physically. X-ray diffraction, chemical analyses or microscopy may be needed when the research question concerns material source or contaminant binding. Laser diffraction is a size-distribution method. It does not secretly measure every other property associated with sediment.
An optical distribution can be a useful monitoring signal after environmental change, but only if protocols are comparable across dates. A shift in Dv50 from 35 to 50 micrometres may reflect true coarsening, different sample layers, different dispersion, changed calibration or a seasonal shift in flocculation. A robust temporal study controls or reports these alternatives. Making the instrument part of the explanation protects the scientific value of the observed change.
Independent transfer questions
Question one. A cumulative volume curve reports 8 percent finer than 2 micrometres and 64 percent finer than 63 micrometres. Using these boundaries, estimate clay-size, silt-size and sand-size volume percentages. Verify that they close to 100 percent and explain what the calculation does not reveal about mineralogy.
Answer one. Clay-size is 8 percent, silt-size is 64 − 8 = 56 percent and sand-size is 100 − 64 = 36 percent. Their sum is 100 percent. These are equivalent-diameter classes on a stated distribution basis. They do not prove that all particles below 2 micrometres are specific clay minerals, nor that all coarser material is quartz. Composition requires additional evidence.
Question two. Dv10 = 6 micrometres, Dv50 = 60 micrometres and Dv90 = 420 micrometres. Calculate the span and say whether its unit is micrometres. Then state whether the values alone prove the distribution contains two peaks.
Answer two. Span = (420 − 6)/60 = 414/60 = 6.9, which is dimensionless. The three percentiles show a broad distribution but do not prove bimodality. Two different distribution shapes can share these percentiles. Inspect the original frequency or cumulative curve and assess sample preparation and optical inversion before inferring distinct particle populations.
Question three. A researcher reports Dv50 = 100 micrometres from a fully dispersed laboratory sample and claims that all naturally settling flocs have a median physical size of 100 micrometres. What is wrong? Identify the physical quantity the instrument more directly estimates and an alternative observation needed for the floc claim.
Answer three. Dispersion may have destroyed the aggregates that existed in the water column. The laser result describes a modelled scattering-equivalent size distribution of the treated suspension. To characterise natural flocs, one needs measurements designed to preserve or directly observe aggregates under relevant conditions, with documented aggregation and transport context. A laboratory mineral-grain median cannot simply substitute for in-situ floc size.
Question four. One sample’s Dv50 is 20 micrometres and another’s is 200. Can the combined Dv50 be declared exactly 110 for equal sample masses? State the missing information and a defensible method of combination.
Answer four. No. The combined percentile depends on both full size distributions and their compatible weighting, including density and the volume-versus-mass basis. Construct a properly weighted combined cumulative distribution and locate its 50-percent threshold. Arithmetic averaging of two medians is not generally valid. Equal masses do not necessarily mean equal particle volumes when material densities differ.
Question five. A cumulative plot uses a logarithmic horizontal axis. Its curve crosses 50 percent between 16 and 64 micrometres, but the actual intervening bins are unavailable. Can an exact median be recovered? Explain why arithmetic interpolation and log interpolation disagree.
Answer five. There is no unique exact median without the underlying distribution or a justified interpolation model. A straight interpolation in physical diameter and a straight interpolation in log diameter represent different assumptions about how cumulative volume changes within the interval. They yield different estimates because logarithm is nonlinear. Reporting either as a measured exact Dv50 would overstate the available evidence.
Question six. An instrument reports Dv50 in millimetres but a map labels it as micrometres without changing the numbers. What is the systematic error? Give an example and describe why this error may be detected by a simple plausibility check.
Answer six. The labels introduce a factor-of-one-thousand error. A value of 0.150 millimetres is 150 micrometres, not 0.150 micrometres. A map claiming sand is submicrometre clay-like material may be inconsistent with sediment photographs, sieving results or known instrument size ranges. Check the physical unit before comparing any trend or ecological interpretation.
Frequently asked questions
Does Dv50 mean the average grain size? It means the median of a specified volume-weighted equivalent-diameter distribution, not necessarily the arithmetic mean. An ordinary word such as average can conceal which statistic was used. Good technical translation uses median when a median is intended and states the weighted basis.
Can a laser report identify the mineral responsible for a contaminant? No. The size distribution may help investigate sorption and transport, but mineralogical identity and contaminant concentration require other analyses. A fine fraction might concentrate a metal under some conditions, yet that is an empirical relationship to measure rather than a guaranteed property of all fine sediments.
Does more sonication always produce a more accurate result? No. Sonication changes particle aggregation and may alter fragile material. It can be useful for a mineral-particle question and destructive for an aggregate-transport question. The right level should be justified by the intended measurement definition and a stability or sensitivity check.
Why do two laboratories disagree on clay-size percentage? Sample splitting, pretreatment, optical assumptions, inversion settings, instrument range and classification boundaries can differ. Some of the discrepancy may be a systematic method difference rather than poor operation. Compare protocols and raw distributions before alleging that either laboratory is wrong.
Can a Dv90 value prove that the largest grain has that diameter? No. Dv90 is the threshold below which ninety percent of the chosen weighted distribution lies. Ten percent lies above it. A maximum diameter is a different statistic and may also be limited by the instrument’s measuring range.
How does terminology affect data translation? Terms like particle, grain, aggregate, floc, size fraction, median, volume and concentration each describe different aspects of the method. The eduKate Vocabulary Learning System
A defensible interpretation in one paragraph
A careful statement might say: After the specified dispersion and optical analysis, this sample had a modelled volume-median equivalent particle diameter of approximately the reported Dv50, with the stated coarse and fine fractions according to declared size boundaries. The value characterises the treated aliquot under the laboratory protocol. It does not independently establish mineral identity, natural floc size, transport velocity or the history of a seabed layer.
An effective reading workflow moves from sample provenance to preparation, from preparation to optics, from optics to cumulative distribution, from distribution to the correct percentiles and from percentiles to a narrowly justified environmental interpretation. The discipline is simple: never ask one instrument to answer a question it did not measure. That prevents both careless units and confident but unsupported stories about marine sediment.
Primary sources and methodological references
ISO 13320:2020 Particle size analysis — Laser diffraction methods
USGS Bascom Laser Diffraction Sedimentology Laboratory: analytical principles and equipment
USGS Woods Hole Coastal and Marine Science Center Sediment Laboratory: instrumentation
USGS grain-size data release: analytical checks and rounding caveats
A deeper worked case: sampling a seabed transect
Imagine three fictional seabed stations surveyed along a coast, with modelled volume medians of 18, 74 and 260 micrometres. Station A is a sheltered muddy setting, B lies near a channel margin, and C is an exposed shallow sand bed. The ordering is consistent with the idea that hydrodynamic sorting can produce coarser sediment in more energetic environments. But the medians are observations from laboratory-treated subsamples, not independent measurements of wave energy. The interpretation needs location, depth, sampling date, full distributions and physical current or wave evidence.
Suppose the sites differ in dispersion procedures: A was treated with dispersant and sonication, B was lightly stirred, and C was sieved before laser analysis. A’s low median may partly reflect disaggregated mud, while B’s median might retain flocs and C’s distribution might exclude measured coarse sand unless the fraction was properly recombined. Before making a coloured map, harmonise the methods or quantify their effects. A map can be visually compelling yet scientifically inconsistent if methodological differences masquerade as geographic variation.
A suitable cross-station comparison should include the fraction below 2 micrometres, the fraction below 63, median diameter, coarse tail and descriptions of each sample’s preparation. The map legend should say laser-diffraction volume-equivalent size after specified treatment, not simply true seabed grain diameter. Where stations are sparse, label extrapolation rather than filling every empty area with certainty. Unmeasured zones do not become observed merely because a computer smoothly interpolates colours between a few points.
Now imagine a repeated visit to station B six months later showing Dv50 decreasing from 74 to 49 micrometres. The change may reflect finer deposition, different weather before sampling, a different sediment layer, altered organic flocculation or a laboratory change. A convincing temporal interpretation would demonstrate comparable field sampling depth, preparation and optical configuration, with replicates sufficient to separate natural variation from instrument differences. A single changing median is a lead for investigation, not a complete history.
The full curve matters more than a headline median
Suppose two samples each have Dv50 of 60 micrometres. Sample X has a narrow, single-peaked distribution around that diameter, while sample Y combines very fine clay-size particles and coarse sand in approximately balanced volume fractions. Their medians may match even though their water retention, permeability, contaminant sorption and erosion behaviour differ. This illustrates why the median is useful for concise summary but inadequate as a full physical description.
Distribution tails can also reflect meaningful changes that the median barely registers. A small addition of extremely coarse shell fragments may increase Dv90 while leaving Dv50 almost constant. A modest fine-clay fraction may alter sediment cohesion and surface chemistry while contributing too little total particle volume to shift the median substantially. If the research concerns sediment fate or environmental exposure, report the fraction relevant to that process rather than relying only on the most familiar percentile.
A careful comparison of two curves asks whether they were normalised to the same total, whether bins use consistent widths and whether the displayed ordinate is frequency per bin or frequency per logarithmic size interval. A peak in a histogram with wide bins can look larger simply because more sizes were grouped inside it. Some software reports distributions as volume density per log diameter, which should not be mistaken for a fraction in one equal-size bin. Axis labels are scientific information, not decoration.
A cumulative curve avoids some visual binning artefacts but can conceal a multimodal shape. The derivative of a cumulative distribution with respect to an appropriately defined size coordinate recovers a density-like curve, subject to smoothing and noise. Readers need not calculate that derivative to understand the point: the same cumulative thresholds can arise from different mixtures. Keep both the cumulative and differential views available when transport or source interpretation depends on distribution shape.
Model equivalence and method comparison
Laser diffraction and sedimentation-based methods do not necessarily produce identical clay fractions even on the same mineral material. Settling methods infer an equivalent diameter from fluid dynamics and particle movement; diffraction methods infer an optical equivalent diameter from scattering. Platy clay particles violate the ideal spherical assumptions in different ways. One method can be repeatable and internally valid while still disagreeing systematically with the other. This is method dependence, not an automatic fraud finding.
A comparative study should ideally analyse the same well-homogenised samples under both methods, with transparent pretreatment and an uncertainty analysis. If a calibration relationship is developed, its validity should be assessed across mineralogy and size distribution rather than applied as a universal factor. Converting one clay percentage into the other by a convenient constant can fail when particle shapes or optical contrasts vary across sites. The original measured quantity should remain available alongside any harmonised estimate.
A sediment core may include carbonate shells that are larger and optically different from the fine siliciclastic matrix. Removing carbonate to study the siliciclastic fraction can be appropriate for provenance questions, but then the resulting distribution no longer describes the original total sample. If the study examines habitats or total seabed texture, carbonate removal may erase an ecologically important component. The measurement question must determine the preparation, and the published description must name that choice.
The same reasoning applies to organic matter. Oxidising organic material may free mineral particles for size analysis but destroy natural aggregation. For an investigation of how mud binds pollutants, both untreated and pretreated distributions might be useful because their contrast reveals aggregate sensitivity. For a transport model, in-situ particle imaging and settling experiments could complement laser diffraction. A single method need not do everything, but every method’s boundary should be explicit.
A numerical check on sample fractions
Suppose a dried sediment sample weighs 100 grams. Sieving shows 25 grams in a coarse fraction that the laser instrument cannot measure reliably; 75 grams are in the fine fraction analysed by laser. The instrument reports 40 percent of its analysed fine-fraction volume below a selected diameter. It would be wrong to announce that exactly 40 percent of total sample mass is below that diameter. The instrument result is volume based within a subset, while the whole-sample split is mass based.
If coarse and fine fractions had identical particle density and the whole 75-gram fine fraction were represented by the laser sample, the corresponding estimated whole-sample volume fraction below the threshold would be 0.75 × 0.40 = 0.30, or 30 percent. But this simplified conversion depends on equal density and compatibility of the separation boundary. If the coarse fraction is dense mineral sand and the fine fraction includes low-density organic aggregates, mass fractions and volume fractions differ. A report should explain the assumptions before combining them.
If the same sample instead has a measured coarse dry mass of 10 grams, fine mass 90 grams and the same 40-percent fine-fraction volume percentage, a naive equal-density approximation would give 36 percent of total volume below the threshold. The change from 30 to 36 percent comes entirely from the coarse-versus-fine mass split, not from a change in the laser curve. This is why a fine-fraction analysis must never be silently presented as if it represented the unsieved whole sample.
Even when density assumptions are justified, the retained coarse fraction needs its own size distribution to reconstruct the complete upper tail. Treating every sieved grain as one diameter invents a false mode. A full-sample cumulative curve is built by merging compatible measured contributions, preserving which range each method covers and checking that the final fractions total 100 percent. A credible dataset is a traceable combination of measurements, not merely a pasted chart.
Final diagnosis of apparently contradictory laboratory results
If a sediment is described as clayey in a field log but the laser instrument reports a sand-dominated volume distribution, first ask what clayey meant in the field: stickiness, colour, plasticity or a formal laboratory size class. Then check whether the aliquot was representative, the instrument measurement range captured coarse material, the sample was overdispersed or insufficiently dispersed, and the graph was read on the correct basis. Apparent contradictions often come from comparing unlike definitions.
If two versions of the same report disagree by a factor of one thousand in median size, inspect micrometre-to-millimetre conversion first. If a graph’s cumulative percent falls rather than rises as diameter increases, check whether it displays percentage coarser rather than percentage finer, or whether the x-axis is phi with reversed physical size direction. If sand, silt and clay sum above 100 percent, inspect overlapping boundaries, rounding and whether fractions refer to separately normalised subsamples.
If the Dv50 is stable but Dv90 changes dramatically after mild sonication, a coarse aggregate population may be breaking apart while the bulk median remains steady. The statement no change in sediment would then be too strong; an important tail changed. Conversely, if only the fine tail changes near the instrument limit when optical parameters are adjusted, the inference about newly deposited clay may be fragile. Sensitivity tests are useful precisely because they identify which part of a distribution is method-dependent.
The best translation preserves both the result and its scope: this is a distribution of modelled scattering-equivalent particle sizes for a sample prepared in a documented manner. From that statement, a reader can move responsibly toward classification, sediment transport, pollution studies or engineering questions, but each additional inference needs its own evidence. Clear language prevents the laboratory’s strongest measurement from being stretched into its weakest environmental claim.