A fluorescent particle is a marker of movement
Why would a scientist sprinkle brightly coloured particles across a dark seabed? The luminous particles, called luminophores, can make sediment mixing visible over time. Researchers place a known tracer layer or distribution, allow natural animals and physical processes to redistribute the particles, then recover images or depth slices. The central result is where the marked particles travelled, how much stayed near the surface, and whether the distribution changed in a way consistent with a specific particle-transport model. A glowing band is a record of tracer position, not a direct measurement of microbial respiration.
Bioturbation is the disturbance and redistribution of sediments by living organisms. Deposit feeding, burrow construction, locomotion and sediment expulsion can move solids vertically or horizontally. Fluorescent particles can help distinguish these motions from ordinary sediment deposition, but only if their behaviour resembles the grains of interest. A laboratory study involving one species cannot be translated into the typical mixing rate of all seabeds. The report must preserve organism, particle type, elapsed time, sediment setting, recovery method and model assumptions before its numerical coefficient means anything.
Particle mixing differs from solute exchange
Marine animals can rearrange sediment particles and also pump water through burrows. The first is sediment reworking or particle mixing; the second is often described as bioirrigation and can accelerate exchange of dissolved substances. A luminophore particle tracer primarily observes solid-particle relocation, not necessarily the water flux driven by a burrow. The two processes may occur together, but a colour image of transported grains does not by itself quantify oxygen, nitrate or sulfate transfer in porewater. Those solute pathways require other measurements or coupled models.
Translation mistakes arise when the general phrase mixing is used without a material basis. Moving one gram of labelled sand into a deeper layer is not the same as introducing a litre of oxygenated water there. A worm may move particles in one direction and ventilate a burrow in another. A robust explanation identifies whether the measured object is a solid grain, a solute, an animal body or a tracer model parameter. This discipline lets readers connect biological activity with chemical consequences without claiming they were directly measured by the same instrument.
Why a zero-depth line matters
A common setup begins with a layer of marked particles at the sediment surface. Time zero is the moment the tracer distribution is established, not necessarily the start of animal activity or the time a photograph is taken. Depth zero may refer to the sediment-water interface, a measured image datum or the centre of the upper sampling slice. If the surface is uneven, shifting these definitions can alter the apparent distance a tracer has travelled. A depth plot should therefore preserve the reference, the thickness of each slice and the uncertainty in locating the interface.
Imagine a picture showing particles five centimetres below a plotted line. Before saying they travelled five centimetres, ask whether their starting layer was at the line, whether the line moved as sediment eroded or accumulated, and whether some tracer was introduced below the surface during setup. If the bed surface changes, net burial can imitate a downward movement in a fixed coordinate frame. Comparing tracer positions requires a common geometric basis. A precise number with the wrong datum is less informative than a correctly qualified observation.
Colour is detection, not a direct mass measurement
A fluorescent camera registers light intensity or bright pixels, while a microscope may count individual particles. Some studies infer tracer quantity from calibrated brightness, others from counts after separating material by depth, and still others from a digital threshold applied to images. These are different observables. More bright pixels can indicate more fluorescent material, but the relation also depends on exposure settings, particle overlap, optical attenuation and background reflections. The measurement must be calibrated before a brightness profile becomes a reliable tracer-mass profile.
If image processing uses a binary threshold, the selected cutoff can change which dim particles are counted. A photograph taken under different lamp intensity may contain an apparent loss or gain of tracer even when the grains have not moved. A good methods section reports illumination, detection settings, background corrections and calibration against known tracer quantities. When interpreting a colour figure, check whether its scale is particle count, fluorescence signal, fractional recovery or predicted tracer concentration. Never replace these labels with one vague word such as abundance.
Biological particle transport has several mechanisms
A burrowing animal can mix grains slowly through repeated small movements, carry a packet of sediment into a burrow, ingest particles at one depth and eject them at another, or deposit excavated material on the surface. These routes create different tracer patterns. Small local displacements may produce a smooth broadening of the original layer; rare large movements can create isolated particles far below the bulk distribution. The fact that a particle ends deep does not prove it passed through every shallower layer in sequence.
This is why a simple particle-diffusion analogy is useful but incomplete. The analogy turns many small motions into a continuous-looking spreading process, yet real sediment organisms operate with feeding strategies, burrow geometry and discrete events. The model cannot be assumed merely because a best-fitting curve looks smooth. The scientist should examine whether the tracer profile shows a secondary deep maximum, a long tail, abrupt jumps or organism-specific structures that a purely local diffusion model struggles to explain.
An operational biodiffusion coefficient
Many studies report a biodiffusion coefficient, usually written D_b, with units of area per time such as square centimetres per year. This number describes the effective spreading rate in a specified mathematical model of particulate reworking. It is not the molecular diffusion coefficient of oxygen or a fixed property of a biological species. Its value may depend on observation duration, tracer characteristics, depth, organism density and the model’s treatment of boundaries. The same bed can support different fitted values under different experimental assumptions.
The simplest one-dimensional relation writes the time change in tracer concentration as D_b multiplied by its second depth derivative, assuming constant D_b and omitting other transport terms. The mathematics describes redistribution down gradients in the tracer profile. But an animal picking up grains at the surface and ejecting them into a deep gallery is a nonlocal process; a model requiring only adjacent-depth exchange may represent it poorly. When reporting a fitted D_b, retain the model definition and avoid presenting the coefficient as directly measured animal velocity.
The units of a mixing coefficient
Square centimetres per year is a spreading unit, not a distance and not a percentage. A value of 10 square centimetres per year cannot be described as ten centimetres of downward burial per year. For an ideal unbounded, one-dimensional random walk with a constant diffusion coefficient, mean squared displacement grows as twice D times elapsed time. A root-mean-square displacement has length units and involves a square root. This relation supplies intuition, but sediment surfaces, particle selection and episodic events often make its assumptions questionable.
For example, a hypothetical Gaussian random walk with D_b of 0.10 square centimetres per day over 20 days has mean squared displacement of four square centimetres and an RMS displacement of two centimetres. That is not a forecast that each tracer grain will travel exactly two centimetres down. Some would move very little and some much farther, and a reflecting sediment surface modifies the distribution. The numerical exercise shows how units connect in an ideal case; actual luminophore profiles require suitable geometry and model fitting.
A profile broadens, but what does the shape show?
A time-series plot may show a sharply concentrated surface tracer layer initially, followed by a wider distribution with depth. This broadening can be consistent with particle mixing, but its shape gives clues about the mechanism. A near-Gaussian profile suggests a different process from one with an isolated deep pocket or a surface layer that barely changes while deep burrows gain tracer. Always inspect the residual between observations and model predictions, not just the reported mixing coefficient. Two models can produce similar overall spreads while assigning different mechanisms.
A colour map can be visually persuasive even when most of the tracer remains close to the interface. Read its normalisation: a relative concentration plot may make a tiny deep fraction look prominent if the colour range is adjusted. Conversely, a mass-weighted figure may hide rare deep movements that are ecologically important. The best interpretation reports both the total recovered fraction and the vertical distribution at each sampling time. The first checks conservation and recovery; the second describes the pathway by which particles changed position.
Surface loss is not necessarily downward mixing
Particles initially placed at the surface can leave the observed system through resuspension, grazing, lateral transport, washout or handling loss. A decreasing surface count alone does not prove that the lost particles entered deeper sediment. Search for corresponding tracer recovery at depth, in overlying-water collectors or in other mapped locations. If only a small fraction of the original label is recovered, a fitted profile of the surviving particles may not describe the whole experiment. Mass balance is a more informative quality control than appearance alone.
Imagine a core with 1,000 labelled particles initially and 600 recovered after a tide-exposed deployment. Two hundred are detected in deeper slices, while four hundred remain near the original layer. Without accounting for the missing four hundred, it is wrong to say that 60 percent of all tracer was biologically mixed or that all missing particles moved below the sampled depth. A carefully worded report states recovery and the measured subsurface fraction separately. This protects environmental inference from an attractive but incomplete picture.
Luminophore size and density can change the experiment
Marked grains are useful when their size, density and surface characteristics resemble the sediment fraction of interest. A very coarse fluorescent sand may not be handled like naturally fine organic-rich particles by deposit feeders. Some animals select food-rich grains or ingest particles according to size. A tracer that is visually easy to count may therefore represent a restricted class of physical particles rather than general sediment mixing. Reports should identify grain size range, material, density where known and any treatment used to make the grains fluorescent.
A difference in mixing between two particle classes can be a genuine biological finding rather than an error. For example, if one animal transports finer particles deeper than coarse grains, the experiment may reveal selective transport. But it cannot be condensed to a single universal D_b without losing that result. Before comparing studies, ask whether one used coarse dyed quartz grains and another used smaller organic-associated material. Model coefficients derived from differently selected tracers can differ even in the same community.
Sampling depth intervals control what a profile can resolve
A core divided into two-centimetre slices yields a profile whose basic spatial support is two centimetres per datum. A different analysis using two-millimetre slices can observe thinner gradients, but it may also have fewer tracer particles per slice and greater counting uncertainty. The centre of a slice is a label for the interval, not an exact location of every particle inside it. Interpolating a finely spaced curve through coarse data does not create sub-slice resolution.
If a deep particle occurs in one coarse layer, the observer may know its depth only within that interval. When fitting a model, predictions should ideally be compared with integrated amounts over the same slice boundaries rather than point concentrations sampled only at centres. A thick slice may blend a narrow burrow accumulation with background material. The spatial scale of the analysis must be explicit in a meaningful statement such as tracer extended to the six-to-eight-centimetre interval.
Time-lapse images can reveal individual events
A single end-point profile describes where tracer resides after an experiment, not necessarily how it travelled. Repeated images, when properly registered, can show sudden jumps, episodes of surface excavation or the formation of a new gallery. The 2004 work by Solan and colleagues combined fluorescent sediment profile imaging with model analysis to illustrate the value of observing discrete bioturbation events. In such a setting, a time-resolved trace may help distinguish continuous local mixing from an animal-driven nonlocal transfer.
However, camera time resolution is not a guarantee that every physical event is visible. Particles can move outside the image plane, become obscured or cross the field between frames. An animal may disturb a layer without exposing the marked grains to the optical window. Interpreting an event requires a stable camera geometry, a clear observation frame and an honest account of occlusion. A fluent translation should preserve the difference between observed movements and modelled movements inferred from the final profile.
A smooth model can hide nonlocal relocation
The local biodiffusion model represents mixing as many small exchanges between neighbouring depths. Burrow transport can instead relocate material from the surface to a deeper location in one event, producing a separate deep peak. If a local model is forced onto that pattern, it may yield a large effective D_b or systematic residuals rather than a physically meaningful description of the process. A nonlocal transport model introduces terms representing jumps or exchange between separated regions. The models answer different questions about how particles are moved.
Do not describe a nonlocal reworking constant as if it were another simple diffusion coefficient with the same units. Its dimensions depend on how the specific equation defines transport frequency and depth distribution. Similarly, a fit that uses two mechanisms may explain a profile without uniquely identifying the organism responsible. Biological observation, such as imaging or fauna surveys, is needed to connect movement mechanics to taxa. Mathematics helps quantify the pattern, but does not turn a tracer silhouette into a species identification.
A model residual is evidence, not embarrassment
After a diffusion or nonlocal model predicts a tracer distribution, the difference between predicted and observed values provides a practical diagnostic. Consistent underprediction at deeper layers can indicate that a local model misses long-distance transport. Repeated overprediction near the surface may point to an incorrect boundary or starting condition. Residuals can also expose image errors, depth misregistration or an unsuitable normalization. Report the shape and units of those differences rather than citing only a goodness-of-fit score.
Many fitted models minimise a weighted sum of squared residuals. The weighting matters: an interval with few particles may have greater relative counting uncertainty than a dense layer, yet both can be ecologically important. A model that fits the large surface peak very well may still fail to reproduce rare deep transfer. Readers should examine the raw profile, fitted curve and residual pattern together. One coefficient without these supporting checks is too compressed to support a strong mechanistic claim.
Physical disturbance can impersonate an animal
Waves, currents, dredging, core collection and handling can redistribute tracer grains independently of benthic animals. A field experiment may deliberately study the interaction between biological mixing and storms, but the mechanisms must be distinguished before attributing all movement to fauna. The recent 2026 study titled Worms and storms examined combined sources of sediment mixing, illustrating why an apparently biological profile can have a physical component. Controls without animals, hydrodynamic records and independent disturbance measures can help test attribution.
A tracer layer placed on the bed before a high-energy event may be buried beneath new sediment even if its grains are not vigorously moved through the preexisting deposit. That is a vertical-coordinate and deposition problem as much as a bioturbation problem. The conclusion must state whether burial, erosion and genuine particle exchange were separated. Calling every downward shift mixing overlooks the possibility that the sediment surface itself moved. A thoughtful report gives biological behaviour and sediment transport their own evidentiary roles.
Normalisation by recovered particles
A histogram of tracer by depth may show particle counts or a fraction of the recovered total. If an experiment starts with 1,000 grains and recovers only 800, a slice containing 200 grains represents 25 percent of the recovered grains but 20 percent of the initial addition. Both percentages are mathematically valid when labelled correctly; they answer different questions. The former describes the vertical distribution conditional on recovery, while the latter contributes to the overall tracer budget. Omitting the denominator hides information about loss and potential selection.
When two studies display percentages at depth, check whether their profiles sum to 100 percent of material recovered, material deposited, or material visible to the camera. Surface erosion can change the recovered total without necessarily changing the profile of the surviving particles. A normalised graph may therefore look very similar between treatments while one treatment lost half the tracer. Preserve both the absolute recovery and the normalised spatial profile whenever possible. This prevents false equivalence between a stable sediment bed and an actively resuspended one.
A particle image is not an independent replicate
A single field image can contain hundreds of tracer grains, but those grains usually share the same experimental environment and animal community. They are repeated observations within one treatment unit rather than hundreds of independent replicates of the habitat. If investigators compare sediment treatments, the replicate may be the entire core, tray or deployed camera, depending on the design. Counting all pixels or grains as independent samples can produce artificially small error bars and exaggerated significance.
A good figure shows the number of independent cores or deployments and the spatial hierarchy of sampling. Multiple depth slices from the same core are not independent sites; multiple frames from a time-lapse sequence are not independent start-to-finish experiments. Repeated observations are still valuable, especially for identifying time-dependent change, but their correlation should be respected. Readers need the experimental unit to interpret p-values or confidence intervals. The numerical resolution of an image is different from the strength of replication.
The duration of the experiment changes the meaning of D_b
A coefficient estimated over a few hours may be dominated by a few discrete movements, while one estimated over several weeks may average many feeding and burrowing episodes. In some systems particles experience a long rest period punctuated by sudden displacement. This violates the simplest assumption that effective mixing is stationary and diffusive at every timescale. A short-term D_b and a long-term D_b can therefore differ without either one being a calculation error. The observation window is part of the reported parameter.
Suppose one animal moves a surface layer once early in a twenty-day experiment and remains inactive thereafter. A snapshot at day one and another at day twenty might show the same depth distribution, yet a naïve fit dividing the total spread by elapsed time could yield very different coefficients. The biology is episodic, not continuously diffusive. Time-lapse observations and models allowing nonlocal events can be preferable. Always report the actual observation interval and avoid translating fitted D_b into a timeless species trait.
An ideal random-walk exercise
Consider an idealised unbounded one-dimensional tracer distribution with a measured mean squared displacement of 6 square centimetres after ten days. Under the simple relation mean squared displacement equals 2 D_b times time, the implied D_b is 6 divided by 20, or 0.30 square centimetres per day. Converting with 365 days per year gives about 109.5 square centimetres per year. This example demonstrates the geometry and units of the diffusion analogy, not a field estimate of the mixing rate at a real seabed.
Now ask what changes when a physical sediment surface prohibits upward travel beyond the interface. The unbounded random-walk assumption no longer applies directly, and a reflecting or eroding boundary alters the predicted distribution. Burrow transport can also introduce large jumps that do not obey the small-step picture. A calculation giving 109.5 does not validate the model itself. Check the starting distribution, domain boundaries, particle recovery and temporal stationarity before accepting a coefficient from a fit.
Translating the difference between rate and extent
A tracer extending eight centimetres below the surface reveals a maximum observed extent within the sampled image or core, not eight centimetres per year of mixing. If the experiment lasted ten days, dividing eight by ten gives a simple average displacement of a furthest observed particle, but that is neither the distribution’s mean velocity nor a defensible biodiffusion coefficient. A few rare deep particles can establish an impressive maximum depth while most grains remain in the top centimetre.
A clearer summary reports the median tracer depth, perhaps a chosen percentile, the deepest resolved positive interval and the overall recovery where those metrics are appropriate. Their values capture different properties. The median describes a typical recovered tracer position; a deep percentile describes a tail; the maximum is sensitive to one particle or image artifact. Readers should not substitute one for another simply because all are expressed in centimetres. Measurement purpose determines which statistic is most useful.
Particle type must be matched to biological behaviour
Marine animals may sort particles by size, density, food quality or attachment to organic matter. A fluorescent quartz grain can behave differently from a flocculated organic particle even at the same nominal diameter. When comparing tracer studies, identify the material’s properties and whether the organism is a surface feeder, a conveyor-belt feeder or a burrower. A coefficient derived from coarse labelled sand may describe how that size fraction moves, not the fate of carbon-rich fine particles or contaminants associated with another phase.
A useful experimental strategy, where scientifically justified, compares distinct tracer fractions under the same conditions. If finer particles move deeper more often than coarse particles, this may reveal selective handling rather than a simple difference in overall animal activity. But each tracer type still needs its own count calibration and recovery. A blended result may obscure precisely the selectivity the experiment could detect. Preserve explicit particle-size and material qualifiers in both the body and figure captions.
Deposition, burial and mixing must not be confused
New sediment accumulating over an original tracer layer shifts its depth relative to the current seabed. That shift can occur without the marked particles spreading into neighbouring layers. True mixing changes their relative positions within the sediment matrix, while deposition alters the vertical reference and burial. Erosion can remove overlying sediment and make a previously buried layer appear shallower. A depth-time profile should therefore be interpreted with independent evidence of surface elevation, accumulated material or erosion.
Imagine a thin fluorescent layer placed on a bed that later receives two centimetres of clean sediment. The layer now lies two centimetres below the new surface, even if it remained intact. Reporting that animals transported it two centimetres downward would be false. The more accurate statement is that the tracer was buried beneath newly deposited material. If the layer also broadens, physical or biological mixing may have occurred, but the two processes must be disentangled. This is a small distinction with major consequences for sediment budgets.
Environmental and laboratory studies answer different questions
Controlled laboratory cores can isolate a known animal density and grain type, but their walls, feeding conditions and hydrodynamics differ from the sea. In situ deployments better represent the field but may be affected by unknown fauna, tides, waves and lateral transport. Neither approach is automatically superior for every question. A controlled core can identify a mechanism, while a field deployment can test its significance under more realistic but less controlled conditions. The interpretation should match the evidentiary strength of the design.
Reporting one field-derived D_b alongside one laboratory coefficient without caveats can create a misleading ranking. The two values may reflect different experiment lengths, boundary conditions, particle properties and model choices. To compare them, first align units and time support, then inspect how the fitted model represented nonlocal movements and sediment accumulation. A useful conclusion might identify a consistent pattern across settings while acknowledging unresolved environmental controls. The result is more defensible than claiming a universal biological mixing constant.
Microbial and chemical changes are downstream inferences
Bioturbation can redistribute reactive organic matter, mineral surfaces and particle-bound contaminants, creating opportunities for altered microbial activity and redox chemistry. But a luminophore tracer by itself does not measure oxygen consumption, sulfate reduction or metal desorption. Those processes should be tested with chemical profiles, flux measurements or targeted assays. A tracer analysis can provide the physical transport constraint used in a coupled model, yet the chemical model has its own parameters and uncertainty.
This distinction matters in a public explanation. Saying that animals stirred marked particles across the upper sediment is a measured transport statement. Saying the same movement increased oxygen uptake is a mechanistic hypothesis unless independently observed or modelled under validated assumptions. Linking two good measurements can be powerful, but the conclusion should show the link rather than assume it. When writing for readers, preserve the chain from observed particle movement to possible consequences and the additional evidence needed for each step.
A second worked example: unequal slice thickness
Suppose a recovered tracer profile has 100 grains in the zero-to-one-centimetre layer, 200 grains in the one-to-three-centimetre layer and 150 in the three-to-six-centimetre layer. The raw counts suggest the middle layer is most populated, but the layers differ in thickness. Dividing by thickness gives 100, 100 and 50 grains per centimetre of depth respectively, before adjusting for cross-sectional area or extraction recovery. This is not a concentration in grams of sediment; it is a depth-normalised count under stated geometry.
The middle slice contains twice as many grains as the upper slice largely because it is twice as thick. If results from another experiment use one-centimetre slices everywhere, direct comparison of unsmoothed totals will be misleading. A scientifically honest graph can show counts per interval together with interval boundaries, or an appropriately normalised distribution and a clear unit. For sophisticated analysis, model predictions should be integrated over the actual slice widths rather than compared at an arbitrary point.
A third worked example: recovered fraction versus initial dose
A team adds 2,000 fluorescent grains and later recovers 1,500. Of those, 300 are below five centimetres. The deep fraction of recovered material is 300 divided by 1,500, or 20 percent. The deep fraction of the originally added material is 300 divided by 2,000, or 15 percent. Neither answer proves that the missing 500 grains escaped into the water column; they might have been lost during handling, buried beyond the sampled depth, obscured from detection or physically exported.
If another treatment recovers 1,900 of 2,000 grains with 300 deep, its deep fraction of recovered material is about 15.8 percent, despite the same observed count at depth. A percentage plot alone would exaggerate the difference in deep transport if recovery processes differ. A responsible comparison reports absolute deep counts, recovery fractions, relevant sample volumes and the statistical uncertainty of independent replicates. The arithmetic is simple; the denominator choice determines the meaning.
Transfer problem: change the model duration
Assume an ideal unbounded random-walk model with D_b equal to 0.20 square centimetres per day. What is the predicted mean squared displacement after five days? It is 2 times 0.20 times five, or two square centimetres. The root-mean-square displacement is the square root of two, about 1.41 centimetres. After twenty days the mean squared displacement is eight square centimetres, and the RMS displacement is about 2.83 centimetres, not four times 1.41.
This is an important teaching result: under ordinary diffusion, typical displacement scales with the square root of time, not directly with time. The value does not say every grain moved that far, nor whether it moved up or down. In a real core with a surface boundary and nonlocal burrows, the simple formula may be inappropriate. The exercise tests algebra and model assumptions together rather than treating an ideal equation as a field law.
Transfer problem: detect an unsupported model
A depth profile has a narrow peak at the surface and a separate band of labelled grains at eight centimetres, with very little tracer at intermediate depths. Is constant local biodiffusion an obvious mechanistic explanation? No. A local diffusion model can broaden a profile, but a separate deep band with a gap suggests examining nonlocal transport, burrow deposition, physical disturbance or a sampling artifact. The correct answer is a list of testable alternatives, not a fabricated coefficient.
Ask what additional evidence would help. Time-lapse observations could show a sudden burrow deposition event. Photographs of the core and the animal community could identify structures consistent with excavation. Independent recovery and depth-registration checks could rule out a sample-cutting error. Additional replicate profiles could show whether the deep band is repeatable. A good scientific response follows the observed pattern into a discriminating measurement, rather than using a flexible fitted curve to make every explanation appear equally true.
Questions readers ask about luminophore mixing
Is a deep fluorescent grain proof that a worm moved it? No; physical disturbance and handling remain possibilities. Is D_b an animal speed? No; it is a model-dependent effective spreading coefficient. Is a brighter image necessarily more tracer mass? Only with suitable calibration and control of illumination. Can bioirrigation be measured by luminophores alone? Not directly. Does a Gaussian-looking curve prove random mixing? No; multiple mechanisms can approximate a similar end-point distribution.
Can a single core provide hundreds of independent replicates because it contains many particles? No; replication follows the experimental unit. Does greater mixing depth always mean greater total flux? Not necessarily; depth, frequency and affected particle quantity are different. Must bioturbation be represented by one coefficient? No; nonlocal and depth-dependent approaches can be needed. Why retain recovery fractions? Because a normalised profile of surviving particles can conceal loss from the system. These distinctions make the result useful outside its specialist audience.
How to describe a profile without overstating what it shows
Begin with the initial tracer arrangement and the depth coordinate system. State the particle size and fluorescent detection method. Give the experiment duration, the independent replication and the total recovery. Describe whether the profile broadened smoothly or contained separated peaks, and report the model and parameters only after explaining their assumptions. Where a biodiffusion coefficient is supplied, include units and model boundaries. Distinguish observed spatial distribution from inferred movement mechanisms and from possible biogeochemical consequences.
A defensible reader-facing conclusion might say that labelled particles became redistributed across specific depth intervals during a defined observation period, with evidence that local mixing and occasional deeper transfers both contributed. Whether that pattern reflects particular animal activities, physical disturbance or their combination depends on additional observations. This is more informative than a statement that the bed was strongly mixed. It identifies what was measured, how it was transformed and which uncertainties remain relevant to future research.
Original methods and related reading
Solan et al. (2004), fluorescent time-lapse sediment profiling and mixing models
Maire et al. (2008), sediment reworking rate methods review
Boudreau et al. (2001), model of deposit-feeder particle diffusion
Worms and storms (2026), biological and physical mixing