The sampling site is located in the western Dutch Wadden Sea, near the northern tip the island of Texel (Fig. 2). We selected two adjacent intertidal locations: a muddy site (53°9.074^′ N, 4°53.67^′ E; -0.055 m relative to Dutch ordinance datum (NAP) – approximately mean sea level), with visible benthic activity (small cast mounds, burrows), and a sandy site (53°9.013^′ N, 4°53.931^′ E; 0.004 m), with presence of large lugworms (large cast mounds and depressions). The sites are microtidal, with a mean tidal range of 1.43 m at the nearest measurement station (Oudeschild). This implies that during high tide the maximum water depth is 0.7 m (0.8 m during spring tides). The sites are located approximately 5 km south-eastwards of the tidal inlet and are relatively sheltered from the prevailing south-westerly winds. Our findings apply therefore to similar microtidal sites.

To describe the benthic community, we sampled both sites using a hand corer (surface area of 0.0177 m2), which was pushed into the sediment until encountering the first shell layer (depth approximately 0.25 m). The core sample was sieved over 1 mm mesh size and the remaining coarse fraction including biota was conserved in 4 % formaldehyde, buffered with borax and stained with Rose Bengal (CAS Number: 4159-77-7). In the laboratory, samples were sorted and biota were identified. Next, dry weight (DW) was measured after drying at 60 °C for 2–3 d and ash free dry weight (AFDW) was measured after combustion at 560 °C for 5 h, to yield biomass per taxon. Due to the limited area and depth of our samples and the lower density of larger Arenicola marina individuals, we did not capture this species in our samples. From the field observations of large A. marina casts at the sandy location, we expected this species to dominate bioturbation. Therefore, we also derived A. marina density based on cast counts from pictures from both locations (4 pictures per site). To estimate individual biomass, we actively collected A. marina from both sites with a spade (N=17), which were dried in the same way as the benthic community samples. We derived bioturbation traits and living depth from a trait database (Beauchard et al., 2023) and calculated the total biomass per trait group.

Two replicate sediment cores were collected at each site in August 2023 for grain size analysis, geochemical composition (X-ray fluorescence, XRF), 210Pb and luminescence analysis. The cores were obtained by hammering two PVC tubes (inner diameter 96 mm) side by side vertically into the tidal flat deposits. After extraction from the tidal flat, the cores were sealed and transported to a dark container where under subdued amber-light conditions a core slicer was used to split each core into two halves. One of the cores halves was kept for luminescence and 210Pb sampling, the other was kept for XRF scanning and sediment composition. We subsampled luminescence and 210Pb samples by cutting 25 mm thick slices from the top to the bottom of each core, of which we selected eight per core for luminescence analysis (see Fig. 2 and Table A1 for intervals). The inner portion of each slice was used for luminescence equivalent dose (De, Sect. 2.6.1) estimation and 210Pb analysis, while the outer material was reserved for determining the environmental dose rate (D⋅, Sect. 2.6.2). The sediment was wet-sieved in a dark room environment to separate the grain-size range of 212–250 µm, used for luminescence analysis, from the fraction smaller than 63 µm, used for 210Pb analysis. We measured 210Pb for five additional samples of each core, amounting to a total of 13 210Pb samples per core.

For grain size distribution, half of each long core was subsampled in 25 mm slices. These samples were freeze-dried, sieved over 1000 µm and measured by laser diffraction (Mastersizer 2000, Malvern Instruments). We determined the volumetric content of sediment size fractions from mud to coarse sand, and the median grain size (d50), and 10th and 90th percentile of the grain size distribution (d10 and d50).

We measured bulk elemental composition over the entire depth of the long cores at 10 mm resolution by X-ray fluorescence, using an AvaaTech XRF core scanner at the Royal Netherlands Institute for Sea Research (NIOZ) (Richter et al., 2006). Each section was exposed for 10 s with a power supply of 10 kV/0.60 mA, or 30 kV/1.6 mA. This yielded bulk elemental composition as element intensities in total counts or counts per second. We selected the following elements and ratios as environmental proxies: Si:K (sand indicator), Cl:Fe (saline pore water indicator, thus proxy of porosity); Ca:Fe (shell content indicator; Croudace et al., 2006), Rb (silt/clay mineral indicator; Guyard et al. (2007); note that the XRF detector window contains rubidium crystals, explaining high counts for this element); and Br (organic content indicator; Ziegler et al., 2008).

The downcore distribution of 210Pb in the mud fraction (< 63 µm) of the sediment was determined indirectly through alpha spectrometry measurement of its grandchild isotope polonium-210 (210Po) (de Stigter et al., 2011). To this end, 0.5 g of dried and homogenized mud, wet sieved from the bulk sediment sample, was spiked with 1 mL of a standard solution of 209Po in 2 M HCl, and then leached for 6 h in 10 mL of concentrated HCl, heated in a hotblock to 85 °C. After diluting the fluid with 45 mL of Milli-Q water and adding 5 mL of an aqueous solution of ascorbic acid (40 gL-1), natural 210Po and added 209Po were collected from the fluid by spontaneous electrochemical deposition on silver plates, with the fluid kept at 75 °C in a hotblock for 16 h. For subsequent alpha-spectrometry, Canberra Passivated implanted Planar Silicon detectors were used. 210Pb activity was calculated from the measured 210Po activity, assuming secular equilibrium and correcting for the time elapsed since collection of the samples.

The luminescence samples were prepared at the Netherlands Centre for Luminescence Dating (NCL) under amber light conditions.

Simultaneous to the long core sampling, we started a luminophore tracer experiment adjacent to both sampling locations. Fluorescent coated luminophores (red/pink tracer, size range: 0–125 µm) were mixed with sieved native sediment from each site (ratio 1:10 for the muddy sediment and 1:17 for the sandy sediment) and applied to the sediment surface as 15 cm × 15 cm × 0.5 cm frozen tiles. Per site, three experimental plots were created by aligning four of these tiles – thus each covering a total area of 30 cm × 30 cm. These were sampled on day 2, 16 and 30 by pushing a PVC tube (diameter 36 mm) at least 20 cm deep into the sediment. The resulting sediment cores (one per plot) were frozen upright. In order to find evidence of recent bioturbation of the fine particles, the frozen cores were sliced in length direction with a tile saw. The cutting surface was scraped clean in lateral direction until the still frozen sediment was reached. If cutting had resulted in visible artefacts in the sediment (cavities or disturbed sediment surface), the damaged core or half core was excluded from analysis, leaving two replicate cores per site per time point. Next, both halves of the core were photographed in a dark room under ultraviolet light, with a ruler for scale reference. The images were processed using ImageJ (Schneider et al., 2012): first, we adjusted the display settings Level ≈ 50 and Window ≈ 100. Red and orange pixels were then identified per 2 cm depth slice by extracting RGB values of each pixel, and applying thresholds (R > 150, G > 40, G < R–30, B < 100). The total count of pixels meeting these criteria was recorded as the red–orange pixel count. The average number of pixels per depth interval were averaged over both core halves.

Chlorophyll a cores were collected next to the luminophore tracer experiment. Cores (inner diameter 36 mm) were collected on 22 August and 5 October 2023 (day 1 and day 45 of the luminophore experiment, respectively 2 and 3 cores per site). The cores were pushed into the sediment as deep as possible, or at least 15 cm deep. We sliced the cores every 0.5 cm down to 3 cm depth, then every 1 cm down to 15 cm depth. The samples were stored at -80 °C and freeze-dried for 72 h before analysis. Next, we homogenised and weighed an aliquot of the sediment (approximately 1000–1500 mg for the sandy site samples and 300–500 mg for the muddy site samples). We extracted Chl a by adding 5 mL 90 % acetone in 50 mL centrifuge tubes and running them in a bullet blender (BB50-AUCE, Next Advance) for 15 min, after which 5 mL 90 % acetone was added again and extraction was repeated for another 15 min. The extracts were measured in a spectrophotometer (Specord 210, Analytik Jena) at 630, 647, 664, 665 and 750 nm. From the measured extinction values, we calculated Chl a concentration following Ritchie (2006).

Macrozoobenthic community biomass and abundance (excluding Arenicola marina) were higher on the muddy site (approximately 19 gAFDWm-2) than on the sandy site (3.8 gm-2; Appendix A3, Table A2). Both locations contained a large fraction of taxa which could be classified as biodiffusers (Appendix A3, Fig. A1). The sandy site contained a larger fraction of downward conveyors, while more upward conveyors inhabited the muddy site. Density of downward conveying lugworms (A. marina) was higher on the muddy site (98 ± 42 individualsm-2 versus 44 ± 11 individualsm-2 on the sandy site), while worms were larger on the sandy site (0.16 gindividual-1 versus 0.05 gindividual-1 on the muddy site). We estimate total A. marina biomass therefore to be slightly higher on the sandy site (6.9 gAFDWm-2 versus 5.1 gAFDWm-2), and to form a substantial part (approximately 60 %) of the total biomass at the sandy site, compared to 20 % at the muddy site.

On the sandy site, the top ∼ 0.3 m of the sediment consists mainly of fine sand, with a mud content under 10 % (Fig. 3a). From here, the coarse sand fraction gradually increases to ∼ 40 % at a depth of 0.45 m. Below 0.45 m, the sediment is less well-sorted, containing both a 20 %–40 % mud fraction, as well as coarse and very coarse sand and a variable d90 (Fig. 3b) We observed shell layers between 0.27–0.29 and 0.44–0.47 m, also reflected in increased Ca:Fe ratios from the XRF measurements. Below the second shell layer, the marine pore water (Cl:Fe) decreases and organic matter content (based on the Br indicator) increases (Fig. 3e–i).

At the muddy site, the top 0.15–0.20 m of the sediment is less well sorted: a large mud fraction (10 %–25 %), as well as a substantial fine and medium sand fraction are present (Fig. 3c and d). A shell layer was present between 0.18–0.34 m. Between 0.2–0.45 m, some layers contain coarse particles (500–1000 µm), which is also reflected in peaks in d90. The amount of marine pore water (Cl:Fe) slightly increases downcore, whereas sand content (Si:K), mud content (Rb) and organic matter (Br) remain relatively constant (Fig. 3j–n). A brick present in the shell layer gives evidence of anthropogenic influence and suggests a relatively young deposit.

Above the first shell layer at the sandy site, 210Pb activity was lowest in the surface slice (0–0.025 m depth), and increased to a subsurface peak at 0.15 m (Fig. 4a). Below the first shell layer (< 0.29 m), activity decreased, to a nearly constant level below 0.5 m. At the muddy site, 210Pb activity slightly decreases down to 0.15 m, peaked in the first shell layer at 0.225 m (Fig. 4b). Until down to 0.4 m, below the second shell layer, 210Pb activity remained high. Activity then decreased to an apparently constant level of ∼ 22 mBqg-1 between 0.5–0.6 m.

In discussing the luminescence ages, we focus on the quartz OSL ages obtained through the iterated age model; given that quartz OSL resets most rapidly upon light exposure this provides the best estimate of the last light exposure of the majority of the grains. Quartz bootMAM ages indicate the last time of introduction of light exposed grains through bioturbation. For feldspar, heterogeneous bleaching is to be expected due to slower resetting of the luminescence signals and therefore we focus on the interquartile range of the age distribution and bootMAM derived ages to infer information on light exposure and thus process information.

The quartz OSL iterated ages suggest last light exposure well within the past 100 years for sediments within the top 0.4 m at the sandy site (Fig. 4c) and 0.15 at the muddy site (Fig. 3i) (Appendix A4, Table A3a). A jump in age is observed between the samples at 0.4 and 0.5 depth at the sandy site, with an age of 430 ± 40 years for the latter according to the iterated model. For the muddy site, older ages (440 ± 30 to 640 ± 40 years) were obtained for the part below the shell layer.

The feldspar IR50 and especially the pIRIR175 signals were not fully bleached in a large fraction of the grains in the top 0.2 m of both cores. At the sandy site, grains with a low pIRIR175 age were only present at 0, 0.15 and 0.325 m; at the muddy site such young grains could be identified at 0.05 and 0.15 m. The ages obtained using the bootMAM approach on the IR50 signal tend to be a few decades older than those obtained through quartz OSL iterated mean, while bootMAM derived pIRIR175 ages are significantly older (typically by a few centuries, Table A3b and c). An exception is formed by the samples from the deeper parts, where IR50 results tend to be younger than quartz OSL, while pIRIR175 agree with quartz OSL. The single-grain age distributions obtained on both feldspar signals are also broader than those obtained through quartz OSL.

Both short-term tracers show a high spatial variability. Whereas two muddy cores and one sandy core contained sub-surface maxima of Chl a concentration (Fig. 5a and b; at 3, 15 and 6–8 cm depth, respectively), in the other cores the Chl a concentration showed a more exponential decrease with depth. At the bottom of the cores in all sites, still some Chl a was present above the analytical detection limit (0.2 µgg-1), which could indicate the supply of Chl a younger less than 3 months ago, or reflect Chl a adsorption to sediment.

Some luminophore particles were already worked down to 18 cm (sandy site) and 20 cm (muddy site) within 2 d (36 h) after initiation of the experiment (Fig. 5c and d). In these cores, the reworking could be clearly linked to bioturbation, as two bivalves were found in the cores with increased luminophore concentration around their burrows (Appendix A5, Fig. A2; Scrobicularia plana at 12–15 cm depth at the sandy site and Cerastoderma edule at 0–3 cm depth at the muddy site). Similar high subsurface luminophore concentrations were not observed in the cores collected at later time points, although some luminophores could be found at greater depths. We saw similar between-core and thus spatial variability as in the Chl a measurement.

In the sediment composition, we did not observe any clear layering in the top decimetres of sediment at both sites which could be attributed to storm deposition. This may be due to an overly coarse vertical sampling resolution, even though our resolution of 25 mm should have been sufficient to capture storm deposited layers, which can have a thickness of centimetres (Zhou et al., 2022). At the sandy site the top of the profile was relatively homogeneous. The top decimetres of the sandy core also appeared to be better sorted than the top part of the muddy core (Fig. 3a–d). Generally, hydrodynamic reworking would be expected to result in more uniformly sorted sediment, whereas biota can create localised patches of finer or coarser material (Aller and Cochran, 2019; Bentley and Nittrouer, 2012) thus resulting in variable sorting of individual sediment samples (Rhoads and Stanley, 1965). Furthermore, the presence of shell layers may indicate biotic influence, since bioturbators such as A. marina are unable to ingest coarse particles like shells, which then remain at the feeding depth (Reise, 2002; van Straaten, 1952), while the ingested sand will be deposited on the sediment surface and subsequently mixed into the sediment.

The tracers targeting the fine-grained sediment fraction showed clear signs of bioturbation. During the luminophore experiment, no storms or strong winds occurred, therefore vertical luminophore displacements can solely be attributed to bioturbation. Both the short-term (luminophores and Chl a) and the long-term (210Pb) tracers showed sub-surface peaks, illustrating that advective vertical transport of particles occurred. This is coherent with the observed macrozoobenthic species that are known non-local mixers (see Sect. 4.3). In addition, the gradual vertical gradients of 210Pb, and of Chl a till 0.15–0.2 m point to substantial diffusive bioturbation in the top decimetres of the sediment, or rapid deposition within the previous months.

The luminescence data gives more details on the deposition and reworking of sand grains. The narrow age distributions obtained through quartz OSL indicate that this signal was well-bleached. The bootMAM and iterated age estimates are very similar, both likely reflecting the time of deposition. The narrow distributions in the top layer (0.4 m at the sandy site and 0.15 m at the muddy site) imply that the vast majority of grains were recently (within the past few decades) exposed to sufficient light to fully reset quartz OSL signals. However, the older ages obtained through feldspar IRSL, even for the relatively fast bleaching IR50 signal, indicate that for a large fraction of grains, full light exposure during and after deposition was limited. Especially old (> 500 years) pIRIR175 measurements in the top 0.10 m at both sites indicate that light exposure prior to burial was insufficient to bleach this signal. It is less likely that these grains are worked upwards by bioturbators from deeper deposits, as other grains of this age were only found below the shell layer at 0.35–0.5 m depth. Both feldspar signals from the top decimetres were not as homogeneously bleached as would be expected with intensive reworking. We conclude that the luminescence ages primarily reflect the time of deposition, rather than the time of bioturbation. This conclusions holds for all luminescence signals used. However, quartz OSL ages provide most reliable depositional ages while the feldspar IRSL and pIRIR results overestimate the burial age due to incomplete resetting of these signals upon deposition and burial.

The luminescence signals gave a few hints of bioturbation as well. Firstly, at both sites, young grains with nearly fully bleached pIRIR175 signals are present at larger depths. At the sandy site, such young grains can be found at 0.15 m, coinciding with the subsurface 210Pb peak. At the muddy site, the young grains are only present above the shell layer, at 0.05 and 0.15 m. This could point at the advective transport of young sand grains to this depth, or good bleaching conditions during deposition of these layers. Given the combination of bioturbation evidence from the other tracers, and the lack of sedimentary structures, we deem bioturbation to be the most likely process. Secondly, feldspar signals in the very top layer (0–0.05 m) at both sites were less well bleached and had wider distributions than those at 0.1–0.2 m deep. Possibly, grains that are recently deposited under low bleaching conditions and are buried close to the surface, are not always directly resurfaced by bioturbators. These grains require a longer period of bioturbation to be bleached further. This hypothesis is supported by Bentley et al. (2014), who described a shallow (1–4 cm) stratified top layer, which was not yet bioturbated, above a more intensely bioturbated zone in a muddy intertidal environment. Secondly, the luminescence data provides some evidence for exchange of grains between stratigraphic layers of different depositional age. Some grains in the luminescence age distribution had ages similar to those of the underlying deposit, even if a shell layer is present in between. At the muddy site, a high level of 210Pb was also present in and below the shell layer. From these observations combined, we conclude that transport of fine particles can sometimes extend beyond the shell layer.

Madsen et al. (2011) also reported quartz OSL profiles on an intertidal flat and attributed a change in apparent sedimentation rate to bioturbation. Their site was not very dynamic, and bioturbation was more likely to be the main mixing process present. The combined evidence of small aliquot quartz OSL and single grain feldspar IRSL50 and pIRIR175 suggests that deposition at our site was episodic and that bioturbation was less efficient in bleaching these signals by resurfacing the grains. Furthermore, given that deposition was recent, it is not possible to clearly differentiate bioturbated grains from those grains that were last light-exposed prior to deposition as is done in terrestrial soils (e.g., Reimann et al., 2017). Hence, we cannot deduce bioturbation rates from our data using approaches as suggested by van der Meij et al. (2025).

We could not derive sediment accumulation rates, because our combined signals pointed to a combination of sediment mixing and episodic deposition. Therefore, assuming the age-depth profiles to be the result of gradual sedimentation alone would be incorrect in this dynamic environment. Quartz OSL showed a hiatus in age between the top and the middle part of the core at the muddy site but not at the sandy site, while 210Pb also indicated a hiatus in activity at the muddy site only (Fig. 4a). Such age gaps can be caused by erosion followed by rapid accumulation, mounding by bioturbators (Kristensen et al., 2015; van der Meij et al., 2025), or switching between a depositional and an erosional regime with bioturbation penetrating the older deposits (Madsen et al., 2011). The opposing gaps of the radionuclide and luminescence tracer profiles could be an effect of an age underestimation of 210Pb compared to OSL of older sediments (Madsen et al., 2007). However, the 210Pb activity just below the shell layer in the muddy site is too high to be fully explained by such an inaccuracy. Instead, we expect differential transport of fine and coarse particles through the shell layer to play a role.

Combining the information from luminescence signals, deposit composition and 210Pb, we propose the following possible scenarios for sedimentation history (Fig. 6):

On the sandy site, rapid sediment deposition took place approximately 400 years ago. The original sediment surface was covered with a new thick layer of sand, and the benthic fauna recolonised the new layer. The quartz signals in this new deposit were then uniformly bleached by bioturbation and physical resuspension and deposition, although incidentally some grains were still exchanged with the bottom layer. Through bioturbation, also the first shell layer was formed. Recently (< 100 years ago) another large, rapid deposition event happened. This shifted the bioturbation depth upwards, forming a new shell layer higher up. Still some fines (210Pb) moved downward through the shell layers, but most accumulated on feeding depth of bioturbators.

The sandy site was relatively more exposed to energetic hydrodynamic conditions than the muddy site. This may have caused the surprisingly low surface 210Pb activity, as waves and tidal currents may resuspend and wash away the recently settled 210Pb-rich material at each tide. The Chl a concentrations were also lower at the sandy site than on the muddy site, but this might be due to differences in degradation rates of Chl a (Sun et al., 1993). The bleaching potential of the luminescence signals could differ between the muddy and sandy tidal flats (Pannozzo et al., 2023). When submerged, higher loads of suspended sediment will decrease subaqueous light penetration during transport and (re)deposition on the muddy site (de Boer et al., 2026b). The sandy site is more elevated and therefore slightly more often exposed to subaerial bleaching than the muddy site due to its lower inundation times. These differences in bleaching potential and probability may have resulted in the lesser bleaching and wider distributions (Fig. 4j) of the luminescence signals on the muddy site.

The sandy and muddy site also differed with respect to the density, biomass and composition of the benthic community. The muddy site contained a relatively higher abundance and biomass of taxa that could be classified as biodiffusers and surficial modifiers. This leads to a higher contribution of diffusive mixing and could also partly explain the larger interquartile ranges of the luminescence signals at the muddy site, compared to more effective bleaching through advective vertical bioturbation at the sandy site (van der Meij et al., 2025). Subsurface maxima of fine-grained tracers, pointing to advective transport, were more pronounced in the sandy site. One of the main likely bioturbating agents was conveyor-belt reworking polychaete A. marina. While the muddy site contained a higher density of small individuals, this species comprised only a fifth of the biomass here. The sandy site was inhibited by fewer, but larger lugworms, so that this species was dominant in biomass here. Such large individuals live in 0.2–0.4 m deep burrows and often dominate sedimentary processes at sandy intertidal flats. Their presence leads to 4–400 times more likely particle reworking rates and an estimated full overturning of a 0.1–0.6 m layer of sediment per year (Cadee, 1976; Riisgård and Banta, 1998; Volkenborn et al., 2007b). These lugworms were probably responsible for the subsurface accumulations of 210Pb at feeding depth, as the 210Pb-rich organic material is filtered from the ingested sand (Longbottom, 1970).

Other bioturbators might also have contributed to non-local transport of fine tracers at both sites, either by active particle transport or through passive processes such as burrow infiltration. Ragworms (family Nereididae), such as Hediste diversicolor, were present at both sites. These worms build a gallery network and can cause diffusive mixing of the surface layer, and advective downward transport, accumulating particles at the lower end of the gallery network due to passive sloping, burrow infilling and during rapid retraction (Foster-Smith, 1978; Herringshaw et al., 2010; Ouellette et al., 2004). Another group that we deem responsible for non-local transport of fines are tentaculate polychaetes, which enhance fine particle transport from the water column into the sediment (Frithsen and Doering, 1986). Pygospio elegans, a tube-building worm present in high densities at both sites, constructs tubes 0.15 m into the sediment, and enhance transport of fine particles down to 3–5 cm depth (Bolam and Fernandes, 2003). Lanice conchilega feeds on organic particles from the sediment surface or the water column and builds long tubes, down to 0.3–0.4 m into the sediment (Carey, 1987). Although this species was only present in the benthos core sample at the muddy site, we observed their tube endings, indicating their presence, on the sediment surface at both locations (Fig. 2b and c).

Scrobicularia plana is described as a surficial modifier in trait databases (Queirós et al., 2013). We however observed that this species also contributes to non-local mixing, as suggested by luminophores that had moved rapidly to the living depth of this bivalve (Fig. A2). Similar discrepancies between “theoretical” bioturbation traits and field observations were reported by (Morys et al., 2017). Therefore, a single trait-based community index for bioturbation (such as BPc), could potentially be informative on larger spatial scales when a more extensive dataset is available, but such an index may be less relevant for a local study.

The differences between tracers are likely the result of particle size-specific physical and biological sorting mechanisms. Fine particles are more easily eroded and transported from the sediment surface by currents and waves. They are therefore less prone to remain on the surface for a prolonged period, whereas coarser sand grains are more stable on the sediment surface. Fines are also more likely to move down into the sediment matrix by passive as well as active processes. In sandy, permeable sediments, fine particles can infiltrate with pore water flux (reviewed by Huettel et al., 2014). Above that, fine and organic particles are more often selected by bioturbators, after which they may accumulate in deeper layers of the deposit. This particle selectivity is common among benthic taxa (Self and Jumars, 1988).

Although the grain size range we selected for luminescence measurements (212–250 µm) falls outside of the feeding range of many polychaetes (Jumars et al., 2015), A. marina, expected to be a dominant bioturbator especially on our sandy site, does feed non-selectively on sediment particles below a size of 1000 µm (Baumfalk, 1979). Bioturbation of this species was shown to still result in accumulations of finer (130 µm) as well as coarser (500–1000 µm) particles at its feeding depth (0.1–0.14 m) (Gebhardt and Forster, 2018). Bivalves can also exhibit size- and particle-type specific selection, but this usually takes place post-capture (reviewed by Ward and Shumway, 2004). Their burrowing movements rather than their feeding activity may thus vertically displace sand grains. A. marina as well as bivalves are therefore expected to have reworked sand grains and affected the luminescence tracer profiles.

It has been demonstrated that benthic fauna select for particle quality as well and prefer organic over inorganic particles (Taghon, 1982). Although we saw some advective transport of luminophores (Fig. 5c, d and A2), we did indeed not observe large sub-surface accumulation over time. One explanation might be that these inorganic particles do not accumulate at the organism's feeding depth in the same way as the organic matter-associated tracers that are retained in the digestive system. This is contrary to Gerino et al. (1998), who observed subsurface peaks of luminophores after 15 d, and additionally modelled Chl a subsurface mixing rates to be in the same order of magnitude as the mixing of non-organic particles in a shallow, muddy environment. In our dynamic intertidal environment, luminophores might have more easily eroded from the sediment surface, resulting in a decrease in surface supply over time, which possibly underlies the lack of later subsurface accumulations (Fig. A3).

A tracer can give meaningful information on mixing processes when the timescale of a tracer (e.g., radionuclide decay rate) matches the timescale of the mixing rate: the longer the tracer timescale, the more particle displacements will have contributed to the signal distribution (Meysman et al., 2003). The luminescence signal is of a different nature than that of radionuclides, as it builds up after burial but resets at each sufficiently long or intense light exposure, rather than gradually declining over a known time. Signals that are more difficult to bleach (e.g. feldspar pIRIR175) require more light exposure to be reset compared to fast bleaching signals (e.g. quartz OSL). They thus require a higher frequency of reworking events to be fully bleached. If sediments are however too young or intensely mixed, no recently reworked grains can be distinguished, as the sediment will be uniformly bleached. Under the conditions at our intertidal study sites, only the feldspar IR50 and pIRIR175 bleached slowly enough to be not fully reset during deposition or by bioturbation. Therefore, these signals can show subsurface supply of young sand grains (Fig. 4e and h) and incomplete bleaching during deposition (Fig. 4d, e, g, and h). In this relatively young and more intensely reworked sediment, where no “old”, never-mixed, sediment fraction appeared to be present, single-grain level measurements cannot be used for determining the bioturbation effectiveness (Reimann et al., 2017). Spatial scale also plays a role, especially for the short-term tracers. The small sampling surface of the Chl a and luminophore cores and the patchy distribution of macrozoobenthos resulted in high variability between cores, and a higher number core replicates may therefore have increased consistency of tracers profiles for these short-term tracers. For the long-term tracers, the number of replicates was lower, however because the distribution of these tracers captures the integrated activity of the macrozoobenthic community over a longer time period, we would expect similar results even if we had used a higher number of replicates.

Although our findings on sediment reworking are site-specific, the tracers used here are more broadly applicable, provided that the limitations of each tracer are considered. A combination of tracers best represents the intricate reality of physical and biotic sediment mixing processes. This corroborates Gerino et al. (1998), who showed that to accurately model bioturbation modes, measurements of multiple tracers with different characteristics are required. A multi-tracer experiment is however time- and resource intensive, and aligning tracer sampling protocols to improve intercomparison is not always feasible. Therefore, we propose a framework for deciding on the type of tracer to use, depending on the research question at hand (Fig. 7). The process timescale, grain size fraction of interest and spatial density need to be considered. Depending on the rate and dominant mode of bioturbation, a short- or long-term tracer may be more appropriate (Lecroart et al., 2010; Meysman et al., 2010). Also, the potential targeting of organic food particles (e.g., Chl a, 210Pb) may be considered. Furthermore, the physical conditions at the field site determine the suitability of a tracer. If the field site is erosive or exposed to higher dynamics, fine luminophores might not be representative over a longer period. With low reworking or light exposure (e.g., long inundation and turbid water), hard-to-bleach luminescence signals are less likely to be bleached by bioturbation.

Feldspar single-grain luminescence is a promising tracer for deposition and mixing processes in intertidal environments. Even though luminescence signals at our study sites pointed to high dynamics and rapid deposition, rather than being all fully bleached by bioturbation, they did give more information on long-term deposition and reworking of sand grains than other existing tracers can. We recommend testing this method in a less dynamic setting, where bioturbation dominates reworking, to improve interpretation of luminescence profiles and further explain the limited feldspar bleaching that we observed. The aim of this research was a qualitative rather than a quantitative description of mixing processes. Given the young age of the sediments and the dynamics of the environment, a quantification of bioturbation as attempted by van der Meij et al. (2025) may be challenging. Nevertheless, single-grain and slow-to-bleach luminescence signals are a promising tool to further test and develop, for ultimately deriving bioturbation rates of sand grains in intertidal settings.

Our results highlight differences between the behaviour and bioturbation of sand and mud within the sediment matrix. This distinction matters for stratigraphy: fines are more mobile within the sediment column, while sand grains have greater potential to preserve layered bedding. Although bioturbation is often assumed to homogenise stratigraphic records (Wheatcroft, 1990), our feldspar results suggest that the sand fraction is not intensively reworked by the present bioturbating community. We also observed exchange of particles through the shell layer, suggesting that such layers do not necessarily form a hard boundary for bioturbators. Furthermore, the grain size-specific reworking can inform process-based sediment transport models that quantify the mixing of specific sediment fractions within the bed (Armanini, 1995). Bioturbation increases the exchange of particles between the active (transportable) and substrate layer of the sediment, with higher exchange probability for small than for large particles.

These findings also bear ecological implications. Fine, organic-rich tracers were rapidly carried below the aerobic zone and accumulated at depth, whereas young sand grains were not. This indicates that bioturbation affects reactive compounds more strongly than coarse particles. Its influence on sand transport may thus be smaller than its role in sediment biogeochemistry and carbon storage. By modifying sediment mud content, bioturbators can modify their own habitat conditions. In sandy sediments, preferential reworking of fines can increase subsurface mud content. Such “muddification” by cockles was demonstrated in a flume experiment (Soissons et al., 2019) and by Ensis leei in a field experiment (Witbaard et al., 2017). On the other hand, resurfaced fines may be mobilised from muddy sediments, decreasing the sediment mud content (Montserrat et al., 2009; Volkenborn et al., 2007a; Wendelboe et al., 2013). These two-way processes affect large-scale intertidal flat morphology and sediment composition (Brückner et al., 2021; Le Hir et al., 2007). Finally, the integrated contribution of bioturbation and storms may be comparable when event energy and frequency are considered together. Whereas bioturbation is a continuous process of small particle displacements, storms are infrequent, but high energy events (Fig. 1b). In effect, benthic animals influence daily sediment properties, while storm events “reset” the system. Because most species have a relatively wide sediment tolerance (Kooistra et al., 2025), such episodic resetting does not necessarily result in a major community shift, and during calmer periods, bioturbators can once again modify sediments at smaller scales, potentially enhancing their own habitat suitability.