Pyrite-lined shells as indicators of limited oxygen exposure time and inefficient bioirrigation in the Holocene-Anthropocene stratigraphic record

Although the depth of bioturbation can be estimated on the basis of ichnofabric, the time scale of sediment mixing and irrigation by burrowers that affects carbonate preservation and biogeochemical cycles is difficult to estimate in the stratigraphic record. However, pyrite linings on interior of shells can be a signature of slow mixing and irrigation rate 10 because they indicate that shells of molluscs initially inhabiting oxic sediment zones were immediately and permanently sequestered in reduced microenvironments where molluscan biomass and associated microbial coatings stimulated sulfate reduction and pyrite precipitation. A high abundance of pyrite-lined shells in the stratigraphic record can thus be diagnostic of limited net exposure of labile tissues to O2 even when the seafloor is inhabited by abundant burrowing infauna as in the present-day northern Adriatic Sea. Here, we reconstruct this sequestration pathway (1) by assessing preservation and 15 postmortem ages of pyrite-lined shells of the hypoxia-tolerant bivalve Varicorbula gibba in sediment cores and (2) by evaluating whether an independently-documented decline in bioturbation, driven by eutrophication and seasonal hypoxia during the 20 th century, affected the frequency of pyrite-lined shells in the stratigraphic record of the northern Adriatic Sea. First, at prodelta sites with high sedimentation rate, linings of pyrite framboids form rapidly in near-surface sediment zones as they appear already in interiors of shells and in intra-shell conchiolin layers younger than 10 years and occur 20 preferentially in well-preserved and articulated shells with periostracum and relatively high concentrations of amino acids. Second, increments deposited in the early 20 th century contain <20% of shells with pyrite at the Po prodelta and 30-40% at the Isonzo prodelta, whereas the late 20 th century increments possess 50-80% of shells with pyrite at both locations. At sites with slow sedimentation rate, the frequency of pyrite linings is low (<10-20%). Third, the upcore increase in the frequency of pyrite-lined shells positively correlates with an abrupt increase in maximum shell size and biomass of V. gibba. Therefore, 25 the upcore increase in the frequency of pyrite-lined shells indicates that sediment mixing and bioirrigation rates declined during the 20 th century, leading to higher sequestration of pyrite-lined shells during the late 20 th century. We hypothesize that the permanent preservation of pyrite linings within the shells of V. gibba in the subsurface stratigraphic record was allowed by slow recovery of infaunal communities frequently interrupted by seasonal hypoxic events, leading to the dominance of surficial sediment modifiers with low irrigation potential. Abundance of well-preserved shells lined by pyrite exceeding 30 ~10% per assemblage in apparently well-mixed sediments in the deep-time stratigraphic record can be an indicator of short net exposure of shells to O2 and inefficient bioirrigation. Fine-grained prodelta sediments in the northern Adriatic Sea deposited since the mid-20 th century, with high preservation potential of reduced microniches, can represent taphonomic and early-diagenetic analogues of deep-time skeletal assemblages with pyrite linings.

from Po 3, in 243 valves from Po 4, in 311 valves from Panzano, and in 232 valves from Piran, and were presented with age calibrations of aspartic acid D/L by 14 C-dated valves by Tomašových et al. (2017Tomašových et al. ( , 2018 and Mautner et al. (2018). To 200 constrain the rate and the location of pyrite formation on the basis of postmortem age data, we evaluated (1) whether valves of the same postmortem age (binned to 10 years at Po stations and to 50 years at Panzano), with and without pyrite linings, differ in their mean stratigraphic depth or whether pyrite-lined valves are located deeper, (2) whether the rate of valve-loss from the mixed layer (by burial or by disintegration) differs between valves with and without pyrite linings, and 3) whether valves of the same age with and without pyrite linings differ in their content of amino acids. We assess the differences in 205 depth and age between valves with and without pyrite by comparing their mean values and 95% bootstrapped confidence intervals. We estimate loss rates of valves from the mixed layer by fitting age distributions of valves with and without pyrite linings from Po and Panzano to a simple model with temporally-constant loss rate that can be well-fitted by the exponential distribution (disintegration-burial model) and to a more complex model where loss rate declines with postmortem age at some sequestration rate and the resulting age distributions are heavy-tailed, typically caused by exhumation of older valves 210 to sediment surface (Tomašových et al., 2014). We use the Akaike Information Criterion corrected for small sample size to assess the relative fit of these two models. The higher support for the disintegration-burial model can indicate that the both types of valves undergo a simple, temporally-constant loss from the mixed layer. We use independent estimates of net sedimentation rate based on 210 Pb data to assess whether this loss-rate parameter corresponds to disintegration or burial. The higher support for the more complex model could indicate that pyrite-lined valves were exhumed from deeper zones to the 215 mixed layer.

Taphonomic scoring
To assess the nature and types of pyrite linings, we evaluated preservation of all specimens of V. gibba at light-microscope scale at 10-20x magnification. We investigated several specimens with scanning electron microscope (SEM) and with 220 backscattered electrons (BSE), using electron probe microanalyzer at 100-1,000x magnification. The chemical composition of pyrite and iron oxides was validated with energy-dispersive X-ray spectroscopy. All valves of V. gibba that were dated by amino-acid racemization at four sites (Po 3, Po 4, Panzano M28, and Piran M53) were scored. At Po 3 (replicate cores M12, M13 and M14), Po 4 (cores M20 and M21), Panzano (cores M28 and M29), all additional specimens were scored under a light microscope. At Brijuni, all specimens of V. gibba from every second increment were scored. Nine alteration variables 225 were scored on specimens of V. gibba at light-microscope magnification (10-20x): (1) pyrite presence (black grains and framboids on interior valve surfaces), (2) loss of periostracum and/or of the internal conchiolin layer (if either external periostracum and/or internal conchiolin layer are visible, conchiolin is scored as being preserved), (3) disarticulation, (4) internal fine-scale surface dissolution, (5) internal bioerosion (generated by algae and sponges, excluding predatory drilling), (6) external encrustation, (7) intense surface wear (loss of external ornamentation), and (8) penetrative dark gray staining 230 (induced by nanopyritic inclusions that fill valve microporosity).

Geographic and stratigraphic differences in the frequency of pyrite-lined valves
To evaluate stratigraphic and geographic changes in preservation of V. gibba at the increment scale, we compute relative frequencies of specimens with a given alteration relative to the total number of specimens in 4 cm (pooling 2 cm increments 235 in the upper 20 cm in each core) and 5 cm-thick increments at five stations. The frequency of disarticulated shells is Pyrite linings can be associated with internal fine-scale dissolution and can co-occur with reddish grains of Fe 280 oxides at all sites. However, the dissolution of valves at Po and Panzano is minor, and is still associated with some surviving portions of periostracum or conchiolin layer. Specimens from Piran and Brijuni are bored and dissolved but are rarely lined by pyrite framboids (Fig. 5). In contrast to pristine valves without any framboids on interiors (Fig. 6A), the pyrite linings can be detected in thin sections as linings formed by framboids dispersed on interior valve surfaces (black grains in Fig. 6B -D) and in the conchiolin layer (Fig. 6E). Valves at Piran and Brijuni contain dispersed framboids that are mainly concentrated in 285 borings or are associated with secondary sediment fillings of borings (Fig. 6F). Some relatively pristine valves from Po and Panzano show a blueish-colored staining on the interior and exterior, still with well-preserved external ornamentation . In contrast, gray-stained valves at Piran and Brijuni are worn, degraded by bioerosion, and affected by internal encrustation (Fig. 5). In thin sections, macroborings produced by sponges are very rare in valves from Po and Panzano, and bioerosion is primarily limited to simple borings that are few microns thick. In contrast, valves from Piran and Brijuni 290 exhibit dense borings (Fig. 6F).

Preservation of pyrite at 100-1,000x magnification
On one hand, BSE images show that pyrite framboids preserved on the interiors of well-preserved valves vary in shape, size (5-10 μm) and packing, ranging from densely-and regularly-packed microcystals within spherical framboids up to loosely 295 and irregularly-packed microcrystals within irregularly-shaped framboids ( Fig. 7A-G). They coalesce into continuous agglomerations at some places. Framboids also partly fill pores between aragonite nodules (arrows in Fig. 7I) within the conchiolin layer (that is ~10-100 μm thick) where they can be partly altered by Fe oxide rims ( Fig. 7H-I). Larger euhedral pyrite crystals were not observed on valve interiors. At Piran and Brijuni, pyrite framboids located in borings (secondary linings) occur within strongly bored and stained valves (Fig. 7L). On the other hand, in addition to framboidal pyrite that 300 forms primary or secondary linings, another type of pyrite preservation can be detected at high magnification and is characteristic of stained valves. The valves that are stained show disseminated or dispersed inclusions of nanopyrite (< 1 μm) in BSE images. These inclusions are located in nanoscale-dissolution pits and microborings located in the inner or the outer layer; they are not located on valve surfaces or within the conchiolin layer ( Fig. 7J-L).
SEM images show that pyrite framboids on surfaces of non-bored or unworn valves at Po and Isonzo prodeltas are 305 attached to the original interior surface (primary linings). Dispersed microcrystals uniformly cover inner, well-preserved and smooth surface or fill irregularities close to the groove around the termination of the internal conchiolin layers (Fig. 8A-D).
Dispersed microcrystals and subspherical to spherical framboids consisting of microcrystals co-occur together and can be distributed across the whole interior surfaces in clusters (Fig. 8E), in strings ( Fig. 8F) and locally co-occur with Fe oxides and gypsum (Fig. 8F). Pristine valves show no or weak bioerosion by larger borers (sponges), although they can be 310 penetrated by simple non-branching borings with micrometric diameters. In contrast, pyrite framboids are rare on interiors of valves that are affected by dense borings produced by sponges, forming complicated galleries, with or without sediment infill ( Fig. 8G-H). SEM and BSE observations show that pyrite framboids directly attached to interior surfaces of wellpreserved and weakly-bored valves represent primary linings, whereas pyrite framboids occurring on highly-altered valves are located in borings and represented secondary linings. 315 https://doi.org/10.5194/bg-2021-153 Preprint. Discussion started: 23 June 2021 c Author(s) 2021. CC BY 4.0 License.

Age and depth distributions
Age distributions of valves with and without pyrite linings show right-skewed shapes in the mixed layer at prodelta sites, with median age equal to 7-10 (without pyrite) and 7-18 years (with pyrite linings) at Po (in the upper 20 cm) and to 11 (without pyrite) and 15 years (with pyrite linings) at Panzano (in the upper 6 cm), respectively ( Fig. 9A-F). They are 320 dominated by recentmost cohorts younger than 10 years. In the mixed layer, valves with and without pyrite linings exhibit the same distribution shape well-fitted by the exponential distribution, with similar loss-rate parameters within each site ( Fig.   9A-F). Therefore, the model with constant burial and disintegration outperforms or is equally efficient than more complex sequestration models both for valves with and without pyrite in the mixed layer at Po and Panzano (black lines in Fig. 9A-F, Table A1). The time to valve loss from the mixed layer is similar to 210 Pb-based estimates of sediment accumulation (1-2 325 cm/y at Po and 0.2 cm at Panzano), indicating that loss rates primarily correspond to burial rates. The pyrite framboids thus form at yearly scales and valves with and without pyrite linings are buried below the mixed layer at comparable rates. Age distributions of V. gibba valves without pyrite linings in the mixed layer at Piran are right-skewed but fat-tailed and contain valves that are older than 1,000 years (median age =1,100 years; only one 140 years-old valve was lined by pyrite, figure not shown). At the scale of the whole cores, median age of pyrite-lined valves and valves without pyrite are also comparable at 330 Po (27 and 33 years with and without pyrite, respectively) and Panzano (100 and 124 years with and without pyrite linings, respectively) ( Fig. 9G-L). However, the whole-core age distributions of valves show signs of multimodality, and the frequency with pyrite-lined valves peaks at 1986 AD at Po and at 1963 AD at Panzano. Equally-old pyrite-lined valves and valves without pyrite tend to be located at similar depths within the mixed layer ( Fig. 9M-O). The mean depth of valves younger than 10 years is 20 cm for specimens both with and without pyrite at Po, and pyrite-lined valves are located deeper 335 (at 60 cm) than valves without pyrite (at 40 cm) in the 20-year cohort. site gradient in sedimentation rate (associated with the gradient in grain size and carbonate content): V. gibba valves at sites 355 with high sedimentation are well preserved and rarely bored, encrusted, abraded or stained (Fig.4). In contrast, V. gibba valves at sites with slow sedimentation show a broader range of preservational signatures, with high abundance of bored, encrusted, worn and stained specimens (Fig. 5). Valves with periostracum or conchiolin layer are abundant at Po and Panzano (with 10-80% specimens per increment), whereas such specimens are extremely rare at Piran and Brijuni, with almost 100% specimens without periostracum. Principal coordinate analysis of increments from all sites shows a major 360 separation in preservation of V. gibba among increments deposited at low and high sedimentation rate ( Fig. 12A-B). The first PCO axis correlates negatively with frequencies of pyrite linings and positively with all other alteration variables (with the exception of fine-scale dissolution).

Stratigraphic trends in the frequency of pyrite-lined valves within the mixed layer 365
The frequency of pyrite-lined valves within the mixed layer is highly irregular at Po prodelta. Valves with pyrite increase in abundance below the mixed layer at the scale of 10 cm increments from 15-24% at 10-20 cm to 35-50% at Po 4. There is no clear downcore trend within this interval at Po 3, and the uppermost increments already contain 40% of pyrite-lined valves.
The abundances of pyrite-lined valves within the mixed layer increase from 10-20% at 4-8 cm to 55-70% at 12-20 cm at Panzano, with the increase coinciding with the base of the 210 Pb-defined mixed layer at 6 cm. At Piran and Brijuni, the 370 frequencies within the mixed layer also do not show any obvious trend.

Stratigraphic trends in the frequency of pyrite-lined valves below the mixed layer
At sites with high sedimentation rate, the frequency of pyrite-lined valves is highest (50-80%) in increments with the smallest bioturbation and the highest organic content (at 20-80 cm at Po and at 8-20 cm at Panzano, Fig. 12D), coinciding 375 with the late 20 th century eutrophication and hypoxia. This maximum declines downcore to 20% at Po prodelta (at 80-150 cm) and to 30-40% at Panzano (at 20-150 cm) in units deposited prior to the late 20 th century. The frequencies of pyrite-lined valves do not change markedly at sites with slow sedimentation and are typically less than 20% at Piran and Brijuni. The frequencies of pyrite-lined valves at Piran and Brijuni further decline when they are limited to valves not affected by bioerosion (black circles in Fig. 12D). Focusing on these sites with high sedimentation rate, PCO shows consistent 380 separation between the latest HST and the late 20 th century increments (Fig. 12C), reflecting the chronological increase in the frequency of pyrite linings. This within-core increase in the frequency of pyrite linings is coupled with an upcore increase in the frequency of valves without the periostracum (Fig. 13A). The frequencies of articulated valves and valves with fine-scale dissolution also increase upcore ( Fig. 13B-C). At sites with slower sedimentation and coarser, carbonate-rich sediments, the frequency of pyrite-lined valves remains constant upcore (<20%). The frequency of pyrite-lined valves not affected by 385 borers is <10%. The frequency of bored specimens gradually increases from 20-30% in the lower (transgressive systems tract) increments to 70-80% in the upper (late-highstand systems tract) increments at Piran (Fig. 13). In contrast, the frequency of bored specimens gradually declines from 90-100% in the lower (transgressive systems tract) incremens to ~40% in the upper highstand increments at Brijuni (Fig. 13).
Per-increment maximum shell size of V. gibba correlates positively with its proportional abundance (r = 0.46, p < 0.0001) and absolute abundance (r = 0.46, p < 0.0001) across all sites. Proportional abundance of V. gibba per increment correlate positively but rather weakly with the per-increment frequency of valves lined with pyrite framboids (r [prop. abundance] = 0.21, p = 0.016). Absolute abundance does not correlate with the per-increment frequency of pyrite-lined valves across all 395 sites (r [abs. abundance] = -0.001, p = 0.99, Fig. 14). However, in contrast to regional-scale relationships that are affected by between-core differences in time averaging, the positive relationships between proportional abundance and the frequency of

Effects of net sedimentation rate on preservation of pyrite linings
Our observation that the pyrite-lined valves are preserved at high abundance in the subsurface stratigraphic record at Po and 420 Isonzo prodeltas whereas they are infrequent at Piran and Brijuni indicates that the condition for their permanent sequestration below the mixed layer is enhanced when net sedimentation rate is high. Some pyrite framboids on valves at sites with slow sedimentation probably represent secondary linings because they directly occur within borings in heavilybored valves (i.e., the difference between frequencies of pyrite linings in all and non-bored valves is large at Brijuni, Fig.   12D). This preservation contrasts with well-preserved pyrite-lined valves at sites with high sedimentation where exclusion of 425 bored valves does not reduce their frequency (i.e., the difference between frequencies of pyrite linings in all and non-bored valves is small at Po and Panzano, Fig. 12D). The preservation of primary pyrite linings is thus largely limited to sites with high sedimentation rate (> 0.2 cm/y). https://doi.org/10.5194/bg-2021-153 Preprint. Discussion started: 23 June 2021 c Author(s) 2021. CC BY 4.0 License. sediment organic enrichment and high oxygen demand of sediments, thus perpetually reducing the extent of O 2 sediment 510 penetration, and/or (ii) when recovery of infaunal communities in the wake of anoxic or hypoxic events in environments affected by eutrophication (as in the northern Adriatic Sea) is slow and in hysteresis (Kemp et al., 2009;Borja et al., 2010;Duarte et al., 2015). Depending on the frequency and duration of sediment organic enrichment, two preservation scenarios can be envisioned (Fig. 15): (i) If the frequency of anoxic or hypoxic events is low and benthic communities rapidly recover, articulated shells will disarticulate and disintegrate when exposed to scavengers, borers and degradation of organics in TAZ. 515 For example, maceration of the conchiolin layer triggers delamination of valves into inner and outer layers (Fig. 3E-H). In such conditions, O 2 penetration and the thickness of the ferruginous suboxic zone will increase (van de Velde and Meysman, 2016), and the thickness of the aerobic zone and the whole mixed layer will increase (Fig. 15A). Pyrite framboids will not nucleate on valves anymore if the labile biomass and microbes coating the decaying tissues were degraded during the earlier phase of decomposition. Therefore, even when sedimentation rate is relatively high and associated with bacterial sulfate 520 reduction in microniches surrounded by iron-dominated pore waters, bioirrigation will catch up with pyrite-lined shells prior to their deep burial and thus inhibit their subsurface sequestration by oxidation. (ii) If the frequency of hypoxic events is high relative to the recovery time of the burrowing community, some subset of pyrite-lined shells can remain preserved if recovery of bioirrigation-inducing burrowers is slow and O 2 penetration depths remains close to the sediment-water interface (Fig. 15B). These conditions with reduced bioturbation and limited iron recycling are typical of oxygen-depleted or strongly-525 eutrophied environments (Karlson et al., 2007;Lehtoranta et al., 2009). Although the present-day estimates of the mixedlayer depth in the northern Adriatic Sea do not indicate any apparent limitation on bioturbation (see below), we suggest that such conditions also characterized benthic habitats with fine-grained sedimentation in the northern Adriatic Sea in the late 20 th century. If the nutrient-fueled eutrophication or other sources of sediment organic enrichment lead to permanent oxygendepletion and to sulfidic sediments at the sediment-water interface (and the formation of dead-zones), the concentrations of 530 pyrite framboids within shells or within intra-skeletal pores will be limited by iron availability, and most pyrite will be disseminated rather than clustered.

Temporal changes in bioturbation generate stratigraphic trends in frequency of pyrite-lined valves
Within-core trends in the frequency of pyrite-lined valves at prodelta sites and the modality of whole-core age distributions 535 of pyrite-lined valves hint to chronological changes in the preservation of pyrite linings. First, in addition to a general feature of the whole-core age distributions dominated by valves younger than 10 years, generated under active input of new dead shells into the mixed layer, older modes in these distributions are indicative about temporal changes in conditions that favored or inhibited the preservation of pyrite linings. At Po, pyrite-lined valves show a secondary mode at ~30-50 years that is formed by cohorts from the late 20 th century, and most valves older than ~100 years do not have pyrite linings (Fig. 9E-F). 540 Similarly, at Panzano, pyrite-lined valves form a mode that is represented by cohorts from the late 20th century (Fig 9G-H).  (Tomašových et al., 2018(Tomašových et al., , 2020. These levels correspond to the mid-20 th century at both prodeltas, and coincide with an increase in nutrient load and in seasonal hypoxia and mucilage frequency in the northern Adriatic Sea (Marchetti et al., 1989;Justic, 1991). Several mutually not exclusive factors probably contributed to the stratigraphic increase in the frequency of pyrite-lined valves, including (1) a temporal increase in hypoxia coupled with an increase in sediment organic enrichment, and (2) a temporal decline in sediment mixing and bioirrigation. Although the initial formation of pyrite framboids in reduced microniches in near-surface sediments can naturally occur in fine-grained 555 and poorly-permeable sediments, it was probably also enhanced by multiple processes that occurred during or in response to hypoxic events and accentuated seasonal shallowing of the oxic-anoxic sediment boundary (Middelburg and Levin, 2009): by a decline in dissolved oxygen concentrations in the overlying water column (reducing sediment penetration by O 2 although permanent anoxia underlain by sulfidic sediment can trigger iron escape from sediment, Pakhomova et al., 2007), by decay of planktonic and benthic phytoplankton or mucilages that increased oxygen demand of sediments as they 560 accumulated on the sediment-water interface (Herndl et al., 1987(Herndl et al., , 1989, or by decay of high-biomass epifaunal communities during late-summer mass mortality events (Stachowitsch, 1984;Nebelsick et al., 1997). The frequency and magnitude of all these effects probably increased in the late 20 th century in the northern Adriatic Sea, thus increasing the likelihood of initiation of reducing microniches within articulated shells of bivalves (rather than being exposed to disarticulation by predators or scavengers). 565 Regardless of initial pathways that led to the formation of pyrite linings in reduced microenvironments, hypoxic and mucilage events that are seasonal or inter-annual in frequency in the northern Adriatic Sea tend to be followed by recovery of infaunal species under oxygenated conditions (leading to iron recycling and pyrite oxidation), and oxygen depletion of near-surface sediments is thus not permanent if infaunal communities fully recover. However, seasonal anoxic or hypoxic events have lasting effects on mixing and irrigation rates because benthic communities tend to recover from such events at 570 slow rate (Bentley and Nittrouer, 1999;Solan et al., 2004) and can remain in a hysteresis state (Duarte et al., 2015). This delay, with benthic recoveries occurring over several years (Borja et al., 2010), was also observed in the northern Adriatic Sea (Stachowitsch, 1991). The decline in mixing and bioirrigation in the wake of hypoxic events (that became more frequent during the late 20 th century in the northern Adriatic Sea) was probably crucial for transfer of pyrite linings into the subsurface stratigraphic record, as envisioned in the scenario in Fig. 15B. The present-day bioturbation rates are thus 575 probably lower relative to those that characterized the northern Adriatic Sea prior to the late 20 th century. This inference is supported by three observations. First, the positive relation between V. gibba size and biomass on one hand and abundances of pyrite-lined valves on the other hand support the hypothesis that eutrophication-driven regime shift in the functioning of benthic communities in the late 20 th century reduced biomixing and bioirrigation. Second, at Po and Isonzo prodeltas, the late 20 th century infauna is dominated by shallow-burrowing deposit-and detritus-feeders that modify surface sediments, 580 including Owenia fusiformis, Varicorbula gibba or Ampelisca diadema (Occhipinti-Ambrogi et al., 2005;Solis-Weiss et al., 2007;N'Siala et al., 2008). These patterns differ from early 20th century ecological surveys (Vatova, 1949;Crema et al., 1991;Schinner, 1993;Schinner et al., 1997)  slopes (interspersed with laminated sediments, Meadows et al. 2000;Smith et al. 2000;Levin et al. 2003). However, the estimates of the mixing depths on the basis of 210 Pb profiles indicated mixing depths < 10 cm in the late 20 th century at Po prodelta (Frignani and Langone, 1991;Frignani et al., 2005;Alvisi et al., 2006Alvisi et al., , 2009 and the thickness of the mixed layer at 595 the Po prodelta as estimated on the basis of sediment cores probably exceeded 20 cm prior to the late 20 th century (Tomašových et al., 2018). In addition to changes in sediment mixing, the nature of benthic communities indicates that irrigation was slower and that iron and sulfide recycling was less efficient in the late 20 th century at Po and Isonzo prodeltas than earlier. To conclude, these observations support the hypothesis that the stratigraphic shift towards higher pyrite frequency of pyrite-lined shells coincides with changes in the composition of macrobenthic communities that reduced their 600 mixing and irrigation efficiency during the 20 th century.

Implications for the fossil record: inferring slow and patchy bioturbation and limited residence time
The pyrite framboids lining intra-skeletal pores (originally filled with organic tissues) in well-preserved shells represent a unique indicator of slow irrigation and their permanent transfer below the oxic-anoxic sediment interface. Such conditions can be produced by delayed recoveries from hypoxic events and/or by community states with low bioirrigation potential that 605 are unable to recover anymore even when the frequency of hypoxic events returns to pre-impact levels (Steckbauer et al., 2011). In permanently-normoxic environments with intense bioturbation where most labile biomass degrades within the aerobic zone and any early pyrite is oxidized, the frequency of shells with shell-lined pyrite transferred into the permanent record will be negligible. This index can be also used to track the net O 2 exposure of skeletal remains and recycling efficiency of iron and sulfides in the deep-time stratigraphic record because pyrite-lined shells represents a distinct 610 taphonomic and diagenetic signature of fossil assemblages preserved in fine-grained sediments (Kobluk and Risk, 1977;Hudson, 1982;Bjerreskov, 1991;Underwood and Bottrell, 1994;Farrell et al., 2009;Brett et al., 2012a), especially in organisms with internal skeletal cavities that do not immediately open after their death (Hudson, 1982;Fisher, 1986;Loope and Watkins, 1989;Jin et al., 2007). On one hand, preservation of well-preserved, frequently articulated skeletal remains of organisms with otherwise fragile elements, coupled with pyrite linings, is a typical taphonomic feature of assemblages 615 preserved in fine-grained sediments in the Paleozoic (Brett et al., 2012a, b) or in the Mesozoic successions (Hudson, 1982;Fernández-López et al., 2000;Paul et al., 2008;Reolid, 2014). On the other hand, actualistic studies assessing skeletal alteration of molluscs, brachiopods or echinoderms rarely record this type of preservation in surface Holocene sediments. It is possible that the concentration of actualistic studies on the taphonomic processes in the mixed layer and on environments with slow sediment accumulation rates underestimate this type of preservation. However, here, we suggest that prodelta 620 sediments in the northern Adriatic Sea affected by the late 20 th century eutrophication can represent analogue conditions that lead to preservation of well-preserved and pyrite-lined shells in the deep-time stratigraphic record. In contrast to models invoking rapid episodic burial to explain the initial sequestration of shells so that they do not disarticulate and decay under anaerobic conditions, the pathway observed in prodelta sediments in the northern Adriatic Sea probably occurs without episodic burial. Although background sedimentation rates need to be sufficiently high so that the time for burial of skeletal 625 remains is shorter the time to mix and irrigate the mixed layer, the effect of sedimentation rate is critical for minimizing the potential for reoxidation rather than for initial sequestration of shells in microniches.

Conclusions
Preservation of pyrite-lined shells as a function of rapid and permanent sequestration below the taphonomic active zone is an 630 indicator of inefficient mixing and bioirrigation, thus documenting limited iron and sulfur recycling at sites with high https://doi.org/10.5194/bg-2021-153 Preprint. Discussion started: 23 June 2021 c Author(s) 2021. CC BY 4.0 License. sedimentation rates in the northern Adriatic Sea. The preservation pathway that leads to primary pyrite linings and their long-term preservation is indicative of permanently-limited depths of O 2 penetration and bioirrigation that can be difficult to detect on the basis of trace fossils and ichnofabric only. Pyrite-lined valves thus represent a unique type of alteration that contrasts with other types of alteration whose incidence increases with residence time in the taphonomic active zone. We 635 suggest that the increase in the frequency of valves with pyrite below the mixed layer at 80-90 cm at the Po prodelta and at 12-20 cm at the Isonzo prodelta represents a temporal signal of the decline in the rate of mixing and bioirrigation in muddy sediments of the northern Adriatic Sea driven by a late 20th century increase in the frequency of seasonal or inter-annual hypoxia and organic enrichment that delayed the recovery of infaunal communities. Although the rates of pyrite formation can also vary over long time scales owing to long-term changes in seawater chemistry (Leavitt et al., 2013;Algeo et al., 640 2015) and depend on the supply of organic matter and iron availability (Goldhaber et al., 1977;Berner and Raiswell, 1983;Berner, 1984;Berner and Westrich, 1985;Kershaw et al., 2018;Wignall et al., 2005Wignall et al., , 2010, we hypothesize that the frequency of pyrite-lined shells (belonging to organisms that inhabit oxic sediment zones) can improve inferences about mixing and irrigation and about the net exposure time of skeletal particles to irrigation in the mixed layer.             grains, interpreted to reflect the response of ecological and taphonomic processes as responding together to higher hypoxia frequency and a narrower extent of aerobic respiration in sediment, possibly linked by higher input of organics at times of mass mortalities and reduced mixing during the recovery. Within-core relationships show stronger relationships than 1200 regional-scale relationships that are affected by between-core differences in time averaging and thus absolute and proportional abundances are noisy.
https://doi.org/10.5194/bg-2021-153 Preprint. Discussion started: 23 June 2021 c Author(s) 2021. CC BY 4.0 License. Figure 15 -Two pathways characterized by differences in the frequency of sediment organic enrichment and hypoxic events and/or by differences in recovery rate of infaunal communities with high mixing and irrigation potential. Both pathways promote the initial formation of reduced microniches when shallow-infaunal shells decay in sediments with iron-dominated pore water. A. In the aftermath of hypoxia, reduced microniches are formed initially, but these are recycled under recovery of infaunal communities that oxidize sediment and contribute to skeletal disintegration of initially articulated shells. B. If the 1210 frequency of hypoxic events associated with higher oxygen consumption is high, the oxic-anoxic interface repeatedly shallow within the sediment. The recovery of burrowing infauna is slow and bioirrigation remains patchy, and some subset of reduced microniches is thus not oxidized. Therefore, shells with pyrite linings can escape the mixed layer in the subsurface stratigraphic record even when the sediment fabric is bioturbated. In the historical layer, valves permanently remain in reducing conditions. 1215 https://doi.org/10.5194/bg-2021-153 Preprint. Discussion started: 23 June 2021 c Author(s) 2021. CC BY 4.0 License. Table A1 -The results of fitting age-frequency distributions (with valves with and without pyrite) from the mixed layer at three stations to a simple disintegration model (with temporally-constant loss rate from the mixed layer) and to a sequestration model with three parameters. AICc -Akaike Information criterion corrected for sample size. 1220 Simple disintegration model (1 parameter