Age structure, carbonate production and shell loss rate in 1 an Early Miocene reef of the giant oyster Crassostrea 2 gryphoides 3

Abstract


Introduction
The genus Crassostrea Sacco, 1897 comprises numerous commercially exploited species.The modes of growth and population structures of extant Crassostrea species are of paramount importance for oyster fishery (FAO, 2015).Consequently, a wealth of data exists on the ontogeny, biology and ecological requirements.These data, in turn are a valuable base for the interpretation of the autecology of fossil congeners.Extant Crassostrea are sessile bivalves adapted to estuarine and intertidal environments where they have to cope with high environmental stress.Whilst physicochemical stress is managed by genetic response (Zhang et al., 2012), the formation of thick shells is a strategy against predation (Lombardi et al., 2013;Robinson et al., 2014).The largest and fastest growing Crassostrea species flourished during the Miocene and Early Pliocene and became replaced by comparatively smaller and thinner species thereafter (Kirby and Jackson, 2004).Among these, Crassostrea gryphoides (Schlotheim, 1813) is the largest, attaining shell lengths of up to 80 cm and individual ages of more than 40 years (Harzhauser et al., 2010).The Pleistocene-Holocene extirpation of large and thick-shelled Crassostrea species was explained by a shift from shallow-marine towards estuarine-intertidal habitats to escape from predation (Kirby, 2000(Kirby, , 2001)).At least for C. gryphoides this model does not fit because the oyster lived as secondary soft-bottom dweller in the intertidal zone of estuaries along the circum-Tethyan coasts (Laurain, 1980;Schultz, 2001;Mandic et al., 2004).This species evolved during the Late Oligocene in the European Paratethys Sea and became ubiquitous in the Paratethys Sea, the proto-Mediterranean Sea and the eastern Atlantic throughout the Early to Late Miocene.It might even have entered the Indian Ocean during the Miocene (Newton and Smith, 1912) and reached the North Sea during the Middle Miocene (Schultz, 2001).Pliocene records from the eastern Atlantic and North Africa, however, may need verification (see Schultz, 2001 for a detailed list of occurrences).This species became extinct around the Miocene/Pliocene boundary or with the onset of the Pliocene cooling 3 Ma ago at the latest.
Studies on growth in fossil Crassostrea species (and other oysters) and the resulting carbonate production were published by Chinzei (1982Chinzei ( , 1986Chinzei ( , 1995)), Chinzei and Seilacher (1993), Kirby (2001) and Kirby and Jackson (2004).These studies are based on collection material comparing species and specimens from different stratigraphic horizons.No study, however, tried to capture the size and age structures of a fossil Crassostrea reef, presumably representing a real population of coeval specimens.The lack of such studies is clearly linked to the fact that fossil shell beds usually represent time-averaged assemblages (Kidwell, 1986(Kidwell, , 1991)), which only vaguely reflect original community structures.Although some in-situ preserved fossil Crassostrea reefs are known (e.g.Hoşgör, 2008;Ragaini and Di Celma, 2009;Chinzei, 2013) no population data exist, which would allow a comparison with modern oyster reefs.Herein, we analyze an Early Miocene Crassostrea shell bed covering an area of 459 m², which is permanently exposed at a geoment park in Austria.The shells are concentrated in a sheet-like, c. 20-cm-thick layer, which was formed by a major storm or tsunami, amalgamating a single oyster reef in an event bed (Fig. 1A-C).Although, the hydrodynamic process will have biased the original structure to some degree, our data suggest that no out-of-habitat transport occurred and the shell bed still reflects the original composition and thus the population structure of an oyster reef from the onset of the Miocene Climatic Optimum (Zachos et al., 2008;Goldner et al., 2014).
Crassostrea reefs flourished during the Miocene within the tropical reef belt (Mandic et al., 2004) but were also successful in more northern latitudes (Wiedl et al., 2013;Harzhauser et al., 2010).Therefore, the main purpose of this paper is to quantify the growth performance of the Miocene giant oyster and to reveal its significance as part of the Miocene "carbonate factory".Moreover, we test if size frequency data deduced from fossil oyster shells allow a comparison with community structures of extant Crassostrea reefs.

A taxonomic note
Although the type of Crassostrea gryphoides (Schlotheim, 1813) was described from the mid-Miocene of Romania, the name is also used in biological literature for an extant backwater species from India and Pakistan (Newton and Smith, 1912;Chatterji et al., 1985;Nagi et al., 2011;Afsar et al., 2014;Trivedi et al., 2015).The taxonomic status of the extant species is unclear; based on molecular data, Reece et al. (2008) considered C. "gryphoides" to be closely related with C. belcheri (Sowerby, 1871), whilst molecular data of Trivedi et al. (2015) suggest a close relation with C. cuttakenses (Newton and Smith, 1912), which was originally described as subspecies of C. gryphoides.Whatever the taxonomic and systematic status of the recent species may be, it is most probably not conspecific with the European fossil species.It differs in its more regular and elongate-ovoid outline (Durve and Bal, 1960), the short and bean-shaped adductor muscle scar (Durve, 1974;Siddiqui and Ahmed, 2002) and it is more inequivalved.Overall, the recent species is clearly smaller with the largest specimen documented so far attaining 508 mm in length (Mahar and Awan, 2012) but usually ranging around 50-120 mm (Chatterji et al., 1985;Nagi et al., 2011).
Nevertheless, the Miocene species is closely related to the extant Asian-Pacific Crassostreinae species, which show a large genetic difference from Atlantic species (Littlewood, 1994;Ó Foighil et al. 1995;Wang et al., 2004;Ren et al., 2010;Salvi et al., 2014) (Wang et al., 2004;Ren et al., 2010;Salvi et al., 2014).The divergence between both groups seems to have happened already in Cretaceous times and the diversification of the Asian Pacific group started during the Eocene (Ren et al., 2014).Based on this genetic evidence and biogeographic separation, Salvi et al. (2014) introduced the genus Magallana Salvi, Macali and Mariottini, 2014 for the Asian Pacific species group with Ostrea gigas Thunberg, 1793 as type species.Salvi et al. (2014), however, did not provide a formal "description or definition that states in words characters that are purported to differentiate the taxon" (ICNZ Art.13.1.1).The authors only refer to a description of the type species.Therefore, Magallana is formally unavailable (see also Marshall, 2015) and we use Crassostrea in its traditional sense.

Geological setting and paleoenvironment
The investigated oyster shell bed was excavated at Stetten in Lower Austria (48° 22' 03.33 N, 16° 21' 33.22 E).It is part of the about 600-m-thick lower Miocene (upper Burdigalian, = Karpatian regional stage) siliciclastic succession of the Korneuburg Basin, which is a 20 km long and 7 km wide halfgraben within the Alpine-Carpathian thrust-belt (Fig. 2) (Wessely, 1998).The basin fill is dominated by an alternation of shoreface and tidal flat deposits, which were formed in an embayment of the Paratethys Sea (Zuschin et al., 2014).More than 650 species-level taxa have been described from the area (Sovis andSchmid, 1998, 2002) and the paleoenvironments are well understood (Harzhauser et al., 2002;Zuschin et al., 2014).The oyster reef flourished in an estuary fringed by salt marshes, Taxodiaceae swamps and scattered Avicennia mangroves (Harzhauser et al., 2002;Kern et al., 2010).The pollen record of the Stetten section documents a warm subtropical climate with marked seasonality (Kern et al., 2010).A warm and wet summer season with c. 204-236 mm precipitation during the wettest month was alternating with a rather dry winter season with precipitation of c. 9-24 mm during the driest month.The mean annual temperature ranged between 15.7-20.8°C, with about 9.6-13.3°C during the cold season and 24.7-27.9°C during the warmest month.
These data suggest similarities with the modern "Cwa" climate of Koeppen (1936).Today this climate covers parts of northern India extending into south-eastern Asia (south Nepal, Myanmar, northern Thailand) to East China and in central south Africa (east Angola, Zambia, north Zimbabwe, north Mozambique) (Peel et al., 2007;Kern et al., 2010).The distinct seasonality was also revealed in sclerochronologic analysis of one of the Crassostrea shells collected from the shell bed (Harzhauser et al., 2010).This shell exhibits a regular annual rhythm of at least 11 seasons with a temperature range of 9.8°C.Thus, the paleoclimatic and paleoenvironmental frame of the C. gryphoides shell bed is comparable to the settings of modern Crassostrea reefs in the subtropical parts of the Asian Pacific.
The fossil shell bed was excavated in a 3-months-campaign by the Natural History Museum Vienna in 2008.The oyster bed was covered by up to ten meters of silty sand and clay, which was successively removed.Due to the largely unconsolidated state of the surrounding silty sand, the excavation of the shell bed could be done manually with steel gravers and brushes; no water or any chemicals were added and all shells and fragments remained in their original position.The oyster shells themselves are well preserved and robust.Therefore, no artificial fragmentation occurred during the excavation.
Originally, the shell bed was nearly flat at the time of deposition but has now an undulate surface due postsedimentary tectonic activity.This tectonic phase occurred during the Middle Miocene at least 1-2 million years after deposition and caused a tilting of the units of ca.25° in western direction.During that tilting, a NW-SE trending fault system developed that caused the current relief.Locally, the displacement by the faults is in the range of few cm.

Harzhauser et al. (in press
) describe the complex taphonomy of the shell bed, which was formed by a tsunami or an exceptional storm and represents an event deposit sensu Einsele et al. (1991) and Kidwell (1991).As discussed by Harzhauser et al. (2015), the assemblage is not monospecific but contains about 46 molluscan species of which Crassostrea gryphoides predominates in individual numbers (79.4%).The species, such as the potamidid gastropod Ptychopotamides papaveraceus and the venerid bivalve Venerupis basteroti, lived partly within the oyster reef or were admixed from adjacent mudflats and shallow sublittoral habitats.As shown by Harzhauser et al. (2015), the fossil bed is parautochthonous.Although the oyster shell bed is clearly not in-situ but reworked, the original community structure still seems to be reflected, which is the basic working hypothesis of this paper.Lack of sorting is indicated by the accumulation of very small and very large shells.Similarly, the equal contribution by left and right valves points to the preservation of the primary composition and contradicts the hypothesis of hydrodynamic sorting and selective transport.Despite the bias of post-event processes, when the exhumed shells were exposed for few years on the seafloor, the rapid subsequent burial preserved most of the original distribution patterns.

Data acquisition by Terrestrial Laser Scanning and orthophotos
Terrestrial Laser Scanning has triggered a revolution in topographic terrain capturing, especially in the generation of digital terrain models.Methods for generating such models from laser scanning data are discussed by Kraus & Pfeifer (2001) and references therein.
Terrestrial Laser Scanning was applied to document the site as georeferenced 3D point cloud (Otepka et al., 2013).A Faro Focus Laser scanner with a nominal point measurement accuracy of 1 mm (std.dev.) in each coordinate and a sampling distance of approximately 1 mm was used.The individual point clouds of each scan were transformed first into one common coordinate system and then georeferenced by control points to Universal Transverse Mercator (UTM) coordinates (resolution below 2 mm).A robust filter (pre-processing) was applied to reduce measurement noise while preserving surface structures like sharp edges (Nothegger and Dorninger, 2009).The surface triangulation is based on the Poisson surface reconstruction method (Kazhdan et al., 2006).The points of this triangulation are used for interpolating a regular grid of heights above the plane of the shell bed using the scientific software OPALS (Pfeifer et al., 2014).In addition, a Canon 60D with a Canon EF 20 mm f2.8 was used to capture more than 300 photos from a moving platform.The camera was placed approximately orthogonal to the fossil bed.From the photos with a nominal ground resolution of approximately 0.6 mm per pixel an orthophoto mosaic was generated with a resolution of 0.5 mm per pixel.To detect patterns in the distribution and composition of shells two transects (N-S, W-E) were defined, each represented by 42 m 2 with a central overlap (Fig. 1).
All objects within this area were manually outlined on the digital surface model and cross checked based on the high-resolution orthophotos (Fig. 3).
Manual outlines are vector datasets in form of manually digitized polygons representing the boundaries of the identified specimens.Note that the attribute fragmentation does not contain any information on abrasion.The fragments usually show sharp fractures and therefore, the main cause for fragmentation seems to be predatory and hydrodynamic breakage.The ratio between left and right valves is balanced (0.98).The distribution of the shells is not uniform, occasionally featuring areas of higher shell densities, which seem to reflect former colony-like concentrations.

Shell length and area
Crassostrea gryphoides shows a very broad range of morphologies, ranging from elongate shells to strongly curved and sigmoidal shapes (Fig. 4).Therefore, measuring shell length as straight line, as done in other extant and fossil Crassostrea species, is inadequate.To overcome this problem, we evaluated shell length based on the 2D center line.Centre line length is the term used in photogrammetry and aims for capturing the real shell length as far as possible.Here it is an imaginary curved line spanning the maximum length of the shell.
The advantage of this method is that the center line will approximate the "real" lengths of the curved and irregularly shaped shells much better than any manual attempt in the field.
For the automatic determination of the center line we used the shell margins, which comprise about 1000 points on average.For easier calculation the outline point number was reduced to 100 and then filtered to points with close to even spacing.In the next step, a Delaunay triangulation was calculated between the filtered outline points (Delaunay, 1934), constrained by the edges between the outline points.To find the center line for each oyster outline, the Voronoi diagram was formed (Voronoi, 1908) from the triangulation.The edges between neighboring Voronoi vertices within the boundary are the medial axis transform (MAT) for the oyster outline (Aichholzer et al., 1996).The longest 2D path in this tree was found using Dijkstra's algorithm between MAT end points (Kirk, 2015).
The center line lengths of 1121 complete Crassostrea shells, rounded to the nearest mm, range from 48 mm to 602 mm with a mean of 237 mm (σ=89 mm) (Fig. 5A).The data distribution displays a positive skewness of 0.52 and the Shapiro-Wilk test excludes normal distribution for raw data and log10-transformed measurements.Area data range from 1708 mm 2 to 56755 mm 2 with a mean of 16983 mm 2 (σ=8414 mm 2 ) (Fig. 5B).These data show also a positive skewness (0.83) and normal distribution is rejected by the Shapiro-Wilk test.
Based on the manual outlines, the exposed shell area can be deduced directly.Area data are slightly underestimated because shells are not always exposed parallel to the bedding plane but may be somewhat oblique.Despite the fact, that area data are somewhat biased by oblique shells, the correlation between center line lengths and areas is highly significant (raw data: r = 0.92, p < 0.001; log10-transformed: r = 0.93, p < 0.001) (Fig. 6).

Length frequency data
Non-normal distribution of length-frequencies is a common pattern in extant Crassostrea reefs (e.g.Coakley, 2004;Baqueiro Cárdenas and Aldana Aranda, 2007;Harding et al., 2008;Nurul Amin et al. 2008;Ross and Luckenbach, 2009;Goslier et al., 2014).It results from seasonal and/or annual recruitment with distinct cohorts (sensu Powell et al., 2006Powell et al., , 2015;;Southworth et al., 2010).For instance, the (sub)tropical C. madrasensis and C. rhizophorae display a distinct annual recruitment peak (Nurul Amin et al., 2008;Mancera and Mendo, 1996).In multiannual communities this results in a right-skewed distribution due to the loss of old specimens by natural mortality and shell loss.For extant Crassostrea reefs, the analysis of the cohorts is routinely performed using Bhattacharya's model or the EM-Algorithm of Dempster et al. (1977), which tries to detect normal distributions within the length-frequency data.Consequently, in order to test for cohort mixing, lengths of C. gryphoides were subjected to mixture analysis, a maximum-likelihood method for estimating the parameters of two or more univariate normal distributions, based on a pooled univariate sample (Hammer, 2015).Statistical analyses were performed in PAST versions 2.17c and 3.06 (Hammer et al. 2001).Akaike's Information Criterion (AIC) was used to test the goodness of fit of the maximum likelihood estimates to the length-frequency data.
In log10-transformed length frequency diagrams, the maximum likelihood based analysis reveals lowest AIC values for four or five cohorts.Similarly, log-transformed area data have lowest AIC values if four or five cohorts are detected.Assuming more groups does not lower the AIC or the computed cohorts comprise unrealistic narrow cohort ranges, which are nested within larger ones.

3.4
Growth model Kirby et al. (1998), Kirby (2000Kirby ( , 2001) ) and Kirby and Jackson (2004)  According to these authors, a shell from Stetten attained 43 cm in length at an age of 11 years and the second one from a slightly younger horizon was 63 cm long at an age of 16 years.
These values document an average growth rate of ~3.9 cm per year and might serve as a base for rough age estimates for the complete adult shells from the Stetten shell bed.
For juvenile shells, this estimate would be wrong due to the non-linear mode of growth known from extant Crassostrea species, which show very high initial growth rates (Kennedy et al., 1996).Similarly, growth rates of Crassostrea gryphoides seem to decline in very old and large specimens as shown for a 78-cm-long and 41-year-old shell from the upper Langhian of Austria (Harzhauser et al., 2010).To cope with the non-linear growth, growth curves of extant Crassostrea species are routinely calculated with the von Bertalanffy equation (von Bertalanffy, 1934).This equation is SL t = SL max (1-e -k(t-t0) ), where SL t is shell length at time t, SL max is the asymptotic shell length, t 0 is the size at time 0, and k is a rate constant.Herein, we used the length of the center line of each shell as SL t as this measure captures the real growth length of the partly strongly curved or sigmoid specimens.For SL max we used the size-to-age data of the 78-cm-long shell, which is the largest individual known so far.

Ratio between chalky and foliate layers
The calcitic Crassostrea shells consist of two structures: thin but densely spaced foliate layers separated by thick layers of light-weight chalky material (Stenzel, 1971;Higuera-Ruiz and Elorza, 2009).This fast growing structure is interpreted to be a major adaptive advantage of Crassostrea to impede drilling predation and to prevent from sinking in the soft bottom (Seilacher, 1984;Chinzei, 1995;Kirby, 2001;Vermeij, 2014).In fossil shells the chalky layer is completely recrystallized and has the same density as the foliate layer.Nevertheless, it is optically easily recognized by its lighter color and the nearly opaque appearance.A polished longitudinal section of an articulated C. gryphoides shell (providing data for left and right shells) was scanned and the ratio between both shell structures was quantified by image analysis (Fig. 7).This method is only an approximation to the true value, as the ratio may vary locally (Durve and Bal, 1960;Chinzei, 1995), but it is clearly an improvement compared to former studies that used only linear transects or sectors within the shell (Durve and Bal, 1960;Chinzei, 1995;Kirby, 2001).
The image analysis of the cross-section documents proportions of 64% chalky to 36% foliate layers for the right shell and of 61% to 39% for the left shell.The density of the chalky layer when wet ranges around 1.15-1.32g cm -3 (Chinzei, 1995) and the density of the foliate layer ranges around 2.2-2.5 g cm -3 (Chinzei, 1995;Yoon et al., 2003) and has a clear upper limit by the density of calcite (2.7 g cm -3 ).Using 1.2 g cm -3 for the chalky layer and 2.2 g cm -3 for the foliate layer as rough estimates, mean density results in 1.84 g cm -3 and 1.81 g cm -3 for the right and left valves, respectively.Shell density in Crassostrea species is independent from age and size (Lombardi et al., 2013) and therefore the density estimates can be applied to the entire data set.

Shell volume
The volume of nine individual shells was determined using close range laser scanning technology, which provides high resolution models with sub-mm resolution.The specimens were selected from the collections of the Natural History Museum and vary in center length from 141 mm to 406 mm; four of these shells represent left-right shell pairs of an individual (Fig. 4).The data were captured with a measuring arm (METRIS MCA, 3600 M7).A handhelp triangulation laser scanner (a laser plane and a camera) was mounted at the end of two arms of fixed length with flexible joints.The laser scanner takes measurements with a maximum scan rate of 80 stripes per second with a strip-width of about 200 mm; the camera has a resolution of 1000 dots per strip.
In the first step, more than half of each shell was scanned and in the second step the other half.The overlap between both parts was more than 70%, which was sufficient for successful registration of the scanned parts.During this registration process, the geometric transformation is determined, which puts the two 3D laser point clouds together based on the points in the overlapping part.This procedure is done using the iterative closest point (ICP) algorithm (Glira et al., 2015).The resulting point cloud is analyzed in order to reduce noise and thus improve the surface description.Outliers (wrongly determined points not on the surface) were manually eliminated.Additionally, the raw point cloud (over 1.5 million points per shell) was uniformly sub-sampled to allow interactive handling.The final resolution is better than 0.18 mm (i.e.around 25 points per mm²).For volume calculation the point cloud has to be transformed into a closed mesh.Remaining holes, non-manifold surfaces and additional not connected components were identified and removed.Finally, the surface area of the mesh and its volume were computed using the algorithm of Mirtich (1996).
Based on the 9 measured shells a relation between center line length and volume can be deduced.The largest shell measured is 406 mm long but no empirical volume data are available for larger shells, because these cannot be removed from the site.Therefore, the von Bertalanffy equation would not be applicable for shells larger than ~40 cm.Consequently, we chose a logistic function to approximate the inverse von Bertalanffy equation: , where v is the volume in mm 3 and SL the center line length in mm (Fig. 8).Applying this equation to all shells results in a total volume of 393273 cm 3 with a mean shell volume of 350.8 cm 3 (σ=313.7).These values do not change significantly if a nonlinear Gompertz growth model is assumed as frequently done for Crassostrea (Lopes et al., 2013;Ginger et al., 2013).The respective equation results in a total volume of 398474 cm 3 .Applying the above discussed average shell gravity of 1.82 g cm -3 results in a total carbonate mass of ~715 kg (logistic) to ~725 kg (Gompertz) for all 1121 shells.Thus, based on the age models of the shells, the annual carbonate production per shell can be calculated, which ranges from 74 g yr -1 (σ=2.9)(Gompertz) to 83 g yr -1 (σ=2.8)(logistic), accounting for ~150 g yr -1 per living oyster individual (= two valves).

From lengths to cohorts
Based on the assumption that the size-frequency groups represent age classes it is apparent to establish length-at-age relationships.Applying the von Bertalanffy equation to the length data reveals a strongly right-skewed distribution with 50% of the shells ranging between 3 and 6 years (Fig. 9A).The frequency of specimens between 6 and 9 years decreases rapidly and the contribution by shells older than 9 years is subordinate although outliers with up to 16 years occur.Again the non-normal distribution of the von Bertalanffy growth model data suggests cohort mixing and the mixture analysis assumes at least four significant cohorts with low AIC value.Due to the rareness of large and aged shells, the fourth cohort displays a rather low amplitude and we assume that at least two natural cohorts may be amalgamated in this group.
This suggests more or less continuous recruitment accentuated by very successful settlement peaks every 2 or 3 years.Similarly, comparable patterns in extant Crassostrea reefs are linked to fluctuating mortality rates and changing recruitment success (Southworth et al., 2010).The data show that old and large shells are rare.The reason for this may be a generally high mortality during the early years of growth, resulting in low survival rates and few old specimens.The high amount of fragments of large shells, however, suggests that a distinctly higher proportion of large shells existed but became successively destroyed.

Shell loss and mortality
The rapid decline of old shells is a combination of two processes.First, natural mortality will result in a tail of large shells.Second, shells of died-off Crassostrea are known to have surprisingly short half-lives ranging from few years to few decades (Powell et al., 2006(Powell et al., . 2015;;Waldbusser et al., 2011).Natural degradation processes, such as fragmentation, dissolution and hydrodynamic export may account for 30 % loss of shells per year (Southworth et al., 2010).The high amount of fragments with sharp edges suggests that fragmentation is a major factor in our case.The importance of hydrodynamic export cannot be evaluated due to the limited outcrop area.Dissolution, however, is a minor factor as the well preserved shellsurfaces lack any signs of chemical degradation.Based on the declining amplitudes of the cohorts the total shell loss can be computed as exponential decay function (Fig. 9B) revealing initial half-lives of less than 4 years.The high proportion of fragmented, abraded and/or bioeroded oyster shells (fragments: n = 7047) would balance the "missing frequencies" easily.
We have no evidence for age-specific shell destruction and assume that all age classes were equally affected by fragmentation.
Overall, under the environmental conditions as present in the Early Miocene estuary, a diedoff Crassostrea gryphoides reef would have been fully degraded within one or two decades if not buried below the taphonomically active zone (Olszewski, 2004).The high sedimentation rates in the rapidly subsiding basin with rates (0.6m/kyr, Zuschin et al., 2014), warranted a rapid burial of the reef, thus capturing the population structure.

Crassostrea as carbonate factory?
The image analysis based estimation of chalky and foliate layers shows that the values for chalky layers are slightly lower than reported for other fast-growing Crassostrea species.The volume of chalky layers in left shells of the extant Crassostrea "gryphoides" (sensu Newton and Smith, 1912) is ~70% (Durve and Bal, 1960).The shells of the Miocene Crassostrea titan (Conrad, 1853) and C. gravitesta (Yokoyama, 1926) were reported to comprise even up to ~90% of chalky deposits (Chinzei, 1995;Kirby, 2001) but this may be a slight overestimation, neglecting the high amount of dense calcite in the hinge areas.A re-evaluation of the illustrated specimen of C. gravitesta in Chinzei (1995) using our method reveals a somewhat lower but still high value of 77% chalky layer.The gravity of the fast growing C. titan, C. gravitesta and C. cahobasensis (Pilsbry and Brown, 1917) was estimated to range around 1.35-1.40g cm -3 (Chinzei, 1995;Kirby and Jackson, 2004).These values are very close to the empirical data on the fast growing C. ariakensis (Fujita, 1913) (1.44 ± 0.12 g cm -3 ) measured by Lombardi et al. (2013).Slightly higher values are given for C. gigas (1.63 ± 0.35 g cm -3 , mean = 1.58) by Chinzei (1995).Generally, species with low amounts of chalky layer have much higher densities; e.g., C. virginica shells range around 2.18-2.35g cm -3 (Kirby and Jackson, 2004;Lombardi et al., 2013).Our gravity estimate of ~1.82 g cm -3 for C. gryphoides is thus somewhat higher than expected for such a growth type.In fact, large parts of the shells are very light-weighed and fit well to the patterns discussed by Chinzei (1995) and Kirby (2001).The main difference is the large proportion of heavy shell material in the huge umbos and hinges.
A comparison of the growth curve of C. gryphoides with the von Bertalanffy growth models of fossil and extant Crassostrea species (Fig. 10) reveals this oyster as outstandingly fast growing species.Thus, C. gryphoides was an important carbonate producer in Neogene estuaries and lagoons where it lived as a secondary soft-bottom dweller in dense colonies in a mixed mode of shell-supported reclining and mud sticking (sensu Seilacher et al. 1985;Seilacher and Gishlick, 2014).Therefore, dense populations with more than 100 individuals per m 2 can be expected.Even within the shell bed, which is clearly not in-situ, the average density is 129 shells (~64 individuals) per m 2 (including also moderately fragmented shells).
This would point to a hypothetic annual carbonate production of up to 15 kg m -2 with the oyster reef.Although this calculation is just a very rough estimate, it indicates that the carbonate production is in the range of fast growing coral reefs with productions of 6-10 kg m -2 yr -1 (Montaggioni, 2005;Jones et al., 2015).A major difference, however, is the rapid shell loss in Crassostrea reefs, which prevents the formation of rigid and stable structures comparable to coral reefs.

Conclusion
Crassostrea gryphoides was the fastest growing and largest Crassostreinae species known so far.Despite the fact, that this species could attain outstanding individual ages of four decades (Harzhauser et al., 2010), the bulk of specimens analyzed herein lived less than 10 years, typically growing up to about 300 mm in length.
The non-normal distribution in the size, area and age frequency data are best explained by the presence of distinct recruitment cohorts, comparable to modern oyster reefs.About four cohorts are detected by mixture analysis and the rapidly decreasing amplitudes of frequency of these cohorts is interpreted to reflect the combined effect of mortality, the declining life expectancy with age, and the shell loss by biotic and physical factors.As no accumulation of large and aged shells occurred, whilst the amount of fragments is high, we assume that shell loss is an important factor to explain the strongly right-skewed distribution.Shell half-lives ranged around 2-4 years and within less than two decades the seemingly rigid and persisting structure of a Crassostrea gryphoides reef could have been completely degraded.This may explain the rareness of in-situ C. gryphoides reefs in the fossil record although the shells are frequent and ubiquitous.
The significant growth rate is clearly boosted by the formation of up to 64 % percentage of fast-growing and lightweight chalky material.The subtropical climate with warm winter temperatures above c.10°C and the nutrient-rich setting in an estuary will additionally have supported the excessive growth.

Figure 1 .
Figure 1.Part of the digital surface model of the shell bed; the white cross indicates the area, within which all objects were digitally outlined and evaluated.It contains 1121 complete shells of C. gryphoides and 7047 fragments of that species (A).outcrop pictures of the shell bed illustrating the density and extent of the shell bed (B-C), B: width c. 2 m; C: see workers as scale).

Figure 2 .
Figure 2. Geographic position of the Stetten site within the Korneuburg Basin north of Vienna in Austria; shaded areas represent pre-Miocene basement; dashed lines are major faults (modified from Dellmour and Harzhauser, 2012).

Figure 3 .
Figure 3. Examples of the data acquisition: orthophoto and digital surface model (DSM) are used to define shell outlines manually.Together with various attributes, such as degree of fragmentation and taxon ID, these data are georeferenced in an ArcGIS database.Yellow lines in the DSM are examples of center lines.

Figure 4 .
Figure 4. Models of 6 shells based on high resolution laser scanning data of shells from the collections of the Natural History Museum.These specimens document the broad range of morphologies and were used for volume calculations; L = left shell, R = right shell.

Figure 5 .
Figure 5. Size frequency diagrams for center line length and area data (log transformed) with cohorts (dashed lines) as detected by mixture analysis (A-B).Box-plot illustrating the strongly right-skewed distribution for fragments (n=7047) and a clear separation from the size distribution pattern of complete shells (n=1121) (C).

Figure 6 .
Figure 6.Regression analysis revealing a significant correlation between length and area of complete shells.

Figure 7 .
Figure 7. Longitudinal section through C. gryphoides from the oyster reef site showing the high amount of chalky layers (grey) and the low amount of foliate layers (black).

Figure 8 .
Figure 8. Logistic function showing the relation between center line length and shell volume based on empirical measurements of 9 shells (dots); L = left shell, R = right shell; shell pairs are linked; numbers correspond to specimens illustrated in Fig. 4.

Figure 9 .
Figure 9. Age frequency data and box-plot of the shells based on center length data transformed with the von Bertalanffy growth equation.Four cohorts are detected by mixture analysis (A).Combined effect of natural mortality and shell loss based on an exponential decay equation derived from the amplitudes of detected cohorts (dashed lines represent 95% confidence intervals) (B).

Figure 10 .
Figure 10.Comparison of von Bertalanffy growth model of Crassostrea gryphoides with selected fossil and extant Crassostrea species.1: C. titan (Conrad, 1853), Miocene, . The Asian-Pacific group (e.g.: C. gigas, C. plicatula, C. ariakensis) and Atlantic group (e.g.: C. virginica, C. They are created as thematic layer in an ArcGIS environment.The polygon is defined by features such as points (i.e.vertices connected with lines).Each polygon is a 2D visual representation of the manually digitized specimen from the adequate orthophoto and its corresponding digital surface model.Further, manually digitized data are organized into a table.This tabular structure has its elements, i.e. numerical and descriptive attributes.For instance, numerical attributes are ID, length, orientation, etc.The outline data are composed of about 1000 virtual points (nodes) on average per object and are also stored in the georeferenced ArcGIS database.These allow an automatic calculation of the surface area of each object by using The Calculate Geometry tool.
In total, 10,284 objects were defined.Of these, 8169 objects were identified as Crassostrea gryphoides of which 1121 are complete shells (see supplementary table); 86% of the specimens display various degrees of fragmentation and are excluded from the size-frequency analysis.Four categories of fragmentation were used: complete shells are fully preserved or display only minor damage, which might have occurred already during the life of the animal (n = 1121).The category low fragmentation comprises shells in which not more than ¼ of the assumed length is missing (n = 951).Moderate fragmentation is defined by representing at least ½ of the original shell lengths (n = 1638).The category high fragmentation comprises 4458 specimens of strongly damaged shells representing less than ¼ of the complete shell.