Post-mortem enamel surface texture alteration during taphonomic processes—do experimental approaches reflect natural phenomena?

Experimental design

Generally, the experimental designs follow from practicality and availability of equipment and specimens, but are also deliberately simplified to deal with a reduced number of variables and thus enable recognition of simple cause and effect relationships. However, it has to be noted that they represent more extreme but also simplified approximations of natural conditions, and can therefore not be immediately compared to naturally occurring post-mortem wear features. How well the simplified approach reflects naturally occurring alteration is one question addressed in this study. The experiments as presented in this study are the first ones of this extent, approaching different post-mortem alteration scenarios (e.g., fluvial transport, aeolian sediment transport and acid etching due to stomach acid), under different conditions and for several taxa with diverse dental morphologies. Several variables influence post-mortem alteration, such as grain size fraction of colliding particles during fluvial transport and make its effects highly variable (Gümrükçü & Pante, 2018). This hinders the understanding of the underlying processes and the recognition and interpretation of observed wear marks on dental surfaces. The simplified experiments with partially extreme conditions were chosen to ensure distinct post-mortem alteration on the dental surface. Furthermore, the experimental designs allow replicability and the best-possible control of the influencing variables, and therefore comparison between different experimental setups and species.

Tumbling experiment

The tumbling experiment was designed to simulate the basics of fluvial transport, i.e., contact between river sediment of different grain size fractions and teeth. To ensure a constant movement of sediment and tooth specimens, commercial hobby rotary tumblers TYP TRO 2 A (Otto Eigner e.K., Industriebedarf & Hobbyschleifmaschinen, Idar-Oberstein, Germany) were employed at University Mainz. Similar to Böhm et al. (2019), two different siliciclastic sediments, a quartz-dominated (∼60%) sand and a fine to medium grained gravel, were sieved into four different grain size fractions for tumbling (Tables 1 and S1): very fine sand (51–168 µm), fine sand (112–292 µm) and medium sand (221–513 µm), and fine to medium-grained gravel (2–8 mm). Experimental tumbling barrels were filled with a mixture of 300 g of sediment and 500 g of tap water, tumbling machines run at a constant speed of 45 rpm. The high mixing ratio of sediment and water was chosen to ensure a constant contact between the sediment and tooth and a constant suspension of sediment in the water. This, together with the constant speed of the rotary tumblers is necessary to ensure similar experimental conditions for all grain size fractions and species during the whole experimental duration. The dataset includes dental surface measurements of modern cheek teeth from three different large and small mammals (zebra (Equus sp.), roe deer (Capreolus capreolus) and African vlei rats (Otomys sp.)), one archosaur (Nile crocodile, Crocodilus niloticus) and a cartilaginous fish (sandbar shark, Carcharhinus plumbeus, Table S2). Three teeth per species and grain size fraction were used for the experiment, resulting in a total 12 teeth per species, and a total of 60 tooth specimens from 5 species (Table S2). Each tumbling experiment was performed with only one tooth per barrel to avoid tooth-tooth contact. Each tooth was tumbled for 336 h (= 2 weeks) and surface texture scans were measured on the same, pre-experimentally marked area of the tooth enamel surface both before and after the experiment. The duration of 336 h of tumbling was chosen based on previous experiments (Böhm et al., 2019), to ensure post-mortem wear feature production. No transport distance was calculated, since rotary tumblers would produce abrasion of the dental surface more rapid than natural fluvial setting, making a direct comparison between artificial and natural transport distances inaccurate (Shipman & Rose, 1988; Gümrükçü & Pante, 2018). It has previously been observed that even 10 km of fluvial transport in a tropical montane system reduced the mass of volcaniclastic pebbles to two thirds of their original mass and resulted in an exponential decline in pebble diameter (Miller et al., 2014). A continuous transport over 336 h represents a very extreme and in natural settings unlikely case of constant movement in a fluvial system and therefore the upper end of possible taphonomic alterations due to fluvial transport. This experimental duration was chosen to maximise abrasion, produce identifiable grain size specific post-mortem wear features during the experiment and to assess post-mortem surface modifications.

Enamel surface scans were obtained directly on the original dental surface, except for, one specimen of roe deer (tx right M3, Table S2), which was moulded using high-resolution silicone-Vinylpolysiloxane precision impression material Provil novo Light regular set EN ISO 4823, type 3, light (Heraeus Kulzer GmbH, Dormagen, Germany). The occlusal surface of this specimen was to oblique to be measured directly on the original dental surface. Scans on original and moulded dental surfaces are considered to have similar quality and contain the same information on wear features (Mihlbachler, Foy & Beatty, 2019). Sample preparation of mammalian teeth is described in detail in Böhm et al. (2019). Data of roe deer, zebra, and Otomys teeth tumbled in the three sand-sediment grain size fractions have originally been published in Böhm et al. (2019) and are included as comparison to non-mammalian vertebrates. Four different wear facets of left and right upper premolar and molar teeth were analysed for the large mammals (roe deer and zebra), while enamel lamella L1 and L2 were analysed for the Otomys teeth (Table S2, compare Böhm et al., 2019). For sharks, the dental surfaces of the lingual side were analysed (compare Weber et al., 2021a) and on the buccal side for the crocodile teeth. For both crocodiles and sharks, measurements were taken from positions close to the apex and towards the tooth centre. To ensure measurement of the same enamel surface areas after the tumbling experiment, mammalian teeth were marked with cutmarks on enamel bands while crocodile and shark teeth were marked with a small drill hole on the centre of the dental surface. The dental surface was aligned to the reference cutmark/drill hole and measured at the same distance as in the previous scans. Four non-overlapping scans were taken of all experimentally altered teeth (for details see Table S2).

In comparison to this experimentally altered set of dental surfaces, we also analysed cheek teeth of the fossil horse Hippotherium primigenium from the Dinotheriensande, Eppelsheim, Germany (Table 1). The Dinotheriensande are Late Miocene (∼10.5 Ma) fluviatile deposits of the ancient Rhine River (Steininger et al., 1996; Andrews & Bernor, 1999). Mammalian fossils in this locality are almost entirely restricted to a siliciclastic, sand- and gravel-dominated horizon close to the base of the sedimentary sequence (Tobien, 1983). The similarity in grain size spectrum of these deposits containing fluvially transported isolated teeth makes them suitable for comparison with our experimentally altered large mammal teeth.

Sandblasting experiment

The sandblasting experiment was designed to simulate post-mortem alteration due to aeolian sediment transport, using a sandblasting machine (basic quattro IS; Renfert, Hilzingen, Germany) to blast the mineral dust vertically onto the sample from a distance of 30 cm and with a pressure of 3.5 bar for 7–14 s. These experimental conditions were chosen for practical reasons, 3.5 bar is the lowest pressure setting on the sandblaster and a seven-second interval is the shortest human reaction time we could set for turning the sandblaster on and off. The pressure of 3.5 bar represents an approximate wind speed of 100 m/s when the particles hit the dental surface, which is about 3 to 4 times higher than gusts of wind in very strong naturally occurring dust-storms (e.g., up to 30 m/s measured in Chinese dust-storms, Gengsheng, Honglang & Wanquan, 2001). Two different natural sediments were used: Late Pleistocene airborne loess (Schwalbenberg outcrop; Remagen, Germany) as a very common sediment deposited by aeolian transport and Oligocene fluvial fine sand (Feinsand 12c; Dörentrup Quarzwerke, Germany) as a comparison for the larger, sand-sized grain size fraction (Table 1). The silt-sized loess was sieved through a 300 µm mesh and had a mean grain size of 36.87 µm (range: 6.3 to 200 µm, data from laser diffraction size analysis, Table S1), with quartz (37.3 wt.-%), calcite (15.7 wt.-%), illite and mica (13.8 wt.-%), plagioclase (9.8 wt.-%), and dolomite (5.4 wt.-%) as main mineral phases. The fine sand 12c had a mean grain size of 166.01 µm (range: 63 to 630 µm, data from laser diffraction size analysis, Table S1) and was composed of nearly pure quartz (99.5 wt.-%). As sample specimen a single fossil tooth of Mammuthus sp. from Pleistocene Rhine River deposits was cut several times parallel to the occlusal surface to produce several slices comprised of multiple enamel and dentine bands. The slices were polished using silicon carbide (SiC) of different grain size fractions down to 6.5 µm, to produce large, smooth enamel surfaces analogous to “blank pages”, where new taphonomic wear features could easily be identified. Four non-overlapping scans were measured on three different enamel bands before and after sandblasting with each of the two sediment types. The experimental duration for sandblasting with the fine sand was 7 s, while sandblasting with loess was prolonged to 14 s after no alteration was observed when initially blasting for 7 s.

For comparison to the artificially produced post-mortem features, naturally “sandblasted” (i.e., eroded by windblown sand) fossil antelope teeth from the Late Miocene Toros-Menalla fossiliferous area in Chad, approximate 500 km NE of the present-day Lake Chad, were analysed (Tables 1 and S2). The fossil-bearing siliciclastic sediments consist of clay and sandstone (Lebatard et al., 2008) with a sediment grain size comparable to that of the fine sand used in our experiment. Most present-day wind-speeds for dust-storms in the Chad are 2 to 6 m/s (Cowie, Knippertz & Marsham, 2014), and thus below our experimental conditions. All specimens were measured on the original dental material.

Acid etching experiment

The acid etching experiment was designed to simulate acid etching from stomach acid during digestion, using hydrochloric acid to achieve predator stomach specific pH-values, and warming the solution to match temperature conditions experienced in the gastro-intestinal tract of different predators (Tables 1, 2 and S2). A mixing ratio of 10 ml acid to 1 g dental material was applied, and the acid was renewed after 24 h to ensure persistence of the target low pH. The pH-value was kept constant during the whole experimental period. Teeth of Otomys sp., Crocodilus niloticus and Carcharhinus plumbeus were in-vitro exposed to acid etching after being tumbled for 336 h in the fine sand fraction (112–292 µm). The samples were chosen for practical reasons, as they were available for destructive experiments, and there were visually few signs of alteration after tumbling, which was also confirmed as we found no significant statistical alteration of the post-tumbling surface compared to the pre-tumbling surface. The measurement positions of the enamel surfaces were similar to the ones from the tumbling experiment. Otomys teeth were kept in trace-metal grade 0.1 mol/L hydrochloric acid of pH ∼1.3 for 12 h at a temperature of 45 °C to simulate the stomach conditions of a barn owl prior to the regurgitation of bone and tooth material in the form of pellets (Grémillet & Plös, 1994; Beasley et al., 2015). Crocodile teeth were kept under crocodile stomach conditions, while shark teeth were kept under simulated shark stomach conditions, to simulate stomach conditions of a carnivorous reptile and predatory fish (Table 2). These large predators frequently replace their teeth, and potentially swallow some of them. Fossil shark and crocodile teeth may therefore have passed through their respective stomachs. For both crocodile and shark teeth, we used 0.05 mol/L hydrochloric acid (pH ∼2) at 25 °C for 48 and 72 h to simulate passage times through the gastrointestinal tract of each respective taxon (Pooley & Gans, 1976; Hutton, 1987; Papastamatiou, Purkis & Holland, 2007, Table 2).

After the experiment, all teeth were thoroughly rinsed with Milli-Q water (18.2 M Ω) and dried at room temperature. Two different sets of modern Otomys teeth from the Cradle of Humankind World Heritage Site, South Africa (collected in 2017) were used as natural reference samples for comparison to those from the acid etching experiment. The first sample set (Otomys sp.) of teeth was extracted from barn owl pellets and served as a direct comparison to the African vlei rat teeth from the etching experiment (Tables 1 and S2). The second sample set of teeth (Otomys irroratus) was extracted from skulls of live-trapped animals and provided the natural baseline for non-altered, ingesta-related dental surface textures from the same geographic setting and species. For all specimens DMTA were performed on the original dental material.

Dental microwear texture analysis

All samples were cleaned prior to moulding or scanning using water to remove adherent dirt, ethanol, and acetone to remove lipids and varnishes or glue from the dental surface. Surfaces were scanned using the high-resolution disc-scanning confocal 3D-surface measuring system µsurf Custom (NanoFocus AG, Oberhausen, Germany) with a blue LED (470 nm) and high-speed progressive-scan digital camera (984  × 984 pixel). Four non overlapping square areas of 160  × 160 µm were scanned using a ×100 long distance lens (numerical aperture 0.8) with a resolution in x, y = 0.16 µm, and z = 0.06 µm on each sample. Measurements with an oblique orientation too significant for measurement were rejected (displacement range δz > 60 µm). Following Böhm et al. (2019) for the experimental settings, a rotation along x-, y- and z-axis of the teeth between individual measurements of 5–10° was accepted. We used 3D (enamel) surface texture analysis (see Schulz, Calandra & Kaiser, 2010, Schulz, Calandra & Kaiser, 2013; Schulz et al., 2013; Calandra et al., 2012) and scale-sensitive fractal analysis (SSFA) with parameters after Ungar et al. (2003) and Scott et al. (2006), implemented in MountainsMap Premium v. 7.4.8676. We used three customized templates in MountainsMap Premium v.7.4.8676 software to generate an S-L surface (noise, waviness, and form removed; for more details see Schulz, Calandra & Kaiser, 2013). A detailed description of the templates including the pre-processing procedures are given in the electronical supplements. Subsequently a set of 46 surface texture parameters were quantified on the S-L surface using the following analyses: (1) the ISO 25178, (2) motif, (3) furrow, (4) direction, (5) isotropy, and (6) ISO 12871 (flatness) analysis (Table S3). In this work, it is focussed on the expression of ante- and post-mortem wear features, such as hills and dales. Dales are shallow or deep, roundish “holes” in the dental surface (Fig. 1), with a lowest point (pit); while hills have a highest point (peak); both are defined in the software MountainsMap Premium v.7.4.8676 by watershed algorithms setting its limits along the course and ridge line including the saddle points. Ante-mortem dales are generally short, shallow, and multiple a combination of multiple dales result in an open motif (furrow-like), that is orientated in the same direction, and might be produced during the chewing process. Contrary, post-mortem open motifs are often longer, deeper, and crosscut ante-mortem wear features (Fig. 1). All scans are accessible via figshare.

Statistics

The open-source software R v.4.0.0 (R. Development Core Team, 2020) with the packages xlsx (Dragulescu, 2014), rJava (Urbanek, 2016), doBy (Højsgaard & Halekoh, 2016), R.utils (Bengtsson, 2016), factoextra (Kassambara & Mundt, 2017), plyr (Wickham, 2011), ggplot2 (Wickham, 2016), Rcpp (Eddelbuettel & Francois 2011, Eddelbuettel, 2013), rela (Chajewski, 2009) and WRS version 0.12.1 (Wilcox & Schönbrodt, 2010) were used for data processing and all statistical analyses. A median for each DMT parameter was calculated from four non-overlapping scans. From this median, a mean and standard deviation (1SD) was calculated for all experimental datasets for t₀ (before each experiment as a baseline) and t_i (after each experiment) prior to analysis (Table S4–S6). In general, we employed pair-wise comparisons for the following datasets:

1. Pairwise comparison for dependent data of pre- and post-tumbling surfaces in all four grain size fractions, five species in the tumbling experiment (Table S7).

2. Pairwise comparison for dependent data of pre- and post-sandblasted surfaces in the two grain size fractions used in the sandblasting experiment (Table S7).

3. Pairwise comparison for dependent data of pre- and post-etching surfaces in all three species in the acid etching experiment (Table S7).

4. Pairwise comparison of independent data from Hippotherium primigenium and all tumbled species (Table S8), and from trapped animals (Otomys irroratus), Otomys sp. extracted from owl pellets and experimentally altered Otomys teeth before and after acid etching (Tables S2 and S9).

For pair-wise comparison of dependent data the Wilcoxon signed rank test with continuity correction was employed (Wilcoxon, 1945). This test indicates whether the dental surfaces after the experiment differ significantly from the dental surface before the experiment. Pairwise comparison of independent data was subjected to the statistical test routine after Calandra et al. (2012) and Schulz, Calandra & Kaiser (2013a). We applied the robust heteroscedastic Welch-Yuen omnibus test (Welch, 1938; Yuen, 1974) and a heteroscedastic pairwise comparison test (“Lincon test”, analogous to Dunnett’s T3 test (Dunnett, 1980)), according to the procedure of Wilcox (Wilcox, 2003; Wilcox, 2005; Wilcox, 2012). We performed a data trimming of 15% to remove outliers and compensate for non-normality. Finally, we applied the robust heteroscedastic rank-based test after Cliff (Cliff, 1996; Schulz, Calandra & Kaiser, 2013). For principal component analyses (PCA) a set of 12 parameters was selected according to the following reasons: best separating parameters (mammals: Schulz, Calandra & Kaiser, 2013; reptiles: Winkler et al., 2019), independent parameters from different parameter categories (Table S3) and no missing values. Before applying PCAs, we use the Kaiser-Meyer-Olkin test for sampling adequacy (p-value > 0.5) and Bartlett’s test of sphericity (<0.05) to validate the application of a principal component analysis (PCA) is justified on our data. PCAs were conducted using the built-in function prcomp(). Principal components (PCs) with standard deviation over 1.0 are discussed in the following. Factor loadings were cut at a value of 0.4 (Budaev, 2010), describing the best separating parameters in the different PCs (Table S10).

Tumbling experiment

Very fine sand (51–168 μm)

The first column of Fig. 2 shows exemplary 3D enamel surface models of experimentally altered dental surfaces before (first row) and after the experiment (second row) for each of the five used species. Experimental alteration of teeth with very fine sand led to a higher number of new wear features, such as new small and large dales, which crosscut ante-mortem wear features and show no preferred direction (larger anisotropy values of parameters, except for Otomys, Figs. S1A and S2A). The post-tumbling dental surface of Carcharhinus plumbeus shows additionally a large number of new dales (Fig. 2). This coincided with an increase in overall complexity of the surface (larger Sdr, Asfc values, except for Otomys). For the large mammal teeth (Equus and Capreolus, Table S4), an increase in height and volume parameters was found. The p-values from the Wilcoxon signed rank test indicate significant differences between the complete dental surface before and after the tumbling experiment for Capreolus capreolus, Crocodilus niloticus, Carcharhinus plumbeus and Otomys sp., while Equus sp. did not show significant surface modifications (Table S7). The PCA displays separation of dental surface before and after tumbling for all five different species along PC1–PC3 (Figs. 3A and 3SA), whereby the isotropy (IsT), complexity (Asfc), area (mea), density (Spd) and plateau size (Smr) show the highest factor loadings and therefore explain more variance in the PCA (Table S10).

Hippotherium primigenium teeth from the Dinotheriensande

Figure 1 shows photosimulations of Hippotherium primigenium enamel surface scan from the Dinotheriensande Eppelsheim, Germany, a fluvial deposit with siliciclastic, sand- and gravel-dominated dominated sediments (Tobien, 1983). The surface scans show clear differences in wear features to ingesta-related dental surfaces of modern grazing and browsing species (Fig. 1), i.e., Connochaetes taurinus, Giraffa camelopardalis, Equus grevyi (Schulz, Calandra & Kaiser, 2013). Hippotherium primigenium teeth show a large number of dales, together with a large number of long furrow-like features (Figs. 3 and S2), however, a very low number of narrow dales (scratch-like features). Dales are aligned in the same direction, crosscutting and overwriting ante-mortem wear features (Fig. 1). Generally, the dental surfaces appear to be very rough, compared to the ingesta-related dental surfaces of modern herbivorous ungulates (Fig. 1). However, Hippotherium primigenium shows only a few significant differences to Equus sp. before the experiment, while the number of significant differences increases after the experiment for the fine and medium sand fraction and the fine gravel fraction (Table S8). Capreolus capreolus differs significantly in one to three parameters before the experiment in the sand fractions, and in 20 parameters in the fine-gravel fraction. After the tumbling experiment, the number of significant differences decreases for the sand fractions slightly and increases for the fine to medium gravel fraction. Statistical comparisons between Otomys sp. and Hippotherium primigenium were only possible for the sand fractions and reveal an increase of significant differences for the very fine and medium sand and a decrease for the fine sand fraction. Crocodilus niloticus and Carcharhinus plumbeus show both a high number of significantly different DMT parameters compared to Hippotherium primigenium before the experiment, and a strong decrease in number of significant differences after the tumbling with the very fine sand and the fine gravel fraction. After tumbling with the fine and medium sand fractions, the number of significant differences is increasing for Crocodilus niloticus and decreasing (fine sand) or remains stable (medium sand) for Carcharhinus plumbeus.

Sandblasting experiment

Sandblasting the Mammuthus sp. polished surface with loess did not change the surface texture pattern; at least not after 14 s of blasting time (Figs. 4A and 5; Tables S5 and S7). The impact of the fine sand 12c however, was very prominent, and resulted in a significant difference between dental surface before and after sandblasting derived from the Wilcoxon Signed-Rank test, with numerous new large dales comparable to impact craters, formed on the surface of the section (Figs. 4B and 5, Table S7). The biplot of medf—mean density of furrows vs. metf—mean depth of furrows displayed strong separation of surface parameter values before and after the experiment with strongly increasing depth of the furrows (Fig. 5A). The biplot Asfc vs. epLsar also showed this strong separation for sandblasting with fine sand, with strongly increasing fractal complexity (Asfc, Fig. 5B). PCA revealed strong separation of dental surface before and after the sandblasting (Fig. 5C), with material volume (Vm) and depth of the furrows (metf) having the highest factor loadings.

Additionally, we compared the experimentally sandblasted mammoth dental surfaces with potential post-mortem features of naturally “sandblasted” fossil antelope teeth from the Late Miocene Toros-Menalla, Chad. Generally, the antelope teeth show no sign of ingesta-related ante-mortem wear compared to ante-mortem dental surface of other modern large ungulates (Fig. 1). The dental surface of the fossil specimens shows a rough dental surface with small to large dales, however, not as large as the enamel surface experimentally altered with the fine sand 12c. The biplots of medf vs. metf show high medf values, while metf is similar to the mammoth dental surface before and after sandblasting with loess (Fig. 5A). The PCA reveals a strong separation of antelope and the experimentally altered mammoth teeth (Fig. 5C).

Acid etching experiment

The acid etching experiment was performed with three different species, Otomys sp., Crocodilus niloticus and Carcharhinus plumbeus at different pH-values and temperatures. Overall, it is evident that the outermost layer of enamel (consisting of hydroxylapatite) and enameloid (consisting of fluorapatite) is significantly affected by etching during the experiment (Fig. 6) and thus the original dental surface is completely overwritten. The dental surface is very rough without clear signs of specific post-mortem features, such as narrow furrow-like features. The greatest alteration is visible for Carcharhinus plumbeus, where the height of the surface increases strongly (Fig. 6). The Wilcoxon Signed-Rank test indicates significant textural surface modification after the acid etching for all three species (Table S7). The biplot medf vs. metf as well as Asfc vs. epLsar displayed a strong separation in all three species with increasing density and depth of furrows, as well as increasing complexity of the dental surface after the etching (Figs. 7A and 7B). PCA revealed a strong separation (Figs. 7C and S5) with the highest factor loading in the parameters density of the peaks (Spd), number of motifs (nMotif), depth of furrows (medf) and mean area (mea) parameters (Table S10).

Additionally, we compared two sets of different natural African vlei rat tooth samples—one live-trapped and the other digested and extracted from barn owl pellets—with each other and with the experimentally altered Otomys teeth. The two natural samples are barely distinguishable from each other according to all surface texture parameters (Figs. 6D and 6E, Fig. 7, Table S10). Even though Dunnett‘s T3 test displays significant differences between the two natural samples sets as well as the experimentally altered Otomys teeth, the overlap of the parameter values is very large and therefore none of the differences were confirmed by Cliff’s method (Table S10). The biplot medf vs. metf and Asfc vs. epLsar, as well as the PCA displayed strong separation of the two natural Otomys sample sets from the experimentally altered Otomys samples after the acid etching (Fig. 7, Fig. S5).

Fluvial transport in a sediment suspension simulated via a tumbling experiment

Tumbling of mammal, reptile and fish teeth shows that mechanical alteration of dental surfaces as characterised by dental microwear texture analysis seems to be more dependent on the sediment particle size than on the type of enamel(-oid) microstructure, tooth morphology or size (Fig. 2). The dental surfaces of different species are strongly affected by sediment abrasion, especially by the fine gravel fraction. Very fine sand and gravel have the strongest alteration effects on abrasive texture patterns over the tumbling two-week time interval. The surfaces altered by the fine gravel fraction look quite similar across species (Fig. 2) and they share a similar parameter space in the biplot metf vs. medf (Fig. S2), as well as in the PCA (Fig. 3). Ante-mortem wear features are completely overwritten with post-mortem wear features, i.e., small and large dales. However, no post-mortem furrow-like features are added. The alteration effect of the very fine sand grain size fraction is characterized by the formation of numerous long narrow and deep furrows crosscutting ante-mortem wear features and each other. Additionally, the formation of post-mortem dales, connected to narrow post-mortem furrows, is detectable, however, fewer than during the experiment with the fine gravel fraction (Fig. 2). In contrast, tumbling with the fine and medium sand had a slight polishing effect resulting in only a few post-mortem scratch-like features. The effects of the tumbling experiment with these grain size fractions are mostly undetectable by statistical analyses (Table S7). In a direct comparison of the pre-tumbling with the post-tumbling surfaces, a significant separation was found in the very fine sand and the fine gravel fraction. The separation was visible for the biplots as well as in the PCA; however, still with a substantial overlap. Volume and overall height parameters seem to be least affected. This is not surprising, since parameters of these groups describe the whole surface and are therefore less affected by addition of singular wear features. Parameters from these groups are among the best for distinguishing between different feeding strategies (e.g., browser-grazer, Schulz, Calandra & Kaiser, 2013; herbivorous, carnivorous, frugivorous and molluscivorous lepidosaurs, Winkler et al., 2019) and less affected by simulated diagenetic alteration (fluvial transport, Böhm et al., 2019). However, diagenetic alteration caused by fluvial transport can be critical for dietary reconstructions using DMTA and requires careful sample selection prior to scanning to exclude taphonomical altered teeth or surface areas. Dental specimens from fossil assemblages found in river sediments with small gravel fractions might show very rough dental surfaces with a low number of dales and furrows, which look very similar between species. Specimens from a more sandy environment might express more scratch-like features or slightly abraded dental surfaces, depending on the grain size fraction. The comparison to the fossil equid teeth Hippotherium primigenium from the Dinotheriensande, which were fluvially transported to a certain extent and deposited in sand-gravel sized siliciclastic sediments (Tobien, 1983), reveals similarities between the natural and the experimental surface alteration features from the fine gravel fraction (Fig. 3D). Generally, our tumbling experiments revealed that fossil teeth transported in fluviatile settings with gravel or sand, could be altered and that ante-mortem wear features might be erased or at least modified mechanically during transport due to post-mortem alteration. Therefore, a proper training in recognising various taphonomic alterations helps to serve this purpose. We recommend that such training should include experience in recognizing various taphonomic alterations, which should ideally be calibrated to a database of surface textures characteristic of poorly preserved enamel surfaces as described in Weber et al. (2021b). Here, we provide a first set of post-mortem wear feature examples produced by artificial fluvial transport with specific grain size fractions. Similar post-mortem wear features could be expected on potentially fluvial transported fossil specimens.

The experimental setup as presented in this study has advantages as well as limitations. The advantage is the guaranteed constant transport of sediment and teeth of different species in the same time interval, and therefore for the same distance. This ensures a replicability and comparability across different species, independent of tooth weight or size (Shipman & Rose, 1988). This would not be possible for experiments in flow channels, for instance. Therefore, for a simplified, replicable, and controllable experimental design, rotary tumblers are the ideal tool (Gümrükçü & Pante, 2018). The disadvantage of the simplified tumbling experiment is that it is not able to simulate the very complex process of fluvial transport in a river system. Fluvial transport is potentially short, depending on different factors, such as size of the fossil and flow rate of the river system, e.g., smaller sediments could be transported over longer distances, in contrast to larger and heavier grains or fossils, where the flow rate must be much faster to transport the material (Harms, Southard & Walker, 1982). The alteration process could be variable and might include movement of sediment around the fossil, reworking or discontinuous draining and flooding processes, leading to not only various post-mortem alterations, but that might also affect the proportion of species in a fossil assemblage (Dodson, 1973; Russel et al., 1996; Läng et al., 2013; Pesquero, Alcala & Fernandez-Jalvo, 2013). The experimental design of this study provides the basic knowledge of the expression of ante-mortem wear features formed by different grain size fraction for a very specific and constant speed and duration of transport. The interaction between different parameters could then be simulated in experimental setups which simulate natural conditions more precisely, e.g., in flow channel experiments.

Aeolian abrasion simulated via sandblasting experiment

Loess is a typical silt-sized aeolian sediment predominantly deposited during glacial periods in many areas and could potentially have altered post-mortem teeth of the contemporaneous fossil fauna. However, despite the high air pressure and fast speed of sediment particles hitting the dental surface, short duration silt-sized (Ø 36 µm) loess blasting experiments do not lead to any visible alteration and formation of post-mortem wear features or changes of surface texture parameter values (Figs. 4 and 5). Fossil teeth from natural loess deposits are therefore likely usable for diet reconstruction employing DMTA, since natural wind speed are much lower than the one applied in this experiment. In contrast, sandblasting of dental surfaces with larger, sand-sized (Ø 166 µm) quartz grains caused major damage to and abrasion of the enamel surface together with the formation of large post-mortem dales, comparable to impact craters (Fig. 4). This leads to strongly increasing complexity and volume parameters (Fig. 5). Fossil tooth specimens that were naturally or artificially sandblasted using sand-sized sediment grain size fractions seem to be affected by mechanical surface modification and showing comparable post-mortem wear features, i.e., large dales (Fig. 4).

Our sandblasting experiment demonstrated the difficulties of such simplified experimental approaches for testing the impact of aeolian sand on dental surfaces exposed to post-mortem or post-depositional weathering by airborne mineral dust. The air pressure applied in our experiment was very high (e.g., at least 20 times higher than present-day natural wind-speeds in the Chad, Cowie, Knippertz & Marsham, 2014) and constant, leading to excessive alteration of the enamel surface by the quartz fine sand. However, it was also demonstrated, that despite the high air pressures, loess-sized grain size fractions appear not to form post-mortem wear features during natural or artificial aeolian sand transport. Due to our experimental design using only two grain size fractions, i.e., loess (Ø 36 µm) and sand-sized (Ø 166 µm) quartz, it remains unclear which particle size starts to cause the formation of dales on the dental surface, and under which wind speeds.

Additionally, besides the implication of the experimental results on erosion due to aeolian sediment transport, there are implications on the effect of sample preparation using air-abrasive techniques, i.e., sandblasting during preparation of fossils from the embedding sediment/rock. In general, various materials, such as Aluminium oxide, crushed glass, plastic, but also walnut shells or wheat starch, with different hardness, shapes and particle size are used for sample preparation (for detailed information see Graham & Allington-Jones, 2018). The commercial sandblasters used for sample preparation are comparable to the one used in the performed experiment, with similar air-pressure, however, even closer distance to the sample. Therefore, formation of post-mortem wear features, as seen in this study, especially for preparation using larger grain size fractions, seems to be plausible. Air-abrasive techniques should thus be avoided when dealing with dental surfaces for later dietary analyses using DMTA, however, if it is not possible to avoid this, special care needs to be taken to protect the dental surfaces from the blast. Specimens prepared with air-abrasive techniques, especially larger and harder abrasive powders comparable to the quartz finesand (Feinsand 12c) from this study, should be evaluated carefully for potential post-mortem alteration. Suspicious wear features, i.e., large dales, comparable to the features produced using the quartz fine sand (Feinsand 12c), should be expected and therefore avoided.

Stomach digestion alteration effects simulated via in-vitro acid etching experiments

Acid-etching experiments of mammal, reptile, and fish teeth with diluted HCl alter the original dental surface of enamel (hydroxylapatite: Otomys sp., Crocodilus niloticus) and enameloid (fluorapatite: Carcharhinus plumbeus) by dissolution processes. The ante-mortem wear features are completely overwritten during the experimental duration (Fig. 6). The complexity and the roughness of the dental surface increase strongly and the textures of the dental surfaces are significantly different before and after the etching experiment (Fig. 7, Table S7).

The in-vitro stomach acid experiment design performed with isolated teeth directly immersed only in hydrochloric acid do not account for the role of digestive enzymes or buffering effects by the carcasses and therefore resemble more closely the digestive system of crocodiles, where stomach fluids consist almost entirely of hydrochloric acid with almost no enzyme activity (Fisher, 1981). A similar experiment performed by Fernández-Jalvo et al. (2014) showed great similarities between the surfaces of bones and teeth of small mammals immersed exclusively in hydrochloric acid and bones digested by crocodiles. Additionally, they found differences between surfaces treated with hydrochloric acid immersion alone and natural digestion of avian predators as well as mammalian carnivores, which we can confirm with our comparison of experimentally altered versus natural Otomys samples from owl pellets. We find the surface textures of the experimentally altered teeth to be very different from those of small mammal teeth digested by avian predators. While the ante-mortem wear of experimentally altered Otomys samples is completely dissolved and significantly different from both, the dental surface before the experiment and the two sets of non-experimental Otomys samples, no significant differences in DMT parameters could be detected between non-digested and digested teeth, i.e., from barn owl-pellets (that were regurgitated after a digestion for at least 10–12 h (Grémillet & Plös, 1994)) and live-trapped specimens, both from the same area in South Africa. For avian predators and mammalian carnivore’s digestion is driven by the combination of stomach fluids and enzymes. Modifications in form of corrosion and decalcification on bones and teeth of small mammals can be reproduced with experimental approaches, but to a lesser degree and at a slower alteration rate than during natural digestion (Fernández-Jalvo et al., 2014). Our results further support the findings of Fernández-Jalvo et al. (2014), that hydrochloric acid erodes the outer tooth enamel layer almost completely, especially the most exposed areas, such as cusps or lamella of molars. The non-existent difference between digested and non-digested teeth of Otomys (compare Fig. 6) is in line with findings that in barn owl stomachs, only 1% of ingested molars evidence digestion-related acid-etching and supports the categorization of barn owls in the weakest predator category 1 (i.e., exhibit no or minimal acid digestion, Fernández-Jalvo et al., 2016). We conclude that small mammal teeth extracted from barn owl pellets can be used for diet reconstruction studies using DMTA, however, we did not test if this is also the case for other owl species. It is likely that, at least for birds of prey of predator category 1 (see Fernández-Jalvo et al., 2016), ingesta-related enamel surface textures of prey vertebrates can survive the weak acid attack. Further study of DMTA from small mammal teeth recovered from owl pellets of other species can hopefully confirm our initial results, thus confirming the suitability of such material for dietary reconstruction of small mammals.