Drivers of beta diversity in modern and ancient reef-associated soft-bottom environments

Beta diversity, the compositional variation among communities, is often associated with environmental gradients. Other drivers of beta diversity include stochastic processes, priority effects, predation, or competitive exclusion. Temporal turnover may also explain differences in faunal composition between fossil assemblages. To assess the drivers of beta diversity in reef-associated soft-bottom environments, we investigate community patterns in a Middle to Late Triassic reef basin assemblage from the Cassian Formation in the Dolomites, Northern Italy, and compare results with a Recent reef basin assemblage from the Northern Bay of Safaga, Red Sea, Egypt. We evaluate beta diversity with regard to age, water depth, and spatial distance, and compare the results with a null model to evaluate the stochasticity of these differences. Using pairwise proportional dissimilarity, we find very high beta diversity for the Cassian Formation (0.91 ± 0.02) and slightly lower beta diversity for the Bay of Safaga (0.89 ± 0.04). Null models show that stochasticity only plays a minor role in determining faunal differences. Spatial distance is also irrelevant. Contrary to expectations, there is no tendency of beta diversity to decrease with water depth. Although water depth has frequently been found to be a key factor in determining beta diversity, we find that it is not the major driver in these reef-associated soft-bottom environments. We postulate that priority effects and the biotic structuring of the sediment may be key determinants of beta diversity.


Part II: Inferred water depth of the Cassian samples
Each of the criteria from Fürsich & Wendt (1977) was rated and the values summed to rank water depth among the samples. Positive values refer to factors predicting an origin among deeper water depths, negative values refer to a shallower setting. The following factors were analyzed semiquantitatively, the range of possible values is provided in parentheses: the ratio of suspension to deposit feeders (-2 to 2), the ratio of carnivores to grazers (-2 to 2), the proportion of articulated bivalves (-1 to 1), the abundance and diversity of gastropods (-1 to 1), the encrustation of specimens (-1 to 1), and the presence of coral, sponge, and echinoderm fragments (-1 to 1). Table S1 shows the inferred values and subsequent ranking along the reef basin.

Part III: Reduced versus complete dataset, Bay of Safaga
The Safaga dataset was reduced to the ten most abundant species for comparability with the Cassian dataset. To test whether results are robust, the analyses using the complete Safaga dataset are provided here.

Environments
Environments and depth from which the Safaga samples were taken are recorded and alpha diversity calculated (Table S2). Samples from the reef slope and from sand between coral patches show highest evenness; samples from sand between coral patches show lowest dominance. There is also no significant correlation between alpha diversity, either measured as dominance or evenness, and depth in the Safaga samples. The range of evenness is slightly higher than in the Cassian samples.

Beta diversity
Overall beta diversity of the modern Safaga dataset is lower (0.80 ± 0.03; range: 0.09-0.99; Table S3) than in the Triassic Cassian Formation. When samples taken from the same site in Safaga Bay are combined (=by-site dataset), we measure a beta diversity of 0.85 ± 0.03. Null models created for each dataset from the gamma species pool yield much lower beta diversity, with a mean of 0.18 ± 0.0001 for the by-site Safaga dataset (8 samples) and 0.20 ± 0.0001 for the by-sample dataset (13 samples) (Fig. S4).
At Safaga, there is no clear pattern of beta diversity related to depth (Fig. S6). Samples taken from the same environment, locality, and depth (only several meters apart) are very similar. Samples taken from deep, muddy settings exhibit the highest mean beta diversity (0.95 ± 0.02). Otherwise there is no relationship between sedimentary attributes and mean dissimilarity. By grouping the samples into depth ranges, we cover several environments for each range. Grouping the localities into two, three, or four depth ranges, we generally find that samples from shallower environments have a slightly more similar community composition than samples from deeper environments (Fig. S6). Dissimilarity between the four shallower and the four deeper samples is 0.72 and therefore lower than other values measured within depth ranges. A visualization of community dissimilarity using NMDS does not show strong relationships among localities from similar depths (Fig. S7).

CONCLUSIONS
Overall beta diversity of the complete dataset (0.80 by-sample, 0.85 by-site) is slightly lower than beta diversity for the dataset containing only the ten most abundant species per sample (0.82 by-sample, 0.89 by-site). This corroborates results from Roden et al. (2018), who found beta diversity estimates based on the five or ten most abundant species per sample to be statistically indistinguishable from estimates using the complete dataset. Beta diversity for most datasets -including the same dataset from the Bay of Safaga -is only slightly higher when only abundant species are included. Null models yield higher beta diversity for the complete dataset.
Since the two Safaga datasets differ in size (the complete dataset contains 23 190 specimens and 639 species, the reduced dataset contains 16 328 specimens and 59 species), the higher beta diversity in the Cassian null model (2901 specimens and 50 species) is probably due to increased randomness by sampling fewer specimens from a smaller species pool. Other results (distance decay, mean PPD vs. depth, NMDS) are very close to results from the reduced dataset. Therefore, overall results are robust.

Part IV: Results from Safaga dataset, by sample
When samples from the same site are not combined (=by-sample dataset), overall beta diversity in the Safaga dataset is lower (0.82 ± 0.04; range: 0.05-1.00) than in the by-site dataset. A null model created from the gamma species pool of the by-sample dataset yields much lower beta diversity, with a mean of 0.12 ± 0.0001 (Fig. S8). Samples taken from same site (only several meters apart) are pooled in the by-site dataset (main text); their composition is very similar (mean PPD samples 94-1-a to -d: 0.12 ± 0.01, PPD samples 94-3-a and -b: 0.16, PPD samples 94-4-a and -b: 0.06). There is a low correlation (rho: 0.25, p-value: 0.02; Fig. S9) in the by-sample Safaga dataset, but this is attributed to very high similarities between samples from the same site.     Most of the specimens from the analysed bulk samples are juveniles, fragments, or (more rarely) isolated protoconchs. Larger specimens with all ontogenetic stages preserved are rare. This represents a major handicap for the identification, especially because, by far, most of the previously described species are based on relatively large type specimens lacking the early ontogenetic shell including the protoconch. Therefore, linking juvenile specimens with the historic taxa is problematic in many cases. For instance, Zygopleuridae can, for the most part, only be identified if the protoconch is known. However, most of the species assigned to this family are based on type material that consists of teleoconch fragments. Many of such taxa will turn out to be nomina dubia. Other groups such as Domerionina and Sinuarbullina are seemingly diverse, but the differences between species are subtle and some of the species involved suffer from insufficient first descriptions and documentations. Thus, species identification is far from being trivial.
For the purpose of the paper, we made a great effort to at least be internally consistent in species identification. It seems to be crucial to illustrate at least the most abundant species so that our identifications become explicit and falsifiable. In the long run, all species must be characterized based on well preserved specimens representing all growth stages, including a re-study of the type material. However, this exceeds the scope of this study.