Taxonomic Sufficiency in a Live-Dead Agreement Study in a Tropical Setting

Palaeogeography, Palaeoclimatology, Palaeoecology 449 (2016) 341–348

Contents lists available at ScienceDirect

Palaeogeography, Palaeoclimatology, Palaeoecology

journal homepage: www.elsevier.com/locate/palaeo

Taxonomic sufﬁciency in a live–dead agreement study in a tropical setting

Paolo G. Albano a,⁎, Adam Tomašových b, Michael Stachowitsch c,MartinZuschina a Department of Paleontology, University of Vienna, Althanstrasse 14, A-1090 Vienna, Austria b Earth Science Institute, Slovak Academy of Sciences, Dúbravská cesta 9, 84005 Bratislava, Slovak Republic c Department of Limnology and Bio-Oceanography, Center of Ecology, University of Vienna, Althanstrasse 14, A-1090 Vienna, Austria article info abstract

Article history: Community studies in paleontological research often rely on identification at taxonomic levels higher than Received 9 October 2015 species (mainly genus or family). Several studies have been conducted on paleo- and modern communities to Received in revised form 1 February 2016 identify the higher level of taxonomic identification that still depicts ecological patterns: genus- and, to a lesser Accepted 14 February 2016 extent, family-level identification are frequently sufficient. The use of higher taxonomic levels has not yet been Available online 22 February 2016 explored in the context of studies comparing living and death assemblages (so-called “live–dead agreement studies”), notwithstanding their interest to quantify the fidelity of the fossil record and for environmental assess- Keywords: Taxonomic sufficiency ment. We conducted such exploration in a highly diverse tropical marine setting, targeting shelled molluscs. Our Paleoecology results suggest that the common practice of genus-level identification of paleontological samples allows for a Live–dead agreement proper reconstruction of the original biological community (e.g., variation in richness, ecological or taxonomic Tropical marine molluscs similarity) at the species level because (1) fidelity at the species and genus levels is very similar and (2) genera are sufficient to characterize between-habitat differences in composition and diversity. Live–dead agreement becomes even better at family and higher taxonomic levels for some metrics, but between-habitat differences in composition become weaker above the family level. However, at the genus and family levels, between-habitat differences are equally strong as at the species level. Genus-level identification may provide more robust results when one of the assemblages is dominated by a single species, because differences in abundance can be compensated by co-generic species. Moreover, in death and fossil assemblages, diagnostic characters get lost easily because of taphonomic processes such as abrasion, dissolution, and fragmentation; nonetheless, genus- and family-level identifications can still be reliable because the most conspicuous diagnostic characters which characterize higher taxa are more persistent after the death of the mollusc. © 2016 Elsevier B.V. All rights reserved.

1. Introduction family-level identification produced results similar to genus-level identification in only approximately 50% of cases (Forcino et al., 2012). Community ecology studies in paleontological research often rely on The fidelity of ecological information that can be preserved in fossil identification at taxonomic levels higher than species (mainly genus or assemblages can be quantified by comparing living (LAs) and death as- family) because (1) identification to generic level tends to be more ro- semblages (DAs), i.e. so-called “live–dead (LD) agreement studies”.This bust when preservation of species-level morphological characters is approach assumes that DAs are reasonable proxies for fully buried fossil variable or not optimal, or (2) genera are geologically long-lived assemblages (Kidwell, 2013). The use of supraspecific taxa has not been and thus allow tracing community types through time (even when spe- explored so far in this context. How does ecological or taxonomic simi- cies identity changes) (e.g., Patzkowsky, 1995; Holland et al., 2001; larity or variation in richness of living assemblages transfer to death as- Olszewski & Patzkowsky, 2001; Scarponi & Kowalewski, 2004; semblages at higher taxonomic levels? And, conversely, if the analysis of Redman et al., 2007). This approach has been tested: the differences in death and fossil assemblages is conducted at supraspecific levels, is it patterns between the species and the genus levels tend to have weak possible to assess the extent to which the results faithfully represent effects on paleoecological interpretations (Pandolfi, 2001), while the original species-level biological community? Furthermore, recent evidence suggests that average LD fidelity is significantly poorer in areas of human impact (Kidwell, 2007), opening interesting perspectives for the application of LD agreement studies for ⁎ Corresponding author. Tel.: +43 1 427753562. environmental assessment and identification of ecological baselines E-mail addresses: [email protected] (P.G. Albano), [email protected] š (A. Tomašových), [email protected] (M. Stachowitsch), (Kidwell, 2009; Kidwell & Toma ových, 2013; Weber & Zuschin, 2012; [email protected] (M. Zuschin). Korpanty & Kelley, 2014; Leshno et al., 2015; Zuschin & Ebner, 2015;

Albano et al., 2015b; Negri et al., 2015). In parallel with studies that 2. Materials and methods addressed the comparability between species and higher taxa in fossil assemblages, multiple ecological studies tested whether higher taxa 2.1. Dataset can detect pollution gradients in living assemblages. Ellis (1985) formalized the concept of “taxonomic sufficiency”: “in any project, or- Samples containing LAs and DAs were collected in 1999 off the ganisms must be identified to a level (species, genera, family, etc.) United Arab Emirates, around two oil infrastructures in the Umm Al which balances the need to indicate the biology (including for example Dalkh field (UA, 25 km north-west of Abu Dhabi) and Zakum field such matters as diversity) of organisms present with accuracy in (ZK, 84 km north-west of Abu Dhabi) (Fig. 1). Sediments showed very making the identifications”. Numerous studies reported that higher tax- low contamination levels in both fields, and the LAs showed little distur- onomic levels tend to be sufficient to detect natural spatial patterns bance (Albano et al., 2015a). Moreover, a comparison between LAs and (e.g., Dethier & Schoch, 2006; Terlizzi et al., 2008; Bevilacqua et al., DAs showed no evidence of community shifts (Albano et al., 2015b). 2009) and disturbance to ecosystems due to anthropogenic pressures Therefore, these samples can be regarded as coming from undisturbed in marine ecosystems (e.g., Warwick, 1988a, 1988b; Ferraro & Cole, environments. The seafloor at UA is a mixture of sand and shells, lies be- 1990, 1992, 1995; Vanderklift et al., 1996; Olsgard et al., 1998; Terlizzi tween 17 and 22 m depth, and is mostly inhabited by a small, sand- et al., 2008). Similar studies assessed sufficiency in freshwater dwelling fauna. The seafloor at ZK consists of the so-called cap rock (e.g., Wright et al., 1995; Heino & Soininen, 2007; Jiang et al., 2013; topped by a thin (ca. 1–5 cm) layer of sand, lying between 6 and 17 m Müller et al., 2013) and terrestrial habitats (e.g., Groc et al., 2010; depth. The fauna is characterized by less frequent sand-dwelling species Rosser & Eggleton, 2012; Timms et al., 2013). In most cases, family- and more frequent large-sized organisms, including scattered coral level identification was sufficient, although results may vary among colonies. phyla (Heino & Soininen, 2007, Bevilacqua et al., 2009). At each location, a grid of sampling stations was designed over an A high correlation is expected between species-level and area of ca. 45 km2 in UA (13 stations) and ca. 240 km2 in ZK (16 sta- supraspecific taxon indices or ordinations in datasets with low species/ tions). Sampling for sediments and benthic organisms was carried out higher taxon (S/T) ratios. Therefore, datasets, or taxa, dominated by by divers with large aluminum scoops used in earlier benthic surveys monotypic higher taxa are likely to produce a high level of taxonomic in the area (Coles & McCain, 1990). The size fraction analyzed sufficiency (e.g., Olszewski & Patzkowsky, 2001). However, these ratios here was sieved with minimum 2 mm mesh size and maximum will tend to increase with the spatial and temporal extent of datasets 5 mm mesh size, and then sorted under dissecting microscopes. The and will be especially high in the tropics, where rates of evolutionary minimum mesh size helped avoid the collection of juvenile speci- diversification are high (Krug et al., 2008). The ability of higher taxa to mens which would have proved daunting to segregate and identify capture compositional and environmental gradients will also depend in this diverse and taxonomically poorly known tropical area. The on whether species belonging to higher taxa respond to gradients simi- maximum mesh size helped remove large but patchily distributed larly (ecological conservatism) and on their taphonomic similarity. individuals which would be undersampled in the DAs due to the lim- We test the hypothesis that within-phylum live–dead agreement ited sampled area. Moreover, the species in this size range are ex- follows the general patterns already described for paleocommunities pected to have more comparable “taphonomic half-lives” (Kidwell, and modern living assemblages, i.e., there is a good correlation in com- 2002) and therefore reduce the effect of differential destruction munity patterns based on species-level and lower supraspecific taxa. rates among species. Mismatch between LAs and DAs was assessed by measuring similarity All living molluscs were extracted from samples, while the extrac- in taxonomic composition, rank–order agreement in relative abundance tion of empty shells from DAs was carried out until at least 1000 skeletal of taxa, taxonomic richness, evenness, and by conducting multivariate elements were collected at each station. In total, the study is based on analysis. Our case study is a highly diverse subtidal molluscan assem- 523 living individuals and 50,000 skeletal elements. To avoid counting blage in the Persian (Arabian) Gulf. This dataset is characterized by a skeletal element (whole shells for gastropods and scaphopods, valves high species richness (ca. 350 species) and by one bivalve species that for bivalves and polyplacophorans) twice, only fragments larger than dominated the DA but was absent from the LA, causing a lower-than- half a whole skeletal element were considered. Skeletal element counts expected fidelity. were divided by the number of skeletal elements of each taxon

Fig. 1. The study area showing the two offshore oilﬁelds (Umm al Dalkh and Zakum) in which the platform structures were sampled. P.G. Albano et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 449 (2016) 341–348 343

Table 1 Number of taxa (S) and ratio between the number of species and of supra-speciﬁc taxa (S/T) in Umm Al Dalkh (UA) and Zakum (ZK) ﬁelds, for living (LA) and death (DA) assemblages (raw values (obs) and mean values of resampled datasets to standardize sample size (stan)).

UA ZK

LA DA LA DA

Sobs S/Tobs Sobs Sstan S/Tobs S/Tstan Sobs S/Tobs Sobs Sstan S/Tobs S/Tstan Species 42 – 253 43 –– 38 – 160 38 –– Genus 38 1.11 171 34 1.48 1.08 35 1.09 124 32 1.29 1.09 Family 23 1.83 72 24 3.51 1.63 26 1.46 57 25 2.81 1.61 Order 12 3.5 23 13 11 2.83 13 2.92 21 10 7.62 3.32 Class 3 14 4 2 63.25 14 3 12.67 4 2 40 13.81

(following Kowalewski et al., 2003) to reconcile the number of skeletal as replicates in multivariate analyses. To account for sample-size differ- elements with the number of individuals in the LA. ences between DAs and LAs, each DA was resampled without replace- Samples were segregated into morphospecies. The insufficient taxo- ment to the sample size of the corresponding LA with 1000 iterations nomic knowledge of the benthic fauna of tropical seas often hindered before computing univariate indices and conducting multivariate anal- full identification. The morphospecies approach overcomes this taxo- yses. In the latter, the smaller sample size of LAs would have inevitably nomic impediment by segregating species based on the diagnostic char- increased their dispersion (Tomašových & Kidwell, 2009). acters of each family or genus, yet retaining a species-level taxonomic In our dataset, LD agreement results at the species level were heavily resolution. This approach proved to be reliable for tropical molluscs influenced by a single bivalve species, Ervilia purpurea (Lamy, 1914), (Albano et al., 2011), temperate and subtropical intertidal flats which was dominant in the DA (up to 90% at some sites) but totally ab- (Barnes & Barnes, 2012, Barnes, 2013), deep-sea marine biodiversity sent from the LA. The lack of this species in living assemblages was relat- (e.g. Grassle & Maciolek, 1992, Poore & Wilson, 1993) and tropical ter- ed to spatial and temporal volatility in its abundance rather than to restrial arthropods (e.g. Erwin, 1982; Ødegaard, 2000). Samples are de- anthropogenic disturbance or to high preservation potential (Albano posited at the Department of Palaeontology of the University of Vienna. et al., 2015b). Therefore, we compared taxonomic sufficiency with and Taxonomy follows WoRMS (2015), with the exception of the order without Ervilia to explore the robustness of different taxonomic levels Cephalaspidea, for which a comprehensive recent revision is available to such dominance patterns. Results without Ervilia are presented only (Oskars et al., 2015). when significantly different from the complete dataset. All analyses, tests and plots were conducted in the statistical 2.2. Data analysis programing environment R, version 3.1.1 (R Development Core Team, 2009), and using the “vegan” package (Oksanen et al., 2013). To assess the extent to which genera and other supraspecifictaxaare represented by one or few species, ratios between the number of spe- 3. Results cies and higher taxa (S/T) were computed. Mismatch between LAs and DAs was assessed by measuring similarity at different taxonomic 3.1. Species richness and evenness levels in: i) taxonomic composition, based on the Jaccard index – (Jaccard, 1901), which is based solely on species presence absence, LAs had 42 and 38 species in UA and ZK, respectively. DAs were – and the Bray Curtis index (Bray & Curtis, 1957), which also con- much more diverse: 253 species were found in UA and 160 in ZK, but siders abundances; ii) rank–order agreement in relative abundance of taxa, by using the Spearman rank correlation coefficient (after re- Umm Al Dalkh Zakum moving of species not present in both assemblages); iii) taxonomic richness, by computing the difference between logarithmic (base 10) values of taxon richness rarefiedtotheLAsamplesize:Δ S= log10(dead S) − log10(live S) (Olszewski & Kidwell, 2007); and iv) evenness based on the Probability of Interspecific Encounter − −∑S 2 deltaS (PIE = [N / (N 1)] (1 i¼1 pi )) (Olszewski & Kidwell, 2007). The PIE is not biased by sample size, allowing direct comparison of live and dead evenness without the need for sample size correction (Gotelli & Graves, 1996). Live–dead differences in evenness were measured as the difference between the values of dead and live

PIE: ΔPIE = PIEDA − PIELA. Multivariate analysis was conducted with non-metric multidimensional scaling (NMDS) (Kruskal & Wish, 1978)usingBray–Curtis distances on square-root transformed proportional abundances. Differ- ences in composition between LAs and DAs and between the two ﬁelds deltaPIE (separately for LAs and DAs) were assessed by permutational multivar-

iate analysis of variance (PERMANOVA, McArdle & Anderson, 2001; −0.4 −0.2 0.0 0.2 0.4 −0.4 −0.2 0.0 0.2 0.4 −0.4 −0.2 0.0 0.2 0.4 −0.4 −0.2 0.0 0.2 0.4 Anderson, 2001). Results at different taxonomic levels were compared class class order by performing Mantel tests with 999 permutations (Mantel, 1967). order family family genus genus species Guillot & Rousset (2013) suggested that Mantel tests should not be species used when both variables are spatially autocorrelated. The Mantel-test correlations between Bray–Curtis distances and spatial Euclidean Fig. 2. Live–dead agreement in taxon richness (ΔS) and evenness (ΔPIE) at different taxonomic levels in Umm Al Dalkh (UA) and Zakum (ZK) fields; mean values and distances (separately for LAs and DAs) are low and insignificant. confidence intervals were generated after resampling 1000 times DA to LA sample size. fi Analyses were computed at the eld scale, pooling station abun- Taxon richness fidelity is similar at all taxonomic levels; the pattern of evenness is dances into field abundances, for univariate indices, and using stations similar although a major drop in fidelity is evident in UA at the order level. 344 P.G. Albano et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 449 (2016) 341–348

Umm Al Dalkh Zakum E. purpurea was retained or not in the dataset (Fig. 4). Significance values are, however, above 0.05, with the single exception of the order level in UA. The species and genus levels showed very similar but small (~0) values in the Zakum field. The NMDS (Fig. 5) shows that LAs and DAs were well separated in two-dimensional space at the species level. At the genus level, segrega- Jaccard tion was also good. Differences between LAs and DAs within each field

LA ZK LA UA Stress=0.14 DA ZK DA UA Species Bray-Curtis −2 −1 0 1 2 −3 −2 −1 0 1 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 Stress=0.17 class class order order family family genus genus species species

Fig. 3. Jaccard (above) and Bray–Curtis (below) similarity indices between LAs and DAs at the field scale at different taxonomic levels in Umm Al Dalkh (left) and Zakum (right) fields; mean values and confidence intervals were generated after resampling 1000 Genus times DA to LA sample size. Taxonomic similarity was low at the species to the order level, and peaked at the class level because it contains few taxa. −2 −1 0 1 2 −2 −1 0 1 2 became equally diverse when size-standardized (43 species in UA and 38 in ZK). Although the species/genus ratio (S/G) of raw DAs Stress=0.20 (1.29–1.48) exceeds S/G of LAs (~1.1), this ratio is close to 1.1 in standardized DAs (Table 1). The species/family ratio was also similar between LAs and standardized DAs. At the order and class levels, species were more evenly distributed among taxa. Δ Δ Fidelity in species richness ( S) and evenness ( PIE) was compara- Family ble at all taxonomic levels (Fig. 2), with the single exception of a drop in fidelity at the order level in UA. As could be expected, datasets that did not contain the dominant E. purpurea in DAs had a higher fidelity in −2 −1 0 1 2 Δ − evenness ( PIE ~0 rather than 0.2). −2 −1 0 1 2

– 3.2. Live dead differences in composition Stress=0.22

Jaccard and Bray–Curtis similarity between LAs and DAs increased from the species to the order level and peaked at the class level (Fig. 3). Results with the dataset without the dominant E. purpurea show almost identical trends and very similar values (not illustrated). Order The rank–order correlation in abundance between LAs and DAs gradually increased with taxonomic level, irrespective of whether −2 −1 0 1 2 −2 −1 0 1 2 Umm Al Dalkh Zakum

Stress=0.11 Class Spearman r −0.5 0.0 0.5 1.0 −0.5 0.0 0.5 1.0 −2 −1 0 1 2 class class −2 −1 0 1 2 order order family family genus genus species species Fig. 5. Non-metric multidimensional scaling plots of LAs (empty symbols) and DAs Fig. 4. Spearman rank-order correlation between abundance in LAs and DAs at the field (solid symbols) of both Umm Al Dalkh (UA, gray) and Zakum (ZK, black). LAs and DAs scale at different taxonomic levels in Umm Al Dalkh and Zakum fields; mean values and were well separated in two-dimensional space at the species and genus levels, while confidence intervals were generated after resampling 1000 times DA to LA sample size. they started losing differences at the family level, and almost completely overlapped at A steady increase in agreement is evident, but p-values tend to be insignificant (p N 0.05). the order and class levels. P.G. Albano et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 449 (2016) 341–348 345

Umm al Dalkh Zakum p b 0.05, Fig. 9). The difference then declined progressively and classes no longer detected differences between the two fields. Results using datasets without the dominant E. purpurea were very similar (not illustrated). Although Mantel-test correlation between compositional distances among LAs and among DAs did not decline at higher taxonomic levels, between-field differences became smaller relative to F value F value p value p value within-field differences, and PERMANOVA thus becomes insignificant

23456 at higher levels. Dissimilarity matrices at the species level were very similar to those 23456

1 b 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 at the genus level (Mantel test, mean r = 0.9, p 0.05); similarity decreased at the family level (r = 0.8, p b 0.05) and further at the class class order order

family family ≤ b genus genus order and class levels (mean r 0.5, p 0.05) (Fig. 10). Datasets without species species E. purpurea showed the same decline in similarity (Fig. 10). Fig. 6. Boxplots of PERMANOVA R2 (solid circles) and F (gray triangles) values showing differences between LAs and DAs in Umm Al Dalkh (left) and Zakum (right) at multiple taxonomic levels, with corresponding p-values. Differences between LAs and DAs are 4. Discussion fully significant up to the order level. Gray line: 0.05 significance level. 4.1. Are within-habitat properties preserved similarly in LAs and DAs at further declined with decreasing taxonomic resolution, although the different taxonomic levels? differences were significant up to the order level (PERMANOVA, p b 0.05, Fig. 6). Mantel tests within each field showed similarly positive Most genus-level metrics of live–dead agreement in composition correlation (~0.4) between dissimilarity matrices of LAs and DAs up to and in compositional gradients (i.e., fidelity of among-sample similari- the order level (Fig. 7): compositional differences among LAs are cap- ties) led to similar results as the species-level metrics. Moreover, tured among corresponding DAs with decreasing taxonomic resolution. distance matrices at the genus level were very similar to those at the Results were affected by dominance patterns: datasets that retained the species level. This implies that paleoecological information obtained dominant E. purpurea had a weaker correlation (lower correlation coef- from fossilized molluscan death assemblages identified at the genus ficient and higher p-values). level enable reconstructing the original biological signature in the Dissimilarity matrices of LAs and DAs of single fields at the species same way as identification at the species level. level were very similar to ordinations at the genus level (Mantel test, Moreover, the genus and family levels seem to be more robust to ex- mean r = 0.99 and 0.88 in UA and ZK, respectively, p b 0.05); similarity treme dominance patterns such as those present in our dataset, where slightly decreased at the family level (mean r = 0.90 and 0.84 in UA and the bivalve E. purpurea made up on average 30–50% of the DAs at the ZK, respectively, p b 0.05) and was lower (r b 0.6) at the order and class field scale whereas it was completely missing from the LAs (Albano levels (Fig. 8). Datasets without E. purpurea showed a similar trend et al., 2015b). For example, LD agreement in composition was much (Fig. 8). higher at the genus than at the species level in ZK, because living individuals of the co-generic Ervilia scaliola Issel, 1869 compensated for the lack of E. purpurea in the living assemblage (PERMANOVA, Fig. 5). 3.3. Live–dead differences in compositional gradients However, this averaging out of within-genus variation in live–dead agreement in abundance at the species level can be expected only NMDS ordinating both fields shows that differences between UA when genera are not monospecific and co-generic species have similar and ZK are preserved by DAs at the genus level (Fig. 5; PERMANOVA, environmental preferences.

Umm al Dalkh Zakum with Ervilia purpurea without Ervilia purpurea with Ervilia purpurea without Ervilia purpurea Mantel r 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 p value 0.0 0.1 0.2 0.3 0.4 0.0 0.1 0.2 0.3 0.4 0.0 0.1 0.2 0.3 0.4 0.0 0.1 0.2 0.3 0.4 class class class class order order order order family family family family genus genus genus genus species species species species

Fig. 7. Boxplots of Mantel-correlation tests between Bray–Curtis distances among LAs and among DAs at different taxonomic levels in Umm Al Dalkh (left) and Zakum (right) fields. Correlation values are moderately and significantly high up to the order level, and more significant in datasets without the dominant Ervilia purpurea. Gray line: 0.05 significance level. 346 P.G. Albano et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 449 (2016) 341–348

Umm al Dalkh Zakum with Ervilia purpurea without Ervilia purpurea with Ervilia purpurea without Ervilia purpurea Mantel r 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 genus family order class genus family order class genus family order class genus family order class

Fig. 8. Boxplots of Mantel-correlation tests between species and supraspecificlevelBray–Curtis distances of LAs and DAs in Umm Al Dalkh (left) and Zakum (right) fields. Correlation is very high at the genus and family levels and is much lower at the order and class levels. Correlations are always significant (p b 0.05).

Identification at the family level also provided results similar to the represent effective taxonomic surrogates in detecting β-diversity species level for several metrics and analyses. Among the exceptions, patterns in diverse marine settings (Terlizzi et al., 2008; Bett & the rank–order correlation in abundance between LAs and DAs steadily Narayanaswamy, 2014). increased with increasing taxonomic level; this result was expected because dominance patterns are preserved when pooling abundances at higher taxonomic levels. The reduction of the number of taxa also ex- 4.3. Why do genera and families work? plains the pattern of the Bray–Curtis similarity index, which gradually increased with taxonomic level. These results imply that trends in The reason for the sufficiency of genus- and family-level identifica- rank–order of taxa abundance and in Bray–Curtis similarities in the tions is partly related to the ratio between the number of species and original biological community at the species level cannot be inferred genera/families: it was below 2 in our datasets, while species/order by a paleocommunity analysis conducted at the family (or higher) and species/class ratios were above 2.5. Therefore, our results confirm level. In contrast, some metrics (such as taxonomic richness and even- that ratios between 2 and 2.5 may be thresholds beyond which multi- ness) also gave results similar to the species level at high taxonomic variate analyses at coarser levels of resolution may not be well correlat- levels, such as order and class. ed with those performed at the species level (Timms et al., 2013). Moreover, similar species/higher taxon ratios in corresponding LAs – 4.2. Are between-habitat differences preserved similarly in LAs and DAs at and DAs can contribute to the surrogacy of higher taxa in live dead different taxonomic levels? agreement comparisons, because they imply the same degree of information loss with increasing taxonomic level. Such similar ratios also de- Time-averaged molluscan DAs do capture compositional and pend on homogeneity of shell durability within higher taxa: if a taxon environmental variations among samples of the living assemblage includes both fragile and robust taxa, then the former may be more (e.g., Tomašových & Kidwell, 2009; Albano & Sabelli, 2011; Weber & easily lost in DAs, causing lower S/T ratios than in LAs. Zuschin, 2012; Albano, 2014; Zuschin & Ebner, 2015). The two areas Closely related organisms also tend to be ecologically similar studied in our work are tens of kilometers apart. They share similar (e.g., Cavender-Bares et al., 2009). Accordingly, environmental gradi- bathymetric ranges, but have different substrates. At this spatial scale, ents (e.g., differences in composition and structure between different differences between areas detected by LAs were detected by DAs up to habitats) can be expected to be preserved at higher taxonomic levels. the order level (PERMANOVA, Fig. 6). These results are in accordance For example, invertebrate tolerance within orders is less variable than with other studies that found, for example, that family data still among orders in freshwater habitats: all species belonging to sensitive higher taxa disappeared from impacted sites (Wright et al., 1995).

Living assemblages Death assemblages 654 654321 with Ervilia purpurea without Ervilia purpurea p value p value F value F value 3 Mantel r 2 1 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 class class order order family family genus genus genusfamily order class genusfamily order class species species

Fig. 9. PERMANOVA R2 (solid circles) and F (grey triangles) values showing differences Fig. 10. Boxplots of Mantel-correlation tests between species and supraspecificlevelBray– between UA and ZK in LAs (left) and DAs (right), with associated p-values. Differences Curtis distances of LAs and DAs of Umm Al Dalkh (left) and Zakum (right) fields (see also between UA and ZK DAs are fully significant up to the family level. Grey line: 0.05 NMDS in Fig. 5). Correlation is very high at the genus and family levels and is much lower significance level. at the order and class levels. Values are always significant (p b 0.05). P.G. Albano et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 449 (2016) 341–348 347

4.4. Taxonomic sufﬁciency helps when diagnostic characters get lost References

The good sufficiency results at the genus and family levels have im- Albano, P.G., 2014. Comparison between death and living land mollusk assemblages in six forested habitats in Northern Italy. Palaios 29, 338–347. portant operational consequences. The loss of taxonomically important Albano, P.G., Sabelli, B., 2011. Comparison between death and living molluscs assem- diagnostic characters is common after the death of organisms due to blages in a Mediterranean infralittoral off-shore reef. Palaeogeogr. Palaeoclimatol. abrasion, dissolution, fragmentation and other damaging processes act- Palaeoecol. 310, 206–215. š fi Albano, P.G., Filippova, N., Steger, J., Kaufman, D.S., Toma ových, A., Stachowitsch, M., ing on skeletal remains. The identi cation at the species level may also Zuschin, M., 2015b. Oil platforms in the Persian (Arabian) Gulf: living and death as- be inaccurate or imprecise if carried out by different operators. In con- semblages reveal no effects. Cont. Shelf Res. http://dx.doi.org/10.1016/j.csr.2015.12. trast, genus-, and especially family-level, identification is certainly 007i. Albano, P.G., Filippova, N., Steger, J., Schmidbaur, H., Tomašových, A., Stachowitsch, M., more accurate, even if conducted by non-specialists. Taxonomy at the Zuschin, M., 2015a. Contamination patterns and molluscan and polychaete assem- family level is based on more relevant characters which can be more blages in two Persian (Arabian) Gulf oilfields. Mar. Ecol. http://dx.doi.org/10.1111/ easily retained in specimens subject to taphonomic damage; it is also maec.12370. more stable in time and among taxonomic schools, allowing a better Albano, P.G., Sabelli, B., Bouchet, P., 2011. The challenge of small and rare species in marine biodiversity surveys: microgastropod diversity in a complex tropical coastal comparability of results among different works. environment. Biodivers. Conserv. 20 (13), 3223–3237. Anderson, M.J., 2001. A new method for non-parametric multivariate analysis of variance. Austral Ecol. 26, 32–46. Barnes, R.S.K., 2013. Spatial stability of macrobenthic seagrass biodiversity. Mar. Ecol. 4.5. Extrapolation to other phyla Prog. Ser. 493, 127–139. Barnes, R.S.K., Barnes, M.K.S., 2012. Shore height and differentials between macrobenthic These results on molluscan assemblages are not necessarily transfer- assemblages in vegetated and unvegetated areas of an intertidal sandflat. Estuar. – fi Coast. Shelf Sci. 106, 112 120. rable to other phyla because levels of taxonomic suf ciency of different Bertrand, Y., Pleijel, F., Rouse, G.W., 2006. Taxonomic surrogacy in biodiversity assess- organisms are variable. Firstly, different ratios between species and ments, and the meaning of Linnean ranks. Syst. Biodivers. 4, 149–159. supraspecific taxa for different phyla may play a role (e.g., Heino & Bett, B.J., Narayanaswamy, B.E., 2014. Genera as proxies for species a- and b-diversity: tested across a deep-water Atlantic–Arctic boundary. Mar. Ecol. 35, 436–444. Soininen, 2007). Secondly, although families generally represent fairly Bevilacqua, S., Fraschetti, S., Musco, L., Terlizzi, A., 2009. Taxonomic sufficiency in the de- homogeneous groups of organisms sharing basic functional traits tection of natural and human-induced changes in marine assemblages: a comparison (Warwick, 1993), the level of such intra-family functional redundancy of habitats and taxonomic groups. Mar. Pollut. Bull. 58 (12), 1850–1859. fi Bray, J.R., Curtis, J.T., 1957. An ordination of upland forest communities of southern could depend on the classi cation criteria adopted and on taxonomic Wisconsin. Ecol. Monogr. 27, 325–349. traditions (Bertrand et al., 2006), which may not be consistent across Cavender-Bares, J., Kozak, K.H., Fine, P.V.A., Kembel, S.W., 2009. The merging of commu- phyla (Warwick & Somerfield, 2008). nity ecology and phylogenetic biology. Ecol. Lett. 12, 693–715. Coles, S.L., McCain, J.C., 1990. Environmental factors affecting benthic infaunal communities of the Western Arabian Gulf. Mar. Environ. Res. 29, 289–315. Dethier, M.N., Schoch, G.C., 2006. Taxonomic sufficiency in distinguishing natural spatial 5. Conclusions patterns on an estuarine shoreline. Mar. Ecol. Prog. Ser. 306, 41–49. R Development Core Team, 2009. R: A Language and Environment for Statistical Comput- ing. R Foundation for Statistical Computing, Vienna (www.R-project.org). The results of our case study suggest that the common practice of Ellis, D., 1985. Taxonomic sufficiency in pollution assessment. Mar. Pollut. Bull. 16, 459. genus-level identification of paleontological samples allows for a proper Erwin, T.L., 1982. Tropical forests: their richness in Coleoptera and other arthropod reconstruction of the original biological community at the species level, species. Coleopt. Bull. 36 (1), 74–75. fi fi Ferraro, S.P., Cole, F.A., 1990. Taxonomic level and sample size suf cient for assessing because delity at the species and genus levels is very similar. Family- pollution impacts on the Southern California Bight macrobenthos. Mar. Ecol. Prog. level identification may also be sufficient. Live–dead agreement Ser. 67, 251–262. becomes even better at family and higher taxonomic levels for some Ferraro, S.P., Cole, F.A., 1992. Taxonomic level sufficient for assessing a moderate impact on macrobenthic communities in Puget Sound, Washington, USA. Can. J. Fish. metrics, but between-habitat differences in composition become Aquat. Sci. 49, 1184–1188. weaker above the family level. At the genus and family levels, however, Ferraro, S.P., Cole, F.A., 1995. Taxonomic level sufficient for assessing pollution impacts of between-habitat differences are equally strong as at the species the Southern California Bight macrobenthos—revisited. Environ. Toxicol. Chem. 14 – (6), 1031–1040. level. The effect of some species that reduce live dead agreement Forcino,F.L.,Stafford,E.S.,Leighton,L.R.,2012.Perception of paleocommunities at differ- (i.e., overwhelmingly abundant in the DA and absent from the LA) is ent taxonomic levels: how low must you go? Palaeogeogr. Palaeoclimatol. Palaeoecol. averaged out at the genus level by abundances of co-generic species 365–366, 48–56. that occur in the LA. Although caution should be exercised when extrap- Gotelli, N.J., Graves, G.R., 1996. Null Models in Ecology. Smithsonian Institution Press, Washington D.C. olating results to other phyla, it is likely that the same results can be ob- Grassle, J.F., Maciolek, N.J., 1992. Deep-sea species richness: regional and local diversity tained when analyzing communities with expected low (b2) ratios estimates from quantitative bottom samples. Am. Nat. 139 (2), 313–341. between species and higher taxon numbers, and with similar ratios in Groc, S., Delabie, J.H.C., Longino, J.T., Orivel, J., Majer, J.D., Vasconcelos, H.L., Dejean, A., 2010. A new method based on taxonomic sufficiency to simplify studies on Neotrop- living and death assemblages. ical ant assemblages. Biol. Conserv. 143, 2832–2839. Guillot, G., Rousset, F., 2013. Dismantling the Mantel tests. Methods Ecol. Evol. 4, 336–344. Heino, J., Soininen, J., 2007. Are higher taxa adequate surrogates for species-level as- Acknowledgments semblage patterns and species richness in stream organisms? Biol. Conserv. 137, 78–89. Field work was conducted in the framework of the “Marine environ- Holland, S.M., Miller, A.I., Meyer, D.L., Dattilo, B.F., 2001. The detection and importance of ” fi subtle biofacies within a single lithofacies: the Upper Ordovician Kope Formation of mental monitoring programme of Zakum and Umm al Dalkh oil elds the Cincinnati, Ohio region. Palaios 16, 205–217. by Texplor G.m.b.H. on behalf of the Zakum Development Company Jaccard, P., 1901. Distribution de la flore alpine dans le Bassin des Drouces et dans (ZADCO). Ernst Geutebrück, Texplor, kindly consented to the re- quelques regions voisines. Bull. Soc. Vaudoise Sci. Nat. 37 (140), 241–272. Jiang, X., Xiong, J., Song, Z., Morse, J.C., Jones, F.C., Xie, Z., 2013. Is coarse taxonomy suffi- analysis of samples and the dissemination of the results. Reinhard cient for detecting macroinvertebrate patterns in floodplain lakes? Ecol. Indic. 27, Kikinger and M.S. took and initially processed the benthic samples on 48–55. board. Antonio Bonfitto, Bruno Dell'Angelo, Manfred Herrmann, Kidwell, S.M., 2002. Time-averaged molluscan death assemblages: palimpsests of – Graham P. Oliver, and Bruno Sabelli are thanked for support in species richness, snapshots of abundance. Geology 30, 803 806. Kidwell, S.M., 2007. Discordance between living and death assemblages as evidence segregation and identification. Jan Steger, Mirko D. Gennaro and for anthropogenic ecological change. Proc. Natl. Acad. Sci. U. S. A. 104 (45), Nadezhda Filippova helped with sorting and identification of samples. 17701–17706. fi P.G.A. was funded by a Marie Curie fellowship by the European Commis- Kidwell, S.M., 2009. Evaluating human modi cation of shallow marine ecosystems: mis- fi match in composition of molluscan living and time-averaged death assemblages. sion (PIEF-GA-2012-326089). A.T. thanks the Slovakian Scienti cGrant In: Dietl, G.P., Flessa, K.W. (Eds.), Conservation Paleobiology: Using the Past to Agency for support (VEGA 2/0136/15). Manage for the Future. The Paleontological Society Papers 15, pp. 113–139. 348 P.G. Albano et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 449 (2016) 341–348

Kidwell, S.M., 2013. Time-averaging and fidelity of modern death assemblages: building a Pandolfi, J.M., 2001. Numerical and taxonomic scale of analysis in paleoecological data taphonomic foundation for conservation palaeobiology. Palaeontology 56 (3), sets: examples from neo-tropical Pleistocene reef coral communities. J. Paleontol. 487–522. 75 (3), 546–563. Kidwell, S.M., Tomašových, A., 2013. Implications of time-averaged death assemblages for Patzkowsky, M.E., 1995. Gradient analysis of Middle Ordovician brachiopod biofacies: ecology and conservation biology. Annu. Rev. Ecol. Evol. Syst. 44, 539–563. biostratigraphic, biogeographic, and macroevolutionary implications. Palaios 10, Korpanty, C.A., Kelley, P.H., 2014. Molluscan live–dead agreement in anthropogenically 154–179. stressed seagrass habitats: siliciclastic versus carbonate environments. Palaeogeogr. Poore, G., Wilson, G., 1993. Marine species richness. Nature 361, 597–598. Palaeoclimatol. Palaeoecol. 410, 113–125. Redman, C.M., Leighton, L.R., Schellenberg, S.A., Gale, C.N., Nielsen, J.L., Dressler, D.L., Kowalewski, M., Carroll, M., Casazza, L., Gupta, N.S., Hannisdal, B., Hendy, A., Krause, R.A., Klinger, M.K., 2007. Influence of spatiotemporal scale on the interpretation of LaBarbera, M., Lazo, D.G., Messina, C., Puchalski, S., Rothfus, T.A., Sälgeback, J., paleocommunity structure: lateral variation in the imperial formation of California. Stempien, J., Terry, R.C., Tomašových, A., 2003. Quantitative fidelity of brachiopod– Palaios 22, 630–641. mollusk assemblages from modern subtidal environments of San Juan Islands, USA. Rosser, N., Eggleton, P., 2012. Can higher taxa be used as a surrogate for species-level data J. Taphonomy 1 (1), 43–66. in biodiversity surveys of litter/soil insects? J. Insect Conserv. 16, 87–92. Krug, A.Z., Jablonski, D., Valentine, J.W., 2008. Species–genus ratios reflect a global history Scarponi, D., Kowalewski, M., 2004. Stratigraphic paleoecology: bathymetric signatures of diversification and range expansion in marine bivalves. Proc. R. Soc. B Biol. Sci. 275 and sequence overprint of mollusk associations from upper Quaternary sequences (1639), 1117–1123. of the Po Plain, Italy. Geology 32, 989–992. Kruskal, J.B., Wish, M., 1978. Multidimensional Scaling, Sage University Paper Series on Terlizzi, A., Anderson, M.J., Bevilacqua, S., Fraschetti, S., Wlodarska-Kowalczuk, M., Quantitative Applications in the Social Sciences, 07-011. Sage Publications, Beverly Ellingsen, K.E., 2008. Beta diversity and taxonomic sufficiency: do higher-level taxa Hills and London. reflect heterogeneity in species composition? Divers. Distrib. 15 (3), 450–458. Lamy, E., 1914. Révision des Mesodesmatidae vivants du Muséum National d'Histoire Timms, L.L., Bowden, J.J., Summerville, K.S., Buddle, C.M., 2013. Does species-level resolu- Naturelle de Paris. J. de Conchyliologie 62, 1–74. tion matter? Taxonomic sufficiency in terrestrial arthropod biodiversity studies. Leshno, Y., Edelman-Furstenberg, Y., Mienis, H., Benjamini, C., 2015. Molluscan live and Insect Conserv. Divers. 6, 453–462. dead assemblages in an anthropogenically stressed shallow-shelf: Levantine margin Tomašových, A., Kidwell, S.M., 2009. Fidelity of variation in species composition and of Israel. Palaeogeogr. Palaeoclimatol. Palaeoecol. 433, 49–59. diversity partitioning by death assemblages: time-averaging transfers diversity Mantel, N., 1967. The detection of disease clustering and a generalized regression from beta to alpha levels. Paleobiology 35 (1), 94–118. approach. Cancer Res. 27, 209–220. Vanderklift, M.A., Ward, T.J., Jacoby, C.A., 1996. Effect of reducing taxonomic resolution on McArdle, B.H., Anderson, M.J., 2001. Fitting multivariate models to community data: a ordinations to detect pollution-induced gradients in macrobenthic infaunal assem- comment on distance-based redundancy analysis. Ecology 82 (1), 290–297. blages. Mar. Ecol. Prog. Ser. 136, 137–145. Müller, M., Pander, J., Geist, J., 2013. Taxonomic sufficiency in freshwater ecosystems: Warwick, R.M., 1988a. Analysis of community attributes of the macrobenthos of effects of taxonomic resolution, functional traits, and data transformation. Freshw. FrierfjordI Langesundfjord at taxonomic levels higher than species. Mar. Ecol. Prog. Sci. 32 (3), 762–778. Ser. 46, 167–170. Negri, M.P., Sanfilippo, R., Basso, D., Rosso, A., 2015. Comparison of live and dead mollus- Warwick, R.M., 1988b. The level of taxonomic discrimination required to detect pollution can assemblages suggests recent human-driven decline in benthic diversity in effects on marine benthic communities. Mar. Pollut. Bull. 19, 259–268. Phetchaburi (NW Gulf of Thailand). Cont. Shelf Res. 111, 9–30. Warwick, R.M., 1993. Environmental impact studies on marine communities: pragmatical Ødegaard, F., 2000. How many species of arthropods? Erwin's estimate revised. Biol. considerations. Aust. J. Ecol. 18, 63–80. J. Linn. Soc. 71, 583–597. Warwick, R.M., Somerfield, P.J., 2008. All animals are equal, but some animals are more Oksanen, J.F., Blanchet, G., Kindt, R., Legendre, P., Minchin, P.R., O'Hara, R.B., Simpson, G.L., equal than others. J. Exp. Mar. Biol. Ecol. 366, 184–186. Solymos, P., Stevens, M.H.H., Wagner, H., 2013. Vegan: Community Ecology Package. Weber, K., Zuschin, M., 2012. Delta-associated molluscan life and death assemblages in R package version 2.0-10. http://CRAN.R-project.org/package=vegan. the northern Adriatic Sea: Implications for paleoecology, regional diversity and Olsgard, F., Somerfield, P.J., Carr, M.R., 1998. Relationships between taxonomic conservation. Palaeogeogr. Palaeoclimatol. Palaeoecol. 370, 77–91. resolution, macrobenthic community patterns and disturbance. Mar. Ecol. Prog. WoRMS, 2015. Mollusca. (Accessed through: World Register of Marine Species at http:// Ser. 172, 25–36. www.marinespecies.org/aphia.php?p=taxdetails&id=51 on 2015-07-11). Olszewski, T.A., Kidwell, S.M., 2007. The preservational fidelity of evenness in molluscan Wright, I.A., Chessman, B.C., Fairweather, P.G., Benson, L.J., 1995. Measuring the impact of DAs. Paleobiology 33, 1–23. sewage effluent on the macroinvertebrate community of an upland stream: the effect Olszewski, T.D., Patzkowsky, M.E., 2001. Measuringrecurrenceofmarinebioticgradi- of different levels of taxonomic resolution and quantification. Aust. J. Ecol. 20, ents: a case study from the Pennsylvanian–Permian midcontinent. Palaios 16, 142–149. 444–460. Zuschin, M., Ebner, C., 2015. Compositional fidelity of death assemblages from a coral Oskars, T.R., Bouchet, P., Malaquias, M.A.E., 2015. A new phylogeny of the Cephalaspidea reef-associated tidal-flat and shallow subtidal lagoon in the Northern Red Sea. Palaios (Gastropoda: Heterobranchia) based on expanded taxon sampling and gene markers. 30, 181–191. Mol. Phylogenet. Evol. 89, 130–150.