Taxonomic Composition
The PCoA shows that the majority of taxonomic variation can be explained by one axis (68.8% for field data, 52.0% for the full OBIS dataset: Fig. 3). Latitudinally, this axis represents a broad northern and southern cluster with a transition in between (Fig. 3). This biogeographic pattern is present regardless of data subsetting, but clustering is more apparent in the field data than in any OBIS dataset. This is consistent with results based on family level data (Fig. S3). When comparing family-level changes in the field data, northern sites have higher proportions of Cardiidae, Psammobiidae, Tellinidae, and Veneridae (Fig. 4). Southern sites are represented by higher proportions of Carditidae and of smaller families (Fig. 4).
When biogeographic variables were compared to environmental variables in a series of spatial autoregressions, temperature was consistently a significant predictor of biogeographic structure across all datasets (Table S3).
Discussion
A strong latitudinal diversity pattern can be seen for marine bivalves along the eastern coastline of Australia. This pattern is consistent regardless of which diversity estimators are used and whether field sites are subdivided along a midpoint.
Strong latitudinal gradients were found regardless of the diversity metric used. Depending on the method used, however, the intensity of the latitudinal gradient changed significantly. Estimated richness values (S2/m) were the most compelling in the field data, which showed weaker gradients using other metrics. OBIS data showed a consistent gradient using all methods but failed to show consistency when the dataset was reduced to measure richness at a local scale, presumably due to insufficient sample sizes.
Including observational data in the OBIS datasets resulted in the largest shift in gradient strength. Observational records made up less than 20% of the data, but their inclusion massively reduced the latitudinal diversity signal across all the analytical treatments (Fig. 2, Table 1). Previous studies have shown large discrepancies in sampling effort across space (Brown et al. , 2000, Pressey, 2004), with a higher degree of observations often resulting in apparent undersampling (Geldmann et al. , 2016). Observational values are typically included in global studies, where they are often subject to screening based on species ecology (Gagné et al. , 2020), but as shown here, they may present issues at the regional scale. Compiling citizen science information presents a similar challenge to that of using composite datasets: the sampling method often highly varies (Pocock et al. , 2017) and very few schemes operate at scales large enough to measure latitudinal patterns – especially for invertebrate groups.
Despite inconsistencies in sampling effort and methods present in large datasets such as OBIS, they do suggest latitudinal diversity patterns in marine bivalves that are broadly consistent with those demonstrated by field collections arrayed at a regional scale. This fact indicates that OBIS data, previously used in many global studies for marine diversity and biogeographic patterns (Chaudhary et al. , 2016, Costelloet al. , 2017, Miller et al. , 2018, Menegotto and Rangel, 2018, Gagné et al. , 2020), are at least minimally suitable for studying diversity dynamics at regional scales – even in underrepresented groups such as bivalved molluscs. Additionally, temperature was found to be the main predictor of taxonomic diversity in the field data, with spatial autoregression explaining 67% of the variation, consistent with previous studies (Barneche et al. , 2019, Saeedi et al. , 2019, Gagné et al. , 2020). Abiotic variables could not predict diversity patterns to the same extent in the OBIS data: the variation explained in models based on those data was half as great. This is likely due to the smaller individual sample sizes in cells for the OBIS dataset. Our field dataset may be a better reflection of true latitudinal diversity patterns because our sampling effort was uniform and intense.
Neither the OBIS data nor the field data are consistent with latitudinal patterns for bivalves seen in other continents (Roy et al. , 1994) or in global reconstructions (Chaudhary et al. , 2016), with a much smoother gradient and no stepwise change that matches a provincial boundary. On the other hand, the biogeographic gradient seen in the field and OBIS data is broadly consistent with published regional biogeographic schemes (Ebach et al. , 2013, Wilson and Allen, 1971), including two provinces, and is similar to Australian provincial patterns shown in global schemes (Costello et al. , 2017). The biogeographic interpretation in our data is that bivalves do not form clear clusters along the Australian coastline, but a long transition that spans a biogeographic boundary.
The biogeographic transition is robust at both species and family level (Fig. 3, Fig. S3), with northern and southern provinces having distinct proportional composition (Fig. 4) despite most families being present at every field site. Ebach et al. (2013) list the upper limit of the Peronian (a province containing NSW and Victoria) as -32.7º, which falls at the start of the transition zone seen here in both datasets. However, little attention has been drawn to the presence of the transitional gradient itself, and both global and local assessments tend to agree on a two-cluster scheme (Ebach et al. 2013, Costello et al.2017).
Historically, transitions have either been recognised as overlapping biotic zones or as mixing zones (Hermogenes De Mendonça and Ebach, 2020). Here, a gradient between two tight clusters can be seen across datasets, with temperature being able to explain it in most cases (Table S3). A change in beach geomorphology along the transition zone (Shortet al. , 2000, Short et al. , 2007) may be a contributing factor, but further research is needed to fully determine the histories of the zones in order to assess the relationship.
Here we show that latitudinal gradients seen in data downloaded from OBIS match those shown in field data – confirming their comparability in diversity studies. OBIS data are less useful, however, for recreating local patterns, where the presence and strength of the gradient are largely dependent on the choice of diversity measure. Adding observational data weakens or removes any clear latitudinal signal, which is likely to be of concern when data sets are largely made up of such information. At the same time, latitudinal and biogeographic patterns were uncovered here using a relatively small number of field sites. Thus, we suggest that regional diversity patterns can be quantified easily using well-spaced, high-intensity sampling to supplement existing databases.
Acknowledgments
This study was funded by an international Macquarie University Research Excellence Scholarship (iMQRES; number 2016343 to MRK). The authors would like to thank Laura Aranda Fernández, Kelly-Anne Lawler, Jim McLean, Joshua Nito, Panayiotis Panaretos, Kathleen Perry, and Aaron Phillips for field assistance. Bonnie Ngata and Sofia Zvolanek provided lab assistance. Fieldwork in national parks and protected areas was carried out in Queensland under a Marine Park Permit (MPP19-002001) and in New South Wales under a Section 37 Collection Permit (P18/0013-1.0).