Biodiversity patterns of epipelagic copepods in the South Pacific Ocean: Strengths and limitations of current data bases

The OBIS data base does not include the observed number of species in each of the surface cur- rent systems (see S1 Table) when compared to total species inventories. For example, the maxi- mum species richness available in OBIS for the East Australian Current (EAC) and the Humboldt Current System (HCS) is 148 and 91, respectively (see Table 1); that account for ca. a quarter of the species inventory compiled by the Banyuls Observatory data base on copepods for those areas [128]. In the other current systems, similar levels of differences between OBIS data and species inventories can be applied. Current data limitations over space and time Sampling biases must be considered when validating the robustness of the derived spatial pat- terns and the role of environmental drivers. For example, it is crucial to consider that due to spatial gaps of data (Fig 1, Table 1), our results may be strongly constrained be sampling cover- age [129, 130]. The spatial sampling effort was evaluated over the SPO through rarefaction/extrapolation curves by grouping the cells into the main surface current systems of the SPO (Fig 5A). It was found that species richness counts did not reach a clear asymptote at any of the defined current systems. Thus, the SPO is under-sampled for copepods. This aspect may have had an influence on Shannon-Wiener index values (Fig 3), as moderate to high values (from 1.7 on) were esti- mated only for a small area of the SPO. Subsequently, large areas exhibiting much lower values

of richness and biodiversity may result from lack of sampling, description, and digitization of copepod data rather than having a relatively low diversity of species on a significant part of its extension. Moreover, the GDM-based spatial analysis done for evaluating species composition was performed with ecological data based on occurrence instead of abundance, which requires the assumption that any location where the species has not been observed can be treated as absence; therefore, making under-sampled locations more problematic [72]. Although inverte- brates are the most common group documented after plants [131], and that an increasing number of data papers over time reflects progress in digitization and online platforms develop- ment for reporting observations that serve for biodiversity research [131, 132], it has been esti- mated that less than 10% of specimens’ registers are digitized [133], with even less data available online [134]. The lack of description and cataloguing of species and the incomplete knowledge regarding their geographic distribution (i.e., Linnean and Wallacean shortfalls, respectively) are among the most common setbacks for biodiversity research using databases [130], representing a practical limit for biodiversity knowledge that may lead to misidentifica- tion of ecological processes [130]. We further addressed this issue by estimating redundancy index, that indicated a middling sampling over the SPO (S6 Fig) with a lower coverage on its eastern side, that highlights the gaps on great extensions and depicts higher values for areas

where a greater sampling effort was seen. Moreover, through Moran’s I index, it is possible to infer that the distribution of richness is more spatially clustered than expected if underly- ing spatial processes were random, as biodiversity aspects of species and communities are linked to complex interactions between physical, chemical, and biological factors. Consider- ing this, and the developments in statistical modeling making possible to account for biases within heterogeneous data [135], we chose to use the RAC approximation [67] for the GAM analyses. This allows the incorporation of another variable into the models to take spatial autocorrelation into account and maintain strong predictive and inferential performance [67]. However, despite the usefulness of these methodologies, more effort is needed for improving the number and quality of distribution records used as primary data in macroeco- logical studies [136], with databasing focused on georeferencing the information obtained from data collections [129]. PLOS ONE Biodiversity of copepods in the South Pacific PLOS ONE | https://doi.org/10.1371/journal.pone.0306440 July 11, 2024 19 / 28