Perception of paleocommunities at different taxonomic levels: How low must you go?

Abstract

To obtain meaningful information from paleocommunities, researchers must collect assemblages that are sufficiently complete to accurately represent the once-living community and large enough to produce statistically robust results. A key decision in community paleoecological research is the level of taxonomic identification. Here, we conducted a meta-analysis of 28 datasets from the Paleobiology Database (PBDB) to determine whether paleocommunity analyses at higher taxonomic levels produce similar results to those analyzed at the generic level. For each dataset, we composed taxon-sample matrices (series of samples containing multiple taxa of varying abundances) at the genus-, family-, order-, and class-levels. We then compared the multivariate paleocommunity results of each of the three higher taxonomic levels (family, order, and class) to the genus-level result. High goodness-of-fit statistics (using three different statistical comparison methods) resulted between comparisons of genus- and family-level taxonomic identifications for 28 paleocommunity datasets. However, 15 of the 28 genus- and family-level taxonomic identification comparisons were determined to produce different paleocommunity results based on qualitative-visual comparisons. Thus, family-level identification of specimens may often lead to the same paleocommunity conclusions as genus-level identification; however, inconsistencies generate enough uncertainty that paleocommunity research would benefit from genus-level identification of specimens. Due to the moderate-to-low goodness-of-fit statistics between genus–order and genus–class comparisons of paleocommunities as well as the clear differences found in the qualitative-visual comparisons, order and class did not reliably reproduce genus-level results. Thus, family-level identifications may be sufficient some of the time for studies employing multivariate statistical methods to compare paleocommunities that would otherwise use the genus level; order- and class-level identifications are probably never sufficient.