Inferring biome associations of recent mammals from samples of temperate and tropical faunas: Paleoecological considerations

Abstract

Reconstruction of past faunal assemblages is of primary concern to paleontologists because it yields important information concerning patterns of community structure, biogeography, and paleoclimatology. Unfortunately, the process of fossilization is serendipitous at best, resulting in fossil beds that only contain a sample of the extant fauna that existed at the time of fossilization. Herein, we examine the degree to which incomplete sampling of four well‐known small mammal faunas gives rise to uncertainty concerning the biome association of each site using neontological data from two tropical (Brazilian Cerrado and Gallery Forest) and two temperate (Kansas Prairie and Kentucky Mixed Deciduous‐coniferous Forest) areas. Two simulation approaches (variable and fixed sample sizes) were applied to each set of data to ascertain what proportion of the mammal fauna must be sampled before inferring the correct biome to which the fauna pertained. In the variable sample size approach, species were selected at random from the species pool until the sample correctly identified the biome from which the sample was obtained; this was repeated 200 times for each of the 4 sites. In the fixed sample size approach, a series of simulations were conducted in which each simulation had a fixed number of species, beginning at 2 and increasing at regular intervals until the entire fauna was included in the sample. The number of species required to correctly associate a site with the proper biome was highly variable for three of the four sites. As a result, the number of species required in a sample to be 95% certain of the biome association of Mixed Deciduous‐coniferous Forest, Prairie, and Gallery Forest sites represented a large proportion of the fauna in each case (78%, 88%, and 89%, respectively). In contrast, only 23% of the fauna needed to be sampled from the Cerrado site to obtain the same level of confidence. Adjusting the Cerrado analysis for the effect of endemic species only resulted in an inflation of this value to 33%. The effects of endemics in reducing the certainty of biome associations are only pronounced when sample sizes are small compared to the total species pool. The accuracy of estimating the correct biome association of a fauna is predicated upon the proportion of the fauna in the sample as well as the particular distributional patterns of the species that compose the fauna. Caution should be exercised when inferring the biome association of fossil sites because neither of these features of the sample are usually known. Moreover, even when these features are know, as in the neontological analyses, a large proportion of the fauna frequently is required to be confident of the correct biome association of the fauna. Results indicate that the body size of species found in a fossil assemblage or their ordinal taxonomic affinities can affect the probability of accurately deducing paleoenvironments. Smaller species, and members of the orders Rodentia or Insectivora, are more stenotopic than larger species, or members of the orders Artiodactyla and Carnivora. Fossil assemblages containing only members of the latter groups will consist of wide‐ranging species of broad habitat tolerance. Thus the paleoenvironment of such a fossil assemblage will be difficult to discern.