GSA Connects 2022 meeting in Denver, Colorado

Paper No. 104-13
Presentation Time: 4:45 PM

SPATIAL STANDARDIZATION TOOLS FOR FAIR COMPARISONS OF BIODIVERSITY ACROSS TIME, CLADES, AND ENVIRONMENTS


ANTELL, Gwen1, BENSON, Roger B.J.2 and SAUPE, Erin2, (1)Department of Earth and Planetary Sciences, University of California, Riverside, Geology 1242, 900 University Ave., Riverside, CA 92521; Department of Earth Sciences, University of Oxford, South Parks Road, Oxford, OX1 3AN, United Kingdom, (2)Department of Earth Sciences, University of Oxford, South Parks Road, Oxford, OX1 3AN, United Kingdom

The published fossil record is unevenly distributed over space and time. Variation arises not only from heterogeneity in original diversity, but also from differential taphonomy, sedimentation, outcrop exposure, collection effort, and research attention. The degree of geographic coverage of fossil occurrences affects estimates of diversity and biogeography parameters in deep time; hence, spatially-uncorrected results may be uninformative at best and confidently wrong at worst. For instance, observed richness correlates with geographic extent of sampling (the species–area effect). Thus, spatially-uncorrected diversity estimates may be larger for an interval with more area or more dispersed area of sampling compared to one with restricted sampling, even if true local and global richness was larger in the undersampled interval. Richness rarefaction methods, while adept at controlling sampling intensity, cannot account for richness differences from taxon turnover across a variable number and dispersion of sites. Therefore, we advocate that analyses should control geographic coverage, irrespective of richness rarefaction, and that spatial standardization should be incorporated from the start of a project workflow, rather than as a refinement after initial exploration.

There are many viable ways to control for unequal spatial coverage across a dataset. One versatile solution is to spatially subsample the data a priori, then estimate parameters and fit models on those subsamples. Several published methods iteratively cut samples from a dataset by first restricting occurrences to a bounding extent, then optionally selecting an equal number of sites within that regional scope. The bounds may have constant shape, diameter, or latitudinal breadth.

We present the R package “divvy,” which implements three customizable methods of spatial subsampling in addition to related analysis functionality. “divvy” operates on any dataset with taxonomic spatial occurrences, such as downloads from the Paleobiology Database, NEOTOMA, or GBIF. We review theoretical background, use cases, and examples of integrating spatial standardization into analytical workflows using the “divvy” package.