Normalization of fossil plant megafossil databases for diversity and palaeobiogeography analyses by filtering taxonomic duplication: Principles, methods, examples, and recommendations

Fossil plants are key to many palaeobiogeographic and deep time diversity studies, but correctly interpreting them can be fraught with problems due to fragmentation in the fossil record. A typical vascular plant comprises 10–12 separate organs depending on its systematic affinity, but complete individuals are exceptionally rare. Fragmentation can result from multiple processes including ontogeny during the plant's life-cycle, or from post-mortem taphonomic processes in fluvial systems. In traditional approaches where raw data is amassed directly in the field, from existing physical collections or electronic databases, duplication is inevitable in that different organs of the same plant species may be inadvertently counted independently, skewing results. Here we outline normalization methods for filtering the palaeobotanical data to remove taxonomic duplications, with examples provided from different types of preservation. We use two case studies to highlight the impact of normalization by analysing raw (unfiltered) versus normalized (filtered) data. The first case study evaluates plant data from the late Permian and Triassic compression/impression floras of South China, focussing on species richness/diversity assessments through the Permian-Triassic mass extinction and its recovery. In this case study, normalization reduced the number of taxa but revealed more detailed evolutionary patterns including the magnitude of floristic turnover, previously obscured by the fragmental preservation typical of plant fossils and nomenclature. The second case study evaluates Carboniferous to Permian anatomically preserved coal-ball floras from Europe, North America and China, focussing on palaeobiogeography and floral provinciality. Normalization reduced the number of coal-ball assemblages when surveyed at both genus and species level but revealed differences in relationships and floristic endemism. We conclude that normalized results should be considered alongside raw data, as they show important and complementary information which can greatly aid in overall interpretation.