Detecting Genuine Versus Pseudo-Absence in The Fossil Record: Applications of Occupancy Modelling For Conservation Palaeobiology

Abstract

Palaeontological data provide a unique avenue to evaluate the impact of climatic, habitat and ecosystem change over longer temporal scales than typically examined in ecology and conservation, contributing critical data on extinction dynamics that can help contextualize the current biodiversity crisis. However, the fossil record is biased by a variety of factors. In particular, the issue of data absence causes a genuine concern when attempting to discern spatial patterns. Does the lack of a fossil occurrence indicate genuine absence or imperfect detection (i.e., pseudo-absence)? Failing to quantify, discern and mitigate both the main drivers and impacts of data absence will have major implications for any attempt to reconstruct past diversity dynamics, limiting the applicability of paleontological data for addressing questions pertaining to present-day biodiversity. Occupancy modelling, a technique commonly applied in the fields of ecology and conservation, provides a novel way to evaluate the impact of both spatial and temporal biases on the fossil record. By distinguishing between true (taxon genuinely absent) and false (taxon present, but not observed) absences, occupancy modelling produces independent and simultaneous probability estimates for both occupancy and detection. Here, we show how paleontological occurrences can be adapted for use alongside relevant modern and paleo covariate data in both single season models run using the R package ‘unmarked’ and dynamic occupancy models using a Bayesian framework. We additionally test the impact of varying spatial scale, as well as uneven numbers of repeated site visits, on model outcomes, and provide recommendations for conservation paleobiologists intending to run these models. Finally, we outline additional benefits of applying occupancy modelling within conservation paleobiology.