Bias, incompleteness and the 'known unknowns' in the Holocene faunal record.

Long-term faunal data are needed to track biodiversity change and extinction over wide spatio-temporal scales. The Holocene record is a particularly rich and well-resolved resource for this purpose but nonetheless represents a biased subset of the original faunal composition, both at the site-level assemblage and when data are pooled for wider-scale analysis. We investigated patterns and potential sources of taxonomic, spatial and temporal bias in two Holocene datasets of mammalian occurrence and abundance, one at the global species level and one at the continental population-level. Larger-bodied species are disproportionately abundant in the Holocene fossil record, but this varies according to trophic level, probably owing to past patterns of human subsistence and exploitation. Despite the uneven spatial distribution of mammalian occurrence records, we found no specific source of sampling bias, suggesting that this error type can be avoided by intensive data collection protocols. Faunal assemblages are more abundant and precisely dated nearer to the present as a consequence of taphonomy, past human demography and dating methods. Our study represents one of the first attempts to quantify incompleteness and bias in the Holocene mammal record, and failing to critically assess the quality of long-term faunal datasets has major implications for understanding species decline and extinction risk. This article is part of a discussion meeting issue 'The past is a foreign country: how much can the fossil record actually inform conservation?'