Revealing the palaeoecology of silent taxa: selecting proxy species from associations in modern vegetation data

AimSpecies severely under-represented in fossil pollen records leave gaps in interpretations and reconstructions of past vegetation. These 'silent taxa' leave little or no trace due to low pollen production, dispersal, preservation and taxonomic resolution. An approach for including them is through associating them with other species with reliable pollen representation. Here, we demonstrate a method for selecting such a proxy species for the Holocene using modern vegetation data.LocationNew Zealand.TaxonBeilschmiedia tawa (A.Cunn.) Benth. & Hook. F. ex Kirk (Lauraceae).MethodsWe used vegetation plot data to perform a pairwise co-occurrence analysis of the New Zealand indigenous forest metacommunity to identify species with a strong positive association with Beilschmiedia tawa (tawa), a common tree severely under-recorded in the pollen record. For those species, we then modelled their realised climatic niches to identify species with high niche overlap. We discuss how well those species could be interpreted from the Holocene fossil pollen record based on the representation of their pollen taxa.ResultsKnightia excelsa (rewarewa; Proteaceae) is a potential proxy for B. tawa in Holocene fossil pollen records, and other, range-limited species may provide community-specific proxies. We show combining resampling with sub-sampling is a robust method for reducing the high false positive rate associated with large co-occurrence analyses (1000+ sites) by limiting the sample size to 100 sites.Main ConclusionsWe show that the palaeoecology of silent taxa can be studied via proxy species, allowing their past distributions to be better understood. We highlight the importance of modelling many aspects of the realised niche to understand the usefulness and limitations of the silent-proxy association. Future research should focus on testing the underlying assumptions of the silent-proxy relationship so that models built on modern data can confidently be applied to palaeoecological data.