Abstract
Antarctic conservation science is crucial for enhancing Antarctic policy and understanding alterations to terrestrial Antarctic biodiversity. Antarctic conservation will have limited long-term impacts in the absence of large-scale biodiversity data, but if such data were available, it is likely to improve environmental protection regimes. To enable the prediction of Antarctic biodiversity across continental spatial scales through proxy variables, in the absence of baseline surveys, we linked Antarctic substrate-derived environmental DNA (eDNA) sequence data from the remote Antarctic Prince Charles Mountains to a selected range of concomitantly collected measurements of substrate properties. We achieved this through application of a statistical method commonly used in machine learning. Our analysis indicated that neutral substrate pH, low conductivity, and certain substrate minerals are important predictors of the presence of basidiomycetes, chlorophytes, ciliophorans, nematodes, and tardigrades. A bootstrapped regression revealed how variations in the identified substrate parameters influence probabilities of detecting eukaryote phyla across vast and remote areas of Antarctica. We believe that our work will improve future taxon distribution modeling and aid in developing more targeted surveys of biodiversity conducted under logistically challenging conditions.