Release the HOGS: An unsupervised marker extraction, classification and georeferencing approach for biodiversity data

Distributional data are essential for understanding species and community responses to environmental and anthropogenic change. Large biodiversity databases provide key information on distributional patterns, but their temporal coverage can be limited. Facilitating access to untapped occurrence data is a pressing need. We present the Historical Occurrence Georeferencing System (HOGS), a Python protocol that isolates occurrence markers from distribution maps, assigns them to taxonomic groups and georeferences each image relative to a baseline coordinate system, producing latitude–longitude coordinates for all mapped marker occurrences. We tested our georeferencing protocol on two historical atlases of over 1000 sub-Saharan bird species, georeferencing 153,052 markers from 751 maps. Our new georeferencing protocol enables rapid generation of occurrence data from historical maps, providing reference or enhanced distribution data for species and inputs for ecological niche modelling in biogeographical, evolutionary and conservation studies.