A local scale evaluation of spatial sampling bias in the Atlas of Australian Birds

The reliability of citizen science datasets where volunteers are free to choose sampling locations is not clear. This study examined local ( patch) scale spatial sampling patterns in the Atlas of Australian Birds and then compared reporting rates, i.e. the proportion of sampling units in which a given species was present, from a sample of atlas points with those from a regular sample. Three sites that have been were surveyed sequentially between January-May 2017: Killawarra Forest, Victoria, Coolah Tops National Park and Pilliga Nature Reserve, New South Wales. Spatial bias in atlas sampling patterns was evident as clusters at tourist areas and special habitat features and linear patterns along roads and creek lines. Atlas samples overestimated reporting rates for species with spatial distributions that were concordant with those sampling patterns and vice versa. At least two-fold differences in atlas/regular sample reporting rate ratios were detected for between 13-15% of non-rare species (with reporting rates [≥] 0.08). Concerns are raised that spatial sampling bias is common in the atlas and affects a variety of species, that popular sites may not be representative of habitat patches and that a large proportion of surveys are being filtered out in data analyses.