Do observer differences in bird detection affect inferences from large-scale ecological studies?

Use of multiple observers in large-scale bird surveys is often unavoidable. But how significant are observer differences in bird detection? Do observer effects significantly influence inferences about environmental factors on birds? We conducted a field experiment to quantify differences between observers in the detection of birds at Booderee National Park, south-eastern Australia. We also re-analysed a large dataset from an observational study where multiple observers had participated in bird surveys. We identified highly significant observer differences for estimates of bird species richness and the probability of detection of three exemplar taxa. We demonstrated that observer effects would not substantially alter inferences we made about relationships between bird species and vegetation type or burning history. We believe that four features of our survey design and protocol limited the magnitude of observer effects on environmental inferences: (1) high levels of replication of classes of field sites critical for relative comparisons of site (vegetation) types; (2) pre-survey screening to ensure that only experienced ornithologists participated in surveys; (3) repeat sampling of field sites by a different observer on a different day to reduce the impacts of observer heterogeneity and 'day' effects; and (4) precluding surveys during poor weather or long after dawn, also to limit 'day' effects.