Large-bodied birds are over-represented in unstructured citizen science data

Area of Science:

• Biodiversity research
• Citizen science
• Ecological modeling

Background:

• Citizen science platforms generate vast biodiversity data annually.
• Data biases in citizen science must be quantified before ecological analysis.
• Unstructured protocols may introduce biases compared to semi-structured or structured ones.

Purpose of the Study:

• To quantify biases in unstructured citizen science data.
• To compare bird observations between iNaturalist (unstructured) and eBird (semi-structured) in the US.
• To identify species traits predicting representation bias.

Main Methods:

• Contrasted bird observations from iNaturalist and eBird for the continental US.
• Tested if body size, commonness, flock size, and color predict species under- or over-representation.
• Analyzed data for species traits influencing bias in unstructured datasets.

Main Results:

• Large-bodied birds were strongly over-represented in iNaturalist data.
• Common species and species in large flocks were moderately to strongly over-represented.
• No evidence suggested colorful species were over-represented in unstructured data.
• Biases are likely linked to detectability and recording processes, e.g., ease of photographing large birds.

Conclusions:

• Unstructured citizen science data exhibit biases compared to semi-structured data.
• Body size is a key trait for understanding and potentially correcting biases in opportunistic species occurrence records.
• Caution is advised when using unstructured citizen science data for ecological modeling.
• Future research should include structured data to compare biases across all platform types.