Hidden observer preferences drive biases in semi-structured citizen science databases

Louis Backstrom

Sightings recorded by volunteer naturalists are an increasingly important source of data used in biodiversity monitoring. These sightings come with several well-recognised issues, however, including variation in observer effort across spatial, temporal, and taxonomic scales. These observer preferences result in most citizen science data being a non-random, biased sample of their population of interest, a problem which has received extensive research effort over the past decade. Despite this, there remain many avenues for further advances in our understanding of how sampling bias manifests in citizen science datasets and the effects that biased data have on biological inference. In this talk, I will present some preliminary results describing a set of observer preferences and resultant biases that have had limited prior recognition in the literature, that I describe as “hidden” preferences. I will give an overview of how these preferences arise because of specific observer behaviours, how they can be identified in datasets, and their influence on ecological models.