Citizen scientists as butterfly predators: using foraging theory to understand individual recorder behaviour

• An optimal foraging model captures recorder behaviour in butterfly citizen science. • Site visits reflect a trade-off between species rarity and travel distance. • Recorders fall on a spectrum from narrow- to broad-ranged foragers, the latter actively seeking ‘interesting’ sites. • The composition of recorders (narrow- vs. broad-ranged) contributing to a given dataset determines the aggregate sampling bias. Citizen science is increasingly important in the collection of biological data. However, to understand the broader utility of the growing number of citizen-derived records, we need to understand exactly how recorder behaviour affects the geographic distribution of records made. Here, we apply an optimal foraging model to citizen science data from the UK to determine how likely a recorder (predator) is to visit any given kilometre square and record a butterfly (prey). By defining the square with the highest density of an individual’s records as their ‘origin’, we show that the probability of visiting a given site depends on its distance from the origin and the rarity-weighted species richness of the species thought to be present. This pattern of behaviour differs between recorders visiting more than or fewer than five squares, termed broad and narrow-range foragers. The model shows that recorder behaviour is driven, in part, by a simple trade-off between distance travelled and the rarity-weighted species richness. This collective behaviour helps explain over-recording by broad-ranging foragers in protected areas at distance and under-recording, by narrow-range foragers, in the wider countryside. It also implies that estimating parameters describing rare species’ distributions (e.g. mean occupancy) will be challenging, since sample inclusion depends on occupancy itself. Mapping rare species’ distributions should be simpler, since the sites at which they can be found tend to be well-sampled, but the same is unlikely to be true of common species, which also occupy areas that are unlikely to be sampled. More work is needed to understand how widely our results can be generalised beyond the UK and the dataset considered.