Species observations collected in a non-systematic manner e.g. through citizen science programs like iNaturalist, eBird, etc. make up a large proportion of occurrence data in GBIF. Providing wide coverage on spatial, temporal and taxonomic scales, opportunistic data, however, can be biased and often lacks information about absences of species.
Using occurrence data on 71 wetland bird species from GBIF publisher Artportalen, a study by Alejandro Ruete (1st prize winner of the 2016 Ebbe Nielsen Challenge) and colleagues introduces a novel dynamic occupancy model that attempts to cope with known sources of bias including lack of absence data and variation in sampling effort.
When applied to the real observations, the model estimates daily occupancy for a given species and site- and when summarized, provides a detailed picture of seasonal site use including within-season population dynamics.
Confirming the robustness of the model using simulated data, the authors conclude that modelling using opportunistic data with multiple replicates can provide more biologically relevant information then traditional annual occupancy models.