Using geolocator tracking data and ringing archives to validate citizen-science based seasonal predictions of bird distribution in a data-poor region

Unstructured citizen-science data are increasingly used for analysing the abundance and distribution of species. Here we test the usefulness of such data to predict the seasonal distribution of migratory songbirds, and to analyse patterns of migratory connectivity. We used bird occurrence data from eBird, one of the largest global citizen science databases, to predict the year-round distribution of eight songbird taxa (Agropsar philippensis, Calliope calliope, Cecropis daurica, Emberiza aureola, Hirundo rustica, Locustella certhiola, Oriolus chinensis, Saxicola torquatus stejnegeri) that migrate through East Asia, a region especially poor in data but globally important for the conservation of migratory land birds. Maximum entropy models were built to predict spring stopover, autumn stopover and wintering areas. Ring recovery and geolocator tracking data were then used to evaluate, how well the predicted occurrence at a given period of the annual cycle matched sites where the species were known to be present from ringing and tracking data. Predicted winter ranges were generally smaller than those on published extent-of-occurrence maps (the hitherto only available source of distribution information). There was little overlap in stopover regions. The overlap between areas predicted as suitable from the eBird data and areas that had records from geolocator tracking was high in winter, and lower for spring and autumn migration. Less than 50% of the ringing recoveries came from locations within the seasonal predicted areas, with the highest overlap in autumn. The seasonal range size of a species affected the matching of tracking/ringing data with the predictions. Strong migratory connectivity was evident in Siberian Rubythroats and Barn Swallows. We identified two migration corridors, one over the eastern mainland of China, and one along a chain of islands in the Pacific. We show that the combination of disparate data sources has great potential to gain a better understanding of the non-breeding distribution and migratory connectivity of Eastern Palearctic songbirds. Citizen-science observation data are useful even in remote areas to predict the seasonal distribution of migratory species, especially in periods when birds are sedentary and when supplemented with tracking data.