The correlation between eBird community science and weather surveillance radar-based estimates of migration phenology

Measuring avian migration can prove challenging given the spatial scope and the diversity of species involved. No one monitoring technique provides all the pertinent measures needed to capture this macroscale phenomenon - emphasizing the need for data integration. Migration phenology is a key metric characterizing large-scale migration dynamics and has been successfully quantified using weather surveillance radar (WSR) data and community science observations. Separately, both platforms have their limitations and measure different aspects of bird migration. We sought to make a formal comparison of the migration phenology estimates derived from WSR and eBird data - of which we predict a positive correlation.

Contiguous United States.

2002-2018.

Migratory birds.

We estimated spring and autumn migration phenology at 143 WSR stations aggregated over a 17-year period (2002-2018), which we contrast with eBird-based estimates of spring and autumn migration phenology for 293 nocturnally migrating bird species at the 143 WSR stations. We compared phenology metrics derived from all species and WSR stations combined, for species in three taxonomic orders (Anseriformes, Charadriiformes and Passeriformes), and for WSR stations in three North American migration flyways (western, central and eastern).

We found positive correlations between WSR and eBird-based estimates of migration phenology and differences in the strength of correlations among taxonomic orders and migration flyways. The correlations were stronger during spring migration, for Passeriformes, and generally for WSR stations in the eastern flyway. Autumn migration showed weaker correlation, and in Anseriformes correlations were weakest overall. Lastly, eBird-based estimates slightly preceded those derived from WSR in the spring, but trailed WSR in the autumn, suggesting that the two data sources measure different components of migration phenology.

We highlight the complementarity of these two approaches, but also reveal strong taxonomic and geographic differences in the relationships between the platforms.