The vinaceous-breasted Parrot (VBP, Amazona vinacea) is an endemic species of the South American Atlantic Forest biome, with considerable uncertainty about its distribution range. The ‘resident’ distribution map published by the International Union for Conservation of Nature (IUCN) comprises five large patches stretching from Argentina to Southeastern Brazil. A recently uploaded ‘possibly extant’ map, also proposed by the IUCN, consists of one single polygon that includes all the ‘resident’ patches. Indeed, VBPs have been frequently recorded outside the ‘resident’ area, emphasizing the need for an updated range map. No single research project has commanded sufficient resources to sample the entirety of VBP’s very large potential range area; however, the same area is well covered by citizen-science platforms, which include hundreds of VBP sightings. We modeled the VBP global distribution range from 2008 to 2018 by integrating research (roost count) data and citizen-science data (from the eBird, Xeno-Canto, and Wikiaves platforms) in a site-occupancy model. Our model accounts for imperfect detection, for the effect of environmental covariates on site occupancy, and for spatial autocorrelation in occupancy. We fit the model to data in a Bayesian framework and draw inference based on MCMC samples from the posterior probability distributions of model parameters.
Results/Conclusions
We mapped the VBP distribution range based on 1,007 VBP sightings from 47,240 samples in four datasets. This range, mapped with our integrated-data occupancy model, is seven and a half times larger than predicted without the citizen-science data. Site occupancy has a strong positive relationship with altitude and a positive but statistically weak relationship with the amount of Atlantic and Araucaria forest covers at each site. No single forest type conditions VBP occupancy throughout the whole study area, but we find unequivocal evidence of spatial autocorrelation in occupancy. Our study reveals a VBP distribution range of approximately 430 thousand km2 over the ten-year period analyzed. This range is subdivided in two large patches, which contrast with the five resident IUCN patches. The integration of data from different sources substantially increased spatial coverage, contributing to the precision of occupancy estimates. Our integration of citizen-science with conventional research data enabled us not only to update the VBP range map but also to quantitatively identify areas of high occupancy uncertainty that should be the target of future fieldwork. This approach could be applied to map distributions of any species.