Belotti, Maria Carolina Tiburcio Dias, authorHorton, Kyle G., advisorBailey, Larissa, committee memberKoons, David, committee memberRoss, Matthew, committee member2025-06-022025-06-022025https://hdl.handle.net/10217/241030During their non-breeding period, many species of swallows and martins (family: Hirundinidae) congregate in large communal roosts, which can gather hundreds of thousands of individuals. These roosts are well-known within local birdwatching communities; however, monitoring them at large spatial scales and with day-to-day temporal resolution is challenging. Due to the high densities of birds within these aggregations, however, their early morning dispersals can be systematically detected by nearby weather radars, which can be used to collect data about roost timing, size, and location. Nonetheless, finding roost signatures amongst the millions of rendered reflectivity images is extremely time-consuming — a fact that has limited the spatial and temporal scopes of previous radar-based studies. We leveraged the recent advances in computer vision and high-performance computing to partially automate this task and build a dataset of 22 years of roost detections captured by 12 radar stations in the Great Lakes region. We first verified that these detections correspond to swallow and martin roost dispersals, comparing the phenology of our findings with that obtained from eBird data. We then describe changes in the roost size distribution throughout the season and discuss the relationship between a roost's size and its persistence in the landscape. We also obtained trends of roost activity, comparing those with breeding population trends from eBird and the North American Breeding Bird Survey. Lastly, we used two years of data (2014-2015) collected by an operational weather radar in Manaus to describe, for the first time, the phenology of swallow and martin aggregations in the Low Negro-Solimões region of the Amazon Rainforest. We compared roosting behavior in the Amazon with what we observed 6,000 km away, in the Great Lakes region, regarding the daily number of birds and the roost size distribution. Our work demonstrates how weather radars offer us a unique opportunity to monitor and study swallow and martin populations at an unprecedented transcontinental scale, something that can rarely be achieved for any other taxa.born digitaldoctoral dissertationsenghttps://creativecommons.org/licenses/by-nc-nd/4.0/legalcodeavian migrationcommunity scienceradar ornithologycommunal roostingapplied statisticsmachine learningTextThis work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 United States License. https://creativecommons.org/licenses/by-nc-nd/4.0/legalcode