Studying nocturnal bird migration is challenging because direct visual observations are difficult during darkness. Radar has been the means of choice to study nocturnal bird migration for several decades, but provides limited taxonomic information. Here, to ascertain the feasibility of enhancing the taxonomic resolution of radar data, we combined acoustic data with vertical-looking radar measurements to quantify thrush (Family: Turdidae) migration. Acoustic recordings, collected in Helsinki between August and October of 2021-2022, were used to identify likely nights of high and low thrush migration. Then, we built a random forest classifier that used recorded radar signals from those nights to separate all migrating passerines across the autumn migration season into thrushes and non-thrushes. The classifier had a high overall accuracy (approximate to 0.82), with wingbeat frequency and bird size being key for separation. The overall estimated thrush autumn migration phenology was in line with known migratory patterns and strongly correlated (Pearson correlation coefficient approximate to 0.65) with the phenology of the acoustic data. These results confirm how the joint application of acoustic and vertical-looking radar data can, under certain migratory conditions and locations, be used to quantify 'family-level' bird migration.This study addresses the challenge of studying nocturnal bird migration, typically hindered by limited taxonomic information from radar data. To enhance resolution, we combined acoustic recordings with vertical-looking radar measurements, focusing on thrush migration. Using a random forest classifier, we achieved a high accuracy in distinguishing thrushes from non-thrushes during autumn migration, relying on key factors like wingbeat frequency and bird size. The estimated thrush migration phenology aligned with known patterns and correlated strongly with acoustic data. Our study provides the first example of combining acoustic and radar data to extract taxonomic information, enabling the quantification of family-level migration from radar data. image
Quantifying nocturnal thrush migration using sensor data fusion between acoustics and vertical-looking radar
Giuntini S.;Martinoli A.;Preatoni D. G.;
2024-01-01
Abstract
Studying nocturnal bird migration is challenging because direct visual observations are difficult during darkness. Radar has been the means of choice to study nocturnal bird migration for several decades, but provides limited taxonomic information. Here, to ascertain the feasibility of enhancing the taxonomic resolution of radar data, we combined acoustic data with vertical-looking radar measurements to quantify thrush (Family: Turdidae) migration. Acoustic recordings, collected in Helsinki between August and October of 2021-2022, were used to identify likely nights of high and low thrush migration. Then, we built a random forest classifier that used recorded radar signals from those nights to separate all migrating passerines across the autumn migration season into thrushes and non-thrushes. The classifier had a high overall accuracy (approximate to 0.82), with wingbeat frequency and bird size being key for separation. The overall estimated thrush autumn migration phenology was in line with known migratory patterns and strongly correlated (Pearson correlation coefficient approximate to 0.65) with the phenology of the acoustic data. These results confirm how the joint application of acoustic and vertical-looking radar data can, under certain migratory conditions and locations, be used to quantify 'family-level' bird migration.This study addresses the challenge of studying nocturnal bird migration, typically hindered by limited taxonomic information from radar data. To enhance resolution, we combined acoustic recordings with vertical-looking radar measurements, focusing on thrush migration. Using a random forest classifier, we achieved a high accuracy in distinguishing thrushes from non-thrushes during autumn migration, relying on key factors like wingbeat frequency and bird size. The estimated thrush migration phenology aligned with known patterns and correlated strongly with acoustic data. Our study provides the first example of combining acoustic and radar data to extract taxonomic information, enabling the quantification of family-level migration from radar data. image
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
