Avian spring migration at the east Adriatic coast: coastal and sea-crossing dynamics of intensity, timing, and flight directions

Abstract

Background: Each spring, migratory birds converge along the Croatian coast from various directions, creating a dynamic intersection of flight paths. Many birds are thought to cross the Adriatic Sea, while others follow a northward route along the coastline. As most migratory birds initiate flight shortly after sunset, we hypothesized that sea-crossing migrants would arrive with a delay at the Croatian coast, compared to migrants that were following the coast, resulting in potentially intricate spatiotemporal patterns that remain poorly understood.

Methods: We deployed four ornithological radar devices along the Croatian coastline: two in southwestern Istria and two in northern Dalmatia. These radars tracked migratory bird activity up to 1000 m above ground, recording intensity and flight directions and their variations across sites, seasons, and individual nights. We conducted an exploratory analysis of these variations, applied functional principal component analysis and hierarchical clustering to summarise within-night activity profiles, and compared these profiles between sites and across the migration season, alongside associated flight direction distributions.

Results: During the early migration season, migration intensity was similar across all technically active sites, but site-to-site variation increased markedly in May. In March, flight directions were predominantly towards NNE, indicating mainly sea-crossing migration throughout the night. In April, NW directions dominated the first half of the night, shifting to scattered N directions later; in Dalmatia, even strong W components were observed early in the night. By May, W to NW movement towards the Italian coast were typical for early-night activity. Later in the night, flight directions shifted towards NE over Istria and N at the Dalmatian site in Zadar, while the site at Vrana exhibited a wide scatter, warranting further discussion. Contrary to our expectations, within-night intensity profiles could not be fully linked to specific directional patterns.

Conclusion: This study points to the complex interplay between coastal and sea-crossing migration along the Croatian coast. Our results demonstrate significant variability in the timing of migration within single nights in the context of aquatic barriers. Crossing such a barrier results in downstream delays compared to birds migrating along the barrier. Between nights, one behaviour or the other may dominate the overall activity, causing the shifts in within-night timing. In addition, barrier crossing is likely to be strongly influenced by weather conditions, contributing to the variability in the within-night timing of migration. However, measured flight directions were not always consistent with within-night timing, highlighting the complexity of avian migration in the context of aquatic barriers. This highlights the need for further research with high temporal resolution to gain a deeper understanding of migration behaviour in response to such barriers.

Keywords: Aeroecology; Bird migration; East Adriatic Coast; Functional data analysis; Ornithological radar.

Fig. 1

The locations of the four radar sites in Rovinj, Palud, Zadar and Vrana, each equipped with a BirdScan MR1 ornithological radar. Broad arrows indicate the expected directions of coastal (reddish) and sea-crossing (blueish) migration

Fig. 2

Left: Timeline plot of the nightly mean migration intensity for the four sites in the 2023 spring season. Right: Summary of migration intensities at each site shown as boxplots, including the median, interquartile range, and outliers

Fig. 3

Nightly flight direction means and interquartile ranges for each radar site during the spring migration season 2023, separated by region. Inset plots show the concentration of flight direction as the mean resultant vector length (MRL); top: Istria; bottom: Dalmatia

Fig. 4

Distribution of recorded flight directions per site and month. Bird echoes have been quantified per 5-degree segment of the circle and standardised for comparability. Colours represent quarters of the night with blue just after sunset, purple before midnight, orange just after midnight, and yellow before sunrise

Fig. 5

Cluster-specific density distribution of the scaled first and second FPC eigenvalues of radar nights. The first FPC represents the amplitude of the migration intensity profiles, the second FPC represents a shift in the timing of the peak intensity. The four lower panels show predicted migration intensity profiles from the centre of a cluster in comparison with the nearest observed radar night profile. (Distribution of the third FPC in the Supporting Information)

Fig. 6

Distribution of recorded flight directions per site and cluster. Bird echoes have been quantified per 5-degree segment of the circle and standardised for comparability. Colours represent quarters of the night with blue just after sunset, purple before midnight, orange just after midnight, and yellow before sunrise. Vrana was excluded because of the radar malfunction (Fig. 4)