Macroecological Analysis of Bird Migration Routes in North America Using eBird Data: Beta Diversity Perspective

Abstract

Bird migration routes across North America, which are often speculated upon at the microscale, are now subject to scrutiny using the ‘eBird’ database, the world’s largest data repository for the North American continent. This database has steadily accumulated data on resident and migratory birds, with annual data on both potentially revealing regional trends. Through analysis of eBird data accumulated since 1980, we identified regional trends indicating bird migration patterns across North America, delineating three primary regions within the North American flyway: the western, central and eastern flyways, with the latter further subdivided into three regions. These findings refine previous categorisations of the four major flyways, deepening our understanding of bird migration dynamics in North America.

1. Introduction

The study of bird migration is paramount in ornithology and involves the use of various methodologies [1,2,3,4]. Bird migration research historically focused primarily on specific species or taxa, neglecting the avian spectrum as a whole [5,6,7,8,9]. Research shows the North American flyway is conventionally divided into the Pacific, Central, Mississippi and Atlantic flyways [10,11,12,13,14,15,16,17] (Figure 1). Despite methodological advancements, such as ring banding and satellite tracking, which have revealed the migratory paths of many bird species, including established routes [17,18,19,20], the results obtained using these methods offer limited perspectives owing to their microecological focus. In particular, satellite tracking is expensive; satellite tracking and ringing fail to encompass entire bird populations, which may number tens to hundreds of thousands [20,21]. Therefore, complementary macroecological studies are imperative for a comprehensive understanding of bird migration dynamics.

Based on data obtained with permission from eBird, the world’s largest avian database, we studied bird migration across the North American continent (US and Canada, excluding Mexico). eBird [22], a citizen science initiative, includes a densely organised network of observation points across the continent. The extensive data repository has facilitated numerous avian studies, including resident and migratory birds, over an extended period [23,24,25].

Figure 1. Establish bird flight path predictions across the North American continent. From left to right: the Pacific (blue), Central (yellow), Mississippi (green) and Atlantic (red) flyways.

Figure 1. Establish bird flight path predictions across the North American continent. From left to right: the Pacific (blue), Central (yellow), Mississippi (green) and Atlantic (red) flyways.

Until now, there has been no discussion from a migration research perspective on how useful big data collected from citizen scientists is and when the collected data should be used to improve statistical resolution.

We hypothesised that data containing resident and migratory bird counts would exhibit regional characteristics, and that macroecological analysis of the relationships among all regions and those with regional characteristics would reveal patterns in bird migration routes. To test this hypothesis, we divided the North American continent into states and conducted a state-level analysis of annual data. We focused on beta diversity, a concept integral to understanding the relationships among all states and states with characteristic traits.

Therefore, we want to share our research results widely to present the connection between big data and migration research for the first time and create a foundation for future researchers to actively utilise big data.

2. Materials and Methods

2.1. Research Design

Beta diversity, which indicates the similarity between regions, becomes closer on the figure when the species composition is similar and becomes farther away as the species composition differs. Because of this characteristic, when analysing beta diversity only for resident birds in various regions, the difference between adjacent regions may be small or almost nonexistent. In this situation, it is migratory birds that cause the differences between each region. Since they move in flocks in their preferred direction, they can be divided into several groups according to their migration route and show differences in beta diversity. Because of this characteristic, by using big data on resident and migratory birds accumulated in each region, resident bird data can establish a basis for positioning on the beta diversity figure, and migratory bird data can show differences between each region and form migration routes on the figure.

2.2. Data Downloading and Pre-Processing

Upon obtaining permission from eBird in 2021, we downloaded the entire 189 Gb dataset up to 2023, which was a process that lasted approximately 3 days. Subsequently, we extracted 538 Gb of UTF-16LE-coded text files in the Windows environment in increments of 1 million lines and created and saved new files [26]. These text files were then transferred to the Linux environment for salting and conversion into the UTF-8 format for the North American countries included in our analysis [27,28]. The ‘data.table’ package was used to import the salted data into the R program, where it was merged by year and state before being pre-processed into a format compatible with the ‘phyloseq’ package for subsequent analysis [29,30,31,32].

2.3. Data Analysis and Visualisation

The ‘tibble’ package was used for conversion to phyloseq objects, and the ‘ggplot2’ package was used for analysis and visualisation [33,34,35,36,37]. We plotted beta diversity using one year of data collected in each region as a single dataset. Each dot on the beta diversity figure represents one year of data for one region. Since data from the distant past, not recent data, were not recorded on-site, we set various time periods to see when data began to provide high resolution for migration paths. Two beta diversity metrics were visualised by pre-processing data for each year as follows: 1910–2023, 1920–2023, 1930–2023, 1940–2023, 1950–2023, 1960–2023, 1970–2023, 1980–2023 and 1990–2023 (Figure 2 and Figure 3).

The ‘microViz’ package was used to visualise beta diversity [38], with two types of beta diversity plots generated based on the (i) Aitchison distance and (ii) redundancy analysis to depict genus level directionality. Given the compositional nature of the total bird population on the North American continent, we employed the Aitchison distance, which is suitable for compositional data analysis [39]. To identify the bird taxa represented in each flyway, we used redundancy analysis [40,41] (Figure 4). For the final flyway plot and redundancy analysis, data from 1990 to 2023 were plotted to present a visibly clear division.

3. Results

3.1. Utility of Accumulated eBird Data

From 1910 to 2023, the cumulative eBird dataset was first organised into nine subsets as follows: 1910–2023, 1920–2023 and 1990–2023. The beta diversity of each subset was then visualised using the Aitchison distance (Figure 2). Results revealed that the inclusion of pre-1980 data led to the geographic locations of all birds being interconnected from Hawaii. Conversely, using 1980–2023 and 1990–2023 data, the plots exhibited separation, closely aligning with the North American continent’s main geographic areas. However, an anomaly arose with the post-1980 data, where plots for Hawaii, located within the Pacific flyway, deviated from their actual geographic locations. Nevertheless, excluding Hawaii, the data are highly valuable for examining migration pathways across continental areas and will inform future analyses.

Figure 2. Beta diversity depicted across nine datasets. (A) 1910–2023, (B) 1920–2023, (C) 1930–2023, (D) 1940–2023, (E) 1950–2023, (F) 1960–2023, (G) 1970–2023, (H) 1980–2023 and (I) 1990–2023 data. Distinct regional trends emerged after 1980.

Figure 2. Beta diversity depicted across nine datasets. (A) 1910–2023, (B) 1920–2023, (C) 1930–2023, (D) 1940–2023, (E) 1950–2023, (F) 1960–2023, (G) 1970–2023, (H) 1980–2023 and (I) 1990–2023 data. Distinct regional trends emerged after 1980.

3.2. Three Primary Migration Pathways in North America

Our analysis delineates three primary migratory pathways on the North American continent, which are distinct from previous descriptions [42] (Figure 3). Starting from Alaska (AK) and ending at the Yukon Territory, bird migration can be divided into three main branches, the western, central and eastern flyways. Of these flyways, the western originates from British Columbia, the central from Saskatchewan (SK) and the eastern from Nunavut (NU), and end at Newfoundland and Labrador. Notably, a bifurcation was observed between the western and central flyways in Alberta (AB), and between the central and eastern flyways in the Northwest Territories and Manitoba (MB).

Although the boundary between the western and central flyways is blurred in existing migration routes [40], our analysis includes a distinct division by state across the North American continent. The central flyway, comprising Saskatchewan (SK)–North Dakota (ND)–South Dakota (SD)–Nebraska (NE)–Kansas (KS)–Oklahoma (OK)–Texas (TX), exhibits a bird composition more similar to that of the eastern flyway than to that of the western flyway. Detailed examination revealed the potential subdivision of the eastern flyway into three travel routes as follows: the known Mississippi flyway (composed of Michigan [MN]–Iowa [IA]–Missouri [MO]–Arkansas [AR]–Louisiana [LA]), the Atlantic flyway formed along the coast and an inland flyway formed between the Mississippi and Atlantic flyways.

Figure 3. Bird flyways illustrated by beta diversity. The flyways of the North American continent are broadly divided into the Western, Central, and Eastern flyways. The Eastern flyway is further divided into three more detailed flyways.

Figure 3. Bird flyways illustrated by beta diversity. The flyways of the North American continent are broadly divided into the Western, Central, and Eastern flyways. The Eastern flyway is further divided into three more detailed flyways.

3.3. Top 20 Genera in Flyways

The resultant plot shows directionality in the North American continent, with Uria and Cepphus dominating in the north, Somateria in the northeast and Perisoreus in the northwest. Morus and Scolopax tended to be more abundant in the east of the continent, whereas Hylocichla, Thryothorus, Cardinalis, Coragyps and Baeolophus were more abundant in the southeast. In the west of the continent, Aechmophorus, Salpinctes, Aeronautes, Nucifraga, Pica, Myadestes, Leucosticte and Cinclus were found in greater numbers, with Pelecanus dominating the south.

4. Discussion

The importance of ‘big data’ is increasing in the modern world, allowing for more diverse and profound analytics through the accumulation of extensive datasets over prolonged periods [43,44,45,46]. The eBird data epitomises this trend by aggregating contributions from countless citizens who report bird sightings from various locales, ranging from backyards to mountains, rivers and coasts, thereby identifying ‘hotspots’ for ongoing data collection [47] and enriching the compositional data of avian populations in different areas [48,49].

We evaluated the completeness of the eBird data by assessing beta diversity patterns, revealing enhanced regional trends using cumulative data extending from 1980 onwards. Our findings can guide future research involving compositional data analysis. Moreover, the continuous accumulation of data from densely populated ‘hotspots’ beyond North America could facilitate validation and analysis in other regions.

Mapping bird migration across the North American continent relied on the most comprehensive and accessible compositional dataset available due to eBird’s widespread adoption by individuals from diverse locations. Rather than relying solely on predictions from existing microecological data for North America, we generated a data-driven map of macroscale bird movements that encompassed all bird species. Our results complement existing bird flyway maps in North America, which often focus on select species and/or taxa.

Although studying the migratory routes of individual species or taxonomic groups typically requires expensive equipment and extensive human resources, our findings highlight the scalability of bird macroecology, including distribution patterns and migration dynamics across North America, using the contributions of a large group of citizen scientists. Integrating micro- and macroecological perspectives is crucial for a comprehensive understanding of bird migration. Moreover, the formation of macroecological communities of ‘big data builders’, such as eBird, in regions beyond North America holds promise for global analyses of bird migration routes based on extensive datasets.

Whether to divide the migratory routes of birds into three or four remains controversial. One limitation of this study was that we could not demonstrate the migratory routes of individual bird species. Each individual bird species may have four, fewer or more migratory routes; therefore, the four previously known migratory routes do not reflect the migratory routes of individual bird species. The results of the present study are different from those of previous studies because they reflect the migration patterns of birds that migrate in large flocks rather than the migration paths of individual species; these results can help in understanding the overall macroscopic flow.

Figure 4. Redundancy analysis at the genus level. Each arrow indicates an increase in the bird population within a given genus as it progresses.

Figure 4. Redundancy analysis at the genus level. Each arrow indicates an increase in the bird population within a given genus as it progresses.

5. Conclusions

We have shown that eBird data accumulated over approximately 100 years show specific macroecological patterns since the 1980s. In addition, using data accumulated since the 1980s, we have shown for the first time that the migration paths of large-scale bird communities can be drawn on beta diversity. Our research provides insight into new understanding of bird migration routes, which can be used to conserve bird migration routes or individual birds.