RESULTS
Spatiotemporal variation in avian diversity. We identify three Principal Components (PCs) that together explain 65% of weekly variance for the five taxonomic and functional diversity metrics across the continental US (Fig. S1). The first Principal Component (PC1; 47% variance explained; Fig. S1), separates the breeding (positive score) from the wintering (negative score) season, with two distinct, lesser peaks likely associated with the temporary addition of transient species during seasonal migration (Fig. 1a). The second Principal Component (PC2, 11% variance explained; Fig. S1) further isolates migration (primarily, spring migration; positive score) from periods of wintering and breeding (negative score; Fig. 1a). The third Principal Component (PC3, 8% variance explained; Fig. S1) further emphasizes the signal of autumn migration (positive score; Fig. 1a). Each subsequent PC explains < 5% of the total variance (Fig. S1) and captures mostly stochastic fluctuations, without a clear seasonal signature (Fig. S3). Seasonal patterns of avian diversity can thus be approximated for each grid cell as the weighted combination of three principal modes: breeding/winter season (PC1), spring (PC2), and autumn (PC3) migration, where the weights (i.e., importance of each PC at different locations) are shown by the PC loading maps (Fig. 1b).
Avian diversity shows clear spatial patterns in the strength (loading) of temporal variation (score) patterns, but there are notable differences among the five diversity metrics in how these patterns are expressed (Figs. 1,2a,S4). Seasonality of species richness demonstrates a strong latitudinal gradient, with the northern US (n=693,160 grid cells, ~74% of study region) characterized by high breeding season SR and low winter SR (strongly positive PC1 loadings; Figs. 1,2a). Along the Gulf of Mexico coast, SR instead peaks during winter (n=48,251, ~5%; strongly negative PC1 loadings). Known migratory staging areas across the South, Southeast, Southwest, and California (n=159,452, ~17%) experience spring and autumn peaks in SR (strongly positive PC2 and PC3 loadings; Figs. 1,2a), with some spatial differences between PC2 (spring) and PC3 (autumn) loading. A few patches in the Southeast (n=32,298, ~4%) instead experience autumn troughs in SR (Figs. 1,2a). Raw functional richness, FRic, expectedly shows fairly strong spatiotemporal congruence with SR, with only small deviations for parts of Texas and the Midwest where FRic peaks during the breeding season and autumn migration, respectively (Fig. 1,2b).
In stark contrast to the strongly correlated SR and FRic, corrected functional richness, cFRic, peaks during winter across most of the East, Midwest, and the Pacific Coast (strongly negative PC1 loadings; n=265,851, ~29%; Figs. 1,2c). This temporal signature implies that the total breadth of functional space occupied by a bird assemblage is higher during winter than would be expected after accounting for seasonal declines in species richness due to migration, but lower during summer when migrants are back on their breeding grounds. In contrast, breeding season peaks in cFRic (strongly positive PC1 loading) are common across northern Michigan, Florida, Texas, the Rocky Mountains, and California’s Central Valley (n=240,285, ~26%; Figs. 1,2c). For Florida and Texas in particular, this suggests that the influx of short-distance migrants from northern latitudes during winter leads to assemblage-wide declines in functional richness. Importantly, passage migrants strongly influence the seasonality of cFRic. Specifically, the high plateaus of the Intermountain West experience spring troughs in cFRic (strongly negative PC2 loadings; n=144,797, ~15%; Figs. 1,2c), while parts of the Southwest, Southeast, and New England see spring and autumn peaks (strongly positive PC2 and PC3 loadings; n=188,824, ~20%; Figs. 1,2c).
Seasonality of functional evenness, FEve, broadly displays an east-west gradient (Figs. 1,2d), with breeding season peaks (strongly positive PC1 loadings) common east of the Rocky Mountains, in parts of the Great Basin, and along the Pacific Coast (n=642,027, ~69%; Figs. 1,2d). Such a temporal pattern indicates an even distribution of species’ abundances in the functional space during the breeding season. In winter, however, species’ abundances amass in a few regions of the functional space. Passage migrants influence seasonality of FEve particularly strongly in high elevation and topographically varying regions. The Intermountain West high plateaus, parts of the Rocky Mountains, and the Sierras experience irregular distribution of species’ abundances within the functional space (low FEve) during spring (negative PC2 loadings; n=108,813, ~12%) and the breeding season (negative PC1 loadings; n=34,122, ~4%), while parts of the Rocky, Chisos, Ozark, and Appalachian Mountains (n=100,862, ~11%) see functional evenness peak in spring and autumn (Fig 1,2d).
Strong breeding-wintering seasonality characterizes functional dispersion, FDis, across most of the continental US (n=745,939; ~80%; Figs. 1,2e), with peaks typically observed during the breeding season (strongly positive PC1 loadings). This temporal signature implies that during the breeding season abundant species are spread further away from the centroid of the functional space relative to rare species, but in winter they are positioned near the centroid. The only regions with winter peaks in FDis (negative PC1 loadings) are the northern Rocky Mountains and lower Peninsular Florida (n=32,980, ~4%; Figs. 1,2e). Passage migrants strongly increase the dispersion of species’ abundances in the functional space in spring and autumn (strongly positive PC2 and PC3 loadings) in the Appalachian Mountains, the Great Lakes region, and upper Peninsular Florida (n=52,619, ~6%), but lower it (strongly negative PC2 and PC3 loadings) in the Intermountain West high plateaus, the Sierras, and the Midwest (n=101,588, ~10%; Figs. 1,2e).
Spatial congruence in seasonality of avian taxonomic and functional diversity. We identify seven distinct spatiotemporal clusters (Figs. 3,S2). Broadly, Clusters 1 (n=140,846, ~15% of continental US) and 2 (n=140,967, ~15%) represent locations where species richness and functional diversity peak during the breeding season and migration, except for cFRic, which alone declines during the breeding season (Fig. 3b). Together, Clusters 1 and 2 cover much of the eastern US (Fig. 3a), with Cluster 1 representing the higher elevation Appalachian region and southeastern plateaus and Cluster 2 representing the low lying plains and prairies of the Lower Great Lakes. Parts of California and the Pacific Northwest also show characteristics of Clusters 1 and 2 (Fig. 3a).
Cluster 3 (n=120,658, ~13%) is broadly defined by breeding season peaks in avian diversity, across nearly all measures, and declines during winter, spring, and autumn (Fig. 3b). cFRic is again an exception to this pattern as it instead peaks in spring and plummets in autumn (Fig. 3b). Cluster 3 comprises high elevation, cold winter, forested regions of the Rocky Mountains, New England, upstate New York, and the Upper Great Lakes (Fig. 3a). Cluster 4 (n=99,342, ~11%) is characterized by low SR but high functional diversity during the breeding season (Fig. 3b) and covers the southernmost regions of Texas, Florida and inland California (Fig. 3a), areas typically characterized by warm winters.
Broadly, Clusters 5 (n=190,523, ~21%) and 6 (n=115,681, ~12%) experience peaks in avian diversity during the breeding season and troughs during winter, spring, and autumn (Fig. 3b) and cover medium (Cluster 5) to high (Cluster 6) elevation plateaus of the western US (Fig. 3a). Finally, Cluster 7 (n=123,113, ~13%) identifies locations where SR peaks during both the breeding and migration seasons and functional diversity peaks in the breeding season (Fig. 3b). Cluster 7 comprises mostly the deserts of the Southwest and extends into the prairies along the Front Range (Fig. 3a). Remarkably, we find a close agreement between these emergent spatiotemporal clusters and Bird Conservation Regions (BCRs; Fig 3a,c)—independently and qualitatively derived regions that are ecologically distinct in terms of their bird communities, habitat types, and resource management issues (Babcock et al. 1998).