Projected changes in wind assistance under climate change for nocturnally migrating bird populations

1 INTRODUCTION

Seasonal bird migration involves the broad-scale movement of bird populations between breeding and nonbreeding grounds as individuals track seasonal variation in resource availability and environmental conditions (Bauer & Hoye, 2014; Ramenofsky & Wingfield, 2007). Many migratory bird species migrate at night, especially species that rely on powered flight (Alerstam, 2009). The timing of migration for these species tends to coincide with the presence of mild winds or winds that provide tailwind support (Åkesson & Hedenström, 2000; Liechti, 2006; Richardson, 1990; La Sorte, Hochachka, Farnsworth, Sheldon, Fink, et al., 2015a). The broad-scale migration strategies displayed by these species are often dictated by seasonal variation in wind speed and direction. For example, long-distance migrants that breed in eastern North America follow looped migration trajectories that promote tailwind support in the spring and headwind avoidance in the autumn (La Sorte et al., 2014). Current climate models and observations indicate that atmospheric circulation is being affected by global climate change (IPCC, 2013), and these changes have the potential to enhance or interfere with existing migration strategies (La Sorte & Fink, 2017b).

Within North America, there are two prominent atmospheric circulation features that shape broad-scale migration strategies within the region. The first is the prevailing westerlies, a circulation feature that is most prominent within the midlatitudes of North America (Randall, 2015). For migrants that encounter these westerly winds during spring migration, there is evidence that they adjust their heading to avoid drift (Horton et al., 2018). During autumn migration, there is evidence migrants time their departures to coincide with lulls in the prevailing westerlies that occur after the passage of frontal weather systems (Able, 1973; Richardson, 1978).

Current evidence suggests the prevailing westerlies will weaken in the autumn under climate change as temperatures in the Arctic increase (Overland & Wang, 2010; Screen et al., 2018; Serreze & Barry, 2011) and as the polar-front jet stream shifts poleward, a phenomenon supported by both observations (Archer & Caldeira, 2008; Fu & Lin, 2011) and simulations (Simpson, Shaw, & Seager, 2014). In North America, models indicate that the decline in the strength of the prevailing westerlies in the autumn will be more pronounced at the midlatitudes in the eastern portion of the continent (La Sorte & Fink, 2017b). Projections outside the autumn tend to be more complex, with some studies suggesting limited changes (Barnes & Polvani, 2013), while others identify equatorial movements and strengthening of atmospheric circulation (Simpson et al., 2014). In the end, several regional processes spanning the Northern Hemisphere affect atmospheric circulation at the midlatitudes, resulting in complex dynamics that are difficult to model and predict across all seasons (Peings, Cattiaux, Vavrus, & Magnusdottir, 2017).

The second major circulation feature in North America is the Great Plains nocturnal low-level jet stream that generates southerly winds through the center of the continent (Bonner, 1968; Parish, 2017; Shapiro, Fedorovich, & Rahimi, 2016). There is evidence that nocturnal migrants select altitudes during spring migration where the tailwinds provided by the low-level jet are strongest, and select altitudes in the autumn that avoid the headwinds from the low-level jet (Wainwright, Stepanian, & Horton, 2016). Current climate observations (Barandiaran, Wang, & Hilburn, 2013) and projections (Bukovsky, McCrary, Seth, & Mearns, 2017; Cook, Vizy, Launer, & Patricola, 2008; Tang et al., 2017) indicate that the Great Plains nocturnal low-level jet will strengthen and expand northward during this century.

How projected changes in atmospheric circulation will affect broad-scale migration strategies in North America has not been fully considered. A recent study examined projected changes in prevailing winds over the Atlantic Ocean within the transatlantic flyway, a migration corridor that provides passage for a handful of species traversing from North to South America during autumn migration (La Sorte & Fink, 2017b). This study found that the strength of the prevailing westerlies over the Atlantic during autumn migration is projected to decline. Examining the magnitude of projected changes in wind speed, however, does not provide a compelling context for interpreting the ecological implications of climate change. To overcome this limitation, one approach is to translate projected changes in wind to projected changes in wind assistance based on observed migration patterns. Wind assistance is a measure of the flow assistance a migrant receives when flying at specific altitudes within the atmosphere (Kemp et al., 2012). Therefore, we can enhance the quality of our interpretation of the implications of climate change for migratory birds by determining how seasonal patterns of wind assistance are projected to be enhanced or reduced under climate change.

Here, we use reflectivity and radial velocity information compiled from 143 weather surveillance radars (WSR) stations within the contiguous United States (Figure 1) to estimate the daily altitude and density of nocturnal migration during spring and autumn migration during the period from 2013 to 2017. We intersect the migration altitude and density estimates with current estimates of wind speed and direction and with projected changes in wind speed and direction derived from 28 atmospheric–oceanic general circulation models (AOGCMs; Table S1). Based on the summary above, we expect migrants at the 143 WSR stations to be associated with weaker westerly winds during autumn migration, especially within the northeastern portion of the continent, and with stronger southerly winds, especially during spring migration within the center of the continent. Further, we expect these associations to translate to enhanced wind assistance during spring migration within the center of the continent, reduced wind assistance during autumn migration within the center of the continent, and reduced wind assistance during autumn migration in the northeast. By testing these predictions, we provide a continental-level assessment of the projected changes in atmospheric circulation at migration altitudes within the contiguous United States. We then use this information to determine how current patterns of wind assistance for nocturnally migrating bird populations may be affected by climate change. Our aim is to expand our understanding of the developing implications of climate change for North American migratory bird species during a critical portion of their annual life cycle.

2 MATERIALS AND METHODS

2.5 Wind assistance calculation and analysis

We identified for each day during spring (1 March–30 June) and autumn migration (1 August–30 November) for each of the 143 WSR stations over the years 2013–2017 the 5 hPa isobaric levels based on geopotential height that occurred within the migration altitudes. We used this information to calculate the average zonal and meridional wind component and the average zonal and meridional wind anomaly across isobaric levels at each of the 143 WSR stations for each day during spring and autumn migration over the years 2013–2017.

We calculated wind assistance (WA) for each day during spring and autumn migration for each of the 143 WSR stations using the EQ^PartialSpeed formulation from (Kemp et al., 2012):

where y is wind speed, θ is the angle between the wind direction and the preferred flight direction, f is the proportion of compensation, and z is the airspeed of the migrating birds. We calculated θ using the observed wind direction for each day and the observed flight direction during each day. We defined z using a constant airspeed of 12 m/s. This value is considered to be representative of the average airspeed of migrating passerine species (Alerstam, Rosén, Bäckman, Ericson, & Hellgren, 2007; Bruderer & Boldt, 2001), which are the dominant nocturnally migrating species in this system (Dokter et al., 2018). We used the same airspeed for both spring and autumn migration because we wanted WA to measure difference in wind and not in adaptive behavior. We used this equation to calculate two versions of WA. The first version is defined by full compensation (f = 1), and the second by partial compensation (f = 0.5). Under full compensation, migrants completely compensate for lateral drift. Under partial compensation, migrants provide a 50% compensation for lateral drift. The first version represents a conservative estimate and the second a more generous estimate of WA.

We calculated projected WA by adding the zonal and meridional wind component anomalies to the current zonal and meridional wind values, and then recalculated WA using the formulation presented above. We then subtracted the current WA from the projected WA to generate an estimate of the projected change in WA. We determined if these values differed from zero during spring and autumn migration within the three migration flyways using t tests applied to the WA differences averaged across days and years for each WSR station weighted by the average VID during that day.

All analysis was conducted in r, version 3.5.0 (R Development Core Team, 2018). We used the integrate_profile function in the bioRad library to vertically integrate the WSR data (Dokter, Desmet, et al., 2019). We implemented the linear interpolation using the approx function in the stats library. The quantile GAMM analyses were implemented using the gamboost function in the mboost library (Hothorn, Buehlmann, Kneib, Schmid, & Hofner, 2017).

3 RESULTS

The altitude above sea level of nocturnal migration across the 143 WSR stations during spring and autumn migration was highly variable, with the greatest density occurring up to 1 km with altitudes approaching 4 km above sea level (Figure 2a). When examined relative to altitude above ground level, the altitude of nocturnal migration was less variable among stations, with the greatest density occurring below 1 km with altitudes approaching 3 km above ground level (Figure 2b).

Migration directions were generally northward during spring migration (Figure 3a) and southward during autumn migration (Figure 3b). The density of migrants was strongest in the central flyway during spring migration (Figure 3a). The density increased during autumn migration across the central and eastern flyways (Figure 3b). Flight speeds during spring migration were 2.5 m/s faster on average in the central flyway (t₄₁ = 10.7, p < 0.001), 1.5 m/s faster in the eastern flyway (t₆₀ = 11.7, p < 0.001), and did not differ between seasons in the western flyway (−0.08 m/s; t₃₉ = −0.3, p = 0.774).

Wind direction and speed at migration altitudes at the 143 WSR stations presented similar patterns during spring (Figure 3c) and autumn migration (Figure 3d). Westerly winds were most prominent in the western flyway and in the northern portion of the eastern flyway, southerly winds were strongest within the southern portion of the central flyway, northerly winds were strongest along the Pacific coast, and easterly winds were most prominent at the southern end of the Florida peninsula (Figure 3c,d). During spring migration, wind speeds did not differ on average among the three flyways (one-way ANOVA, F_2,140 = 0.7, p = 0.498). During autumn migration, wind speeds were faster on average in the central flyway (3.0 m/s) followed by the western (2.6 m/s) and eastern flyways (2.2 m/s; one-way ANOVA, F_2,140 = 4.9, p = 0.009).

The zonal (east–west) wind anomalies at migration altitudes presented contrasting geographic patterns during spring (Figure 4a) and autumn migration (Figure 4b) across the 143 WSR stations. During spring migration, the westerly winds within the northern portions of the central and eastern flyways (Figure 3c) are projected to increase in strength (Figure 4a), and the easterly winds along the Gulf of Mexico (Figure 3c) are projected to increase in strength (Figure 4a). On average during spring migration, the zonal winds are not projected to change in the central flyway (−0.02 m/s; t₄₁ = −0.8, p = 0.401), are projected to become more westerly in the eastern flyway (0.09 m/s; t₆₀ = 4.0, p < 0.001), and are not projected to change in the western flyway (0.01 m/s; t₃₉ = 0.3, p = 0.748). During autumn migration, the westerly winds in the northern portion of the eastern flyway (Figure 3d) are projected to decrease in strength (Figure 4b), and the westerly winds in the western flyway (Figure 3d) are projected to increase in strength (Figure 4b). On average during autumn migration, the zonal winds are projected to become more easterly in the central flyway (−0.10 m/s; t₄₁ = −8.3, p < 0.001), are projected to become more easterly in the eastern flyway (0.11 m/s; t₆₀ = −8.2, p < 0.001), and are projected to become more westerly in the western flyway (0.08 m/s; t₃₉ = 3.5, p = 0.001).

The meridional (north–south) wind anomalies at migration altitudes presented stronger and more consistent geographic patterns that differed less between spring (Figure 4c) and autumn migration (Figure 4d) across the 143 WSR stations. In general, winds are projected to become more southerly across a broad portion of the region during spring and autumn migration. During spring migration, the southerly winds within the southern portion of the eastern and especially central flyways (Figure 3c) are projected to increase in strength, and the westerly winds along the Atlantic coast (Figure 3c) are projected to have a stronger southerly component (Figure 4c). On average during spring migration, the southerly winds are projected to increase in strength by 0.42 m/s in the central flyway (t₄₁ = 13.3, p < 0.001), are projected to increase in strength by 0.25 m/s in the eastern flyway (t₆₀ = 15.9, p < 0.001), and are projected to increase in strength by 0.06 m/s in the western flyway (t₃₉ = 2.8, p = 0.007). During autumn migration, the southerly winds in the central flyway (Figure 3d) are projected to increase in strength (Figure 4d), and the northerly winds along the Pacific coast (Figure 3d) and the westerly winds along the Atlantic coast (Figure 3d) are projected to have a stronger southerly component (Figure 4d). On average during autumn migration, the southerly winds are projected to increase by 0.34 m/s in the central flyway (t₄₁ = 19.0, p < 0.001), are projected to increase by 0.18 m/s in the eastern flyway (t₆₀ = 11.7, p < 0.001), and are projected to increase by 0.22 m/s in the western flyway (t₃₉ = 7.2, p < 0.001).

Current patterns of WA at migration altitudes during spring migration across the 143 WSR stations based on full compensation were strongest on average within the central flyway (4.4 m/s), especially within the southern portion, followed by the eastern flyway (3.6 m/s), where WA was most pronounced in the central portion, and was weakest within the western flyway (1.1 m/s; Figure 5a). Current patterns of WA at migration altitudes during autumn migration were strongest on average within the eastern flyway (2.3 m/s), followed by the central flyway (1.8 m/s), and the western flyway (1.3 m/s; Figure 5b). When compared between spring and autumn migration, WA was greater on average during spring migration within the central flyway (2.5 m/s; t₄₁ = 7.3, p < 0.001), the eastern flyway (1.3 m/s; t₆₀ = 9.3, p < 0.001), and the western flyway (0.7 m/s; t₃₉ = 3.1, p = 0.004). When contrasted with WA based on partial compensation, WA was slightly stronger and presented spatially similar patterns during spring (Figure S1a) and autumn migration (Figure S1b).

When contrasting current and projected WA at migration altitudes across the 143 WSR stations based on full compensation, WA is generally projected to increase during spring migration (Figure 5c) and remained unchanged or decreased during autumn migration (Figure 5d). During spring migration, WA is projected to increase across broad portions of the central and eastern flyways (Figure 5c). On average during spring migration, WA increased by 0.44 m/s (10.1%) within the central flyway (t₄₁ = 17.4, p < 0.001), increased by 0.32 m/s (9.6%) within the eastern flyway (t₆₀ = 23.7, p < 0.001), and did not change within the western flyway (t₃₉ = 0.91, p = 0.367). During autumn migration, WA declined in the northern portion of the central flyway and to a lesser extent across the southern portion of the eastern flyway (Figure 5d). On average during autumn migration, WA declined by 0.32 m/s (19.3%) within the central flyway (t₄₁ = −13.8, p < 0.001) and declined by 0.17 m/s (6.6%) within the eastern flyway (t₆₀ = −14.1, p < 0.001). There was marginal evidence for a decline (0.08 m/s; 14.0%) within the western flyway (t₃₉ = −2.0, p = 0.050). When comparing spring and autumn migration, changes in WA were greater on average during spring migration within the central flyway (0.12 m/s; t₄₁ = 3.2, p = 0.002), were greater on average during spring migration within the eastern flyway (0.15 m/s; t₆₀ = 8.2, p < 0.001), and were greater on average during autumn migration within the western flyway (0.07 m/s; t₃₉ = −2.4, p = 0.024). When contrasted with WA based on partial compensation, very similar results were generated during spring (Figure S1c) and autumn migration (Figure S1d).

4 DISCUSSION

The climate projections examined in this study indicate that atmospheric circulation at migration altitudes in North America is projected to change during this century. The prevailing westerlies are projected to increase in strength during spring migration and decrease in strength to a lesser degree during autumn migration. Southerly winds will increase in strength across the continent during both spring and autumn migration, with the strongest gains occurring in the center of the continent. When examining these projections within the context of WA, our findings suggest current patterns of WA will be enhanced during spring migration and reduced during autumn migration. Specifically, tailwind support during spring migration will increase throughout the central and eastern portions of the continent, and headwind resistance during autumn migration will increase in the central and eastern portions of the continent. These findings suggest that, within the region's dominant migration corridor, the speed and efficiency of nocturnal migration will increase in the spring and decline in the autumn.

If current patterns of nocturnal migration are maintained, our findings indicate that migrants will encounter roughly a 10% increase in WA during spring migration and 19% decrease in WA during autumn migration within the center of the continent. The region's dominant migration flyway occurs in the central portion of the continent, especially during spring migration, and an increase in WA will likely benefit a large number of individuals across many species. These benefits could be improved as migrants adjust existing migration altitudes, routes, or timing to take advantage of additional tailwind support (Wainwright et al., 2016). Current evidence indicates that strengthening tailwinds often results in reduced airspeeds (Hedenström, Alerstam, Green, & Gudmundsson, 2002; Nilsson, Bäckman, & Alerstam, 2014), suggesting the primary outcome of enhanced WA may be energy gains through reduced airspeed. In total, our findings indicate that current associations with WA during spring migration are likely to be maintained with the potential for broad-scale improvements during this century.

During autumn migration, in contrast, our findings indicate that the current associations with WA in the eastern and especially central portions of the continent are likely to become degraded through increased headwind resistance. Migrants that pass through these regions in the autumn, however, may have the option to adjust current migration altitudes, routes, or timing to avoid additional headwinds (Wainwright et al., 2016). This possibility is supported by our finding that migrants were associated with supportive winds during both spring and autumn migration even though the prevailing winds in the autumn are largely opposing. In the end, to better understand the implications of climate change for migratory birds, it would be beneficial to explore how migratory behavior is affected by changing WA and the consequences for time and energy expenditures.

Our results show strong seasonal differences in the broad-scale migration strategies displayed by these species. Migration was most intense within the very center of the continent during spring migration. In the autumn, migration intensity was greatest throughout the central and especially eastern portions of the continent. These seasonal differences reflect the addition of large juvenile populations in the autumn (Dokter et al., 2018) and the use of clockwise looped migration strategies where migration routes shift from the center to the eastern portion of the continent in the autumn (La Sorte, Fink, Hochachka, & Kelling, 2016a). The use of looped strategies promotes associations with tailwinds in the spring and the avoidance of headwinds in the autumn (La Sorte et al., 2014). Our findings suggest that looped strategies will remain relevant under climate change through enhanced tailwind support within the center of the continent in the spring and the need to avoid stronger headwinds within the center of the continent in the autumn.

When examining seasonal migration altitudes, our findings indicate that migrants select altitudes relative to ground level and not based on atmospheric pressure above sea level, a pattern also documented in the northeastern United States (La Sorte, Hochachka, Farnsworth, Sheldon, Doren, et al., 2015b) and Europe (Bruderer, Peter, & Korner-Nievergelt, 2018). There is evidence that nocturnal migrants will gain altitude until they first encounter favorable winds, even when more favorable conditions may exist at higher altitudes (Dokter, Shamoun-Baranes, Kemp, Tijm, & Holleman, 2013; Kemp, Shamoun-Baranes, Dokter, van Loon, & Bouten, 2013; Mateos-Rodríguez & Liechti, 2012). There is also evidence that migrants will climb in altitude when crossing mountain ranges (Bruderer et al., 2018). When considered in combination with our findings, altitude selection during nocturnal migration in North America appears to be driven by the quality of WA within a limited distance above ground level.

When considering the broader implications of climate change for migratory birds, our finding of enhanced WA during spring migration may contain additional consequences. First, enhanced WA may allow migrants to compensate under climate change to advancements in the spring phenology of ecological productivity (Badeck et al., 2004). These phenological changes have broadly affected the timing of spring migration, with short-distance migrants responding more readily than long-distance migrants (Usui, Butchart, & Phillimore, 2017). The use of more rigid endogenous cues to initiate spring departures is thought to place long-distance migrants at a particular disadvantage during spring migration (Åkesson et al., 2017). Our findings suggest enhanced WA may act as a mitigating factor by promoting faster migration speeds in the spring for long-distance migrants. En route variation in migratory behavior tends to be greater for long-distance migrants, especially when considering migration speed (La Sorte & Fink, 2017a), and some migrants have advanced the timing of spring migration by associating with enhanced tailwind support (Sinelschikova, Kosarev, Panov, & Baushev, 2007). Thus, when arriving in North America, long-distance migrants may have the capacity to use enhanced WA to advance arrival times and reduce phenological mismatches. However, adjustments in migratory behavior alone may not be sufficient to compensate for large phenological changes (Schmaljohann & Both, 2017), suggesting the ability of migrants to use enhanced WA to counter these disruptions may be limited.

Another implication of enhanced WA is that long-distance migrants may be able to compensate for the large-scale distributional shifts that are occurring under climate change. There is evidence birds and other taxa over the past several decades have shifted their geographic distributions in the Northern Hemisphere poleward to cooler latitudes (Chen, Hill, Ohlemüller, Roy, & Thomas, 2011). The increasing separation between the breeding and nonbreeding grounds has the potential to increase the time, energy, and risk associated with the migration journey (Howard et al., 2018). Our findings suggest enhanced WA in the spring may allow migrants to more efficiently traverse these longer distances. Conversely, stronger opposing winds during autumn migration may increase the challenges associated with extended autumn migration journeys.

Based on current climate models (Bukovsky et al., 2017) and observations (Barandiaran et al., 2013), the Great Plains low-level jet is expected to strengthen and expand northward. One consequence of this expansion is that spring precipitation in the center of the continent is expected to shift northward. The development of enhanced WA during spring migration may therefore coincide with drier conditions in the south and wetter conditions in the north, which may affect the quality of stopover habitats within the region.

When our findings are considered in combination with other possible outcomes of global change, it is apparent that any positive implications of enhanced WA during spring migration will occur in combination with other less beneficial effects. Enhanced flight speeds in the spring will reduce the time, energy, and risk associated with the journey and improve the chances of survival. These benefits, however, will occur in combination with other factors that are likely to negatively affect survival and reproduction. This includes the loss of suitable habitat on the nonbreeding grounds through land-use change (La Sorte et al., 2017) and the potential for increasing ecological disruptions through the formation of novel climates (La Sorte, Fink, & Johnston, 2018). Both of these processes can affect the quality of stopover habitat, which may require longer or more frequent stopover visits, extending the overall duration of spring migration and reversing any benefits acquired through enhanced WA (Schmaljohann & Both, 2017).

Based on climate models and observations, the frequency and intensity of climate extremes at the midlatitudes in the Northern Hemisphere are expected to increase (Hall, Erdélyi, Hanna, Jones, & Scaife, 2014; Trouet, Babst, & Meko, 2018). Extreme weather events can affect the timing of spring migration and the quality of breeding resources (La Sorte, Hochachka, Farnsworth, Dhondt, & Sheldon, 2016b). Climate models and observations (Knutson et al., 2010) also point toward more intense and frequent Atlantic hurricanes, which has the potential to severely disrupt spring migration (Dionne, Maurice, Gauthier, & Shaffer, 2008). In contrast to changes in WA that will develop over many decades, extreme weather events are immediate and, when combined with long-term climate trends, can result in significant ecological consequences (Harris et al., 2018).

In total, we show that our understanding of the implications of climate change for a natural system can be enhanced by placing climate change projections within a relevant ecological context informed through empirical observations. In this case, we used data from a large collection of WSR stations to generate vertical estimates of nocturnal migration density across North America. This information allowed us to determine how current patterns of WA will be affected by projected changes in wind speed and direction. Our findings point toward a beneficial outcome developing during spring migration and a detrimental outcome developing during autumn migration. Thus, based on projected changes in WA, our findings suggest that climate change will result in both positive and negative implications for nocturnal migratory bird populations in North America. However, these outcomes will develop over several decades and do not consider the implications of other global change processes, and these outcomes are based on migratory species maintaining existing behaviors and strategies. Our findings therefore provide baseline projections for one facet of climate change that will affect migratory birds in flight during spring and autumn migration. It would be valuable to determine how migratory birds are responding to broad-scale changes in WA, and how these changes will interact with other factors to affect the long-term survival of migratory bird populations.