Estimating apparent survival of songbirds crossing the Gulf of Mexico during autumn migration

1. Introduction

Migratory animals, regardless of taxa, are declining [1]. Birds, specifically, face growing challenges during migration, including loss of stopover habitat [2], changes in climate [3] and potential collisions with tall, man-made structures like wind turbines [4] and skyscrapers [5]. Even without these hazards, migration is a challenging period in the annual cycle of many bird species. Migrating birds must find suitable stopover habitat, avoid predators, quickly replenish fat stores and make appropriate decisions in response to weather conditions, all while advancing towards their destinations [6]. Moreover, large, inhospitable geographical features figure prominently in the movement ecology of almost all long-distance migratory birds throughout the world (e.g. Sahara Desert [7], North Sea [8] and Atlantic Ocean [9]). A geographical feature that may present an obstacle for migration of many Nearctic–Neotropical migratory birds is the Gulf of Mexico [10]. Reports of landbirds resting on offshore oil and gas platforms [11], washing ashore in large beach kills [12] and being found in shark stomach contents [13] support that crossing large water bodies such as the Gulf of Mexico poses a risk. However, accepting the potential risks, large geographical features, such as the Gulf of Mexico, may provide corridors allowing individuals to more quickly and safely reach their destination [14].

Although migration is thought to be the most hazardous phase of the avian annual cycle [12,15,16], robust estimates of survival during migration are few [15]. The lack of season-specific survival estimates is primarily owing to the logistical and technological difficulties of gathering data needed to produce robust estimates, especially for small birds during the migratory period. Furthermore, mortality during migration is probably not evenly distributed along the migratory route. Large geographical features like the Gulf of Mexico may have a disproportionate impact on survival. While knowledge of average season-specific survival rates during the annual cycle can assist in conserving migratory bird populations, identification of specific migratory events that exert the greatest influence on survival will focus conservation efforts on areas of greatest importance. Indeed, information is needed to better understand when and where populations become limited during migration, as well as to inform management and conservation efforts [2].

The role that the Gulf of Mexico plays in the survival of small land birds during autumn migration is expected to be dynamic. The greater than 1000 km crossing between the northern Gulf Coast and the Yucatan Peninsula (YP) in autumn requires, on average, 22 h of sustained flight for land birds such as Swainson's thrush (Catharus ustulatus) [17]. Relative to spring, autumn crossing poses unique challenges, including greater likelihood of encountering unfavourable winds [18] and young birds crossing for the first time. Extrinsic factors such as weather and intrinsic factors such as age, sex and fat stores influence survival when crossing large geographical features [7,17]. The ability to determine the favourability of winds and respond accordingly should be adaptive [19,20]; both wind speed and direction should strongly influence whether an individual attempts to cross a large geographical feature [17]. The fat stores that individual birds accumulate before departure are an important determinant of stopover duration for spring- and autumn-migrating birds [21–23], influence departure decisions [17,20,24–26] and surely influence the duration of their migratory flights [27]. Age may also contribute to an individual's probability of survival en route across large geographical features such as the Gulf of Mexico because young birds crossing for the first time may have a lower survival probability than older, experienced individuals [28,29]. Finally, the sex of an individual may influence survival if males experience greater pressure to return early to wintering grounds [30] or because sexually dimorphic wing shape influences flight performance in relation to atmospheric conditions [31]. From a conservation perspective, it is important to consider not only relationships between survival and extrinsic variables, but also how conditions for crossing the Gulf of Mexico varied in the past and are forecast to change in the future, as climate change alters the frequency and strength of favourable winds [32]. Assessing whether these changes will be beneficial or detrimental to migratory birds requires a better understanding of how birds respond to and are influenced by wind when negotiating large geographical features.

We estimated the apparent survival of Swainson's thrushes as they attempted to cross the Gulf of Mexico during autumn migration. Swainson's thrushes are well suited for examining trans-Gulf survival as most individuals probably cross on their way to wintering grounds in South America and are large enough to carry a small transmitter [17]. Additionally, the species' migratory behaviour has been studied for decades [17,28,33–35]. Previous research has shown that Swainson's thrushes typically fly for 7 h each night during overland flights [33,36] and an average of 22 h over the Gulf [17]. Thus, crossing of the Gulf of Mexico requires a prolonged flight three times greater than their average migratory flight over land. We were able to directly monitor the departure of radio-tagged Swainson's thrushes leaving the Alabama coast (northern coast of the Gulf of Mexico) and their arrival at the YP (southern coast of the Gulf of Mexico) using a network of automated radio-telemetry systems (ARTS). We set out to address: (i) how extrinsic (weather and day of year) and intrinsic (fat, age and sex) factors influence survival of Swainson's thrushes crossing the Gulf of Mexico during autumn passage, and (ii) how the favourability of weather conditions for crossing the Gulf of Mexico has changed over the past 25 years.

2. Material and methods

(b) Automated tracking data

An ARTS network consisting of four receiving towers was operated on the FTM Peninsula to estimate the departure date, time and direction for each bird (figure 1a) and seven ARTS across the northern tip of the YP to record the arrival in the Yucatan (figure 1b). The precise locations of the towers on the FTM Peninsula used in this study changed slightly among years depending on site access, although two towers were always within 1.8 km of the banding station. We were able to detect all departures from the study area.

• Download figure
• Open in new tab
• Download PowerPoint

To detect the arrival of birds in Mexico, we established a ‘telemetry fence’ along the northern coast of the YP using seven ARTS (figure 1b), each equipped with two high-gain stacked Yagi antennas. The high-gain antennas were oriented towards the east (90°) and west (270°). All antennas were erected above the vegetation and other surrounding objects approximately 10.6 ± 2.4 m above ground level to improve the detection capabilities of our tracking towers. The flat topography of the northern YP favoured the detection of arriving birds by our tracking towers (assuming transmitters were still functioning); and the fact that, in most instances, arriving birds were detected almost simultaneously on multiple towers suggests we were able to detect all radio-tagged birds passing through the northern YP. Towers in the YP were separated by an average 42 km. The maximum range of detection of a radio-tagged bird is largely determined by the bird's altitude (in relation to offsetting the effect of the Earth's curvature to achieve line-of-sight contact), transmitter power, receiving antenna gain and orientation of the transmitting antenna with respect to the receiving antenna. Theoretically, a bird flying at an altitude of 500 m above sea level could be detected at a range of 80 km using our high-gain antennas, while a bird flying at an altitude of 1000 m could be detected at a range of approximately 115 km [17]. While we do not know the exact altitude of birds arriving in the YP, previous studies suggest that most songbirds migrate at altitudes greater than 500 m above ground level [41]. The ARTS in the Yucatan experienced more periods in which they were not operating compared with Alabama, primarily owing to the occurrence of hurricanes. The percentage of time the automated receiving units were operating was: 2009—100.0%; 2010—97.4%; 2011—76.7%; 2012—94.2%; 2013—98.2%; 2014—100.0%. While some towers may not have been operating, 44% of birds detected in the Yucatan were detected on multiple towers. On average, a detected bird was detected on 1.91 towers and was detected for an average of 54.9 min (minimum 15.0 min–maximum 126.0 min). Also, most of the time when towers were not functional was early in season, while thrush migration was generally later in the season. Further information on the criteria used to identify the arrival of radio-transmittered birds in the YP can be found in the electronic supplementary material [17,42]. Information on calculating the departure directions can be found in the electronic supplementary material [25]. The detection history for these individuals is available at https://doi.org/10.13012/B2IDB-2805211_V1.

3. Results

We captured and radio-tagged 139 Swainson's thrushes (16–32 individuals per year) from 13 September to 28 October 2009–2014. Of these, there were 70 males and 58 females (11 of unknown sex); 91 were classified as young (HY) and 47 as adult (AHY) (one unknown). Fat scores ranged from 0 to 5 (mean = 3.08, s.d. = 1.78), with 55% (n = 76) classified as 4 or 5. Most birds (n = 112; 81%) departed FTM the same day of capture, and 9% (n = 13) stopped over for 3 or more days; 43 birds (31%) were redetected in the Yucatan.

Our AIC approach found the best model for detection (resighting) probability (p) included additive effects of capture date and fat score (table 1) with probability declining with capture date and increasing with fat score, and estimated detection probability per 6 h period ranged from 0.79 to 0.95, with an estimated probability of 0.91 (±0.01 s.e.) for a bird captured on 2 October with a fat score of 3. The best-fitting structure for our temporary emigration parameters (γ″ and γ′) included no covariates, with an estimated probability of emigrating from our system (γ″) of 0.22 (±0.06 s.e.), and estimated probability of staying away given prior emigration (γ′) of 0.78 (±0.06 s.e.).

Both intrinsic and extrinsic factors were well supported as predictors of survival (φ) for thrushes attempting to cross the Gulf of Mexico. The best-fitting model included positive influences of fat score and wind profit with the wind profit effect specific to south-departing individuals, and a quadratic effect of departure date (table 1). Estimated survival was highest for fat, south-departing birds on nights with favourable wind profit (positive values denote favourable tailwinds and negative values indicate headwinds; figure 2), and peaked around 8 October (figure 3). The estimated apparent survival for a south-departing individual leaving on 8 October (the average departure date) with a fat score of 5 was 0.80 (±0.09 s.e.) under average wind conditions (wind profit = −0.7 m s⁻¹); 0.91 (±0.06 s.e.) under favourable wind conditions (upper quartile for wind profit = 0.9 m s⁻¹) and 0.98 (±0.02 s.e.) under the maximum observed wind profit (4.4 m s⁻¹). To illustrate the role of fat score, the apparent survival for an individual with a fat score of 1 was 0.29 (±0.16 s.e.), 0.49 (±0.21 s.e.) and 0.87 (±0.17 s.e.) under average, favourable and maximum wind profit, respectively. We found little support for effects of age, sex or other weather conditions on apparent survival.

• Download figure
• Open in new tab
• Download PowerPoint

• Download figure
• Open in new tab
• Download PowerPoint

Wind profit varied considerably within and among years. Across all years (1990–2014), wind profit during September and October ranged from −25.0 to 11.5 m s⁻¹ with a mean of −1.8 (s.d. = 4.1). Annual mean values for wind profit ranged from −4.3 (1998) to −0.1 m s⁻¹ (2005). Similarly, there was considerable variation within and among years in the number and distribution of nights when conditions would be expected to be favourable (φ ≥ 0.80) for trans-Gulf migration (figure 4). The was no evidence that the proportion of favourable nights for crossing the Gulf changed differently over time for birds in different fat classes (fat × year: F_5,138 = 0.56, p = 0.732), so we dropped the interaction to evaluate the influence of the main effects. The proportion of favourable nights for crossing within each autumn migration season increased from 1990 to 2014 (electronic supplementary material, figure S2; year: F_1,143 = 4.04, p = 0.046; β = 0.0059, 95% confidence interval (CI) 0.0001–0.0118), although there was no evidence of a quadratic effect (year²: F_1,142 = 1.52, p = 0.219).

• Download figure
• Open in new tab
• Download PowerPoint

4. Discussion

The risk to Swainson's thrushes crossing the Gulf of Mexico in autumn is closely tied to the individual's fat stores, wind direction and wind speed. Under ideal conditions, crossing poses little risk to the individual, but under less than favourable conditions, the crossing might be one of the riskiest events of autumn migration and the annual cycle. Survival is unlikely if birds attempt to cross under strong headwinds, regardless of their fat stores. For lean birds, survival remains low even under favourable wind conditions (wind profit = 0.9). Of the 76 lean birds we tagged, only six departed towards the south, out over the Gulf, and only one was subsequently detected in the YP. That individual, however, spent 9 days in the capture area at FTM prior to departure and probably gained fat [28] before departing across the Gulf. While crossing the Gulf of Mexico under unfavourable conditions may be one of the riskiest events during autumn migration, birds presumably have evolved departure decision rules and sensitivity to environmental cues that maximize the probability of survival by coordinating trans-Gulf departures with favourable intrinsic and extrinsic conditions [18].

The apparent survival of thrushes crossing the Gulf was driven by factors birds have some capacity to manage: wind profit, fat load and day of year. Several studies have found that birds prefer to migrate when winds are in their favour [17,19,51–53]. Fat is the primary source of energy for migrating birds [21]; given the long journey in terms of distance, time and likelihood of encountering variable conditions, it follows that individuals with more fat are more likely to survive the crossing. The importance of day of the year was surprising and is probably owing to a combination of several factors leading to greater survival in early October compared with earlier and later in the season. Early October may be the optimal time to migrate because the amount of fruit available to thrushes and other migrants declines throughout autumn [54], while winds generally become more favourable later in the season. Taken together, these factors may define a narrow period in which individuals that leave in early- to mid-October optimize available food resources and the likelihood of encountering favourable winds. Although we accounted for fat, which would be related to fruit abundance and wind, it is possible that there are other aspects of bird health and environmental conditions that were not incorporated into our metrics. Alternatively, there may be other unmeasured intrinsic or extrinsic variables that happen to peak during this period.

This study attempts to quantify the risk small birds face when crossing a large geographical feature, but it is important to emphasize that apparent survival represents a minimal survival estimate. Some thrushes may have bypassed the YP and migrated through Veracruz or Cuba. However, Swainson's thrushes crossing to Veracruz would face an over-water flight of approximately 1500 km from FTM (1.5× greater than crossing to the Yucatan), which would probably reduce survival. Swainson's thrushes are also an uncommon migratory species in Cuba [55], which reduces the likelihood of bypassing the YP to the east. Although migrating across large bodies of water is often considered riskier than over land movement, our minimum estimates of survival for birds crossing the Gulf of Mexico under favourable conditions (φ = 0.9 with fat score greater than or equal to 4 and wind profit greater than 2) argue that this is not necessarily the case. As opposed to a barrier to migration, the Gulf of Mexico may represent a direct corridor for individuals with favourable departure conditions (i.e. high fuel loads and tail winds) to reach the wintering grounds quickly and safely, similar to what has been suggested for northern wheatears (Oenanthe oenanthe) crossing the Atlantic [56].

The decision to depart for the YP and the factors that influence whether an individual survives the journey may require individuals to hedge their bets. Swainson's thrushes are territorial on the wintering grounds [57], which may incentivize them towards early departure. The pressure to arrive on wintering grounds early represents an external force that presumably compels birds to select sub-optimal conditions to cross. Birds may optimize this time–risk trade-off by managing factors over which they exert some measure of control. The overall risk profile associated with crossing the Gulf may be mediated largely by fat acquisition, which makes food availability critical to managing the current risk landscape. For example, increasing fat stores beyond theoretical necessity [32] serves as a buffer that offsets risk posed by less than favourable winds and of course, carrying more fat can also be costly [58]. Our models show that a fat 3 thrush crossing with a 4 m s⁻¹ wind profit experiences similar apparent survival as a fat 5 thrush crossing with a 2 m s⁻¹ wind profit. By increasing their fat stores (a factor they can control), birds in effect increase the number of potential nights with favourable conditions, without excessively compromising risk. In addition, the early arrival on the breeding grounds may facilitate the acquisition of a high-quality wintering territory.

Among all phases of the avian annual cycle, previous studies have found the migratory period to be most limiting [59], and our analysis suggests that surviving a Gulf of Mexico crossing hinges on the availability of wind assistance and adequate fat stores. Climate change may affect wind profit [3], which varied widely within and among years during the study. Optimal wind profits in this study were identified as winds blowing towards the south and are generally associated with weather fronts [60]. Weather fronts out of the north (Nortes) produce strong south-directed winds favourable to birds crossing the Gulf. Pérez and colleagues [32] investigated how Nortes might change under projected climate change scenarios. They predicted an increase in the number of Nortes with strong winds favourable to migrating birds, albeit with shorter lifespans, lasting only a couple of days. Over the 25 years, we investigated change in favourable winds, which are probably associated with Nortes; we found an increase in frequency in favourable winds. However, it should be noted that the increase in the proportion of favourable nights (electronic supplementary material, figure S2) represents only about two more favourable nights across the 25 year period. This pattern, however, suggests a possible benefit of climate-induced changes in wind patterns for trans-Gulf crossing in autumn, albeit a more restricted window of favourability. If the duration of favourable weather systems shortens [32], birds will be pressed to deposit adequate fat stores and to make more judicious departure decisions.

Finally, our findings have clear conservation implications. Management of habitat that allows migratory birds to maintain or increase fat stores is needed. Migrants attempt to maximize fuelling opportunities during passage [61], but stopover habitats vary in quality [22,62]. Creation and protection of new stopover habitats, through restoration and land acquisition as well as management of existing habitats, should focus on vegetation types that increase the availability and diversity of food resources for migrating birds [63]. Our focal species, the Swainson's thrush, has declined throughout North America from 1966 to 2015 at an annual rate of 1.25% (95% CI −2.12 to −0.49) in the Eastern Region of the Breeding Bird Survey. This downward trend is reflective of many of the bird species migrating through our study area [64]. Clearly, thrushes and other species would benefit from conservation efforts directed towards stopover habitats, but protection of natural habitats is often at odds with urban development and other anthropogenic land-use changes common in coastal areas [65]. Although we focused on the Gulf of Mexico, long-distance migratory birds invariably encounter other large geographical features (e.g. oceans, deserts and mountain ranges) during passage and will probably rely on stopover areas where access to food resources and the ability to increase fat stores will create flexibility for minimizing risk in relation to negotiating these features.

Ethics

The capturing and banding of birds in Alabama was approved by the University of Southern Mississippi and the University of Illinois IACUC (no. 15154), the Bon Secour National Wildlife Refuge and conducted under the Federal Bird Banding Permit (21221).

Data accessibility

The capture history and weather data are archived at: https://databank.illinois.edu/; the capture history: https://doi.org/10.13012/B2IDB-2805211_V1; and the weather data: https://doi.org/10.13012/B2IDB-1311182_V1.

Authors' contributions

M.P.W. collected data and co-wrote the manuscript. T.J.B. conducted the mark-recapture analysis and co-wrote the manuscript. J.D. collected data and designed the research. T.J.Z. collected and analysed data. R.H.D. conducted the analysis of weather variables. A.C.-M. collected and analysed data. R.B. developed the weather model. F.R.M. established the long-term banding station and collected data. All authors contributed substantively to drafts of the manuscript.

Competing interests

We have no competing interests.

Funding

This work was supported by the National Science Foundation (NSF) (IOS awards 1146832, 1147096 and 1145952) and National Geographic Society's Committee on Research and Exploration (award 8971-11). T.J.Z. was supported by a NSF GK-12 Fellowship awarded to the University of Southern Mississippi (0947944), the Mississippi-Alabama Sea Grant Consortium (NA14OAR4170089) and the NOAA RESTORE Act Science Program (NA17NOS4510092).