Field methods
We captured 24 whimbrels between 2008 and 2018. Birds were captured on migration staging sites along the lower Delmarva Peninsula in Virginia, USA (n = 6) (37.398° N, 75.865° W), along the coast of Georgia, USA (n = 5) (31.148° N, 81.379° W), along the Acadian Peninsula in New Brunswick, Canada (n = 3) (47.973° N, 64.509° W) as well as on the nesting ground near the Mackenzie River, Northwest Territories, Canada (n = 10) (69.372° N, 134.894° W). All birds were aged as adults by plumage26, 27 and were banded with United States Geological Survey tarsal bands and coded leg flags. Sex of captured birds was not determined.
We fitted all birds with satellite transmitters called Platform Transmitter Terminals (PTTs) using a modification of the leg-loop harness28, 29. Instead of elastic cord, we used Teflon® ribbon (Bally Ribbon Mills, Bally, Pennsylvania, USA) that was fastened with brass rivets or crimps30. We glued transmitters to a larger square of neoprene to elevate it above the body and prevent the bird from preening feathers over the solar panels. The transmitter package was below 3% of body mass (measured at the time of deployment,(bar{x}) = 484.5 ± 17.1) for all individuals tracked in this study. The PTTs used in this study were 9.5 g PTT-100 (n = 14) or 5.0 g PTT-100 (n = 10) solar-powered units produced by Microwave Telemetry, Inc. (Columbia, Maryland, USA).
Tracking
Birds were located using satellites of the National Oceanic and Atmospheric Administration and the European Organization for the Exploitation of Meteorological Satellites with onboard tracking equipment operated by Collecte Localisation Satellites (CLS America, Inc., Largo, Maryland, USA)31. Transmitters were programmed to operate with a duty cycle of 24 h off and 5 h on (n = 9) or 48 h off and 10 h on (n = 15) and collected 1–34 ((bar{x}) = 5.48 ± 0.07) locations per cycle. Locations in latitude and longitude decimal degrees, date, time, and location error were received from CLS America within 24 h of satellite contact with PTTs. Locations were estimated by the Advanced Research and Global Observation Satellite (ARGOS) system (www.Argos-system.org), which uses a Doppler shift in signal frequency and calculates a probability distribution within which the estimate lies. The standard deviation of this distribution gives an estimate of the location accuracy and assigns it to a “location class” (LC): LC3 = < 150 m, LC2 = 150–350 m, LC1 = 350–1000 m, LC0 > 1000 m, LCA = location based on 3 messages and has no accuracy estimate, LCB = location based on 2 messages and has no accuracy estimate, and LCZ = location process failed. We used LC classes 1–3 to determine whimbrel locations.
Migration pathways
We used tracking data to delineate fall migration pathways and, though migration duration can include fueling at breeding territories32, we defined migration duration as the time between departure from the breeding grounds and arrival on winter territory. We identified the source population for all individuals included in this study either by capture on the breeding grounds (n = 10) or by capture within migratory staging sites and tracking birds to the breeding grounds (n = 14). Birds were either from the Mackenzie Delta (n = 13) or Hudson Bay (n = 11) breeding populations. We assessed departure and arrival when birds moved away from or settled into stationary breeding and winter territories respectively. Departure was abrupt and we recorded no “false starts” of birds leaving breeding areas and then returning before resuming migration. We present a stylized map of migration routes that was drawn by hand using the collection of flights recorded to provide a broad overview of routes relative to the distribution of storms.
Trans-atlantic flights
We used tracking data to delineate migration pathways across the Atlantic Ocean (from coast of North America to coast of South America). Most birds departed from coastal staging sites and we considered the last staging location prior to crossing the Atlantic the terminal staging area. Several birds departed from inland locations on James Bay. We only consider the segment of the latter flights that occur over the ocean. We consider the duration of transoceanic flights to be the time interval between emerging from the coast of North America and arriving along the coast of South America. In cases where departure and arrival times occurred outside the radio transmitter’s duty cycle, we drew a straight-line between the last known location on land for departures or the first known location on land for arrivals and the nearest location over water and measured the distance between the in-flight point and the coastline along the line. We then used the mean overall speed between in-flight points for all birds ((bar{x}) = 14.8 ± 0.4 m/s, n = 40) to interpolate the leaving or arrival times. We consider the flight length to be the sum of the distance between consecutive locations along the path taken between the site of emergence along the coast of North America and the site of landfall along the northern coast of South America.
Exposure to tropical cyclones
We examined the distribution of tropical cyclones throughout the Atlantic Ocean using position records (1961–2018) within the revised Atlantic hurricane best tracks from the National Hurricane Center (https://www.nhc.noaa.gov/data/#hurdat), known as the Atlantic HURDAT233. We restricted our analyses to storms classified as tropical depressions or above and HURDAT data collected since 1961, when satellites were first used to monitor tropical cyclone activity34. The database contains the storm category (Saffir Simpson Scale), wind speed (mph) and coordinates recorded for six-hour intervals during the period that each storm existed using standard six-hour intervals which allows for weighting of the storms according to their lifespans and estimating the distribution of probability density. We selected storms (N = 590) that were active between 15 July and 30 November to coincide with whimbrel migration through the region. We mapped all storm observation points (N = 17,637) using a kernel density estimator (KDE) method35 with the “ks” package36 in program R37. We used the normal (or Gaussian) kernel and a smooth cross-validation bandwidth selector38 to map 50% kernel densities. We considered the 50% KDE to be the area of highest storm occurrence and estimated exposure to this region by overlaying whimbrel tracks on the KDE polygon and measuring each whimbrel’s time within the area. Because the first and last points within the polygon occurred when the bird’s transmitter first transmitted the bird’s location within and outside the polygon, rather than when the bird first entered and exited the polygon, we measured the distance between the first point inside the polygon and the previous point outside the polygon and used the mean overall speed between in-flight points for all birds (,(bar{x}) = 14.8 ± 0.4 m/s, n = 40) to interpolate the time that the bird entered the polygon. We used the same method to calculate the time that the bird left the polygon using the last point within the polygon and next point outside the polygon.
Encounters with tropical cyclones
We documented encounters between whimbrels and tropical cyclones within the Atlantic Basin by overlaying migration tracks for individual birds on archives of storm tracks within HURDAT2 for the period (2008–2019) of the tracking study. We considered a whimbrel-storm encounter to have occurred when bird tracks intersected storm tracks during the same time period. For grounded birds, we considered an encounter to have occurred when a storm track moved over the ground position of a bird. For each encounter, we recorded the coordinate of the encounter and the storm intensity. Storm intensities were classified as tropical depressions, (≤ 38 mph), tropical storms (39–73 mph), category 1 hurricane (74–95 mph), category 2 hurricanes (96–110 mph), category 3 hurricanes (111–129 mph), category 4 hurricanes (130–156 mph), and category 5 hurricanes (≥ 157 mph) according to the Saffir–Simpson Hurricane Wind Scale39.
We examined the post-encounter track of birds to categorize the response of birds including none, detour or grounding. We considered birds to exhibit no response to the storm encounter if the migration trajectory was unchanged during or shortly following a storm encounter. We considered birds to have taken a detour in response to a storm encounter if the migration trajectory followed over the previous day was deflected by > 20° during or shortly following an encounter. We considered birds to have grounded if they landed on an island following a storm encounter.
Statistics
We used mixed-effects logistic regressions (R3.6.2: R Core Team 2019) to compare the likelihood of storm encounters between whimbrel populations using tracks as replicate samples. We initially fit models using whimbrel identity and year as random intercepts to account for potential lack of independence for journeys made by the same individuals and journeys made within the same year, but inclusion of bird identity as a random intercept resulted in a singular fit so this variable was excluded from further analysis. We then compared models with year as a random intercept and no fixed effects, year as a random intercept and breeding population (Mackenzie Delta vs Hudson Bay) as a fixed effect, year as a random intercept and journey number (1st, 2nd, or 3rd journey) as a fixed effect, and year as a random intercept with breeding population and journey number as fixed effects. We used Akaike’s information criterion for small sample size (AICc) and selected the model with the lowest AICc score as the best-supported model if no other model was within 2 ΔAICc after removing models with uninformative parameters40. Several birds made more than one transoceanic crossing in different years and we consider these to be independent samples. We used two-tailed t-tests to compare migration lengths and duration between routes. We used g-tests with Yates correction to make frequency comparisons.
Data and ethics statement
This study was conducted in compliance with ARRIVE guidelines. Data used in this manuscript are unique and have not been submitted for publication elsewhere. The authors claim no conflict of interest. This project was reviewed and approved by the William & Mary Institutional Animal Care and Use Committee protocol IACUC-2017-04-18-12065 of The College of William and Mary, Environment Canada Animal Care Committee protocols EC-PN-12-006, EC-PN-13-006, EC-PN-14-006, Mount Allison University Animal Care Committee protocol 15-14, and the Government of the Northwest Territories Wildlife Care Committee protocol NWTWCC2014-007. All Methods were performed in accordance with the relevant guidelines and regulations.
Source: Ecology - nature.com
