Skip to main content
Dryad logo

Migration tactics and connectivity of a Nearctic-Neotropical migratory shorebird

Citation

Herbert, John (2022), Migration tactics and connectivity of a Nearctic-Neotropical migratory shorebird, Dryad, Dataset, https://doi.org/10.5061/dryad.2rbnzs7q3

Abstract

During long-distance spring migrations, birds may rest and refuel at numerous stopover sites while minimizing the time to reach the breeding grounds. If habitat is limited along the migration route, pre-breeding birds optimize flight range by having longer stopovers at higher quality sites compared to poorer quality sites. Stopover duration also depends on distance remaining to breeding grounds, ecological barriers, and individual characteristics.

We assessed spring migration tactics and connectivity of a Nearctic-Neotropical migratory shorebird, the semipalmated sandpiper (Calidris pusilla), at two sites with known relative habitat quality on the Northern Gulf of Mexico (NGOM) coast, the first land encountered after crossing the Gulf of Mexico (GOM).

We used automated radio telemetry (Motus) to estimate stopover duration and probability of departure. Migration speed was estimated for individuals detected at subsequent receivers on the Motus Network. To measure migratory connectivity, we used morphometrics and the Motus network to assign general breeding regions. Additionally, feather stable isotope ratios of C and N provided coarse information about over-wintering regions.

Stopover duration declined with higher fuel loads at capture as expected under a time-minimizing strategy. After accounting for fuel load, stopover duration was approximately 40% longer at the higher quality site. We found no detectable effect of age, sex, or breeding location on stopover behavior. Probability of departure was strongly affected by humidity and also by tailwind and weather conditions. Birds stopping at the higher-quality site had earlier apparent arrival to the breeding grounds. The Louisiana coast is an apparent stopover hub for this species, since the individuals were departing to range-wide breeding regions and isotope values suggested birds were also using widespread wintering regions.

Our study shows how high-quality, coastal wetlands along the NGOM coast serve a critical role in the annual cycle of a migratory shorebird. Stopover behavior indicated that high quality habitat may be limited for this species during spring migration. As threats to the GOM coast increase, protection of these already limited wetlands is vitally important.

Methods

Our two trapping sites were in Louisiana, USA, at Rockefeller Wildlife Refuge (RWR; 29.729845°N, 92.818009°W) and Grand Isle (GI; 29.2585°N, 89.9540°W) (Fig. 1). We have several pieces of evidence that support our premise that RWR was higher quality habitat compared to GI. RWR is managed for waterbirds, comprised of salt marsh, brackish marsh, freshwater marsh, and aquacultures that are passively drained every spring, producing mudflats that are heavily used as foraging habitat by shorebirds (pers. obs.). GI is a barrier island composed of sandy beaches, dunes, and salt marsh. GI has approximately 160 km2 of wetlands in a 20km radius (Motus range), whereas RWR has more than twice as many wetlands (~350 km2) and is adjacent to similar wetland habitat of approximately 1300 km2 (United States Fish & Wildlife Service (USFWS), 2020). Human disturbance can decrease habitat quality (Foster et al. 2009), and public access is limited at RWR with minor human disturbance to birds and GI is a public beach with daily human disturbance (driving, fishing, sunbathing) (pers. obs.). Additionally, GI was severely impacted by the 2010 oil spill, which can have long-term negative effects on habitat quality (Bianchini & Morrissey, 2018, Henkel & Taylor, 2012).

Sampling

Field work occurred during spring migration (May-June) over three years (2017-19). We captured SESA in 30mm mesh mist nets (6-12m length) at foraging locations. Individuals were fitted with aluminum USFWS bands and coded colored leg-flags. We measured tarsus, culmen, flattened wing chord (±.1mm), and mass (±.01g). We examined individuals for body/flight feather molt, and estimated subcutaneous fat score from a 0 (no fat) to 7 (bulbous fat) scale (Meissner, 2009). We extracted approximately 30 μl of blood from the wing vein for sexing. To determine age and estimate wintering region (see below), we removed the sixth primary covert from each wing and conducted Carbon-13 (δ13C) and nitrogen-15 (δ15N) stable isotope analyses, performed at the University of California, Davis, Stable Isotope Facility (Davis, CA, USA) (See Appendix 1 for within-run operational standards). We deployed uniquely coded Lotek (Newmarket, ON, Canada) NTQB-3-2 VHF (166.380MHz) transmitters (nanotags), weighing 0.67g, with varying burst rates (10.1, 10.3, 13.1 seconds), and a battery expectancy of 4-6 months. Nanotags were attached to the skin with super glue on the lower back, above the preening gland after the feathers were trimmed with scissors (Diemer et al., 2014). An array of three Motus towers at each site provided detections of individuals at the site (Fig. 1), and towers throughout the Western Hemisphere detected some individuals during northbound and southbound migration, coordinated by Bird Studies Canada (Port Rowan, ON, Canada, Motus.org).

Field work was conducted under U.S.G.S bird banding permit (#22526), LDWF permit (WDP-19-028), and Tulane University IACUC-0403R.

Predictor Variables

Ten predictor variables (sex, age, breeding population, relative fuel load, tailwind, humidity, weather condition, site, capture date, year) were hypothesized to drive variation in stopover duration and/or probability of departing the stopover site: Sex (male or female) was determined from blood samples by extracting DNA using Qiagen (Hilden, Germany) DNeasy blood and tissue kits and a PCR analysis using 2550F and 2718R primers to assign male or female (Fridolfsson & Ellegren, 1999; Ndlovu, 2018). Age (after second-year ASY or second-year SY) was assigned using stable isotope values. Adult SESA molt their feathers on the wintering grounds, whereas juvenile individuals will not molt until their second winter season (Hicklin and Gratto-Trevor 2020). SY birds, therefore, have feather stable isotope values (delta (δ) notation in per mil (0/00 ) units) from the Arctic (our samples: -17 to -27 0/00 (-22.36 ± 0.55 0/00 (Mean ± Standard Error (SE)) δ13C: 7.3 to 9.7 0/00 (8.62 ± 0.14 0/00) δ15N) and ASY have feather stable isotope values from South America (our samples: -12 to -17 0/00 (-13.81 ± 0.07 0/00) δ13C : 9.5 to 13.8 0/00 (11.63 ± 0.11 0/00) δ15N) (Franks et al., 2009; Hobson et al., 2007; Appendix 1). We categorized all individuals into one of three likely breeding populations (Western/Central/Eastern) following morphometric (bill length) measurements determined by Gratto-Trevor et al. (2012) and Hicklin and Gratto-Trevor (2020). Females have longer bills than males, so depending on sex, we assigned individuals with short bill length (M:15.5-18.1mm; F:16.2-18.4mm) as western (Alaska), medium (M:18.2-19.1mm; F:18.4-19.6mm) as central (central Canada), and long (M:19.5-20.8mm; F:19.5-24.4mm) as eastern (eastern Canada). We calculated relative fuel load, f, at capture, a continuous variable calculated from the equation ƒ=(mcap–m0)/m0, where mcap was the mass at initial capture in grams and lean mass m0 was calculated using data from all captured individuals with a zero fat score (R2=0.76) (m0=(0.16x) + 4.35), where x is flattened wing chord (Anderson et al., 2019, Delingat et al., 2008). Three weather variables tailwind, humidity, and weather condition were calculated for each day from data from the weather stations at Abbeville Airport (Abbeville, LA, USA) and South Lafourche Airport (Galliano, LA, USA), the stations closest to RWR (75km) and Grand Isle (35km), respectively. We calculated tailwind with the formula Vw*cos(β), where Vw is ground wind speed and β the difference of the northbound migratory track and wind direction (Dossman et al., 2016). Relative humidity was categorized as either high (≥70%) or low (≤69%) (Senner et al., 2018). Weather condition (visibility) was categorized as either fair (clear to mostly cloudy) or poor (cloudy to rain/thunder storms) (Becciu et al. 2019; Sjöberg et al. 2015). Our last three covariates were, site of capture (site) (RWR or GI), capture date (Julian day), and year (2017, 2018, 2019). We tested for multicollinearity among predictor variables using the variance inflation factor, and found no collinearity among predictor variables (highest VIF: 1.6).

Response Variables: Stopover Duration and Departure Probability

Stopover duration, represented the minimum stopover duration, was calculated from day of capture to departure date, rounded to the whole day. To estimate departure date, we visually examined signal strength of each nanotag from the array of towers at each site. We had reliable departures with a peak in signal strength, followed by a weakening signal strength from the northern most tower and/or antennae (Dossman et al., 2016; Wright et al., 2018). SESA forage and roost on the ground and were rarely detected on the ground, outside of the antennae’s line-of-sight, so we also had reliable departures when SESA took off from the ground, which had an absence of detections followed by an increasing signal then not detected subsequently (González et al., 2020; Taylor et al., 2017). We removed birds from the analysis for whom we had no detections after initial capture, were detected at other Motus towers in Louisiana outside of the two study sites (local relocation), or with undistinguishable departure signals.

Statistical Analysis: Drivers of variation in stopover duration

We developed a set of five candidate models to test drivers of variation in stopover duration within a Bayesian framework (Wright et al., 2018). Each model contained different combinations of seven predictor variables and interactions. These models were to assess (1) body condition; (2) age, sex and breeding location; and (3) site only. The (4) full model with all interaction terms; and (5) main effects only were also included in the model set (Appendix 2, Table S1). We ran a Markov chain Monte Carlo scheme with 2500 iterations, four chains, burn-in of 1000, and thinned every fourth sample within the R package ‘brms’ (Bürkner, 2017). Models were compared by leave-one-out cross validation (LOO) and evaluated with Bayesian R2 (Gelman et al., 2019). We evaluated the marginal effects for all fixed effects and assessed importance with 95% credible intervals not overlapping zero (Vehtari et al., 2017). 

Statistical Analysis: Probability of Departure

We developed a separate set of seven candidate models to estimate the probability of departure from the stopover site using extended Cox proportional hazard models in the R package ‘survival’ (Dossman et al., 2015; Therneau, 2020; Wright et al., 2018). Each model contained different combinations of 10 predictor variables and interactions. These models were (1) weather and site; (2) body condition; (3) age, sex and breeding location; (4) site only; (5) full model with all interaction terms, (6) main effects only, and (7) null model (Appendix 2, Table S2). On the day of departure, weather variables were estimated from the time of departure for each individual. On days prior to departure, weather was estimated from the mean departure time (17:15) of all individuals (Dossman et al., 2015). We compared models by Akaike information criterion (AICc; AIC adjusted for small sample size) and evaluated with the concordance index.