Marbled Godwit migration characterized with satellite telemetry

Abstract

Marbled Godwits (Limosa fedoa) breed in 3 disparate areas: The majority breed in the prairies of midcontinental North America, but there are also 2 small and widely separated tundra-breeding populations, 1 in eastern Canada and 1 on the Alaska Peninsula, USA. The major winter ranges include the Atlantic, Pacific, and Gulf coasts of the USA and Mexico. Bear River Migratory Bird Refuge at Great Salt Lake, Utah, USA, is a major stopover site, hosting large godwit populations in the spring and fall. Although the distributions of Marbled Godwit populations and their habitats across the landscape are generally known, the linkages between them are not. We tracked 23 Marbled Godwits equipped with satellite transmitters from sites in Utah, Mexico, Canada, and coastal Georgia during 2006–2010. Our goals were to characterize the migration strategy of Marbled Godwit populations and to determine migratory connectivity of major breeding, staging, and wintering areas. We found that: 1) godwits breeding in the western USA and Canada followed an overland route to winter sites in Mexico after departing their Utah stopover site; 2) godwits tagged in eastern Canada migrated across the continental USA and wintered at sites along the Gulf of California, Mexico; and 3) godwits wintering in coastal Georgia bred in North and South Dakota. We believe this to be the first demonstration of a continental “crisscross” migration pattern in a shorebird. We identified differences in migration elements such as distances traveled, timing of migration, duration, residency, and stopover strategy between the subpopulations, but not between males and females.

Resumen

Rastreamos 23 individuos de Limosa fedoa marcados con transmisores satelitales miniatura en Norte América (2006–2010). Nuestros objetivos fueron caracterizar la estrategia de migración de las poblaciones de L. fedoa y poner a prueba las hipótesis sobre rutas migratorias, anidación y destino de las subpoblaciones en época de invernada. Específicamente, pusimos a prueba la hipótesis de que los individuos de L. fedoa que nidifican en el oeste de EUA y Canadá usan puntos de escala en el Lago Great Salt, Utah e invernan en sitios del oeste de México usando una ruta migratoria a lo largo de la costa pacífica. También pusimos a prueba la hipótesis de que los individuos que anidan en la Bahía James, en el oeste de Canadá, invernan a lo largo de la costa atlántica de EUA. En contraste con las predicciones, 1) los individuos de L. fedoa que anidan en el oeste de EUA y Canadá siguen una ruta terrestre hacia sitios de invernada en México después de partir desde su punto de escala en Utah, 2) los individuos marcados en el este de Canadá migraron a través de la región continental de EUA e invernaron en sitios a lo largo del Golfo de California, México, y 3) los individuos que invernaron en la costa de Georgia anidaron en Dakota del Norte y del Sur. Creemos que esta es la primera demostración de migración continental “entrecruzada” observada en aves playeras. Los individuos de L. fedoa del este de Canadá emplearon la costosa estrategia de migración de “omisión,” mientras que los que usaron el punto de escala en Utah emplearon la estrategia de “salto.” Las aves marcadas en el este de Canadá también viajaron una distancia más larga entre los sitios de apareamiento e invernada. Los individuos marcados en sitios de invernada en Georgia tuvieron el periodo de residencia más largo. Proponemos que las diferencias en estrategias migratorias pueden tener efectos diferenciales en la biología de las subpoblaciones.

Palabras clave: distancia, estrategia migratoria, Limosa fedoa, ruta migratoria, sitios de parada, transmisores satelitales

Introduction

Knowledge of how avian populations are geographically linked between the periods of the annual cycle is recognized as crucial information for devising effective conservation strategies for migrants (Martin et al. 2007). Piecing together this migratory connectivity includes identifying specific wintering, breeding, and migratory stopover areas used by each species' subpopulations. The Marbled Godwit (Limosa fedoa) is a large shorebird in the Scolopacidae with 2 recognized subspecies—L. f. fedoa and L. f. beringiae (Gibson and Kessel 1989, Gratto-Trevor 2000)—and is identified as a species of high conservation concern in the United States and Canada (Donaldson et al. 2000, Brown et al. 2001, Melcher et al. 2006). The L. f. fedoa subspecies has 2 breeding subpopulations: the midcontinental North American population (170,000 individuals) breeding in the prairies of western Canada and the USA, and a smaller population (2,000 birds) breeding in southern James Bay, Ontario and Quebec, in eastern Canada (Andres et al. 2012). Some 2,000 L. f. beringiae are estimated to breed on the Alaska Peninsula (Andres et al. 2012). The major winter ranges include the Atlantic, Pacific, and Gulf coasts of the United States and Mexico (Gratto-Trevor 2000). Marbled Godwit migration takes place through the midcontinent and along both U.S. coasts (Skagen et al. 1999). There is evidence that some birds from the midcontinental breeding population winter or migrate along the Pacific Coast into Mexico (Luther 1968, Kelly and Cogswell 1979, Gratto-Trevor 2000, Melcher et al. 2006, Gratto-Trevor 2011). The southern James Bay, Canada, breeding population is thought to migrate through the eastern USA or to fly directly to wintering areas along the U.S. Atlantic Coast (Morrison et al. 1976). Finally, the Alaskan breeding population is believed to winter on the U.S. Pacific Coast from Washington to northern California (Gibson and Kessel 1989, Gratto-Trevor 2000). An additional piece of the migratory connectivity puzzle is that a large proportion of North America's Marbled Godwit population (up to 43,000 birds) travels an undetermined route through the western USA to stop at Bear River Migratory Bird Refuge and adjacent freshwater wetlands located in the northeastern arm of Great Salt Lake in Utah (Olson 2011).

Migration routes, timing, and stopover sites utilized by each of the 3 breeding godwit subpopulations are not well understood. Understanding how Marbled Godwit populations are distributed across the landscape in the nonbreeding season and during migration will help to identify key habitats that support large proportions of each subpopulation. Protection and management of these sites should be included in any landscape-scale species conservation strategy. This is particularly true for the Marbled Godwits breeding in eastern Canada and Alaska, as their small populations make them particularly vulnerable to extirpation.

The primary goals of our project were to characterize the migration strategies of Marbled Godwit populations in North America and test long-held hypotheses about migration routes and wintering destinations of the subpopulations. Using miniature satellite transmitters, our objectives were: 1) to determine routes, distances, and timing of both north- and south-bound migrations; 2) to identify breeding, stopover, and wintering localities and residency periods at these habitats for each subpopulation; and 3) to determine if overall migration strategy varied by subpopulation. We tested the following hypotheses: 1) that the Marbled Godwits stopping in Utah are part of the midcontinental breeding population that migrates along the Pacific Coast to winter in Mexico; and 2) that the Marbled Godwits breeding in James Bay, Canada, migrate along the eastern seaboard or fly directly to wintering areas along the Atlantic Coast of the USA.

Methods

Study Areas

Godwits were captured from 4 locations in North America representing 2 of the 3 breeding subpopulations: midcontinent (prairie), and southern James Bay in Canada. We studied movements of Marbled Godwits captured at: 1) Bear River Migratory Bird Refuge at Great Salt Lake in Utah (41.44°N, 112.22°W), a major staging site; 2) San Blas, Nayarit, Mexico (21.54°N, 105.28°W), a wintering site; 3) Akimiski Island, Nunavut, Canada (53.11°N, 80.96°W), a breeding site; and 4) the Altamaha River Delta in Georgia, on the Atlantic Coast of the USA (31.29°N, 81.28°W), a wintering site.

Capture

We captured Marbled Godwits using mist nets, leg-hold noose mats, and cannon nets. Following capture, godwits were marked with alphanumeric plastic color bands with a duplicate code on each tibia and a federal steel band on the right lower leg (tarsus). A combination of mass (g), culmen (exposed), and wing length (flattened; mm) measurements were used to assign sex following Gratto-Trevor (2000). Bill color and plumage patterns were also used to assign gender at Akimiski Island, eastern Canada.

Telemetry

We used 25 9.5-g (38 mm × 17 mm × 12 mm) and 3 12-g (43 mm × 18 mm × 14 mm) solar platform transmitting terminals (PTTs; Model 100, Microwave Telemetry, Columbia, MD, USA). The transmitters were permanently preprogrammed to turn on for 6–10 hr and to turn off for 24–48 hr, year-round. Transmitters were placed on the backs of birds with beading cord elastic (Stretchrite; 3.15-mm width), using a modified leg-loop system (Mong and Sandercock 2007). The weight of the PTT and harness we used represented 3–5% of a godwit's total body mass. All birds were monitored until the PTT stopped transmitting or sensors indicated lack of activity or movement.

We used the Argos satellite-based location and data collection system (Argos 2008), including multisatellite service with standard and auxiliary location processing, to monitor the godwits. Argos classifies the accuracy of each location. The location classes (LC) are 3 (within 250 m), 2 (within a range of 250–500 m), 1 (within a range of 500–1,500 m) or 0 (range greater than 1,500 m; Argos 2008). No accuracy estimates are provided for poor-quality locations (LC A, B, and Z). Individual transmitter performance measures such as temperature, battery voltage level, and activity counter were also provided. These latter parameters were examined to determine final transmission date.

Argos raw data was manually downloaded from ArgosWeb (https://www.argos-system.org) via the Internet at ≤9-day intervals. Downloaded Argos data usually included >1 useable location per bird in a transmission day. Multiple locations within a transmission cycle are not considered independent (Petersen and Douglas 1995). Therefore, a set of decision criteria (1–6) following Miller et al. (2005) was used to select a single location that best represented each bird each transmission-day (“best of day”) for identification of important habitats. In the decision set: 1) The Argos location(s) with the highest quality precision index (LC 3, 2, 1, and 0) was favored. Where several locations of similar accuracy occurred in a cluster, we favored 2) the one closest to the previously or subsequently selected location, 3) the one with the largest number of positive plausibility checks, 4) the one with the largest number of messages, and 5) the one that was most biologically plausible. We considered 6) poor-quality locations of LC A, B, and Z only if these locations were redundant with previous or subsequent locations and if they were biologically plausible.

Data Analysis

We used ArcMap (Environmental Systems Research Institute, Redlands, CA, USA) to analyze and plot the “best of day” and in-flight migration locations to delineate migration routes and corridors, identify winter, breeding, and stopover locations, and calculate residency and mean departure and arrival dates at important habitats for each Marbled Godwit. We identified and summarized migration routes for each group of tagged godwits (Utah, Canada, Mexico, and Georgia) by latitude and longitude. Distances were calculated between “best of day” locations for each segment over the entire migration route, and the total migration distance was measured as the sum of these segments (Martell et al. 2001). A straight line was assumed to delineate migration routes between actual location points.

As the PTTs did not transmit continuously, departure dates were calculated using the median date (and/or date and hour) between the last signal at the previous location and the first signal on migration or at the new destination. Date of arrival at the next habitat area was calculated using the same method, but if there was more than a 10-day gap between departure and arrival, we did not calculate departure or arrival dates or stopover duration for those individuals (Martell et al. 2001). Residency at an en-route stopover was determined by calculating the time elapsed between the first signal at the stopover and the first signal at a subsequent destination or first signal while in flight.

Statistical Analyses

We used one-way ANOVA to test for group differences, with unpaired (two-tailed) t-tests as post-hoc tests for significant differences between means as well as to test for differences between males and females. Significance was set at P < 0.05. Statistical tests were completed using GraphPad online software (http://www.graphpad.com, San Diego, CA, USA) and Microsoft Excel 2010 (Microsoft, Redmond, WA, USA). Results are presented as mean ± SD unless otherwise noted.

Results

Twenty-eight Marbled Godwits (16 male, 12 female) were captured and equipped with PTTs at 4 locations in North America. Godwits captured included 8 males and 5 females in Utah (April 13 and 19, 2006, August 4 and 6, 2006, April 11–14, 2007, April 12 and 14, 2008, and September 30, 2008), 2 males in Mexico (January 28 and 31, 2007), 4 males and 3 females in eastern Canada (June 1 [Plate 1] and 6, 2007), and May 25–27, 2008), and 2 males and 4 females in Georgia (November 20 and December 3, 2008). We tracked 23 of the 28 godwits, as 1 male from eastern Canada shed the transmitter or died before departure while 4 other birds (3 males, 1 each from Utah, Canada, and Georgia; and 1 female from Georgia) failed to initiate migration during the transmission period. We tracked godwits during 20 northbound migrations (9 between stopover and breeding sites; 6 from winter to breeding sites; and 5 in which PTT transmissions stopped during migration after departing winter sites). We also tracked 24 southbound migrations (4 between breeding and stopover sites; 3 from stopover to winter sites; 13 from breeding to winter sites; and 4 in which PTT transmissions stopped during migration after departing either a breeding or a stopover site). In total, 13,630 locations over 6,090 tracking days, covering every month of the year, were received from PTTs in 2006–2010. Individual godwits were tracked from 67 to 522 days over the course of 1–3 calendar years. Many of the Marbled Godwits with transmitters that spanned more than one annual cycle in our study did not undertake multiple migrations. This precluded year-to-year comparisons.

We received 1,401 (10%) locations of LC 3, 2,414 (18%) of LC 2, 3,650 (27%) of LC 1, 4,134 (30%) of LC 0, 967 (7%) of LC A, 986 (7%) of LC B, and 78 (1%) of LC Z. Location Class Z was not recorded for 6 birds. Twenty PTTs (87%) provided spring data, 19 PTTs (83%) provided breeding season data, 17 PTTs (74%) provided fall data, and 14 PTTs (61%) provided wintering data.

Migration Routes and Destinations

Bear River Migratory Bird Refuge, Utah, USA

Northbound godwits captured in the spring at Bear River Migratory Bird Refuge, Utah, USA (hereafter, Bear River), dispersed to midcontinental prairie breeding locales in Alberta, Canada (n = 4), Saskatchewan, Canada (n = 1), Montana, USA (n = 4), and North Dakota, USA (n = 2; Figure 1).

Marbled Godwits southbound from breeding grounds returned to Bear River/Great Salt Lake using routes similar to the northward routes (Figure 2). After departing Bear River/Great Salt Lake, most southbound godwits followed overland routes through Utah, Nevada, USA, and Arizona, USA (n = 6). These migration routes converged along the Colorado River between southern California and Arizona (35–36°N, 114–115°W; Figure 2). One godwit departed Great Salt Lake at a bearing of 230° with a final transmission in eastern Nevada. This trajectory suggests that the bird may have been on a path to cross the Great Basin to reach the Pacific Coast of California (Figure 2). Final wintering destinations used by godwits tagged in Utah included Salton Sea, California, USA (n = 1; 33.17°N, 115.81°W), Ojo de Liebra/Guerrero Negro wetland complex, Baja, Mexico (n = 4; 27.75°N, 114.18°W), Gulf of California coast, Sinaloa, Mexico (n = 1; 22.85°N, 106.05°W), and Sonora, Mexico (n = 1; 26.72°N, 109.57°W; Figure 2).

Only 2 godwits provided complete northbound migration information from Mexico wintering habitats to breeding grounds. These northbound godwits from Sinaloa and Sonora traveled inland through the western USA to reach breeding grounds in Montana and Alberta, respectively, via a stopover at Bear River/Great Salt Lake (Figure 1). A northbound godwit that wintered at the Ojo de Liebre/Guerrero Negro complex traveled along the Baja, Pacific Coast to Los Angeles, California, before its final PTT transmission (Figure 1). This same godwit used an inland route through Utah, Arizona, and Sonora during its southbound migration. These data suggest that some godwits may use a circular migration strategy or differing southbound vs. northbound migration routes.

Akimiski Island, Nunavut, Canada

Four godwits tagged in eastern Canada traveled south via the midcontinental and southwestern USA to winter destinations in coastal Sonora, Mexico (27–31°N, 111–114°W; Figure 2). Marbled Godwits departing Akimiski Island, Nunavut, Canada, traveled southwest over Ontario, Canada, and continued over southwestern Minnesota, USA, to the first en-route stopovers in South Dakota, USA (n = 4), Nebraska, USA (n = 1), and Kansas, USA (n = 1; Figure 2).

Three godwits originating from eastern Canada used multiple sites in southeastern North Dakota and northeastern South Dakota (44–47°N, 98–101°W) for extended periods (15–34 days) before continuing their southward migrations (Figure 2). Five of six southbound Canada-tagged godwits traveled through a narrow corridor from northeast (39–41°N, 102–104°W) to south-central (37–39°N, 104–106°W) Colorado, USA, after departing fall staging areas in the Dakotas. These godwits then followed the Rio Grande through New Mexico, USA, before crossing Chihuahua and Sonora, Mexico, to reach wintering sites in Sonora (27–31°N, 110–113°W; Figure 2).

Three godwits tagged in Canada provided complete or partial northbound migration data. These birds again traveled through the midcontinental USA to return to Akimiski Island along routes similar to those traveled during their southbound migrations (Figure 1).

San Blas, Nayarit, Mexico

Only 1 of 2 godwits tagged in Mexico initiated northbound migration. This bird migrated northward through the midcontinental USA with several in-flight locations over western Minnesota (Figure 1). The final PTT transmission was from eastern Manitoba, Canada.

Georgia, Atlantic Coast, USA

Northbound Marbled Godwits tagged in Georgia, USA, traveled to apparent breeding grounds in North (n = 2; 46–48°N, 101°W) and South Dakota (n = 1; 44°N, 99°W; Figure 1). One of the North Dakota–bound godwits likely used a stopover for <24 hr at an unknown location between Georgia and northern Illinois, USA, before continuing migration. The second godwit apparently traveled nonstop from the Atlantic Coast to a stopover in South Dakota before continuing to North Dakota. The PTT signal was lost in central Illinois from a third bird that appeared to be on the same bearing as the 2 that traveled to North Dakota (Figure 1). The South Dakota–bound godwit took a more southerly migration route that included a stopover in southern Missouri, USA. Southbound godwits returned to the Atlantic Coast using migration routes similar to those used during northbound migration (Figure 2). The Georgia godwits exhibited winter site-fidelity as they returned to the same approximate location on the Atlantic Coast where they were captured the previous winter. Migration routes of the Georgia-tagged birds overlapped or crossed over those of the eastern Canada–tagged birds. The area of “crisscross” overlap occurred in the midcontinental USA (42–47°N, 94–100°W) during both migration periods (Figures 1 and 2).

Migration Characteristics

Data were pooled across the geographic sampling groups to test for significance between males and females in migration characteristics. No significant differences were detected between the sexes for any migration element. Below we focus on test results by geographic sampling group.

Distance

Marbled Godwits traveled a mean of 2,622 km during northbound migration and 2,821 km during southbound migration for a mean annual total migration distance of 5,210 ± 1,355 km (range: 3,963–7,203 km, n = 5; Table 1). Significant differences were found among sampling groups in total distance traveled during southbound migration (F_2,11 = 17.9, P < 0.001), with eastern Canada–tagged birds traveling a greater distance (3,798 ± 200 km, n = 4) than godwits tagged in Utah (2,533 ± 505 km, n = 7; t₉ = 4.71, P = 0.001) and Georgia (2,190 ± 190 km, n = 3; t₅ = 10.75, P < 0.001; Figure 3). The sample size was too small for statistical analysis of distances traveled during northbound migration.

Duration and strategy

We found a significant difference among geographic sampling groups in the duration of southbound migration (F_2,10 = 8.4, P = 0.007). The 2 ± 0 day (n = 3) duration that the Georgia-tagged godwits spent on southbound migration was significantly shorter than the 51.3 ± 23.8 days (range: 25–83 days, n = 6) and 22.8 ± 10.4 days (range: 15–37 days, n = 4) of the Utah (t₇ = 3.47, P = 0.01) and Canada (t₅ = 3.37, P = 0.02) sampling groups, respectively. The difference in migration duration between Utah and eastern Canada birds was not quite significantly different (t₈ = 2.23, P = 0.06). Northbound migration data were not analyzed due to insufficient sample sizes.

The Canada-tagged birds used an area along the border between North Dakota and South Dakota as a migration stopover. The Utah-tagged godwits from the midcontinental breeding population used Bear River Migratory Bird Refuge as a stopover. Godwits used these areas for short durations during northbound migration (2–5 days), but stayed for extended periods (8–69 days) during southbound migration (Table 2). No migration stopovers were used for more than 2 days by the Georgia sample group.

There were significant differences in the distance traveled to the first stopover site during southbound migration among the sampling groups (F_2,12 = 34.40, P < 0.001; Figure 3). Distance to first southbound stopover ranged from a short hop of 282 km from a Utah-tagged godwit departing Alberta and stopping in Montana to a long jump of 2,451 km from a Georgia-tagged godwit returning directly to the Atlantic Coast from North Dakota. Godwits tagged in Utah traveled a significantly shorter mean distance (670 ± 227 km) to their first southbound stopover habitat than godwits tagged in eastern Canada (1,925 ± 405 km; t₁₀ = 6.63, P < 0.001) and Georgia (2,204 ± 214 km; t₇ = 9.72, P < 0.001). There was no significant difference between Canada and Georgia godwits in distance traveled to first southbound stopover (t₇ = 1.1, P = 0.31).

Timing

Mean onset of northbound migration (winter departure) by Marbled Godwits took place over a month-long period from April 20 to May 19 (Table 3). There was a significant difference in winter departure dates among sampling groups (F_2,7 = 16.6, P = 0.002). The mean winter departure date of May 17 for birds tagged in eastern Canada (n = 3) was 26 and 27 days later than mean departure dates for birds tagged in Georgia (t₅ = 5.5, P = 0.003) and Utah (t₄ = 5.6, P = 0.005), respectively. No significant difference in arrival date at breeding sites was found between the Utah and Georgia birds (t₅ = 0.62, P = 0.55; the sample size for Akimiski Island, Canada, birds arriving at this breeding site was too small for analysis).

Date of departure from the breeding grounds (southbound migration) ranged from an early mean of June 20 for the Georgia godwits to a late August 7 mean by the eastern Canada birds (Table 4). The breeding ground departure dates were significantly different (F_2,14 = 22.8, P < 0.001) among the sampling groups as well as between each paired sample mean. Birds originating in eastern Canada departed breeding sites later than birds tagged either in Utah (t₁₂ = 5.1, P < 0.001) or in Georgia (t₇ = 6.8, P < 0.001). Godwits tagged in Georgia departed earlier than Utah-tagged godwits (t₉ = 2.4, P = 0.04). Subsequently, Georgia godwits returned 64 and 86 days earlier to winter habitats than did godwits tagged in eastern Canada (t₅ = 8.3, P < 0.001) and Utah (t₉ = 4.0, P = 0.003), respectively (Table 4). No significant differences were found between winter arrival dates of birds tagged in eastern Canada and Utah.

Residency

A breeding site arrival date of May 25 was assumed for the eastern Canada (Akimiski Island) population (K. Abraham personal communication). There were significant differences in duration of residency among the geographic sampling groups at winter (F_2,7 = 108.9, P < 0.001) and breeding-habitat localities (F_2,14 = 6.2, P = 0.01). Godwits from Georgia remained on the breeding grounds for a mean of 56 ± 6 days (n = 3), a significantly shorter breeding site residency period than that of godwits tagged in either Utah (74 ± 8 days; t₉ = 3.50, P = 0.007) or eastern Canada (73 ± 9 days; t₇ = 3.12, P = 0.02). Utah-tagged godwits' winter habitat residency duration of 174 ± 18 days was significantly shorter than that of birds from both eastern Canada (269 ± 9 days; t₅ = 9.94, P < 0.001) and Georgia (303 ± 7 days; t₆ = 15.92, P < 0.001). Georgia-tagged godwits had the longest winter residency, staying 129 and 34 days longer than birds tagged in Utah and eastern Canada, respectively. Residency periods at wintering grounds were not calculated for godwits tagged in Mexico, as they were captured mid-winter.

Discussion

We established migratory connectivity across breeding, stopover, and wintering sites for the midcontinental North America and eastern Canada subpopulations of the Marbled Godwit. We rejected the long-held hypothesis that the breeding population of Marbled Godwits in eastern Canada migrates down the Atlantic seaboard or flies directly to wintering areas along the USA's Atlantic coast. Instead, birds tagged on Akimiski Island in eastern Canada crossed the continent and wintered along the Gulf of California in Mexico. Godwits tagged at a winter location along the Atlantic Coast in Georgia migrated to breeding locations in the midcontinental USA. Marbled Godwits tagged at a Utah stopover site were part of the midcontinental breeding subpopulation, as predicted. Contrary to expectations, these birds used an overland route through the southwestern USA, instead of following the Pacific Coast of California, USA, to arrive at wintering sites in Mexico. Migration routes of the Georgia-tagged birds crossed over those of the eastern Canada–tagged birds in the midcontinental USA. We believe this “crisscross” to be a novel shorebird migration pattern, although there are suggestions that it may also occur in the Long-billed Curlew (Numenius americanus; Page et al. 2014). We found differences in migration elements such as total distance traveled, distance to first southbound stopover site, migration duration, and residency, as well as stopover strategy, among geographic sampling groups, but not between males and females. Differences in migration elements were also found among geographic sample populations as well as male and female Long-billed Curlews (Page et al. 2014).

Shorebirds may use a “hop,” “skip,” or “jump” migration strategy, referring to distances moved between successive stopovers (Piersma 1987). Utah-tagged Marbled Godwits appeared to employ a “hopping” migration strategy. They traveled medium distances (<700 km) between successive stops before and after the Bear River Migratory Bird Refuge/Great Salt Lake staging site in order to hop over inhospitable habitats (mountains and deserts) along the migration route. In contrast, the godwits tagged in eastern Canada and Georgia exhibited a “skipping” migration strategy, traveling >1,400 km before making their first stop en route or flying directly to breeding or wintering destinations. “Hopping” can be energetically less costly than “skipping” flights due to the costs of transporting extra fuel for the latter (Piersma 1987). The Akimiski Island breeding subpopulation traveled the longest distance between breeding and wintering grounds compared with the midcontinental breeding birds. These differences imply that the godwits breeding in eastern Canada employ a more energetically costly migration strategy that may have differential effects on survivorship and reproductive outputs.

Godwits stopping in Utah during southbound migration undergo a full body molt (Olson 2011). Godwits present along the Atlantic Coast in the fall have also been observed undergoing molt (B. Winn personal communication). Though we lack information on when the godwits from the southern James Bay/Akimiski Island subpopulation molt, we think it likely that some members of this group molt during migration. Mid-migration molt, an energetically costly process, would explain the use of multiple stopover sites over an extended period at the North and South Dakota borderlands region by southern James Bay/Akimiski Island godwits during southbound migration. In contrast, 2 godwits tagged in eastern Canada and several from the Utah-tagged subpopulation exhibited a compressed southbound migration schedule in our study. This “sprint” vs. “marathon” migration (Martell et al. 2001) may be an individual, alternative strategy to allow arrival at wintering habitats prior to beginning molt.

Events or conditions at one point or time in the avian annual cycle can have carryover effects at subsequent locations or seasons. Winter habitat quality can influence arrival and subsequent reproductive outputs at the breeding grounds (Norris et al. 2004, Gunnarsson et al. 2006), annual survivorship (Studds and Marra 2005), body condition (Bearhop et al. 2004, Burton et al. 2006), and sexual selection (Reudink et al. 2009). Godwits in our study spent a considerable amount of time at winter habitats (48–83% of the annual cycle). We also detected significant differences in winter residency periods among the sample groups. We recommend further investigation into the effects of this variation in winter residency lengths and winter habitat quality on reproductive success to better understand carryover effects in this species.

The Bear River Migratory Bird Refuge/Great Salt Lake wetland complex is used by Marbled Godwits during both northbound and southbound migrations, hosts a high proportion of the continental population predictably, year after year, and has an abundant and reliable food resource (Chironomidae spp.; Paul and Manning 2002, Shuford et al. 2002, B. E. Olson personal observation). These factors satisfy the criteria for recognizing this wetland complex as a shorebird “staging” site (Hands 1988, Melvin and Temple 1982 in Farmer and Parent 1997, Warnock 2010). Avian staging sites are important to identify and protect as part of a migratory species' conservation strategy. These sites are crucial stops along the migration route where birds rest and refuel to meet the energetic and physiological requirements for their next flight. Although we identify this wetland complex as essential for the conservation of the Marbled Godwit, portions remain unprotected. Bear River Bay outside of the boundaries of Bear River Migratory Bird Refuge continues to be vulnerable to development (i.e. mineral extraction) and habitat degradation via water deprivation, as river inputs are increasingly diverted for municipal and agricultural uses.

Though our sample size was small, the results of our study offer some intriguing differences in migration strategy to explore further. These variations are starting points from which to examine how various aspects of migration may differentially affect the biology of distinct breeding subpopulations. Additional telemetry studies of Marbled Godwits would further refine our understanding of migratory corridors and stopover strategies and functions, as well as how these patterns compare across years. Finally, tracking individuals from wintering sites along the Gulf of Mexico and the Atlantic Coast of Florida would allow us to assign these birds to a particular breeding subpopulation.

Acknowledgments

The findings and conclusions in this article are those of the authors and do not necessarily represent the views of the U.S. Fish and Wildlife Service. This project was funded by the U.S. Fish and Wildlife Service, U.S. Geological Survey, Ontario Ministry of Natural Resources, Utah Wetlands Foundation, George S. and Delores Doré Eccles Foundation, The Wilson Conservation Trust, The Nature Conservancy, Microwave Telemetry, ProNatura, Georgia Department of Natural Resources, and The Environmental Resources Network. We thank S. Hicks and A. Trout for moral and financial support at Bear River Migratory Bird Refuge, Utah; Dr. Ken Abraham for logistical and capture support at Akimiski Island, Canada; Xico Vega for logistical support in Mexico; and Brad Winn for logistical support and capture efforts in Georgia. We are grateful to the following individuals for assistance in the field: S. Brown, M. McGarvey, N. Atkins, J. Olson, S. Edler, R. Jacobson, K. Stopher, and T. Woodward.

Literature Cited