INTRODUCTION

The use of light-level geolocation devices (geolocators) to identify migration routes, timing of migration, stopover sites, and breeding or wintering locations for individual birds has recently become widespread in ornithology (Stutchbury et al. 2009, Bridge et al. 2013). Geolocators, composed minimally of a light sensor, a battery, and a microcomputer, can be attached to birds prior to their migration. They function by recording ambient light levels along with time stamps at regular intervals. Times of sunrise and sunset as well as day length are used to estimate the longitude and latitude of the geolocator-tagged bird on a daily basis throughout the annual cycle (Bridge et al. 2013). To obtain data from the devices, birds must be recaptured, usually when they return to the location at which they were originally captured and marked. Miniature geolocators, now weighing <0.4 g, have been used successfully on birds as small as 9-g Golden-winged Warblers (Vermivora chrysoptera) and 12-g Blackpoll Warblers (Setophaga striata; DeLuca et al. 2015, Peterson et al. 2015).

On passerines, geolocators are usually attached over the synsacrum using some form of leg-loop harness (Rappole and Tipton 1991, Bridge et al. 2013, Streby et al. 2015). A variety of geolocator types and harness materials have been used, and return rates have been highly variable across studies (Bowlin et al. 2010, Bridge et al. 2013). Although annual return rates are commonly reported for both geolocator-tagged and color-banded control birds, it is becoming apparent that the effects of these devices are species- (or guild-) specific. For example, geolocators (and other tracking devices) are more of a burden on aerial insectivores (Stutchbury et al. 2009, Bowlin et al. 2010, Gómez et al. 2014; but see Matyjasiak et al. 2016) and may be more detrimental to species that make longer migratory flights (Bridge et al. 2013). Because abiotic (e.g., microclimate, wind, weather events) and biotic (e.g., predators, prey) pressures likely differ among microhabitats (e.g., canopy vs. understory), geolocators may also differentially affect birds that occupy these dissimilar environments.

In addition to the potential for context-dependent impacts on return rates, geolocators may also influence breeding biology (Schmaljohann et al. 2012, Arlt et al. 2013, Bridge et al. 2013, Gómez et al. 2014). Studies of the effects of radio-transmitters on the breeding activities of Golden-winged and Hooded (Setophaga citrina) warblers indicated that there were no negative effects on nestling provisioning rates, incubation times, or seasonal productivity (Neudorf and Pitcher 1997, Streby et al. 2013). However, a meta-analysis and a literature review of the effects of transmitters and related tracking devices suggested that attached devices can have negative impacts on a number of aspects of a bird's life, such as energy expenditure, reproductive success, and the probability of nesting (Calvo and Furness 1992, Barron et al. 2010), and negative effects of tracking devices on birds are likely underreported (Hill and Elphick 2011). Finally, although geolocators have similarities to radio-transmitters in harnessing method (for songbirds), they differ in that their attachment methods are more permanent, and thus they may affect birds differently compared with radio-transmitters (Bridge et al. 2013, Streby et al. 2015).

Here, we evaluated the effects of geolocators on several components of fitness within the breeding season (within-season apparent survival, nestling provisioning rate, and nest survival) and across seasons (return rate) for a declining, canopy-dwelling, Nearctic–Neotropical migrant, the Cerulean Warbler (Setophaga cerulea). Although the benefits of understanding migratory timing, pathways, and connectivity for the conservation of this species (and other songbirds) are substantial (Hostetler et al. 2015, Marra et al. 2015), the potential for deleterious effects caused by the attachment of tracking devices should be evaluated and reported to reduce unintended mortality, limit impacts on reproductive performance, and provide insights that may guide technical improvements in the devices or their deployment.

METHODS

Study Area

We captured male Cerulean Warblers and deployed geolocators across 2 regions within the Cerulean Warbler's breeding range: on the Allegheny Plateau in northwestern Pennsylvania and on the Ozark Plateau in Missouri and Arkansas, USA (Figure 1). In Pennsylvania, sites were located within the State Game Lands No. 86 (41.8°N, 79.3°W) along the Allegheny River and within the adjacent Allegheny National Forest (41.6°N, 79.3°W). In the Ozark region, individual sites were located within the Buffalo National River (36.0°N, 92.6°W), Cherry Bend Recreation Area in the Ozark National Forest (35.7°N, 93.8°W), and along the Eleven Point River in the Mark Twain National Forest (36.7°N, 91.2°W).

FIGURE 1.

Study sites in the Ozarks (O), located in Missouri and Arkansas, and Pennsylvania (P), USA, where we fitted geolocators to Cerulean Warblers to investigate effects on reproduction and survival. O_a = Cherry Bend Recreation Area, Ozark National Forest; O_b = Buffalo National River; O_c = Eleven Point River, Mark Twain National Forest; and P = Pennsylvania State Game Lands 86 and adjacent Allegheny National Forest.

RESULTS

Within-season Survival

Four models were equivalent, with none including ‘group' (Table 1), indicating no support for a geolocator-marking effect on within-season φ. The model-averaged weekly φ, estimated from the top 4 models, was 0.968 ± 0.026 (n = 53 for the combined groups); across the entire breeding season, this was equivalent to an apparent seasonal survival rate of 85% (0.968^{5 weeks}). Note that this is apparent survival rate, including dispersal as well as mortality.

TABLE 1.

Comparison of top equivalent models (ΔAIC_c < 2.00) for apparent weekly survival of male Cerulean Warblers in Pennsylvania, USA, 2014 and 2015, using Akaike's Information Criterion adjusted for small sample size (AIC_c). K = the number of parameters, w_i = model weight, and Dev = model deviance. The top model that included the variable ‘group' (which indicates the influence of geolocators; the 2 groups were geolocator-tagged birds and color-banded control birds) and the null model are shown for comparison.

Nest Survival and Provisioning Rates

The top model describing daily nest survival rate (DSR) was the null model, indicating that geolocators did not affect nest survival. The DSR 95% CI of geolocator-tagged males (0.940–0.978; n = 21) overlapped with that of control males (0.928–0.977, n = 19; Table 2).

TABLE 2.

Summary of performance estimates associated with geolocator-tagged and control male Cerulean Warblers in the Ozarks (O) and Pennsylvania (P) study sites (see Figure 1 for study site locations) in 2014–2016: Mean weekly apparent survival (φ) over the breeding season (Pennsylvania only), daily nest survival (DSR; Pennsylvania only), nestling provisioning rates (NPR; Pennsylvania only; visits per nestling per hr), and annual return rates (Pennsylvania and Ozarks). Weekly φ and DSR estimates were calculated from the top model that included group (see Tables 1 and 3).

TABLE 3.

Best-supported models (ΔAIC_c < 2.00) for return rates of male Cerulean Warblers in Pennsylvania and the Ozarks, USA, for the 2014–2015 and 2015–2016 annual cycles, using Akaike's Information Criterion adjusted for small sample size (AIC_c). The null model is also shown for comparison. K = the number of parameters, w_i = model weight, and Dev = model deviance. The group effect refers to the influence of geolocators; the 2 groups were geolocator-tagged birds and color-banded control birds.

Nestling provisioning rates did not differ between banded (2.37 ± 0.43 visits per nestling per hr) and unbanded control males (2.26 ± 0.34 visits per nestling per hr; t₂₂ = −0.18, P = 0.86), nor between SY (2.50 ± 0.48 visits per nestling per hr) and ASY banded control males (2.26 ± 0.77 visits per nestling per hr; t₁₂ = 0.27, P = 0.80). Thus, to maximize our sample size at 9 geolocator-tagged and 15 control birds, unbanded (age unknown) control birds were included and parental age was not accounted for in our models. The null model was the sole top model (ΔAIC_c = −2.91), providing no support for geolocator-marking influencing provisioning rates (geolocator-tagged birds = 2.47 ± 0.70 visits per nestling per hr; control birds = 2.24 ± 0.27 visits per nestling per hr; Table 2).

DISCUSSION

Our study is among the first to evaluate the effects of geolocators on both reproduction and return rate. We found no evidence to suggest a negative effect of geolocators on apparent within-breeding-season survival, daily nest survival, or nestling provisioning rate, which suggests that reproductive output prior to migration is not likely to be hampered by the use of this technology. However, we found that birds with geolocators were less likely to return to their breeding grounds in the following year than control birds, which is cause for some concern.

Following geolocator deployment, our data and qualitative field observations suggest that Cerulean Warblers quickly made any necessary adjustments associated with carrying a geolocator, resuming normal activities (e.g., singing, territory defense, and nestling provisioning) within 24 hr. Their short-distance flights and foraging efforts did not appear to be affected, as has also been observed in Barn Swallows (Hirundo rustica; Matyjasiak et al. 2016). Although our statistical power was not always high, our results providing no evidence for geolocator effects on breeding activities of Cerulean Warblers are in agreement with those of similar studies that examined the effects on warblers of tagging with radio-transmitters: these devices had no effects on nestling provisioning by Hooded Warblers (Neudorf and Pitcher 1997) or on nest survival (as well as the number of fledglings produced and postfledging survival) of Golden-winged Warblers (Streby et al. 2013).

In studies of the effects of geolocators on passerines besides warblers, a lack of evidence supporting a geolocator-marking effect on apparent within-season survival and fledging success has also been reported (within the season of geolocator deployment) for Tree Swallows (Tachycineta bicolor; >19.5 g; Gómez et al. 2014), which also had unaffected nestling provisioning rates, and Northern Wheatears (Oenanthe oenanthe; 21–24 g; Schmaljohann et al. 2012, Arlt et al. 2013). However, Arlt et al. (2013) documented decreased nest success (as well as a reduced return rate and later arrival and clutch initiation dates) in the subsequent breeding season in geolocator-tagged Northern Wheatears. In our case, we were unable to monitor enough nests during the breeding season following the return of individuals that bore geolocators to have adequate statistical power for a comparison, but the possibility of reduced nest survival and likelihood of mate pairing (potentially due to geolocator-related carryover effects and/or reduced migration speed) is worth future investigation.

In a study that investigated the effects of radio-tagging Great Tits (Parus major), the nestling provisioning rates of radio-tagged birds did not differ from those of control birds (Snijders et al. 2017). However, the authors did report a radio-tagging effect on nest desertion by adults marked during the nestling provisioning period (nest desertion by both parents increased), suggesting that marking with tracking devices during the nestling provisioning period may negatively affect reproductive output in some cases. While we did not find an overall difference in nestling provisioning rates between geolocator and control groups, we did observe 2 instances (11% of 18 nests associated with geolocator-tagged males that had nestlings) of apparent nest abandonment by males. In both cases, the birds were tagged during the nestling provisioning period, and in both cases the female continued to provision the nestlings and each nest fledged at least 2 young. In another experimental study on nonmigratory Great Tits, nestling mass was negatively associated with geolocator-tagged males, marked males increased in body mass but their wing chord decreased over the postdeployment year, and the marked males seemingly exhibited less competitive ability to secure winter roosts (Atema et al. 2016). However, once again, the marked males' survival and reproductive success were unaffected.

A decreased return rate for geolocator-tagged individuals, as found in our study, implies that geolocator-tagging causes one or more of the following: (1) displacement from the previous year's territory, potentially caused by a later arrival date or decreased aggression toward territorial intruders (one geolocator-tagged bird was resighted ∼860 m from the location of his territory in the previous year, and this individual, along with 3 other marked birds, exhibited noticeably less aggression toward song playback upon return); (2) reduced site fidelity, with some birds possibly stopping elsewhere within the breeding range that involves a shorter travel distance from the wintering grounds; or (3) decreased annual survival, potentially due to increased weight or drag during long-distance flights (Bowlin et al. 2010). Although this study was not designed to distinguish among these 3 explanations, it is possible that our results were related to altered breeding dispersal by geolocator-tagged birds. However, we found no evidence that the effects of geolocators differed by site, which may be assumed if geolocator-tagged birds were attempting to reduce the cost of migration. Return rates between control and geolocator-tagged birds differed similarly in the Ozarks, in the southwestern corner of their breeding range, and in northwestern Pennsylvania, toward the northern end of the breeding distribution. However, we may have lacked adequate statistical power to show support for differential fidelity by study site through a significant interaction effect of geolocator status and site on likelihood of return. Alternatively, it is possible that the reduced return rates of geolocator-tagged birds occurred because mortality during the nonbreeding season was greater for these individuals, with the most likely period of differential mortality occurring during one of their long-distance migrations (Sillett et al. 2002, Jones et al. 2004).

Other studies that have explored the possibility that the return rates of small migratory birds are affected by geolocator-tagging have come to a variety of conclusions. In a technical review, Bridge et al. (2013) found decreased return rates for 9 of 24 (38%) bird species weighing 12–80 g, with smaller species more likely to be negatively affected. And, in a meta-analysis across study species, Costantini and Møller (2013) found that birds with small body mass were more likely to be negatively affected by carrying tracking devices. However, small body mass does not always lead to reduced return rates of geolocator-tagged migrants, as some of the smallest species marked to date, Golden-winged Warblers and Aquatic Warblers (Acrocephalus paludicola), were not adversely affected (Salewski et al. 2012, Peterson et al. 2015). Additionally, inclusion of body mass as a variable in our analysis did not improve the best model (geolocator status only), and Barron et al. (2010) found little support for increasingly negative impacts on survival caused by proportionally heavier devices in their meta-analysis of tracking device–related effects on birds.

Instead of mass alone, several other factors could influence the decreased return rates of Cerulean Warblers (and other small migratory birds) fitted with geolocators. Our preliminary analysis of the light-level data from the geolocators carried by returning birds in this study has indicated long migratory flights across the Gulf of Mexico and/or the Caribbean Sea as they travel to and from their winter territories in the northern Andes Mountains. Migration strategy (including single flight distances and land cover during migration) along with weather patterns during this stage could influence whether geolocators reduce survival (Butler 2000). If geolocator weight and drag decrease maximum flight range (Bowlin et al. 2010), small long-distance migrants that make long overwater flights could be more negatively affected by geolocators than species that undertake shorter flights and stop more often (even if the cumulative trip length is equivalent). Additionally, unfavorable weather patterns or storm systems that are encountered during migration, which are more likely to occur during El Niño events (which occurred during both the fall and spring of 2015; NOAA National Weather Service 2016), may further differentially reduce return rates for individuals fitted with geolocators (again, especially for species that make overwater flights; Butler 2000). Therefore, it is possible that weather conditions during migration were more favorable in 2013–2014 (La Niña conditions during migration, and geolocator-tagged Golden-winged Warbler return rates unaffected; Peterson et al. 2015) than in 2014–2015 and 2015–2016 (reduced return rates of geolocator-tagged Cerulean Warblers in this study). As another example of long overwater flights potentially affecting the survival of geolocator-tagged birds more than untagged birds, DeLuca et al. (2015) reported a relatively low return rate for Blackpoll Warblers, which make long trans-Atlantic migratory flights, some flying overwater from the northeastern coast of the United States to stopover locations in the Caribbean Sea (minimum overwater flights of 2,270–2,770 km) before continuing to northern South America. Only 5 of 37 (∼14%) geolocator-tagged Blackpoll Warblers returned to study sites in Nova Scotia or Vermont in 2013, compared with return rates of birds without tracking devices from 3 breeding sites in different years ranging from 27% to 70% in New Brunswick and New Hampshire (DeLuca et al. 2013). Brown et al. (2017) also found variation in the return rates of geolocator-tagged Semipalmated Sandpipers (Calidris pusilla) among breeding sites across the North American Arctic. Collectively, these studies suggest that migration strategy may be a more important predictor of the effects of geolocators on the likelihood of return than the proportional mass of the device, both at the species level and for individuals within species.

From the perspective of global conservation, reduced return rates do not preclude the continued use of geolocators to improve our understanding of the full annual cycle of Cerulean Warblers and other similarly sized birds. Although the potential for increased mortality during the nonbreeding season, or alterations to migratory timing or pathways, is of some concern, the fact that reproduction does not appear to be hampered is encouraging, particularly for a species that is short-lived (Buehler et al. 2013). Spatial and temporal data acquired from the use of geolocators has enormous conservation potential, and these benefits likely outweigh the potential cost of a relatively small number of individuals being negatively affected by carrying geolocators. The information obtained from individuals that are capable of successfully migrating with geolocators allows researchers to identify important migration stopover locations and to link breeding and nonbreeding season geographies (Stanley et al. 2015, Horns et al. 2016). In turn, such knowledge allows researchers and conservation professionals to more accurately develop full annual cycle population models, which can be used to more effectively target species- and region-specific conservation activities (Faaborg et al. 2010, Hostetler et al. 2015).