For this study, we used captive-reared Japanese quail from the long-term research colony at the U.S. Geological Survey’s Eastern Ecological Science Center. We chose Japanese quail as our proxy for juvenile common terns (Sterna hirundo). Japanese quail were selected due to the similarities in adult mass (~ 120 g [for common terns see , for Japanese quail see "Results"]) and because both species are fully covered in down at hatching . Additionally, the slower development (relative to common tern chicks) of the strain of Japanese quail used in this study enables a careful examination of any complications as they emerge. Similarly, while body shape is admittedly different between the two species, the elevated levels of breast expansion seen in Japanese quail relative to common terns enable these results to account for exceptional scenarios and provide greater confidence that selected methods will be less likely to result in unanticipated impacts. While it would be ideal to test attachment methods directly on common tern chicks, this species is protected at the federal level by the Migratory Bird Treaty Act (16 US Code §§703-711) and provided varying degrees of special protections in multiple states along the Great Lakes and coastal regions of the United States , e.g.,  making initial testing when potential injuries are unknown inadvisable (see [35, 45]). Fortunately, Japanese quail are an easily accessible domestic species without conservation concerns and can serve as a surrogate. While it should be noted that these species vary markedly in behavioral characteristics (i.e., terns dive for food and rely more regularly on flight), we believe that Japanese quail allow for a reliable initial examination of potential physiological and point of contact impacts of marking juveniles. This work is not intended to provide a final definitive answer for the best tagging method for use with juvenile terns, but instead to provide guidance on a safe place to begin such investigations and limit potential injuries to wild birds during method development.
Birds were hatched in incubators and transferred to multi-rack brooder towers when they reached 2 days of age. At 18 days old, the study birds were divided among 7 racks with 13 birds per rack. To promote timely development, light exposure varied throughout the study, beginning with constant light exposure (birth until 14 days old) and slowly transitioning towards a more darkness-oriented routine. At 41 days old, in an effort to curtail the development of aggressive behavior , the amount of light was reduced from 13 to 9 h, and the light source was switched from overhead to wall-mounted lights with one of two bulbs from each light fixture in the facility when birds reached 43 days old. Chicks were given food and water ad libitum throughout the study, initially being fed a diet of gamebird starter crumble, but transitioned to a lower-protein maintainer diet at 43 days old. Following the conclusion of the study, colony managers selected individuals to retain as breeders for colony purposes and moved these individuals to breeding towers.
Tag construction and attachment
We constructed transmitter packages consisting of three varying components: the transmitter type, the harness type, and the harness material with each unique combination serving as one treatment in this study (see Fig. 1 for complete breakdown of treatment components). In order to facilitate the examination of any differences in effect based on the type of transmitter attached to juvenile quail, we created mockups of two models of small transmitters suitable for birds of this size, the CTT LifeTag (mass = 0.8 g; Cellular Tracking Technologies, Rio Grande, NJ) and the Lotek NTS-1 solar NanoTag (mass = 1.4 g; Lotek Wireless, Newmarket, Ontario). While these tags are similar in size, they differ in the way the tags attach to harnesses. CTT LifeTags attach to harness material at connection points in a vinyl strip that extends from the main body of the tag, causing the tag to sit below the attachment points (hereafter CTT tags referred to as Rear-weighted; Fig. 1). Conversely, Lotek tags are centered between attachment points on both sides of the tag allowing them to be more centered within the harness (hereafter Lotek NanoTags are referred to as Center-weighted). Unfortunately, we could not use real transmitters in this study due to concerns of inadequate direct light resulting in the tags becoming nonoperational. Mock-ups were custom designed to match the dimensions of the actual transmitters and were 3D printed out of a polylactic acid plastic. Antennas, also made to replicate the features of those found in functional tags, were made of vinyl coated, 26 AWG Poly-STEALTH wire (Davis RF Co., North Haverhill, NH) and secured to the transmitter body by melting the surrounding plastic onto the antenna. In order to mimic the thin vinyl portion of a Rear-weighted tag, sections of vinyl folder were shaped and melted onto the transmitter body. Final mock-ups were within 0.1 g of their respective units (Rear-weighted mock-up = 0.9 g, Center-weighted mock-up = 1.5 g).
In addition to the two transmitter types we used three different materials to make harnesses: ¼” tubular Teflon ribbon made from PFTE (Bally Ribbon Mills, Bally, PA) hereafter “PFTE ribbon”; 3/16″ Conrad-Jarvis automotive ribbon (the smallest width available for this product; Conrad-Jarvis Corp. Pawtucket, RI), hereafter “Automotive ribbon”; and 1 mm Stretch Magic elastic cord (Soft Flex Company, Sonoma, CA), hereafter “Elastic cord”). The PFTE material was relatively thin with a smooth surface and very pliable across both length and width but was non-elastic. Conversely, Automotive ribbon was very elastic and pliable along its length but not width, while being somewhat textured along the surface and presenting a thicker profile. Finally, the Elastic cord was pliable along both length and width while being elastic only along its length. Elastic cord was also smooth along its surface but had a “sticky” texture when stretched. Images of these materials can be found in Fig. 1. These materials were selected based upon communications with researchers with extensive experience tagging tern species based upon what materials they felt had the ability to operate properly based on the physiology and habitat of common terns (pers. comm. E. Craig, P. Loring, D. Lyons, J. Spendelow, and L. Welch).
We used these materials to construct both backpack and leg-loop harnesses. Backpack harnesses were roughly modeled after Thaxter et al.  “wing harness” method (selected due to lower number of sewing points and better ergonomic fit for common terns versus the “body harness” method reported by the same authors) while leg-loop harnesses were a modified version of the Rappole–Tipton Harness . Harnesses made of Automotive ribbon were sewn to the tags using upholstery thread, whereas the Elastic cord harnesses were threaded through tubes or eyelets added to the tags during manufacturing and secured with 2 mm sterling silver crimp beads. The Elastic cord backpack harnesses were then secured in the center, across the breast, using a small section of 2 mm heat shrink tubing. Finally, harnesses made from PFTE ribbon were either sewn to the tag (Rear-weighted) or threaded through eyelets and secured via a knot (Center-weighted) depending on tag type. We also used a 3D-printed harness (Cellular Tracking Technologies, Rio Grande, NJ) with each tag type (material was non-elastic and somewhat stiff but presented a smooth surface). They were secured to tags using small sections of 2 mm heat shrink tubing and ethyl cyanoacrylate glue (Krazy Glue®). Images of all fully assembled treatments can be found in Additional file 4. Only tagging methods suitable for long-term attachment were tested in this study (see "Discussion" for more details).
We attached tags to the quail at 28 days old (hereafter, day 1). This date was chosen to represent the approximate point in the development of an 11-day old common tern chick (slightly earlier than any anticipated tagging of terns). Upon attachment, we recorded mass and wing chord for all birds in the study (both tagged and control birds). All harnesses except for those constructed out of Elastic cord were pre-sized based on the mass of birds, with the goal of similar amounts of tag gap across treatments and harness sizes. Elastic cord harnesses were sized on the bird prior to being locked to size with a crimp bead. Overall, this resulted in fourteen treatment groups (unique combinations of transmitter type, harness type, and harness material) divided among seven 13-bird racks, with two treatments per rack. Each treatment contained five birds, except for the two 3D-printed harness treatments with four birds; this rack had five control birds. A complete breakdown of racks, treatments, and sample sizes is available in Table 1. Control birds (generally n = 3) were included in each rack and were handled and treated in the same manner as treatment birds throughout the experiment.
Monitoring for effects
Following tag attachment, daily checks assessed tag gap, abrasion, or tag damage. Tag gap, or the amount of space between the harness and the bird’s back or rump (for backpacks and leg-loop harnesses, respectively), was scored on qualitatively on a Likert scale with zero being no gap and three being a large gap (equates to a very loose harness, ~ 1 cm). While measuring tag gap, the transmitter was gently pulled up, away from the bird. Enough tension was applied as to take out any slack but not enough to cause noticeable stretching in the harness material. Abrasion was also scored on a Likert scale with one being no abrasion and higher values indicating increasingly severe abrasion. Photographic examples of each abrasion score can be found in Additional file 5. All birds, regardless of treatment, were also weighed each day and wing chord was taken every three days. These metrics were selected as they provide a look at potential adverse impacts of these tagging methods on wild birds. For instance, while a large tag gap could result in the bird snagging the harness on debris and becoming entangles, too small of a tag gap could result I constriction of blood flow. Similarly, abrasion would indicate potential for injury and possibly result in altered behavior. General qualitative observations of mobility were also made during handling. Tags were removed on day 30, at which time a final evaluation was conducted for each bird, and the complete suite of measurements was repeated. All birds selected to be retained as breeding pairs in the colony were also measured at 60 and 87 days after study initiation.
We generated growth curves using control bird mass to visualize the percent of total growth completed at the age at which the birds were tagged. We tested for effects of treatment (each combination of transmitter type, harness type, and harness material) on growth metrics (bird mass and wing cord, independently), tag gap, and skin abrasion using generalized additive mixed models using the function “gamm()” from the package mgcv  in R version 4.0.2 . All models presented only included the 30-day treatment period as no differences in growth were observed between tagged and untagged birds (see "Results").
Full growth models contained the effects of treatment and sex. To examine changes across time, a smooth effect of day was added with a cubic regression spline, as well as smooth interactions between day and treatment and day and sex. To control for correlated errors across time an Ar1 temporal autocorrelation structure was included , as well as the random effect of bird ID nested within rack to control for repeated measures and rack effect, respectively. Models were assessed using a Wald Chi-square test, and we removed non-significant terms through backwards model selection. Similar models were used for changes in tag gap, but with the leg-loop and backpack methods compared in separate models due to the inherent differences in these attachment types. While we acknowledge that our experimental design forces rack effect to be confounded with harness material, the lack of a significant rack effect in growth models (see "Results") suggests differences in tag gap or abrasion are unlikely due to any effect of rack and can likely be attributed to differences in treatments. The 3D harness treatments were not evaluated in this metric due to their dramatic difference in design and function.
Skin abrasion on the thighs and posterior of the wing/body juncture (metapatagium) were only monitored for the attachment types that could damage these respective areas and were thus analyzed separately. Due to a near complete lack of higher abrasion classes in most treatments, abrasion was treated as a Bernoulli variable, where abrasion values greater than 1 were coded as “Abrasion Present” and values of 1 coded as “No Abrasion.” Birds were then pooled by rack, sex, and treatment for binomial regression. The full models contained the same explanatory variables as those in the growth models, excluding the random effect of bird ID as birds were grouped for analysis. All data used in analyses are available at the USGS ScienceBase repository .