Retention of Riveted Aluminum Leg Bands by Wild Turkeys

No loss of marks is a critical assumption of mark–recapture models so they provide unbiased estimates of population parameters. For wild turkeys (Meleagris gallopavo), butt-end aluminum leg bands have typically been used by wildlife biologists to mark birds, but researchers have found these bands to have retention probabilities <1.0 (Diefenbach et al. 2009; Butler et al. 2011). Diefenbach et al. (2009) reported that retention probabilities differed between adults and juveniles and declined to <0.233 after 15 mo. Butler et al. (2011) reported no differences in retention among sex-age classes, but retention probability was 0.864 15 mo after banding.

One approach to account for tag loss is to have a subsample of animals fitted with a permanent mark so that loss rates can be estimated and incorporated into the estimator (Laake et al. 2014). However, a better solution is to use marks that have retention probabilities close to 1.0 as long as they are inexpensive and easy to apply. Diefenbach et al. (2009) and Diefenbach et al. (2012) used rivet bands to mark wild turkeys and assumed the retention probability of these bands was 1.0. Our objective was to estimate the retention of rivet aluminum leg bands on wild turkeys.

The study area encompassed the range of wild turkeys in New York and Pennsylvania. We captured male turkeys throughout upstate New York (except Essex and Rensselaer counties) north of Rockland and Westchester counties as part of a study of harvest rates (Diefenbach et al. 2012). Captures in Pennsylvania occurred throughout the state for a study of male harvest rates (Diefenbach et al. 2012), but most captures (of both male and female wild turkeys) occurred in central Pennsylvania in wildlife management units 2C, 2E, 4A, 4B, 4D, 2F, 2G, and 2H for a study of female harvest rates (Pennsylvania Game Commission 2015).

We baited capture sites with cracked corn and used rocket nets to capture turkeys (Delahunt et al. 2011). During December 2008–March 2009, we captured male turkeys in New York and Pennsylvania; all but one capture occurred January–March. Beginning in August 2012 and ending October 2014, we captured male and female turkeys in central Pennsylvania during August–October and January–March.

We determined age of turkeys (adult [ad]: >1 y old; juvenile [juv]: <1 y old; Delahunt et al. 2011) and fitted them with an aluminum rivet band (National Band and Tag, Newport, KY) below the spur on each leg. We fitted females with rivet band model 1242FR8 and males with model 1242FR9. Each band was imprinted with a unique alphanumeric sequence and listed a toll-free number for reporting recovery of a band.

We conducted analyses with R.3.1.2 (R Development Core Team 2012) and used conditional logistic regression (Laake et al. 2015) to estimate the proportion of rivet bands retained. The analysis was equivalent to distance sampling double-observer surveys where the probability of an object being detected (i.e., each band being retained) is estimated conditional on at least one observer detecting the object (i.e., at least one band being retained; Burnham et al. 2006). We investigated models in which band loss was related to age of turkeys at time of banding (ad and juv), weeks between banding and recovery, and an intercept-only model. We used Akaike's information criterion adjusted for sample size (AIC_c) to identify the best model (Burnham and Anderson 2002).

We obtained recoveries of 188 males (40 ad, 148 juv) and 37 females (19 ad, 17 juv). Recoveries occurred between <1 and 225 wk since banding (mean = 51.2 wk, SD = 44.0). All females were recovered with both rivet bands present so we conducted no further analyses on these data. For males, 6 of 188 turkeys were recovered missing one band 15–91 wk after banding (mean = 42.3 wk, SD = 12.0).

The model with the lowest AIC_c value was the intercept-only model, although models that included age (ΔAIC_c = 1.96) or time (ΔAIC_c = 1.52) were competitive. However, the age model indicated greater retention by adults (slope coefficient = 0.31, SE = 1.11), which was opposite of what was reported by Diefenbach et al. (2012). The time model indicated increased retention over time (slope coefficient = 0.008, SE = 0.011), whereas both Diefenbach et al. (2009) and Butler et al. (2011) reported increased band loss over time. Consequently, we selected the intercept-only model (intercept = 4.105, SE = 0.4149) as the best model because results of the other models were contrary to our expectations (i.e., other studies have found lower retention rates for adults and reduced retention over time) and the slope coefficients were not different from zero. We estimated the retention probability of a single rivet band for males was 0.984 (95% CI = 0.96–0.99).

The loss of butt-end bands on wild turkeys reported by Butler et al. (2011) indicated band-recovery models would exhibit negative bias in survival rates of −4 to −6%. We observed 100% retention in females (n = 37) and 98.4% retention in males (n = 188) and suggest that use of rivet bands on wild turkeys would introduce little bias in dead-recovery models using this marking technique. The highest retention scenario modeled by Nelson et al. (1980), which was lower than what we estimated for rivet bands, indicated bias of −2 to −0.4% depending on the length of the study.

In contrast to previous studies of loss of butt-end bands by wild turkeys (Diefenbach et al. 2009, Butler et al. 2011), the model that estimated band loss as a function of time (weeks) suggested that retention increased over time. Although the effect was small, the model suggested that a rivet band was 1.03 times more likely to be retained for every 4 wk the band was on the turkey (95% CI = 0.99–1.08). Another explanation for this result is that failure in rivet bands occurs soon after being deployed but otherwise they are retained. If this interpretation of the time model is correct, it further supports the use of the estimated loss probability of 0.984 for most mark–recapture studies.

The objectives of a given study, however, will determine an acceptable loss rate for marks. Therefore, if researchers need to further minimize the probability of a wild turkey losing its mark (using a rivet band) then birds could be double-banded. If we assume loss of each rivet band were independent (sensu Diefenbach and Alt 1998), by double-banding wild turkeys the probability of recovering a banded bird with both bands missing would be near zero (0.0003 = [1 − 0.984]²). We were not able to test for independence of loss of bands, such that loss of one band would mean loss of a second band is more likely, but dependence is unlikely to be a concern with so few bands lost (Diefenbach and Alt 1998; Laake et al. 2014).

Butler et al. (2011) suggested that there was enough variability among turkey populations that data from some banding programs based on butt-end bands might be useful for population monitoring. We believe the results of analyses that rely on recaptures of wild turkeys banded with butt-end bands with the assumption that marks are not lost should be interpreted with caution. A better alternative would be to conduct a pilot study and band a sample of birds with rivet and butt-end bands to estimate the loss rate of butt-end bands (e.g., Diefenbach et al. 2009). Such an approach could allow for an ad hoc adjustment for bias introduced in the estimator. A preferred approach is to use estimators that incorporate tag loss directly into the estimator when a subsample is permanently marked (Conn et al. 2004; Laake et al. 2014). However, such models are advantageous only when it is not possible to permanently mark all animals because of cost or logistics.

Please note: The Journal of Fish and Wildlife Management is not responsible for the content or functionality of any archived material. Queries should be directed to the corresponding author for the article.

To cite this archived material, please cite both the journal article (formatting found in the Abstract section of this article) and the following recommended format for the archived material.

Text A1. Data from: Code (for program R using the package mrds) to estimate band retention in wild turkeys (Meleagris gallapavo) double-banded with rivet bands in Pennsylvania, USA, 2008–2014. Archived in Dryad Digital Repository: http://dx.doi.org/10.5061/dryad.8433

Found at DOI: http://dx.doi.org/10.5061/dryad.1032r

Text A2. Data from: Data of band retention for male and female wild turkeys (Meleagris gallapavo) double-banded with rivet bands in Pennsylvania, USA, 2008–2014. Archived in Dryad Digital Repository: http://dx.doi.org/10.5061/dryad.8433

Found at DOI: http://dx.doi.org/10.5061/dryad.1032r

Support for this research was provided by the Pennsylvania Game Commission and the New York Department of Environmental Conservation, Division of Fish, Wildlife, and Marine Resources.

We thank J. L. Laake for suggesting the use of conditional logistic regression and assistance with use of the mrds package. We thank P. Conn, T. Cooper, J.L. Laake, and the associate editor for suggestions that improved the manuscript.

Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.