Mercury Exposure Affects the Reproductive Success of a Free-Living Terrestrial Songbird, the Carolina Wren (Thryothorus ludovicianus)

Transcription

1 Merury Exposure Affets the Reprodutive Suess of a Free-Living Terrestrial Songird, the Carolina Wren (Thryothorus ludoviianus) Author(s) :Allyson K. Jakson, David C. Evers, Matthew A. Etterson, Anne M. Condon, Sarah B. Folsom, Jennifer Detweiler, John Shmerfeld, and Daniel A. Cristol Soure: The Auk, 128(4): Pulished By: The Amerian Ornithologists' Union URL: BioOne ( is a nonprofit, online aggregation of ore researh in the iologial, eologial, and environmental sienes. BioOne provides a sustainale online platform for over 170 journals and ooks pulished y nonprofit soieties, assoiations, museums, institutions, and presses. Your use of this PDF, the BioOne We site, and all posted and assoiated ontent indiates your aeptane of BioOne s Terms of Use, availale at Usage of BioOne ontent is stritly limited to personal, eduational, and non-ommerial use. Commerial inquiries or rights and permissions requests should e direted to the individual pulisher as opyright holder. BioOne sees sustainale sholarly pulishing as an inherently ollaorative enterprise onneting authors, nonprofit pulishers, aademi institutions, researh liraries, and researh funders in the ommon goal of maximizing aess to ritial researh. PersonIdentityServieImpl

2 The Auk 128(4): , 2011 The Amerian Ornithologists Union, Printed in USA. MERCURY EXPOSURE AFFECTS THE REPRODUCTIVE SUCCESS OF A FREE-LIVING TERRESTRIAL SONGBIRD, THE CAROLINA WREN (THRYOTHORUS LUDOVICIANUS) ALLYSON K. JACKSON, 1,4,5 DAVID C. EVERS, 1 MATTHEW A. ETTERSON, 3 ANNE M. CONDON, 2 SARAH B. FOLSOM, 1 JENNIFER DETWEILER, 1 JOHN SCHMERFELD, 2 AND DANIEL A. CRISTOL 4 1 Biodiversity Researh Institute, Gorham, Maine 04038, USA; 2 U.S. Fish and Wildlife Servie, Virginia Field Offie, Glouester, Virginia 23061, USA; 3 U.S. Environmental Protetion Ageny, Offie of Researh and Development, National Health and Environmental Effets Researh Laoratory, Mid-Continent Eology Division, Duluth, Minnesota 55804, USA; and 4 Institute for Integrative Bird Behavior Studies, Department of Biology, College of William and Mary, Williamsurg, Virginia 23187, USA Astrat. Despite mounting evidene of merury aumulation in terrestrial eosystems, few data exist on how environmental merury exposure affets reprodutive suess in free-living songirds. From 2007 through 2010, we monitored reprodutive suess of Carolina Wrens (Thryothorus ludoviianus) reeding along the forest floodplain of two merury-ontaminated rivers in Virginia. Using an information-theoreti approah, we found a 34% redution in nesting suess of Carolina Wrens on merury-ontaminated sites when ompared with referene sites. Blood merury onentration of the attending female was a strong preditor of nest suess. Birds nesting on ontaminated sites were 3 more likely to aandon their nests than irds on unontaminated referene sites. We report a range of effets onentrations assoiated with various levels of reprodutive impairment; for example, a 10% redution in nest suess orresponded with 0.7 μg g 1 merury in the lood, 2.4 μg g 1 merury in ody feathers, 3.0 μg g 1 merury in tail feathers, and 0.11 μg g 1 merury in eggs. This is the first field study to doument the effet of speifi adult songird lood merury onentrations on reeding performane; our results show that free-living songirds an suffer negative reprodutive effets at relatively low merury onentrations. Reeived 13 May 2011, aepted 23 August Key words: Carolina Wren, eotoxiology, merury, nest suess, point-soure pollution, Thryothorus ludoviianus, Virginia. La Exposiión al Merurio Afeta el Éxito Reprodutivo de Thryothorus ludoviianus, un Ave Canora Terrestre Silvestre Resumen. A pesar de una antidad reiente de evidenia sore la aumulaión de merurio en los eosistemas terrestres, existen poos datos sore ómo la exposiión amiental al merurio afeta el éxito reprodutivo de las aves anoras silvestres. Desde 2007 hasta 2010, monitoreamos el éxito reprodutivo de individuos de la espeie Thryothorus ludoviianus que estaan riando a lo largo de la planiie ososa de inundaión de dos ríos ontaminados on merurio en Virginia. Usando un enfoque asado en teoría de la informaión, enontramos una reduión del 34% en el éxito reprodutivo de T. ludoviianus en sitios ontaminados on merurio, omparado on sitios de referenia. La onentraión de merurio en la sangre de las hemra anidantes predijo adeuadamente el éxito de anidaión. Las aves que estaan anidando en los sitios ontaminados tuvieron tres vees más proailidades de aandonar sus nidos que las aves de sitios de referenia no ontaminados. Brindamos informaión sore una serie de onentraiones on efetos asoiados on varios niveles de difiultades reprodutivas; por ejemplo, una reduión del 10% del éxito de anidaión orrespondió a onentraiones de 0.7 μg g 1 de merurio en la sangre, 2.4 μg g 1 de merurio en las plumas del uerpo, 3.0 μg g 1 de merurio en las plumas de la ola y 0.11 μg g 1 de merurio en los huevos. Este es el primer estudio de ampo que doumenta el efeto de onentraiones espeífias de merurio en la sangre de aves anoras adultas sore su desempeño reprodutivo. Nuestros resultados demuestran que las aves anoras silvestres pueden sufrir efetos reprodutivos negativos ante onentraiones relativamente ajas de merurio. Merury is a persistent and dispersive environmental ontaminant found in many eosystems around the world. Merury released from industry often finds its way into aquati systems, where it has long residene times and an ioaumulate in aquati food wes (Evers et al. 2005). Most avian ioaumulation studies have examined fish-eating speies that are diretly linked with aquati eosystems and eat at high trophi levels (Sheuhammer et al. 2007, Seewagen 2010). The Common Loon 5 allyson.jakson@riloon.org The Auk, Vol. 128, Numer 4, pages ISSN , eletroni ISSN y The Amerian Ornithologists Union. All rights reserved. Please diret all requests for permission to photoopy or reprodue artile ontent through the University of California Press s Rights and Permissions wesite, om/reprintinfo.asp. DOI: /auk

3 760 JACKSON ET AL. AUK, VOL. 128 (Gavia immer) is frequently used as an indiator speies eause of its status as an upper-level predator with a wholly aquati diet. Common Loons display ~40% redution in nest suess at lood merury onentrations of 3.0 μg g 1 (wet weight) (Burgess and Meyer 2008, Evers et al. 2008). It has reently eome apparent that merury an ontaminate terrestrial eosystems, where it iomagnifies up the food hain just as in aquati eosystems. Terrestrial songirds that feed primarily on predatory inverterates suh as spiders an ioaumulate merury in their tissues at onentrations similar to pisivorous irds eause of iomagnifiation y inverterates (Cristol et al. 2008). Little is known aout the effets of merury exposure on terrestrial-feeding songirds. Studies looking at the reprodutive effets of environmental merury have foused on Tree Swallows (Tahyineta iolor), whih are a model organism in eotoxiology (Jones 2003). Tree Swallows living at merury-ontaminated sites exhiited moderate redution in reprodutive suess assoiated with merury exposure, more so when stressed y drought or heat (Brasso and Cristol 2008, Hallinger and Cristol 2011). However, Tree Swallows are not representative of most songirds; as migrants they may winter in areas with low environmental merury, allowing them to depurate merury taken in during the reeding season, and as aerial insetivores they do not feed on spiders or other terrestrial predatory insets. On the asis of these oservations, it is not unexpeted that some songird speies ioaumulate merury at higher onentrations than Tree Swallows and may e affeted more. For example, near a meruryontaminated river in Virginia, Carolina Wrens (Thryothorus ludoviianus; hereafter wrens ) had lood and feather merury onentrations similar to those in aquati-feeding irds (Cristol et al. 2008) and had elevated merury levels 140 km downstream from the soure of aquati merury (Jakson et al. 2011). Although eotoxiologial studies foused on Tree Swallows are generally designed to ompare nest suess in treatment groups to ontrols (Custer et al. 2007, Brasso and Cristol 2008), the analysis of nest suess in songird speies other than Tree Swallows rings several added ompliations. First, for speies that do not nest in artifiial avities (e.g., nest oxes), nests are typially disovered at different stages of the nesting yle, making the duration of time that eah nest is under oservation a soure of error originally identified y Mayfield (1961, 1975) likely to influene estimates of suess rates. Seond, eause of logistial onstraints, nests in wild populations are often monitored at irregular intervals, neessitating the estimation of dates of fledging or failure. Finally, suess of nests in natural sites tends to e influened y more fators than those in nest oxes. Daily nest survival rates may vary y date or age of nest (Dinsmore et al. 2002, Etterson et al. 2007) as well as a long list of other fators (Johnson 2007). Few eotoxiologial studies have taken into aount the multiple fators, in addition to the toxin under study, that influene nesting suess in wild irds. Current nest-survival estimation methods used y researhers to examine the influenes of hypothesized ovariates on nest survival impliitly treat all auses of failure as equivalent in their effets (Dinsmore et al. 2002, Shaffer 2004). However, given the diversity of auses of nest failure (Heisey and Patterson 2006) it is logial to expet that ovariates that influene rates of failure due to one ause (e.g., predation) may e different from ovariates influening rates of failure due to another ause (e.g., aandonment or adverse weather; Etterson et al. 2007a, ). Although ornithologists have reognized this issue for some time (reviewed y Johnson 2007), generalized software for modeling these ompeting risks and their relative effets in the presene of disovery ias and irregular monitoring shedules has een laking. We modeled nest suess using a novel software program that an take into aount ovariates thought to affet nest survival in wild populations (e.g., merury exposure, age of nest, time in season, year, and type of nest avity) and doument the relative influene of these fators on various auses of nest failure. Using all of these ovariates set within an information-theoreti approah, our goals were to (1) determine the relative influene of merury exposure ompared to other ovariates and model the differene in nest survival etween ontaminated and referene treatment groups; (2) use female lood merury onentrations to model nest survival ased on individual merury ody urdens; (3) ompare ause-speifi nest failure rates etween merury ontaminated and referene populations; and (4) alulate preliminary estimates of merury onentrations in lood, feathers, and eggs assoiated with various levels of reprodutive failure. METHODS Study speies. Wrens are monogamous and territorial passerines that nest up to three times per year in our study region (Haggerty and Morton 1995). Both males and females remain on their territories year round, making them prime andidates to e indiators of ontaminants on a small geographi sale. Blood merury onentrations reflet reent dietary exposure to merury (Frenh et al. 2010). Feather samples represent exposure to merury >8 months previous eause wrens molt from July through Septemer, after their reeding season is omplete; therefore, with the exeption of aidentally lost feathers that are regrown within the year, feather values reported here likely reflet exposure during the previous year (Haggerty and Morton 1995, Pyle 1997). Most wrens likely experiene similar merury onentrations year round. Study area. We monitored wrens along two meruryontaminated rivers in Virginia: the North Fork Holston and South rivers (Fig. 1). The two rivers are in different watersheds and, historially, oth were ontaminated with merury as the result of industrial disharge. The floodplain forest surrounding these rivers is sujet to seasonal flooding and is lassified as a forested wetland (Cowardin et al. 1979). Given that all monitored territories autted a river, the irds oupying these territories likely were exposed to dietary uptake of merury within the floodplain forest. The North Fork Holston River (NFHR) flows through Smyth and Washington Counties in southwestern Virginia. The NFHR joins the South Fork Holston River to form the Holston River near Virginia s southern order with Tennessee. Contamination ourred etween 1950 and 1972 at a hlor-alkali faility in Saltville, Virginia. We monitored wrens on two sites downstream of the merury ontamination and three referene sites upstream (Fig. 1B). All nests inluded in our analysis were within 50 m of the shore of the river. The South River (SR) flows through Rokingham and Augusta ounties in entral Virginia. The SR joins the North River to form the South Fork Shenandoah River, whih ultimately eomes the Shenandoah River. Contamination ourred etween

4 OCTOBER 2011 EFFECT OF MERCURY ON SONGBIRD REPRODUCTIVE SUCCESS 761 FIG. 1. Carolina Wren study sites ( ) in Virginia. (A) Two river systems where Carolina Wrens were monitored in Virginia; (B) study sites along the North Fork Holston River, where ontamination ourred at Saltville (star); and (C) study sites along the South and Middle rivers where ontamination ourred at Waynesoro (star). Eah river flows in the diretion of the lak arrows. Contaminated sites are shown in lak, and referene sites are shown in gray and 1950 at a textile fatory in Waynesoro, Virginia. We monitored wren territories at 11 sites downstream of the soure of ontamination and 11 referene sites, either upstream of the point soure along the SR itself (three sites), on a triutary of the SR (Bak Creek, six sites), or on the Middle River (two sites) (Fig. 1C). Birds on these referene sites have een shown in other studies to have low levels of merury exposure onsistent with akground atmospheri deposition (Brasso and Cristol 2008, Cristol et al. 2008). All nests inluded in our analysis were within 200 m of the shore of the river. Field methods. We ereted nest oxes ( m) or nest tues (46 10 m diameter) within floodplain forest ontaining resident wrens on eah river to enourage wrens to nest where we ould aess them. We mounted nest oxes on poles ~1.5 m off the ground and fit them with stovepipe predator guards, ut these proved ineffetive in our study. We fashioned nest tues from flexile plasti irrigation pipe with a hole on one end for the ird to enter and a removale over on top to allow nest heks. We attahed the nest tues diretly to trees, ~1 m off the ground, and did not install predator guards. We monitored nest oxes and tues starting on 1 April 2007, 10 April 2008, 7 April 2009, and 15 Marh 2010; we heked the nest oxes and tues weekly until a nest was initiated and then more often to monitor reprodutive suess. We did not atively searh for nests in natural nest avities from 2007 through 2009, ut if we disovered nests in natural avities during nest-ox heks they were also monitored. In 2010, field teams onduted intensive ehavioral oservations to find and monitor natural nest avities. Oservers spent 2 to 3 hours week 1 on eah territory where no nest had een initiated in a nest ox to oserve adult ehavior and loate natural nest avities. We heked nests approximately every 1 to 3 days to reord the state of the nest, inluding determining whether the female was inuating, how many eggs or hiks were present, and the approximate age of the nestlings ased on size and feathering. In 2010, we also used motionsensing infrared video ameras (Bushnell Corporation, Overland Park, Kansas) to determine auses of nest aandonment or predation. These motion ativated ameras were set up after inuation had started and reorded video in 10-s intervals when ativated y movement. Territorial irds were aptured early in the season (efore a nest was found) using playak reordings and identified as male y the presene of a loaal protuerane. Beause we wanted lood and feather samples assoiated with partiular nest attempts we reaptured and sampled oth parents one a nest was found. In order to minimize the risk of nest aandonment we aptured parents at the nest y mist net only after nestlings had hathed (average 7 days posthath; range: day 1 to day 14). We olleted either the two outermost tail feathers or 5 to 10 ody feathers sampled from a similar area on the ak. We used 26- to 28-gauge needles to punture the utaneous ulnar vein of the wing and olleted lood in heparinized apillary tues sealed at oth ends with Critoaps and plaed in 10-m 3 plasti tues for protetion. We plaed samples on ie in a ooler and they were frozen at 25 C within 6 h of olletion. We anded the adult irds with a federal metal and and a omination of one or two olor ands. Males had already een aptured and sexed earlier in the season and we determined females y the presene of a rood path. In 2009 and 2010, we olleted any inviale eggs left after all nestlings hathed or a nest failed, plaed them in glass jars, and stored them on ie until taken to the laoratory. The egg ontents (alumen and yolk) were extrated from the shell, weighed, and frozen at 25 C within 6 h of olletion. We weighed egg ontents efore and after freezing and the loss of weight was not statistially signifiant. Merury. Prior to analysis we sprayed eah feather with deionized water for 1 min to remove surfae partiles and dried them at low humidity for 48 h, therey returning the leaned feather to approximately wet weight. We ut up and homogenized the entire

5 762 JACKSON ET AL. AUK, VOL. 128 feather (inluding rahis) prior to analysis. We analyzed all lood and feather samples diretly from the thawed olletion ontainers without freeze drying; we report oth lood and feather samples as wet weight. We freeze dried egg ontents (alumen and yolk) prior to analysis. Beause eggs lost little moisture during the freezing proess we alulated merury onentrations using the postfreezing wet weight, as this ould e done on the same laoratory alane, reduing instrument-derived variane. Egg merury onentrations are reported as wet weight. All onentrations are for total merury, whih is a proxy for methylmerury eause 95 99% of total merury found in avian lood, feathers, and eggs onsists of methylmerury (Rimmer et al. 2005, Bond and Diamond 2009). Determination of total merury onentration of avian tissues ourred at four laoratories: Trae Element Researh La at Texas A&M University, College Station, using a Diret Merury Analyzer (2007 lood samples; DMA-80; Milestone, Shelton, Connetiut); Center for Environmental Sienes and Engineering at the University of Connetiut (2008 lood samples, using EPA method with a flow injetion merury system; Perkin Elmer, Milford, Connetiut; 2009 and 2010 egg samples, Milestone DMA-80); College of William and Mary, Williamsurg, Virginia (2009 lood samples, Milestone DMA-80); and Biodiversity Researh Institute, Gorham, Maine (2010 lood and feather samples, Milestone DMA-80). Avian tissue samples were analyzed over a 4-year period at these four laoratories and qualityassurane data for the narrow periods when wren samples were run met aeptale standards (Appendix). In general, efore and after every set of 20 samples, two samples eah of two standard referene materials (SRMs, DORM, and DOLT; National Researh Counil, Ontario, Canada), two methods lanks, and two sample lanks were run, and reovery of SRMs was ~100% (Appendix). During the periods of merury determination we spiked samples of lood or egg expeted to have low merury onentrations with SRM to measure reovery in the appropriate matrix. We reovered lose to 100% of the added merury (Appendix). We inluded approximately one pair of samples from the same ird with every 20 samples and otained relative perent differenes etween dupliates that were onsistently <10% (Appendix). All lood and feather samples were well aove minimum instrument detetion limits. Statistial analysis. We used EXCEL (Mirosoft, Redmond, Washington) and JMP (SAS Institute, Cary, North Carolina) for general statistial analysis. We log transformed lood merury values to normalize the data and heked for normality within the ontaminated and referene groups using a Shapiro-Wilk test of normality. We ompared lood merury onentrations for eah treatment (ontaminated vs. referene) and year omination for all adult (after-hath-year, AHY) wrens using a one-way analysis of variane (ANOVA) followed y Tukey s HSD post ho test. Using inviale eggs olleted on the SR in 2009 and 2010, we ompared the relationship etween attending female lood merury onentration and average inviale egg merury onentrations using linear regression. In 2010, we intensified sampling effort and were therefore ale to investigate the relationship etween male and female lood merury onentrations using a linear regression for adult pairs that were aught while attending the same nest. We used the regression alulated from this analysis to reate post ho estimates of female lood merury for nests at whih we had aptured the male ut were unale to apture the attending female. For all aptured adults we explored the relationship etween lood and feather merury, plotting a linear regression etween logtransformed lood merury onentration and log-transformed tail or ody feather merury. All figures show ak-transformed data. As disussed previously, many eotoxiologial studies fail to aount for iologial ovariates known to influene nest suess. Several good omputer programs (White and Burnham 1999, Dinsmore et al. 2002) and ode for statistial software (Stanley 2000, 2004; Shaffer 2004) already exist for estimating daily survival rates as a funtion of ovariates, along with aounting for disovery ias and irregular hek shedules. These programs depend on a inomial distriution of outomes (suess or failure) and fewer methods exist for nests with multinomial outomes (suess, depredated, adverse weather, aandonment, et.), espeially when the latter require analysis of ovariates. When nests are visited daily, multinomial logisti regression may e used (Thompson and Burhans 2004). However, when nest visitation is variale the mathematial aounting requires speial handling. We therefore used a novel program alled MCESTIMATE to estimate daily and overall proailities of nest failure due to speifi auses (ompeting risks) using the Markov hain algorithms desried y Etterson et al. (2007a, ) within a user-friendly graphi user interfae. It is programmed in MATLAB (Mathworks 2009) and ompiled as a stand-alone program. Like other urrent nestsurvival estimation methods, MCESTIMATE is a generalization of Mayfield (1961, 1975) methods for estimating daily proailities of failure. When nest outomes are lassified inomially, the likelihood funtion employed y MCESTIMATE is equivalent to that of Johnson (1979) and Bart and Roson (1982), whih underlies the nest-survival algorithm in Program MARK (White and Burnham 1999, Dinsmore et al. 2002), and to logisti exposure (Shaffer 2004). We used MCESTIMATE within an informationtheoreti framework to analyze the nest survival data from 2007 through 2010, thus determining whether merury ontamination had more effet than other variales likely to affet nest survival. In addition to merury ontamination we identified five variales that might influene nest survival: year, date in season, time sine egg laying, avity type (ox, tue, or natural), and river system (SR or NFHR). We were unale to inlude individual sites within eah river system as a ovariate eause of low sample size of nests at many sites and the fat that site oundaries were aritrary, related to aess and ownership rather than iology (e.g., nests on the periphery of some sites were loser to nests on other sites than they were to nests on their own sites). To determine the relative effet of eah variale we first ran six univariate models and seleted the minimal set of ovariates that aounted for 95% of the model weights. We then used this suset of ovariates to reate a logial model set to use within an information-theoreti framework (Akaike s information riterion orreted for small sample sizes; AIC ). We evaluated eah model within this AIC framework. If multiple models appeared to have similar AIC weights, indiating that more than one model ould explain variation in nest suess, we model averaged aross all models to alulate predited nest survival rates etween treatment and ontrol nests. This

6 OCTOBER 2011 EFFECT OF MERCURY ON SONGBIRD REPRODUCTIVE SUCCESS 763 also required speifying values for the other two ovariates that ourred in this model set: avity type (set equal to natural nest) and date (set equal to 24 May, the average luth initiation date in our sample). We then alulated model-averaged preditions for treatment and ontrol nests using Akaike weights following Burnham and Anderson (2002). We alulated perent redution from the suess of referene nests using the following equation: perent redution = [(suess Ref suess Cont )/ suess Ref ]. To etter understand the effet of avity type we held treatment group (set equal to referene) and date (set equal to 24 May) onstant and model-averaged the effet of avity type aross all models. Effetive sample size was alulated following Rotella et al. (2004). We assumed that all these nests were independent data points eause even for known renesting or seond luth attempts nestox loations had hanged. Other nests may have een renests as well if pairs esaped detetion during their first attempt or moved into the study area, so we made no attempt to distinguish etween nesting attempts and assumed that inlusion of date in season as a ovariate would suffie to apture any effet of renesting or doule luthing. In 2010, we intensified effort to otain lood merury onentrations for eah attending female. The larger sample size allowed us to examine the relationship etween individual merury onentration and reprodutive suess in 2010 as opposed to omparing referene and ontaminated groups. In ases where we were unale to apture attending females we estimated female lood onentrations using the linear regression etween territorial pairs desried aove. We looked at four variales separately avity type (ox, tue, or natural), date in season, time sine laying, and female lood merury onentration and seleted the minimal set of ovariates that aounted for 95% of the model weights. In 2010, we used oservations from nest ameras to lassify nest fate into several ategories: fledged, failed due to predation, and failed due to aandonment. Fledged nests were those oserved ative late in the nesting yle (usually days posthath) and then found empty with no sign of disturane and with parents oserved feeding young fledglings. Nests that were depredated had nestlings or eggs missing, and often the nest was distured. In some ases we oserved the predator on video. Aandoned nests were those in whih eggs or nestlings were found left in the nest, with no evidene of a predator. Aandonment may have ourred for multiple, ut unknown, reasons (e.g., egg infertility, anormal inuation or feeding ehavior, adult mortality, food availaility, weather, or merury toxiity in eggs or nestlings). We used MCESTIMATE to estimate and ompare ause-speifi failure rates for these fates etween ontaminated and referene for 2010 data only. Effets onentrations. We used the MCESTIMATE model that alulated nest suess at various female lood merury onentrations as the asis for our extrapolation of effets onentrations in lood, feather, and eggs. Using the three regression equations alulated for the relationship etween wren lood merury onentrations and ody feathers, tail feathers, or egg onentrations, we alulated the onentration in eah tissue assoiated with different levels of reprodutive failure. We report all error estimates as SE. RESULTS Nest sampling. We monitored 6 wren nests in 2007 along the NFHR and 9 nests in On the SR, we monitored 29 nests in 2009 and 45 in All ut one nest in 2007 through 2009 were in nest oxes or tues; in 2010, we monitored 27 natural nests. Cluth initiation dates ranged from 23 Marh to 7 August. We monitored renesting attempts for the same anded female 19 times (4 in 2008, 2 in 2009, 13 in 2010). Suessful nests did not vary in the numer of fledglings produed etween treatment groups (Kruskal-Wallis, χ 2 = 0.54, df = 1, P = 0.46) or among years (Kruskal-Wallis, χ 2 = 1.75, df = 3, P = 0.63). On average, suessful nests produed 4.0 ± 0.15 fledglings (range: 1 6, n = 49). Merury onentrations. Log transformed lood merury values for all adult irds were normally distriuted within ontaminated and referene populations (Shapiro-Wilk Test, ontaminated: W = 0.990, P = 0.73; referene: W = 0.988, P = 0.58; n = 95). There were signifiant differenes etween treatment groups and year (one-way ANOVA, F = , df = 7 and 174, P < 0.001, n = 87). Blood merury onentrations were signifiantly elevated aove referene in eah year (Fig. 2). NFHR and SR irds had similar annual patterns of lood merury onentrations, with the exeption of a signifiant differene etween lood merury onentrations in 2007 (x = 2.69 ± 1.1 μg g 1, n = 30; all territories were <50 m from river) and 2010 (x = 1.74 ± 1.1 μg g 1, n = 33; some territories were m from river; Fig. 2). We sampled oth sexes in eah year, ut low sample sizes prevented statistial omparison y sex (Tale 1). In 2010 only, we sampled a high proportion of the male wrens assoiated with nests and found a strong relationship etween male and female lood merury at the same nest (F = 385.2, P < 0.001, n = 16; Fig. 3). FIG. 2. Mean ak-transformed adult Carolina Wren lood merury levels for eah treatment group and year (ww = wet weight). The North Fork Holston River was sampled in 2007 and 2008 and the South River was sampled in 2009 and Different letters indiate signifiant differenes etween the groups with analysis of variane (P < 0.05) and numers indiate sample size. Error ars represent ak-transformed 95% onfidene intervals.

7 764 JACKSON ET AL. AUK, VOL. 128 FIG. 3. A plot of the relationship etween lood merury onentrations of adult Carolina Wren pairs at the same nest in 2010 in Virginia (ww = wet weight). Adult Carolina Wren male and female (sampled from the same nest) lood merury onentrations show a strong positive linear relationship (r 2 = 0.97, F = 385.2, P < 0.001). In 2010, we sampled tail and ody feathers (presumaly grown at the end of the previous reeding season) from adult wrens and found a strong relationship etween lood and ody feather merury (r 2 = 0.88, F = 133.1, P < 0.001, n = 21; Fig. 4) and a signifiant ut weaker relationship etween merury in lood and tail feathers (r 2 = 0.56, F = 64.5, P < , n = 53). Although we had a small sample size, we found a strong relationship etween attending female lood merury onentration and average inviale egg total merury (r 2 = 0.89, F = 41.51, P = 0.001, n = 7). Nest survival. Our analysis of nest survival in MCESTI- MATE was ased on an effetive sample size of 1,371 days. Three ovariates aounted for >95% of the model weights: treatment group (w i = 0.81), avity type (w i = 0.11), and date (w i = 0.04). Time sine laying (w i = 0.02), river system (w i = 0.02), and year (w i < 0.001) ontriuted little additional explanatory power and were exluded from the susequent analyses (data not shown). Using the three top ovariates (treatment group, avity type, and date), we designed a set of eight andidate models that we hypothesized ould explain the variation in survival and found support for many models inluding the null model within the AIC framework (Tale 2). Although four models inluding the null model had ΔAIC sores <2, the total weight of models that ontained treatment group was 0.88, more than twie the total weight of models that inluded either date (0.38) or avity type (0.24). After running eah model with these parameters set to average values we model-averaged the effet of treatment aross all models (Burnham and Anderson 2002). The overall 30- day nest survival for referene nests was ± (n = 42), whereas that for nests at ontaminated sites was ± (n = 46), a redution of 34%. Beause avity type was the ovariate with the seond highest weight, we also wanted to test whether there were differenes in survival etween the three avity types. We oserved oth predation and aandonment in all types of nest avities. We saw little differene in overall nest suess among the different avity types, with natural nests (0.602 ± 0.110, n = 28) having slightly higher survival than tue nests (0.534 ± 0.122, n = 14) or ox nests (0.552 ± 0.101, n = 46). We used MCESTIMATE to model nest survival in 2010 on the asis of individual female lood merury onentrations eause that was the only year for whih we had adequate sample size (effetive sample size = 581 days). The female lood merury onentrations used in the model ranged from 0.07 to 3.48 μg g 1 (x = 1.11 ± 0.17, n = 40). The maximum female lood merury onentration for a aptured female in 2010 was 3.22 μg g 1 (n = 30; Tale 1); for nests where we ould not ath the attending female (n = 10) we estimated female lood merury onentrations from that of the male. Four nests were exluded eause of missing lood merury data (two ontaminated and two referene). We again ran univariate models to determine the relative influene of ovariates (female merury, avity type, date, and time sine laying). The top two ovariates aounted for >95% of the model weights (female merury w i = 0.93; avity type w i = 0.03) and so were seleted to e inluded in the model set while the lower-ranked ovariates (time sine laying w i = 0.03; date w i = 0.01) were exluded. From female merury and avity type ovariates we reated four andidate models and found support for the two top-ranked models (Tale 3). Models ontaining female lood merury onentration TABLE 1. Mean lood merury onentrations for female and male adult Carolina Wrens in eah year and treatment group in Virginia. Female Male River and year a Treatment n Mean SD Maximum Minimum n Mean SD Maximum Minimum NFHR 2007 NFHR 2008 SR 2009 SR 2010 C R C R C R C R a NFHR = North Fork Holston River, SR = South River. Treatment group: C = ontaminated, R = referene.

8 OCTOBER 2011 EFFECT OF MERCURY ON SONGBIRD REPRODUCTIVE SUCCESS 765 TABLE 2. AIC ranking for treatment-group andidate model set for Carolina Wren nests monitored etween 2007 and 2010 on the North Fork Holston and South rivers in Virginia. Models are ranked y inreasing ΔAIC sores, showing that the treatment effet appears in the top three ranked models and aounts for >80% of model weights. Model K a ΔAIC w i Treatment Treatment + date Treatment + avity Cavity Null Cavity + treatment + date Date Cavity + date a Numer of parameters. Saled AIC ; ΔAIC = 0.00 is interpreted as the est fit to the data among all models. Weight of evidene interpreted as a proportion. Weights aross all models sum to TABLE 3. AIC model rankings for female lood merury urden analysis for Carolina Wren nests monitored in 2010 on the South River, Virginia. Models are ranked y inreasing ΔAIC sores, showing that the effet of female lood merury appears in the top two ranked models and aounts for >90% of model weights. Model K a ΔAIC w i Merury Merury + avity Null Cavity FIG. 4. Relationship etween untransformed Carolina Wren lood (ww = wet weight) and merury onentrations (ww) of (A) ody feathers (log transformed, r 2 = 0.88, F = 133.1, P < 0.001) or (B) tail feathers (log transformed, r 2 = 0.56, F = 64.5, P < ) olleted from irds in 2010 in Virginia. Equation on eah graph indiates relationship ased on linear regression analysis. aounted for 93% of weights and models ontaining avity type aounted for 24% of weights. Within the top-ranked model (effet of female lood merury alone) we found an effet of female lood merury on daily nest survival (on the multinomial logit sale, β = ± 0.072). We transformed this eta estimate into predited nest suess (assuming a 30-day nesting yle) ased on hypothesized female merury ody urden and found that as female lood merury inreased, modeled nest survival dereased (Fig. 5A). Beause we were interested in how individual lood merury onentrations affeted nest survival in relation to individuals with no merury exposure, we alulated perent redution in nest suess ompared with modeled females with 0 μg g 1 lood merury onentrations and found that nest suess dereased as lood merury onentration inreased (Fig. 5B). Competing risks. In 2010, out of 19 monitored ontaminated nests, 9 fledged, 3 were depredated, and 7 were aandoned. a Numer of parameters. Saled AIC ; ΔAIC = 0.00 is interpreted as the est fit to the data among all models. Weight of evidene interpreted as a proportion. Weights aross all models sum to On referene sites, 21 nests fledged young, 2 were depredated, and 2 others were aandoned. Video revealed predation y Amerian Blak Bears (Ursus amerianus) and ativity around failed nests y other potential nest predators, inluding Northern Raoons (Proyon lotor), Virginia Opossums (Didelphis virginiana), and Coyotes (Canis latrans). We suspet some instanes of snake predation ut did not apture this on video. Of the 9 aandoned nests, 4 ourred during the laying inuation stage and 5 ourred during the nestling stage. In some ases the parents renested later in the season, whih indiates that adult mortality had not aused aandonment, ut we do not have any other information to explain aandonment. By evaluating ompeting risks in MCESTIMATE we found that ause-speifi nest failure differed etween ontaminated and referene sites (effetive sample size = 615 days). The topranked model whih had 3 the support of the seond-ranked model revealed an important effet of treatment on the rate of aandonment ut no important effet of treatment on predation (Tale 4). Within the top-ranked model ontaminated nests were nearly 4 more likely to e aandoned (aandonment Cont = ± 0.111, aandonment Ref = ± 0.081) ut there was little

9 766 JACKSON ET AL. AUK, VOL. 128 TABLE 4. AIC model ranking for Carolina Wren nest survival in 2010 along the South River, Virgina, when nest survival was allowed to vary y ause etween models. Models are ranked y inreasing ΔAIC sores, showing a strong effet of aandonment within treatment groups, ut not predation. Model K a ΔAIC w i Aandonment{treatment} Predation{.} Aandonment{treatment} Predation{treatment} Aandonment{.} Predation{.} Aandonment{.} Predation{treatment} a Numer of parameters. Saled AIC ; ΔAIC = 0.00 is interpreted as the est fit to the data among all models. Weight of evidene interpreted as a proportion. Weights aross all models sum to TABLE 5. Carolina Wren lood, feather, and egg merury effets onentrations (ww = wet weight) assoiated with MCESTIMATE-modeled redution in nest suess. Results ased on data olleted in 2010 from nests along the South River in Virginia. Redution in nest suess a Blood merury (μg g 1, ww) Body feather merury (μg g 1, ww) Tail feather merury (μg g 1, ww) Egg merury (μg g 1, ww) d 10% % % % % % % % 3.8 e % 4.4 e % 5.6 e FIG. 5. The relationship etween MCESTIMATE-modeled Carolina Wren nest survival and female lood merury onentration for nests found in 2010 in Virgina. (A) Predited Carolina Wren nest suess over their 30- day nest yle in relation to female lood merury onentration when other ovariates were held onstant (date = 24 May, nest avity = natural). Error ars indiate SE. Dotted portion of the line indiates model extrapolation past oserved female lood merury onentrations. (B) Perent redution in nest survival (from nest survival at 0 μg g 1 ) in relation to female lood merury onentration. Blood merury onentrations assoiated with 10% inrements of redution in nest suess are shown. differene in predation etween treatment groups (predation Cont = ± 0.061, predation Ref = ± 0.078). Effets onentration. We found that the lood onentration assoiated with 10% redution in nest suess was 0.70 μg g 1, whih orrelated to 2.4 μg g 1 merury in ody feathers, 3.0 μg g 1 merury in tail feathers, and 0.11 μg g 1 merury in egg tissue (Tale 5). Extrapolating our model to higher merury values predits 99% redution in reprodutive suess at lood onentrations of 5.6 μg g 1, ody feather onentrations of 12.8 μg g 1, tail feather onentrations of 19.5 μg g 1, and egg onentrations of 0.97 μg g 1 (Tale 5). a Calulated using MCESTIMATE, omparing proaility of fledging at least 1 young at 0 μg g 1 to the proaility of fledging at least 1 young at eah ontaminated lood onentration. Calulated using the regression equation [ody feather Hg] = [lood Hg] Calulated using the regression equation [tail feather Hg] = [lood Hg] d Calulated using the regression equation [egg Hg] = [lood Hg] e Extrapolation past known lood merury levels using the MCESTIMATE model. DISCUSSION Carolina Wrens reeding in two merury-ontaminated watersheds in Virginia had 34% lower nest suess ompared with neary referene sites without histories of industrial merury ontamination. At an individual level, females with higher lood merury onentrations had lower nest suess; aross the range from 0 to 4.0 μg g 1, modeled suess dereased y 10% on average for eah projeted 0.50 μg g 1 inrease in lood merury onentration. Our analyses onsistently ranked merury ontamination as a leading preditor of nest suess, more so than date in season, year, avity type, age of nest, or river system. By using new software to model nest suess on the asis

10 OCTOBER 2011 EFFECT OF MERCURY ON SONGBIRD REPRODUCTIVE SUCCESS 767 of ompeting risks we were ale to show that nests on ontaminated sites were 3 more likely to fail eause of aandonment, whereas oth treatment groups experiened equal levels of nest predation. We estimated approximately 60% nest suess on referene sites and 40% nest suess on ontaminated sites, oth of whih are higher than the only other known pulished estimate for this speies: 26% of nests in northwestern Alaama produed at least one fledgling (n = 118; Haggerty and Morton 1995). In our female lood merury analysis, we estimate that females without methylmerury in their diet should fledge young ~80% of the time. This is higher than expeted for a songird, ut similar rates have een reported in other avity-nesting speies (Martin and Li 1992, Etterson et al. 2007). Beause we determined the merury onentrations of female parents (or estimated it from male onentrations) at the time of nesting, we were ale to model the response of nest survival to female merury onentration in This is the first study, to our knowledge, that douments a orrelation etween lood merury onentration and redution in nest survival in a free-living ird population. Several reent field studies have ompared the reprodutive suess of irds exposed to environmental merury with that of an unexposed referene group (Custer et al. 2007, Brasso and Cristol 2008). These studies, however, did not onsider nest suess on the asis of individual lood merury onentration. The maximum lood merury onentration for wrens in 2010 was 3.22 μg g 1 (wet weight) for females and 3.65 μg g 1 (wet weight) for males. Although it is possile to extrapolate the model to show almost omplete reprodutive failure (99%) at 5.6 μg g 1 in the lood, these extrapolations should e viewed with aution. It is possile that omplete reprodutive impairment ours at lower onentrations than those predited y our model; individuals with lood merury levels higher than those that we sampled may e unale to estalish territories or find mates, rendering their reprodutive output effetively zero, whih is a finding that annot e shown in our study. In other orders of irds, lood merury onentrations are affeted y depuration of merury into eggs (Beker 1992, Monteiro and Furness 2001, Kennamer et al. 2005, Frenh et al. 2010) or into feathers during molt (Nihols et al. 2010). We elieve that the timing of lood sample olletion within the nesting yle is not likely to have aused variaility in lood merury onentrations in wrens at this heavily ontaminated South River site. For example, merury transfer from lood into eggs y Tree Swallows reeding along a merury-ontaminated streth of the South River did not derease in the amount of merury transferred to eah susequent egg laid, whih indiates that intake of merury through prey was high enough to maintain ody urden despite the depuration into eggs (Brasso et al. 2010). Methylmerury depuration through molt also ompliates interpretation of lood merury levels, ut the wren is a year-round resident and does not start molting until after their nestlings have fledged. Therefore, molt would not have eliminated lood merury to growing feathers until after our sampling efforts. Aandonment rates were ~3 higher at ontaminated than at referene sites. There are several putative mehanisms y whih this aandonment may have ourred (some seen in other merury studies and some speulative), ut we are unale to distinguish etween them without further study. Some parents that aandoned nests may have suumed to lethal effets of merury, although determining whih ehavioral or physiologial effet is not possile. At the sulethal level, adults may have exhiited anormal inuation or feeding ehavior that led to nest loss. Common Loons exposed to environmental merury displayed aerrant inuation ehavior, leaving eggs unattended more often when merury onentrations were elevated (Evers et al. 2008). Nestlings may have ehaved anormally, for example voalizing less efore hathing or egging less, eliminating ues neessary to stimulate parental ehavior. Finally, merury an ause outright emryo mortality in songirds, in whih ase the doumented aandonments would e est explained as a response to egg inviaility rather than a ause (Heinz et al. 2009). Other mehanisms are possile, ut identifying aandonment rather than predation as a differential ause of redued nest survival is an important step forward in fousing future studies of the effets of merury on songird reprodution. Dosing studies may e an important next step, eause they an eliminate the risk of predation to fous on the mehanisms of aandonment. In a dosing study of Amerian Kestrels (Falo sparverius), researhers also found dereased nest suess ut were unale to onlude whether differenes were aused y anormal parental ehavior or disrupted egg and hik development (Alers et al. 2007). Wrens aumulate merury to higher onentrations than other songirds reeding in floodplain forests, likely eause of their heavy reliane on spiders in their diet (Cristol et al. 2008). Beause spiders feed at high trophi levels, they ioaumulate merury to higher onentrations than many other inverterate prey speies. Beause wrens remain on territories year round, they may e one of the more at-risk speies, ut the effets of merury reported here are iologially signifiant enough that migrants, and speies feeding lower on the food hain, may also e affeted. Although the sites studied here were oth industrial point soures, songirds living in areas remote from industry an also aumulate merury onentrations omparale to those that have effets on wren reprodution. These inlude Rusty Blakirds (Euphagus arolinus; Edmonds et al. 2010), Nelson s Sparrows (Ammodramus nelsoni; Winder and Emslie 2011), and Saltmarsh Sparrows (A. audautus; Lane et al. 2011). Conlusions. Our results have important impliations with regard to future regulation of merury pollution and mitigation of previously ontaminated sites. Our finding that terrestrial songirds exposed to merury exhiited high rates of nest aandonment, leading to sustantial redution in return on reprodutive effort, suggests that aquati merury pollution may harm terrestrial songirds in floodplain forest haitats near many of the thousands of water odies sujet to merury fishonsumption advisories in the United States (see water.epa.gov/ siteh/swguidane/fishshellfish/fishadvisories) and worldwide. This study is the first to link speifi merury onentrations in songirds to a quantifiale redution in nest suess, at least in part eause of inreased aandonment of nests. Understanding the ramifiations of merury ontamination for other speies and regions requires further investigation, partiularly for speies found in haitats sensitive to methylmerury prodution or experiening greater-than-usual physiologial demands suh as longdistane migration.

11 768 JACKSON ET AL. AUK, VOL. 128 ACKNOWLEDGMENTS The findings and onlusions in this artile are those of the authors and do not neessarily represent the views of the U.S. Fish and Wildlife Servie or the U.S. Environmental Protetion Ageny. North Fork Holston River researh was funded y the Olin Corporation. South River researh was funded y E.I. DuPont de Nemours and Company, and researh was ompleted with oversight from the South River Siene Team, whih is a ollaoration of state and federal agenies, aademi institutions, and environmental interests. We thank the field rews and numerous landowners who allowed us aess to their property, along with two anonymous reviewers who offered helpful omments on the manusript. LITERATURE CITED Alers, P. H., M. T. Kotera, R. Rossmann, W. A. Link, J. B. Frenh, R. S. Bennett, and W. C. Bauer Effets of methylmerury on reprodution in Amerian Kestrels. Environmental Toxiology and Chemistry 26: Bart, J., and D. S. Roson Estimating survivorship when the sujets are visited periodially. Eology 63: Beker, P. H Egg merury levels deline with the laying sequene in Charadriiformes. Bulletin of Environmental Contamination and Toxiology 48: Bond, A. L., and A. W. Diamond Total and methyl merury onentrations in seaird feathers and eggs. Arhives of Environmental Contaminants and Toxiology 56: Brasso, R. L., M. K. Adel Latif, and D. A. Cristol Relationship etween laying sequene and merury onentration in Tree Swallow eggs. Environmental Toxiology and Chemistry 29: Brasso, R. L., and D. A. Cristol Effets of merury exposure on the reprodutive suess of Tree Swallows (Tahyineta iolor). Eotoxiology 17: Burgess, N. M., and M. W. Meyer Methylmerury exposure assoiated with redued produtivity in Common Loons. Eotoxiology 17: Burnham, K. P., and D. R. Anderson Model Seletion and Multimodal Inferene: A Pratial Information-Theoreti Approah, 2nd ed. Springer-Verlag, New York. Cowardin, L. M., V. Carter, F. C. Golet, and E. T. LaRoe Classifiation of wetlands and deepwater haitats of the United States. U.S. Department of the Interior, Fish and Wildlife Servie, Washington, D.C. Cristol, D. A., R. L. Brasso, A. M. Condon, R. E. Fovargue, S. L. Friedman, K. K. Hallinger, A. P. Monroe, and A. E. White The movement of aquati merury through terrestrial food wes. Siene 320:335. Custer, C. M., T. W. Custer, and E. F. Hill Merury exposure and effets on avity-nesting irds from the Carson River, Nevada. Arhives of Environmental Contamination and Toxiology 52: Dinsmore, S. J., G. C. White, and F. L. Knopf Advaned tehniques for modeling avian nest survival. Eology 83: Edmonds, S. T., D. C. Evers, D. A. Cristol, C. Mettke-Hofmann, L. L. Powell, A. J. MGann, J. W. Armiger, O. P. Lane, D. F. Tessler, P. Newell, and others Geographi and seasonal variation in merury exposure of the delining Rusty Blakird. Condor 112: Etterson, M. A., L. R. Nagy, and T. Rodden-Roinson. 2007a. Partitioning risk among different auses of nest failure. Auk 124: Etterson, M. A., B. J. Olsen, and R. Greenerg The analysis of ovariates in multi-fate Markov hain nest-failure models. Pages in Beyond Mayfield: Measurements of Nest Survival Data (S. L. Jones and G. R. Geupel, Eds.). Studies in Avian Biology, no. 34. Evers, D. C., N. M. Burgess, L. Champoux, B. Hoskins, A. Major, W. M. Goodale, R. J. Taylor, R. Poppenga, and T. Daigle Patterns and interpretation of merury exposure in freshwater avian ommunities in northeastern North Ameria. Eotoxiology 14: Evers, D. C., L. J. Savoy, C. R. DeSoro, D. E. Yates, W. Hanson, K. M. Taylor, L. S. Siegel, J. H. Cooley, Jr., M. S. Bank, A. Major, and others Adverse effets from environmental merury loads on reeding Common Loons. Eotoxiology 17: Frenh, J. B., R. S. Bennett, and R. Rossmann Merury in the lood and eggs of Amerian Kestrels fed methylmerury hloride. Environmental Toxiology and Chemistry 29: Haggerty, T. M., and E. S. Morton Carolina Wren (Thryothorus ludoviianus). In The Birds of North Ameria Online (A. Poole, Ed.). Cornell La of Ornithology, Ithaa, New York. [Online.] Availale at na.irds.ornell.edu/na/speies/188. Hallinger, K. K., and D. A. Cristol The role of weather in mediating the effet of merury exposure on reprodutive suess in Tree Swallows. Eotoxiology 20: Heinz, G. H., D. J. Hoffman, J. D. Klimstra, K. R. Steins, S. L. Kondrad, and C. A. Erwin Speies differenes in the sensitivity of avian emryos to methylmerury. Arhives of Environmental Contamination and Toxiology 56: Heisey, D. M., and B. R. Patterson A review of methods to estimate ause-speifi mortality in presene of ompeting risks. Journal of Wildlife Management 70: Jakson, A. K., D. C. Evers, S. B. Folsom, A. M. Condon, J. Diener, L. F. Goodrik, A. J. MGann, J. Shmerfeld, and D. A. Cristol Merury exposure in terrestrial irds far downstream of an historial point soure. Environmental Pollution 159: Johnson, D. H Estimating nest suess: The Mayfield method and an alternative. Auk 96: Johnson, D. H Estimating nest suess: A guide to the methods. Pages in Beyond Mayfield: Measurements of Nest Survival Data (S. L. Jones and G. R. Geupel, Eds.). Studies in Avian Biology, no. 34. Jones, J Tree Swallows (Tahyineta iolor): A new model organism? Auk 120: Kennamer, R. A., J. R. Stout, B. P. Jakson, S. V. Colwell, I. L. Brisin, Jr., and J. Burger Merury patterns in Wood Duk eggs from a ontaminated reservoir in South Carolina, USA. Environmental Toxiology and Chemistry 24: Lane, O. P., K. M. O Brien, D. C. Evers, T. P. Hodgman, A. Major, N. Pau, M. J. Duey, R. Taylor, and D. Perry Merury in reeding Saltmarsh Sparrows (Ammodramus audautus audautus). Eotoxiology 20: in press.