MERCURY EXPOSURE AFFECTS THE REPRODUCTIVE SUCCESS OF A FREE-LIVING TERRESTRIAL SONGBIRD, THE CAROLINA WREN (THRYOTHORUS LUDOVICIANUS)

The Auk 128(4):759 769, 2011 The Americn Ornithologists Union, 2011. 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 Reserch Institute, Gorhm, Mine 04038, USA; 2 U.S. Fish nd Wildlife Service, Virgini Field Office, Gloucester, Virgini 23061, USA; 3 U.S. Environmentl Protection Agency, Office of Reserch nd Development, Ntionl Helth nd Environmentl Effects Reserch Lbortory, Mid-Continent Ecology Division, Duluth, Minnesot 55804, USA; nd 4 Institute for Integrtive Bird Behvior Studies, Deprtment of Biology, College of Willim nd Mry, Willimsburg, Virgini 23187, USA ABSTRACT. Despite mounting evidence of mercury ccumultion in terrestril ecosystems, few dt exist on how environmentl mercury exposure ffects reproductive success in free-living songbirds. From 2007 through 2010, we monitored reproductive success of Crolin Wrens (Thryothorus ludovicinus) breeding long the forest floodplin of two mercury-contminted rivers in Virgini. Using n informtion-theoretic pproch, we found 34% reduction in nesting success of Crolin Wrens on mercury-contminted sites when compred with reference sites. Blood mercury concentrtion of the ttending femle ws strong predictor of nest success. Birds nesting on contminted sites were 3 more likely to bndon their nests thn birds on uncontminted reference sites. We report rnge of effects concentrtions ssocited with vrious levels of reproductive impirment; for exmple, 10% reduction in nest success corresponded with 0.7 µg g 1 mercury in the blood, 2.4 µg g 1 mercury in body fethers, 3.0 µg g 1 mercury in til fethers, nd 0.11 µg g 1 mercury in eggs. This is the first field study to document the effect of specific dult songbird blood mercury concentrtions on breeding performnce; our results show tht free-living songbirds cn suffer negtive reproductive effects t reltively low mercury concentrtions. Received 13 My 2011, ccepted 23 August 2011. Key words: Crolin Wren, ecotoxicology, mercury, nest success, point-source pollution, Thryothorus ludovicinus, Virgini. L Exposición l Mercurio Afect el Éxito Reproductivo de Thryothorus ludovicinus, un Ave Cnor Terrestre Silvestre RESUMEN. A pesr de un cntidd creciente de evidenci sobre l cumulción de mercurio en los ecosistems terrestres, existen pocos dtos sobre cómo l exposición mbientl l mercurio fect el éxito reproductivo de ls ves cnors silvestres. Desde 2007 hst 2010, monitoremos el éxito reproductivo de individuos de l especie Thryothorus ludovicinus que estbn crindo lo lrgo de l plnicie boscos de inundción de dos ríos contmindos con mercurio en Virgini. Usndo un enfoque bsdo en teorí de l informción, encontrmos un reducción del 34% en el éxito reproductivo de T. ludovicinus en sitios contmindos con mercurio, comprdo con sitios de referenci. L concentrción de mercurio en l sngre de ls hembr nidntes predijo decudmente el éxito de nidción. Ls ves que estbn nidndo en los sitios contmindos tuvieron tres veces más probbiliddes de bndonr sus nidos que ls ves de sitios de referenci no contmindos. Brindmos informción sobre un serie de concentrciones con efectos socidos con vrios niveles de dificultdes reproductivs; por ejemplo, un reducción del 10% del éxito de nidción correspondió concentrciones de 0.7 µg g 1 de mercurio en l sngre, 2.4 µg g 1 de mercurio en ls plums del cuerpo, 3.0 µg g 1 de mercurio en ls plums de l col y 0.11 µg g 1 de mercurio en los huevos. Este es el primer estudio de cmpo que document el efecto de concentrciones específics de mercurio en l sngre de ves cnors dults sobre su desempeño reproductivo. Nuestros resultdos demuestrn que ls ves cnors silvestres pueden sufrir efectos reproductivos negtivos nte concentrciones reltivmente bjs de mercurio. MERCURY IS A persistent nd dispersive environmentl contminnt found in mny ecosystems round the world. Mercury relesed from industry often finds its wy into qutic systems, where it hs long residence times nd cn bioccumulte in qutic food webs (Evers et l. 2005). Most vin bioccumultion studies hve exmined fish-eting species tht re directly linked with qutic ecosystems nd et t high trophic levels (Scheuhmmer et l. 2007, Seewgen 2010). The Common Loon 5 E-mil: llyson.jckson@briloon.org The Auk, Vol. 128, Number 4, pges 759 769. ISSN 0004-8038, electronic ISSN 1938-4254. 2011 by The Americn Ornithologists Union. All rights reserved. Plese direct ll requests for permission to photocopy or reproduce rticle content through the University of Cliforni Press s Rights nd Permissions website, http://www.ucpressjournls. com/reprintinfo.sp. DOI: 10.1525/uk.2011.11106 759

760 JACKSON ET AL. AUK, VOL. 128 (Gvi immer) is frequently used s n indictor species becuse of its sttus s n upper-level predtor with wholly qutic diet. Common Loons disply ~40% reduction in nest success t blood mercury concentrtions of 3.0 µg g 1 (wet weight) (Burgess nd Meyer 2008, Evers et l. 2008). It hs recently become pprent tht mercury cn contminte terrestril ecosystems, where it biomgnifies up the food chin just s in qutic ecosystems. Terrestril songbirds tht feed primrily on predtory invertebrtes such s spiders cn bioccumulte mercury in their tissues t concentrtions similr to piscivorous birds becuse of biomgnifiction by invertebrtes (Cristol et l. 2008). Little is known bout the effects of mercury exposure on terrestril-feeding songbirds. Studies looking t the reproductive effects of environmentl mercury hve focused on Tree Swllows (Tchycinet bicolor), which re model orgnism in ecotoxicology (Jones 2003). Tree Swllows living t mercury-contminted sites exhibited moderte reduction in reproductive success ssocited with mercury exposure, more so when stressed by drought or het (Brsso nd Cristol 2008, Hllinger nd Cristol 2011). However, Tree Swllows re not representtive of most songbirds; s migrnts they my winter in res with low environmentl mercury, llowing them to depurte mercury tken in during the breeding seson, nd s eril insectivores they do not feed on spiders or other terrestril predtory insects. On the bsis of these observtions, it is not unexpected tht some songbird species bioccumulte mercury t higher concentrtions thn Tree Swllows nd my be ffected more. For exmple, ner mercurycontminted river in Virgini, Crolin Wrens (Thryothorus ludovicinus; herefter wrens ) hd blood nd fether mercury concentrtions similr to those in qutic-feeding birds (Cristol et l. 2008) nd hd elevted mercury levels 140 km downstrem from the source of qutic mercury (Jckson et l. 2011). Although ecotoxicologicl studies focused on Tree Swllows re generlly designed to compre nest success in tretment groups to controls (Custer et l. 2007, Brsso nd Cristol 2008), the nlysis of nest success in songbird species other thn Tree Swllows brings severl dded complictions. First, for species tht do not nest in rtificil cvities (e.g., nest boxes), nests re typiclly discovered t different stges of the nesting cycle, mking the durtion of time tht ech nest is under observtion source of error originlly identified by Myfield (1961, 1975) likely to influence estimtes of success rtes. Second, becuse of logisticl constrints, nests in wild popultions re often monitored t irregulr intervls, necessitting the estimtion of dtes of fledging or filure. Finlly, success of nests in nturl sites tends to be influenced by more fctors thn those in nest boxes. Dily nest survivl rtes my vry by dte or ge of nest (Dinsmore et l. 2002, Etterson et l. 2007b) s well s long list of other fctors (Johnson 2007). Few ecotoxicologicl studies hve tken into ccount the multiple fctors, in ddition to the toxin under study, tht influence nesting success in wild birds. Current nest-survivl estimtion methods used by reserchers to exmine the influences of hypothesized covrites on nest survivl implicitly tret ll cuses of filure s equivlent in their effects (Dinsmore et l. 2002, Shffer 2004). However, given the diversity of cuses of nest filure (Heisey nd Ptterson 2006) it is logicl to expect tht covrites tht influence rtes of filure due to one cuse (e.g., predtion) my be different from covrites influencing rtes of filure due to nother cuse (e.g., bndonment or dverse wether; Etterson et l. 2007, b). Although ornithologists hve recognized this issue for some time (reviewed by Johnson 2007), generlized softwre for modeling these competing risks nd their reltive effects in the presence of discovery bis nd irregulr monitoring schedules hs been lcking. We modeled nest success using novel softwre progrm tht cn tke into ccount covrites thought to ffect nest survivl in wild popultions (e.g., mercury exposure, ge of nest, time in seson, yer, nd type of nest cvity) nd document the reltive influence of these fctors on vrious cuses of nest filure. Using ll of these covrites set within n informtion-theoretic pproch, our gols were to (1) determine the reltive influence of mercury exposure compred to other covrites nd model the difference in nest survivl between contminted nd reference tretment groups; (2) use femle blood mercury concentrtions to model nest survivl bsed on individul mercury body burdens; (3) compre cuse-specific nest filure rtes between mercury contminted nd reference popultions; nd (4) clculte preliminry estimtes of mercury concentrtions in blood, fethers, nd eggs ssocited with vrious levels of reproductive filure. METHODS Study species. Wrens re monogmous nd territoril psserines tht nest up to three times per yer in our study region (Hggerty nd Morton 1995). Both mles nd femles remin on their territories yer round, mking them prime cndidtes to be indictors of contminnts on smll geogrphic scle. Blood mercury concentrtions reflect recent dietry exposure to mercury (French et l. 2010). Fether smples represent exposure to mercury >8 months previous becuse wrens molt from July through September, fter their breeding seson is complete; therefore, with the exception of ccidentlly lost fethers tht re regrown within the yer, fether vlues reported here likely reflect exposure during the previous yer (Hggerty nd Morton 1995, Pyle 1997). Most wrens likely experience similr mercury concentrtions yer round. Study re. We monitored wrens long two mercurycontminted rivers in Virgini: the North Fork Holston nd South rivers (Fig. 1). The two rivers re in different wtersheds nd, historiclly, both were contminted with mercury s the result of industril dischrge. The floodplin forest surrounding these rivers is subject to sesonl flooding nd is clssified s forested wetlnd (Cowrdin et l. 1979). Given tht ll monitored territories butted river, the birds occupying these territories likely were exposed to dietry uptke of mercury within the floodplin forest. The North Fork Holston River (NFHR) flows through Smyth nd Wshington Counties in southwestern Virgini. The NFHR joins the South Fork Holston River to form the Holston River ner Virgini s southern border with Tennessee. Contmintion occurred between 1950 nd 1972 t chlor-lkli fcility in Sltville, Virgini. We monitored wrens on two sites downstrem of the mercury contmintion nd three reference sites upstrem (Fig. 1B). All nests included in our nlysis were within 50 m of the shore of the river. The South River (SR) flows through Rockinghm nd August counties in centrl Virgini. The SR joins the North River to form the South Fork Shenndoh River, which ultimtely becomes the Shenndoh River. Contmintion occurred between

OCTOBER 2011 EFFECT OF MERCURY ON SONGBIRD REPRODUCTIVE SUCCESS 761 FIG. 1. Crolin Wren study sites (2007 2010) in Virgini. (A) Two river systems where Crolin Wrens were monitored in Virgini; (B) study sites long the North Fork Holston River, where contmintion occurred t Sltville (str); nd (C) study sites long the South nd Middle rivers where contmintion occurred t Wynesboro (str). Ech river flows in the direction of the blck rrows. Contminted sites re shown in blck, nd reference sites re shown in gry. 1929 nd 1950 t textile fctory in Wynesboro, Virgini. We monitored wren territories t 11 sites downstrem of the source of contmintion nd 11 reference sites, either upstrem of the point source long the SR itself (three sites), on tributry of the SR (Bck Creek, six sites), or on the Middle River (two sites) (Fig. 1C). Birds on these reference sites hve been shown in other studies to hve low levels of mercury exposure consistent with bckground tmospheric deposition (Brsso nd Cristol 2008, Cristol et l. 2008). All nests included in our nlysis were within 200 m of the shore of the river. Field methods. We erected nest boxes (10 10 25 cm) or nest tubes (46 10 cm dimeter) within floodplin forest contining resident wrens on ech river to encourge wrens to nest where we could ccess them. We mounted nest boxes on poles ~1.5 m off the ground nd fit them with stovepipe predtor gurds, but these proved ineffective in our study. We fshioned nest tubes from flexible plstic irrigtion pipe with hole on one end for the bird to enter nd removble cover on top to llow nest checks. We ttched the nest tubes directly to trees, ~1 m off the ground, nd did not instll predtor gurds. We monitored nest boxes nd tubes strting on 1 April 2007, 10 April 2008, 7 April 2009, nd 15 Mrch 2010; we checked the nest boxes nd tubes weekly until nest ws initited nd then more often to monitor reproductive success. We did not ctively serch for nests in nturl nest cvities from 2007 through 2009, but if we discovered nests in nturl cvities during nest-box checks they were lso monitored. In 2010, field tems conducted intensive behviorl observtions to find nd monitor nturl nest cvities. Observers spent 2 to 3 hours week 1 on ech territory where no nest hd been initited in nest box to observe dult behvior nd locte nturl nest cvities. We checked nests pproximtely every 1 to 3 dys to record the stte of the nest, including determining whether the femle ws incubting, how mny eggs or chicks were present, nd the pproximte ge of the nestlings bsed on size nd fethering. In 2010, we lso used motionsensing infrred video cmers (Bushnell Corportion, Overlnd Prk, Knss) to determine cuses of nest bndonment or predtion. These motion ctivted cmers were set up fter incubtion hd strted nd recorded video in 10-s intervls when ctivted by movement. Territoril birds were cptured erly in the seson (before nest ws found) using plybck recordings nd identified s mle by the presence of clocl protubernce. Becuse we wnted blood nd fether smples ssocited with prticulr nest ttempts we recptured nd smpled both prents once nest ws found. In order to minimize the risk of nest bndonment we cptured prents t the nest by mist net only fter nestlings hd htched (verge 7 dys posthtch; rnge: dy 1 to dy 14). We collected either the two outermost til fethers or 5 to 10 body fethers smpled from similr re on the bck. We used 26- to 28-guge needles to puncture the cutneous ulnr vein of the wing nd collected blood in heprinized cpillry tubes seled t both ends with Critocps nd plced in 10-cm 3 plstic tubes for protection. We plced smples on ice in cooler nd they were frozen t 25 C within 6 h of collection. We bnded the dult birds with federl metl bnd nd combintion of one or two color bnds. Mles hd lredy been cptured nd sexed erlier in the seson nd we determined femles by the presence of brood ptch. In 2009 nd 2010, we collected ny invible eggs left fter ll nestlings htched or nest filed, plced them in glss jrs, nd stored them on ice until tken to the lbortory. The egg contents (lbumen nd yolk) were extrcted from the shell, weighed, nd frozen t 25 C within 6 h of collection. We weighed egg contents before nd fter freezing nd the loss of weight ws not sttisticlly significnt. Mercury. Prior to nlysis we spryed ech fether with deionized wter for 1 min to remove surfce prticles nd dried them t low humidity for 48 h, thereby returning the clened fether to pproximtely wet weight. We cut up nd homogenized the entire

762 JACKSON ET AL. AUK, VOL. 128 fether (including rchis) prior to nlysis. We nlyzed ll blood nd fether smples directly from the thwed collection continers without freeze drying; we report both blood nd fether smples s wet weight. We freeze dried egg contents (lbumen nd yolk) prior to nlysis. Becuse eggs lost little moisture during the freezing process we clculted mercury concentrtions using the postfreezing wet weight, s this could be done on the sme lbortory blnce, reducing instrument-derived vrince. Egg mercury concentrtions re reported s wet weight. All concentrtions re for totl mercury, which is proxy for methylmercury becuse 95 99% of totl mercury found in vin blood, fethers, nd eggs consists of methylmercury (Rimmer et l. 2005, Bond nd Dimond 2009). Determintion of totl mercury concentrtion of vin tissues occurred t four lbortories: Trce Element Reserch Lb t Texs A&M University, College Sttion, using Direct Mercury Anlyzer (2007 blood smples; DMA-80; Milestone, Shelton, Connecticut); Center for Environmentl Sciences nd Engineering t the University of Connecticut (2008 blood smples, using EPA method 245.6 with flow injection mercury system; Perkin Elmer, Milford, Connecticut; 2009 nd 2010 egg smples, Milestone DMA-80); College of Willim nd Mry, Willimsburg, Virgini (2009 blood smples, Milestone DMA-80); nd Biodiversity Reserch Institute, Gorhm, Mine (2010 blood nd fether smples, Milestone DMA-80). Avin tissue smples were nlyzed over 4-yer period t these four lbortories nd qulityssurnce dt for the nrrow periods when wren smples were run met cceptble stndrds (Appendix). In generl, before nd fter every set of 20 smples, two smples ech of two stndrd reference mterils (SRMs, DORM, nd DOLT; Ntionl Reserch Council, Ontrio, Cnd), two methods blnks, nd two smple blnks were run, nd recovery of SRMs ws ~100% (Appendix). During the periods of mercury determintion we spiked smples of blood or egg expected to hve low mercury concentrtions with SRM to mesure recovery in the pproprite mtrix. We recovered close to 100% of the dded mercury (Appendix). We included pproximtely one pir of smples from the sme bird with every 20 smples nd obtined reltive percent differences between duplictes tht were consistently <10% (Appendix). All blood nd fether smples were well bove minimum instrument detection limits. Sttisticl nlysis. We used EXCEL (Microsoft, Redmond, Wshington) nd JMP (SAS Institute, Cry, North Crolin) for generl sttisticl nlysis. We log trnsformed blood mercury vlues to normlize the dt nd checked for normlity within the contminted nd reference groups using Shpiro-Wilk test of normlity. We compred blood mercury concentrtions for ech tretment (contminted vs. reference) nd yer combintion for ll dult (fter-htch-yer, AHY) wrens using one-wy nlysis of vrince (ANOVA) followed by Tukey s HSD post hoc test. Using invible eggs collected on the SR in 2009 nd 2010, we compred the reltionship between ttending femle blood mercury concentrtion nd verge invible egg mercury concentrtions using liner regression. In 2010, we intensified smpling effort nd were therefore ble to investigte the reltionship between mle nd femle blood mercury concentrtions using liner regression for dult pirs tht were cught while ttending the sme nest. We used the regression clculted from this nlysis to crete post hoc estimtes of femle blood mercury for nests t which we hd cptured the mle but were unble to cpture the ttending femle. For ll cptured dults we explored the reltionship between blood nd fether mercury, plotting liner regression between logtrnsformed blood mercury concentrtion nd log-trnsformed til or body fether mercury. All figures show bck-trnsformed dt. As discussed previously, mny ecotoxicologicl studies fil to ccount for biologicl covrites known to influence nest success. Severl good computer progrms (White nd Burnhm 1999, Dinsmore et l. 2002) nd code for sttisticl softwre (Stnley 2000, 2004; Shffer 2004) lredy exist for estimting dily survivl rtes s function of covrites, long with ccounting for discovery bis nd irregulr check schedules. These progrms depend on binomil distribution of outcomes (success or filure) nd fewer methods exist for nests with multinomil outcomes (success, depredted, dverse wether, bndonment, etc.), especilly when the ltter require nlysis of covrites. When nests re visited dily, multinomil logistic regression my be used (Thompson nd Burhns 2004). However, when nest visittion is vrible the mthemticl ccounting requires specil hndling. We therefore used novel progrm clled MCESTIMATE to estimte dily nd overll probbilities of nest filure due to specific cuses (competing risks) using the Mrkov chin lgorithms described by Etterson et l. (2007, b) within user-friendly grphic user interfce. It is progrmmed in MATLAB (Mthworks 2009) nd compiled s stnd-lone progrm. Like other current nestsurvivl estimtion methods, MCESTIMATE is generliztion of Myfield (1961, 1975) methods for estimting dily probbilities of filure. When nest outcomes re clssified binomilly, the likelihood function employed by MCESTIMATE is equivlent to tht of Johnson (1979) nd Brt nd Robson (1982), which underlies the nest-survivl lgorithm in Progrm MARK (White nd Burnhm 1999, Dinsmore et l. 2002), nd to logistic exposure (Shffer 2004). We used MCESTIMATE within n informtiontheoretic frmework to nlyze the nest survivl dt from 2007 through 2010, thus determining whether mercury contmintion hd more effect thn other vribles likely to ffect nest survivl. In ddition to mercury contmintion we identified five vribles tht might influence nest survivl: yer, dte in seson, time since egg lying, cvity type (box, tube, or nturl), nd river system (SR or NFHR). We were unble to include individul sites within ech river system s covrite becuse of low smple size of nests t mny sites nd the fct tht site boundries were rbitrry, relted to ccess nd ownership rther thn biology (e.g., nests on the periphery of some sites were closer to nests on other sites thn they were to nests on their own sites). To determine the reltive effect of ech vrible we first rn six univrite models nd selected the miniml set of covrites tht ccounted for 95% of the model weights. We then used this subset of covrites to crete logicl model set to use within n informtion-theoretic frmework (Akike s informtion criterion corrected for smll smple sizes; AIC c ). We evluted ech model within this AIC c frmework. If multiple models ppered to hve similr AIC weights, indicting tht more thn one model could explin vrition in nest success, we model verged cross ll models to clculte predicted nest survivl rtes between tretment nd control nests. This

OCTOBER 2011 EFFECT OF MERCURY ON SONGBIRD REPRODUCTIVE SUCCESS 763 lso required specifying vlues for the other two covrites tht occurred in this model set: cvity type (set equl to nturl nest) nd dte (set equl to 24 My, the verge clutch initition dte in our smple). We then clculted model-verged predictions for tretment nd control nests using Akike weights following Burnhm nd Anderson (2002). We clculted percent reduction from the success of reference nests using the following eqution: percent reduction = [(success Ref success Cont )/ success Ref ]. To better understnd the effect of cvity type we held tretment group (set equl to reference) nd dte (set equl to 24 My) constnt nd model-verged the effect of cvity type cross ll models. Effective smple size ws clculted following Rotell et l. (2004). We ssumed tht ll these nests were independent dt points becuse even for known renesting or second clutch ttempts nestbox loctions hd chnged. Other nests my hve been renests s well if pirs escped detection during their first ttempt or moved into the study re, so we mde no ttempt to distinguish between nesting ttempts nd ssumed tht inclusion of dte in seson s covrite would suffice to cpture ny effect of renesting or double clutching. In 2010, we intensified effort to obtin blood mercury concentrtions for ech ttending femle. The lrger smple size llowed us to exmine the reltionship between individul mercury concentrtion nd reproductive success in 2010 s opposed to compring reference nd contminted groups. In cses where we were unble to cpture ttending femles we estimted femle blood concentrtions using the liner regression between territoril pirs described bove. We looked t four vribles seprtely cvity type (box, tube, or nturl), dte in seson, time since lying, nd femle blood mercury concentrtion nd selected the miniml set of covrites tht ccounted for 95% of the model weights. In 2010, we used observtions from nest cmers to clssify nest fte into severl ctegories: fledged, filed due to predtion, nd filed due to bndonment. Fledged nests were those observed ctive lte in the nesting cycle (usully dys 12 14 posthtch) nd then found empty with no sign of disturbnce nd with prents observed feeding young fledglings. Nests tht were depredted hd nestlings or eggs missing, nd often the nest ws disturbed. In some cses we observed the predtor on video. Abndoned nests were those in which eggs or nestlings were found left in the nest, with no evidence of predtor. Abndonment my hve occurred for multiple, but unknown, resons (e.g., egg infertility, bnorml incubtion or feeding behvior, dult mortlity, food vilbility, wether, or mercury toxicity in eggs or nestlings). We used MCESTIMATE to estimte nd compre cuse-specific filure rtes for these ftes between contminted nd reference for 2010 dt only. Effects concentrtions. We used the MCESTIMATE model tht clculted nest success t vrious femle blood mercury concentrtions s the bsis for our extrpoltion of effects concentrtions in blood, fether, nd eggs. Using the three regression equtions clculted for the reltionship between wren blood mercury concentrtions nd body fethers, til fethers, or egg concentrtions, we clculted the concentrtion in ech tissue ssocited with different levels of reproductive filure. We report ll error estimtes s SE. RESULTS Nest smpling. We monitored 6 wren nests in 2007 long the NFHR nd 9 nests in 2008. On the SR, we monitored 29 nests in 2009 nd 45 in 2010. All but one nest in 2007 through 2009 were in nest boxes or tubes; in 2010, we monitored 27 nturl nests. Clutch initition dtes rnged from 23 Mrch to 7 August. We monitored renesting ttempts for the sme bnded femle 19 times (4 in 2008, 2 in 2009, 13 in 2010). Successful nests did not vry in the number of fledglings produced between tretment groups (Kruskl-Wllis, χ 2 = 0.54, df = 1, P = 0.46) or mong yers (Kruskl-Wllis, χ 2 = 1.75, df = 3, P = 0.63). On verge, successful nests produced 4.0 ± 0.15 fledglings (rnge: 1 6, n = 49). Mercury concentrtions. Log trnsformed blood mercury vlues for ll dult birds were normlly distributed within contminted nd reference popultions (Shpiro-Wilk Test, contminted: W = 0.990, P = 0.73; reference: W = 0.988, P = 0.58; n = 95). There were significnt differences between tretment groups nd yer (one-wy ANOVA, F = 149.97, df = 7 nd 174, P < 0.001, n = 87). Blood mercury concentrtions were significntly elevted bove reference in ech yer (Fig. 2). NFHR nd SR birds hd similr nnul ptterns of blood mercury concentrtions, with the exception of significnt difference between blood mercury concentrtions in 2007 (x = 2.69 ± 1.1 µg g 1, n = 30; ll territories were <50 m from river) nd 2010 (x = 1.74 ± 1.1 µg g 1, n = 33; some territories were 50 200 m from river; Fig. 2). We smpled both sexes in ech yer, but low smple sizes prevented sttisticl comprison by sex (Tble 1). In 2010 only, we smpled high proportion of the mle wrens ssocited with nests nd found strong reltionship between mle nd femle blood mercury t the sme nest (F = 385.2, P < 0.001, n = 16; Fig. 3). FIG. 2. Men bck-trnsformed dult Crolin Wren blood mercury levels for ech tretment group nd yer (ww = wet weight). The North Fork Holston River ws smpled in 2007 nd 2008 nd the South River ws smpled in 2009 nd 2010. Different letters indicte significnt differences between the groups with nlysis of vrince (P < 0.05) nd numbers indicte smple size. Error brs represent bck-trnsformed 95% confidence intervls.

764 JACKSON ET AL. AUK, VOL. 128 FIG. 3. A plot of the reltionship between blood mercury concentrtions of dult Crolin Wren pirs t the sme nest in 2010 in Virgini (ww = wet weight). Adult Crolin Wren mle nd femle (smpled from the sme nest) blood mercury concentrtions show strong positive liner reltionship (r 2 = 0.97, F = 385.2, P < 0.001). In 2010, we smpled til nd body fethers (presumbly grown t the end of the previous breeding seson) from dult wrens nd found strong reltionship between blood nd body fether mercury (r 2 = 0.88, F = 133.1, P < 0.001, n = 21; Fig. 4) nd significnt but weker reltionship between mercury in blood nd til fethers (r 2 = 0.56, F = 64.5, P < 0.0001, n = 53). Although we hd smll smple size, we found strong reltionship between ttending femle blood mercury concentrtion nd verge invible egg totl mercury (r 2 = 0.89, F = 41.51, P = 0.001, n = 7). Nest survivl. Our nlysis of nest survivl in MCESTI- MATE ws bsed on n effective smple size of 1,371 dys. Three covrites ccounted for >95% of the model weights: tretment group (w i = 0.81), cvity type (w i = 0.11), nd dte (w i = 0.04). Time since lying (w i = 0.02), river system (w i = 0.02), nd yer (w i < 0.001) contributed little dditionl explntory power nd were excluded from the subsequent nlyses (dt not shown). Using the three top covrites (tretment group, cvity type, nd dte), we designed set of eight cndidte models tht we hypothesized could explin the vrition in survivl nd found support for mny models including the null model within the AIC c frmework (Tble 2). Although four models including the null model hd ΔAIC c scores <2, the totl weight of models tht contined tretment group ws 0.88, more thn twice the totl weight of models tht included either dte (0.38) or cvity type (0.24). After running ech model with these prmeters set to verge vlues we model-verged the effect of tretment cross ll models (Burnhm nd Anderson 2002). The overll 30- dy nest survivl for reference nests ws 0.602 ± 0.111 (n = 42), wheres tht for nests t contminted sites ws 0.398 ± 0.129 (n = 46), reduction of 34%. Becuse cvity type ws the covrite with the second highest weight, we lso wnted to test whether there were differences in survivl between the three cvity types. We observed both predtion nd bndonment in ll types of nest cvities. We sw little difference in overll nest success mong the different cvity types, with nturl nests (0.602 ± 0.110, n = 28) hving slightly higher survivl thn tube nests (0.534 ± 0.122, n = 14) or box nests (0.552 ± 0.101, n = 46). We used MCESTIMATE to model nest survivl in 2010 on the bsis of individul femle blood mercury concentrtions becuse tht ws the only yer for which we hd dequte smple size (effective smple size = 581 dys). The femle blood mercury concentrtions used in the model rnged from 0.07 to 3.48 µg g 1 (x = 1.11 ± 0.17, n = 40). The mximum femle blood mercury concentrtion for cptured femle in 2010 ws 3.22 µg g 1 (n = 30; Tble 1); for nests where we could not ctch the ttending femle (n = 10) we estimted femle blood mercury concentrtions from tht of the mle. Four nests were excluded becuse of missing blood mercury dt (two contminted nd two reference). We gin rn univrite models to determine the reltive influence of covrites (femle mercury, cvity type, dte, nd time since lying). The top two covrites ccounted for >95% of the model weights (femle mercury w i = 0.93; cvity type w i = 0.03) nd so were selected to be included in the model set while the lower-rnked covrites (time since lying w i = 0.03; dte w i = 0.01) were excluded. From femle mercury nd cvity type covrites we creted four cndidte models nd found support for the two top-rnked models (Tble 3). Models contining femle blood mercury concentrtion TABLE 1. Men blood mercury concentrtions for femle nd mle dult Crolin Wrens in ech yer nd tretment group in Virgini. Femle Mle River nd yer Tretment b n Men SD Mximum Minimum n Men SD Mximum Minimum NFHR 2007 NFHR 2008 SR 2009 SR 2010 C 17 3.38 1.83 8.38 0.62 1 1.07 1.07 1.07 R 9 0.29 0.14 0.52 0.12 5 0.18 0.11 0.33 0.07 C 5 1.96 0.65 2.49 0.84 6 2.40 0.97 3.91 1.04 R 3 0.48 0.41 0.95 0.21 3 0.29 0.11 0.40 0.18 C 9 2.24 1.11 4.77 1.17 6 3.27 1.74 5.81 1.01 R 10 0.38 0.20 0.84 0.08 4 0.34 0.25 0.70 0.11 C 11 2.13 0.67 3.22 0.96 22 1.74 0.68 3.65 0.56 R 21 0.21 0.11 0.55 0.07 24 0.19 0.08 0.42 0.09 NFHR = North Fork Holston River, SR = South River. b Tretment group: C = contminted, R = reference.

OCTOBER 2011 EFFECT OF MERCURY ON SONGBIRD REPRODUCTIVE SUCCESS 765 TABLE 2. AIC c rnking for tretment-group cndidte model set for Crolin Wren nests monitored between 2007 nd 2010 on the North Fork Holston nd South rivers in Virgini. Models re rnked by incresing AIC c scores, showing tht the tretment effect ppers in the top three rnked models nd ccounts for >80% of model weights. Model K AIC c b w i c Tretment 2 0.00 0.39 Tretment + dte 3 0.20 0.32 Tretment + cvity 4 1.00 0.14 Cvity 3 1.99 0.05 Null 1 2.34 0.04 Cvity + tretment + dte 5 2.42 0.03 Dte 2 3.04 0.02 Cvity + dte 4 3.83 0.01 Number of prmeters. b Scled AIC c ; AIC c = 0.00 is interpreted s the best fit to the dt mong ll models. c Weight of evidence interpreted s proportion. Weights cross ll models sum to 1.00. TABLE 3. AIC c model rnkings for femle blood mercury burden nlysis for Crolin Wren nests monitored in 2010 on the South River, Virgini. Models re rnked by incresing AIC c scores, showing tht the effect of femle blood mercury ppers in the top two rnked models nd ccounts for >90% of model weights. Model K AIC c b w i c Mercury 2 0.00 0.71 Mercury + cvity 4 1.21 0.21 Null 1 2.70 0.05 Cvity 3 3.38 0.02 FIG. 4. Reltionship between untrnsformed Crolin Wren blood (ww = wet weight) nd mercury concentrtions (ww) of (A) body fethers (log trnsformed, r 2 = 0.88, F = 133.1, P < 0.001) or (B) til fethers (log trnsformed, r 2 = 0.56, F = 64.5, P < 0.0001) collected from birds in 2010 in Virgini. Eqution on ech grph indictes reltionship bsed on liner regression nlysis. ccounted for 93% of weights nd models contining cvity type ccounted for 24% of weights. Within the top-rnked model (effect of femle blood mercury lone) we found n effect of femle blood mercury on dily nest survivl (on the multinomil logit scle, β = 0.564 ± 0.072). We trnsformed this bet estimte into predicted nest success (ssuming 30-dy nesting cycle) bsed on hypothesized femle mercury body burden nd found tht s femle blood mercury incresed, modeled nest survivl decresed (Fig. 5A). Becuse we were interested in how individul blood mercury concentrtions ffected nest survivl in reltion to individuls with no mercury exposure, we clculted percent reduction in nest success compred with modeled femles with 0 µg g 1 blood mercury concentrtions nd found tht nest success decresed s blood mercury concentrtion incresed (Fig. 5B). Competing risks. In 2010, out of 19 monitored contminted nests, 9 fledged, 3 were depredted, nd 7 were bndoned. Number of prmeters. b Scled AIC c ; AIC c = 0.00 is interpreted s the best fit to the dt mong ll models. c Weight of evidence interpreted s proportion. Weights cross ll models sum to 1.00. On reference sites, 21 nests fledged young, 2 were depredted, nd 2 others were bndoned. Video reveled predtion by Americn Blck Bers (Ursus mericnus) nd ctivity round filed nests by other potentil nest predtors, including Northern Rccoons (Procyon lotor), Virgini Opossums (Didelphis virginin), nd Coyotes (Cnis ltrns). We suspect some instnces of snke predtion but did not cpture this on video. Of the 9 bndoned nests, 4 occurred during the lying incubtion stge nd 5 occurred during the nestling stge. In some cses the prents renested lter in the seson, which indictes tht dult mortlity hd not cused bndonment, but we do not hve ny other informtion to explin bndonment. By evluting competing risks in MCESTIMATE we found tht cuse-specific nest filure differed between contminted nd reference sites (effective smple size = 615 dys). The toprnked model which hd 3 the support of the second-rnked model reveled n importnt effect of tretment on the rte of bndonment but no importnt effect of tretment on predtion (Tble 4). Within the top-rnked model contminted nests were nerly 4 more likely to be bndoned (bndonment Cont = 0.507 ± 0.111, bndonment Ref = 0.131 ± 0.081) but there ws little

766 JACKSON ET AL. AUK, VOL. 128 TABLE 4. AIC c model rnking for Crolin Wren nest survivl in 2010 long the South River, Virgin, when nest survivl ws llowed to vry by cuse between models. Models re rnked by incresing AIC c scores, showing strong effect of bndonment within tretment groups, but not predtion. Model K AIC c b w i c Abndonment{tretment} Predtion{.} 3 0.00 0.74 Abndonment{tretment} Predtion{tretment} 4 1.21 0.22 Abndonment{.} Predtion{.} 2 3.23 0.03 Abndonment{.} Predtion{tretment} 3 4.48 0.01 Number of prmeters. b Scled AIC c ; AIC c = 0.00 is interpreted s the best fit to the dt mong ll models. c Weight of evidence interpreted s proportion. Weights cross ll models sum to 1.00. TABLE 5. Crolin Wren blood, fether, nd egg mercury effects concentrtions (ww = wet weight) ssocited with MCESTIMATE-modeled reduction in nest success. Results bsed on dt collected in 2010 from nests long the South River in Virgini. Reduction in nest success Blood mercury (µg g 1, ww) Body fether mercury (µg g 1, ww) b Til fether mercury (µg g 1, ww) c Egg mercury (µg g 1, ww) d 10% 0.7 2.4 3.0 0.11 20% 1.2 3.4 4.7 0.20 30% 1.7 4.5 6.4 0.29 40% 2.1 5.3 7.7 0.36 50% 2.5 6.2 9.1 0.43 60% 2.9 7.1 10.4 0.50 70% 3.3 7.9 11.8 0.57 80% 3.8 e 9.0 13.5 0.66 90% 4.4 e 10.3 15.5 0.76 99% 5.6 e 12.8 19.5 0.97 FIG. 5. The reltionship between MCESTIMATE-modeled Crolin Wren nest survivl nd femle blood mercury concentrtion for nests found in 2010 in Virgin. (A) Predicted Crolin Wren nest success over their 30- dy nest cycle in reltion to femle blood mercury concentrtion when other covrites were held constnt (dte = 24 My, nest cvity = nturl). Error brs indicte SE. Dotted portion of the line indictes model extrpoltion pst observed femle blood mercury concentrtions. (B) Percent reduction in nest survivl (from nest survivl t 0 µg g 1 ) in reltion to femle blood mercury concentrtion. Blood mercury concentrtions ssocited with 10% increments of reduction in nest success re shown. difference in predtion between tretment groups (predtion Cont = 0.145 ± 0.061, predtion Ref = 0.196 ± 0.078). Effects concentrtion. We found tht the blood concentrtion ssocited with 10% reduction in nest success ws 0.70 µg g 1, which correlted to 2.4 µg g 1 mercury in body fethers, 3.0 µg g 1 mercury in til fethers, nd 0.11 µg g 1 mercury in egg tissue (Tble 5). Extrpolting our model to higher mercury vlues predicts 99% reduction in reproductive success t blood concentrtions of 5.6 µg g 1, body fether concentrtions of 12.8 µg g 1, til fether concentrtions of 19.5 µg g 1, nd egg concentrtions of 0.97 µg g 1 (Tble 5). Clculted using MCESTIMATE, compring probbility of fledging t lest 1 young t 0 µg g 1 to the probbility of fledging t lest 1 young t ech contminted blood concentrtion. b Clculted using the regression eqution [body fether Hg] = 2.1407974[blood Hg] + 0.8531665. c Clculted using the regression eqution [til fether Hg] = 3.3762108[blood Hg] + 0.6427166. d Clculted using the regression eqution [egg Hg] = 0.1748381[blood Hg] 0.007394. e Extrpoltion pst known blood mercury levels using the MCESTIMATE model. DISCUSSION Crolin Wrens breeding in two mercury-contminted wtersheds in Virgini hd 34% lower nest success compred with nerby reference sites without histories of industril mercury contmintion. At n individul level, femles with higher blood mercury concentrtions hd lower nest success; cross the rnge from 0 to 4.0 µg g 1, modeled success decresed by 10% on verge for ech projected 0.50 µg g 1 increse in blood mercury concentrtion. Our nlyses consistently rnked mercury contmintion s leding predictor of nest success, more so thn dte in seson, yer, cvity type, ge of nest, or river system. By using new softwre to model nest success on the bsis

OCTOBER 2011 EFFECT OF MERCURY ON SONGBIRD REPRODUCTIVE SUCCESS 767 of competing risks we were ble to show tht nests on contminted sites were 3 more likely to fil becuse of bndonment, wheres both tretment groups experienced equl levels of nest predtion. We estimted pproximtely 60% nest success on reference sites nd 40% nest success on contminted sites, both of which re higher thn the only other known published estimte for this species: 26% of nests in northwestern Albm produced t lest one fledgling (n = 118; Hggerty nd Morton 1995). In our femle blood mercury nlysis, we estimte tht femles without methylmercury in their diet should fledge young ~80% of the time. This is higher thn expected for songbird, but similr rtes hve been reported in other cvity-nesting species (Mrtin nd Li 1992, Etterson et l. 2007b). Becuse we determined the mercury concentrtions of femle prents (or estimted it from mle concentrtions) t the time of nesting, we were ble to model the response of nest survivl to femle mercury concentrtion in 2010. This is the first study, to our knowledge, tht documents correltion between blood mercury concentrtion nd reduction in nest survivl in free-living bird popultion. Severl recent field studies hve compred the reproductive success of birds exposed to environmentl mercury with tht of n unexposed reference group (Custer et l. 2007, Brsso nd Cristol 2008). These studies, however, did not consider nest success on the bsis of individul blood mercury concentrtion. The mximum blood mercury concentrtion for wrens in 2010 ws 3.22 µg g 1 (wet weight) for femles nd 3.65 µg g 1 (wet weight) for mles. Although it is possible to extrpolte the model to show lmost complete reproductive filure (99%) t 5.6 µg g 1 in the blood, these extrpoltions should be viewed with cution. It is possible tht complete reproductive impirment occurs t lower concentrtions thn those predicted by our model; individuls with blood mercury levels higher thn those tht we smpled my be unble to estblish territories or find mtes, rendering their reproductive output effectively zero, which is finding tht cnnot be shown in our study. In other orders of birds, blood mercury concentrtions re ffected by depurtion of mercury into eggs (Becker 1992, Monteiro nd Furness 2001, Kennmer et l. 2005, French et l. 2010) or into fethers during molt (Nichols et l. 2010). We believe tht the timing of blood smple collection within the nesting cycle is not likely to hve cused vribility in blood mercury concentrtions in wrens t this hevily contminted South River site. For exmple, mercury trnsfer from blood into eggs by Tree Swllows breeding long mercury-contminted stretch of the South River did not decrese in the mount of mercury trnsferred to ech subsequent egg lid, which indictes tht intke of mercury through prey ws high enough to mintin body burden despite the depurtion into eggs (Brsso et l. 2010). Methylmercury depurtion through molt lso complictes interprettion of blood mercury levels, but the wren is yer-round resident nd does not strt molting until fter their nestlings hve fledged. Therefore, molt would not hve eliminted blood mercury to growing fethers until fter our smpling efforts. Abndonment rtes were ~3 higher t contminted thn t reference sites. There re severl puttive mechnisms by which this bndonment my hve occurred (some seen in other mercury studies nd some specultive), but we re unble to distinguish between them without further study. Some prents tht bndoned nests my hve succumbed to lethl effects of mercury, lthough determining which behviorl or physiologicl effect is not possible. At the sublethl level, dults my hve exhibited bnorml incubtion or feeding behvior tht led to nest loss. Common Loons exposed to environmentl mercury displyed berrnt incubtion behvior, leving eggs unttended more often when mercury concentrtions were elevted (Evers et l. 2008). Nestlings my hve behved bnormlly, for exmple voclizing less before htching or begging less, eliminting cues necessry to stimulte prentl behvior. Finlly, mercury cn cuse outright embryo mortlity in songbirds, in which cse the documented bndonments would be best explined s response to egg invibility rther thn cuse (Heinz et l. 2009). Other mechnisms re possible, but identifying bndonment rther thn predtion s differentil cuse of reduced nest survivl is n importnt step forwrd in focusing future studies of the effects of mercury on songbird reproduction. Dosing studies my be n importnt next step, becuse they cn eliminte the risk of predtion to focus on the mechnisms of bndonment. In dosing study of Americn Kestrels (Flco sprverius), reserchers lso found decresed nest success but were unble to conclude whether differences were cused by bnorml prentl behvior or disrupted egg nd chick development (Albers et l. 2007). Wrens ccumulte mercury to higher concentrtions thn other songbirds breeding in floodplin forests, likely becuse of their hevy relince on spiders in their diet (Cristol et l. 2008). Becuse spiders feed t high trophic levels, they bioccumulte mercury to higher concentrtions thn mny other invertebrte prey species. Becuse wrens remin on territories yer round, they my be one of the more t-risk species, but the effects of mercury reported here re biologiclly significnt enough tht migrnts, nd species feeding lower on the food chin, my lso be ffected. Although the sites studied here were both industril point sources, songbirds living in res remote from industry cn lso ccumulte mercury concentrtions comprble to those tht hve effects on wren reproduction. These include Rusty Blckbirds (Euphgus crolinus; Edmonds et l. 2010), Nelson s Sprrows (Ammodrmus nelsoni; Winder nd Emslie 2011), nd Sltmrsh Sprrows (A. cudcutus; Lne et l. 2011). Conclusions. Our results hve importnt implictions with regrd to future regultion of mercury pollution nd mitigtion of previously contminted sites. Our finding tht terrestril songbirds exposed to mercury exhibited high rtes of nest bndonment, leding to substntil reduction in return on reproductive effort, suggests tht qutic mercury pollution my hrm terrestril songbirds in floodplin forest hbitts ner mny of the thousnds of wter bodies subject to mercury fishconsumption dvisories in the United Sttes (see wter.ep.gov/ scitech/swguidnce/fishshellfish/fishdvisories) nd worldwide. This study is the first to link specific mercury concentrtions in songbirds to quntifible reduction in nest success, t lest in prt becuse of incresed bndonment of nests. Understnding the rmifictions of mercury contmintion for other species nd regions requires further investigtion, prticulrly for species found in hbitts sensitive to methylmercury production or experiencing greter-thn-usul physiologicl demnds such s longdistnce migrtion.

768 JACKSON ET AL. AUK, VOL. 128 ACKNOWLEDGMENTS The findings nd conclusions in this rticle re those of the uthors nd do not necessrily represent the views of the U.S. Fish nd Wildlife Service or the U.S. Environmentl Protection Agency. North Fork Holston River reserch ws funded by the Olin Corportion. South River reserch ws funded by E.I. DuPont de Nemours nd Compny, nd reserch ws completed with oversight from the South River Science Tem, which is collbortion of stte nd federl gencies, cdemic institutions, nd environmentl interests. We thnk the field crews nd numerous lndowners who llowed us ccess to their property, long with two nonymous reviewers who offered helpful comments on the mnuscript. LITERATURE CITED ALBERS, P. H., M. T. KOTERBA, R. ROSSMANN, W. A. LINK, J. B. FRENCH, R. S. BENNETT, AND W. C. BAUER. 2007. Effects of methylmercury on reproduction in Americn Kestrels. Environmentl Toxicology nd Chemistry 26:1856 1866. BART, J., AND D. S. ROBSON. 1982. Estimting survivorship when the subjects re visited periodiclly. Ecology 63:1078 1090. BECKER, P. H. 1992. Egg mercury levels decline with the lying sequence in Chrdriiformes. Bulletin of Environmentl Contmintion nd Toxicology 48:762 767. BOND, A. L., AND A. W. DIAMOND. 2009. Totl nd methyl mercury concentrtions in sebird fethers nd eggs. Archives of Environmentl Contminnts nd Toxicology 56:286 291. BRASSO, R. L., M. K. ABDEL LATIF, AND D. A. CRISTOL. 2010. Reltionship between lying sequence nd mercury concentrtion in Tree Swllow eggs. Environmentl Toxicology nd Chemistry 29:1155 1159. BRASSO, R. L., AND D. A. CRISTOL. 2008. Effects of mercury exposure on the reproductive success of Tree Swllows (Tchycinet bicolor). Ecotoxicology 17:133 141. BURGESS, N. M., AND M. W. MEYER. 2008. Methylmercury exposure ssocited with reduced productivity in Common Loons. Ecotoxicology 17:83 91. BURNHAM, K. P., AND D. R. ANDERSON. 2002. Model Selection nd Multimodl Inference: A Prcticl Informtion-Theoretic Approch, 2nd ed. Springer-Verlg, New York. COWARDIN, L. M., V. CARTER, F. C. GOLET, AND E. T. LAROE. 1979. Clssifiction of wetlnds nd deepwter hbitts of the United Sttes. U.S. Deprtment of the Interior, Fish nd Wildlife Service, Wshington, 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. 2008. The movement of qutic mercury through terrestril food webs. Science 320:335. CUSTER, C. M., T. W. CUSTER, AND E. F. HILL. 2007. Mercury exposure nd effects on cvity-nesting birds from the Crson River, Nevd. Archives of Environmentl Contmintion nd Toxicology 52:129 136. DINSMORE, S. J., G. C. WHITE, AND F. L. KNOPF. 2002. Advnced techniques for modeling vin nest survivl. Ecology 83:3476 3488. EDMONDS, S. T., D. C. EVERS, D. A. CRISTOL, C. METTKE-HOF- MANN, L. L. POWELL, A. J. MCGANN, J. W. ARMIGER, O. P. LANE, D. F. TESSLER, P. NEWELL, AND OTHERS. 2010. Geogrphic nd sesonl vrition in mercury exposure of the declining Rusty Blckbird. Condor 112:789 799. ETTERSON, M. A., L. R. NAGY, AND T. RODDEN-ROBINSON. 2007. Prtitioning risk mong different cuses of nest filure. Auk 124:432 443. ETTERSON, M. A., B. J. OLSEN, AND R. GREENBERG. 2007b. The nlysis of covrites in multi-fte Mrkov chin nest-filure models. Pges 55 64 in Beyond Myfield: Mesurements of Nest Survivl Dt (S. L. Jones nd G. R. Geupel, Eds.). Studies in Avin 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. 2005. Ptterns nd interprettion of mercury exposure in freshwter vin communities in northestern North Americ. Ecotoxicology 14:193 221. EVERS, D. C., L. J. SAVOY, C. R. DESORBO, D. E. YATES, W. HAN- SON, K. M. TAYLOR, L. S. SIEGEL, J. H. COOLEY, JR., M. S. BANK, A. MAJOR, AND OTHERS. 2008. Adverse effects from environmentl mercury lods on breeding Common Loons. Ecotoxicology 17:69 81. FRENCH, J. B., R. S. BENNETT, AND R. ROSSMANN. 2010. Mercury in the blood nd eggs of Americn Kestrels fed methylmercury chloride. Environmentl Toxicology nd Chemistry 29:2206 2210. HAGGERTY, T. M., AND E. S. MORTON. 1995. Crolin Wren (Thryothorus ludovicinus). In The Birds of North Americ Online (A. Poole, ED.). Cornell Lb of Ornithology, Ithc, New York. [Online.] Avilble t bn.birds.cornell.edu/bn/species/188. HALLINGER, K. K., AND D. A. CRISTOL. 2011. The role of wether in mediting the effect of mercury exposure on reproductive success in Tree Swllows. Ecotoxicology 20:1368 1377. HEINZ, G. H., D. J. HOFFMAN, J. D. KLIMSTRA, K. R. STEBBINS, S. L. KONDRAD, AND C. A. ERWIN. 2009. Species differences in the sensitivity of vin embryos to methylmercury. Archives of Environmentl Contmintion nd Toxicology 56:129 138. HEISEY, D. M., AND B. R. PATTERSON. 2006. A review of methods to estimte cuse-specific mortlity in presence of competing risks. Journl of Wildlife Mngement 70:1544 1555. JACKSON, A. K., D. C. EVERS, S. B. FOLSOM, A. M. CONDON, J. DIE- NER, L. F. GOODRICK, A. J. MCGANN, J. SCHMERFELD, AND D. A. CRISTOL. 2011. Mercury exposure in terrestril birds fr downstrem of n historicl point source. Environmentl Pollution 159:3302 3308. JOHNSON, D. H. 1979. Estimting nest success: The Myfield method nd n lterntive. Auk 96:651 661. JOHNSON, D. H. 2007. Estimting nest success: A guide to the methods. Pges 65 72 in Beyond Myfield: Mesurements of Nest Survivl Dt (S. L. Jones nd G. R. Geupel, Eds.). Studies in Avin Biology, no. 34. JONES, J. 2003. Tree Swllows (Tchycinet bicolor): A new model orgnism? Auk 120:591 599. KENNAMER, R. A., J. R. STOUT, B. P. JACKSON, S. V. COLWELL, I. L. BRISBIN, JR., AND J. BURGER. 2005. Mercury ptterns in Wood Duck eggs from contminted reservoir in South Crolin, USA. Environmentl Toxicology nd Chemistry 24:1793 1800. LANE, O. P., K. M. O BRIEN, D. C. EVERS, T. P. HODGMAN, A. MAJOR, N. PAU, M. J. DUCEY, R. TAYLOR, AND D. PERRY. 2011. Mercury in breeding Sltmrsh Sprrows (Ammodrmus cudcutus cudcutus). Ecotoxicology 20: in press.