Avian top predator and the landscape of fear: responses of mammalian mesopredators to risk imposed by the golden eagle

A ian op p eda o and he landscape o ea : esponses o mammalian mesop eda o s o isk imposed by he golden eagle Ma i S. Lyly 1 , Alexand e Ville s 1,2 , Elina Koi is o 1 , Pekka Helle 3 , Tuomo Ollila 4 & E kki Ko pim€ aki 1 1 Sec ion o Ecology, Depa men o Biology, Uni e si y o Tu ku, FI-20014 Tu ku, Finland 2 Cen e d’E udes Biologiques de Chiz e UMR 7372, CNRS & Uni e si  e de La Rochelle, 79360 Beau oi su Nio , F ance 3 Finnish Game and Fishe ies Resea ch Ins i u e, Oulu Game and Fishe ies Resea ch, Uni e si y o Oulu, POB 413, FI-90014 Oulu, Finland 4 Finnish Fo es and Pa k Se ices, Na u al He i age Se ices, POB 8016, FI-96101 Ro aniemi, Finland Keywo ds In aguild p eda ion, mesop eda o supp ession, pine ma en, ap o , ed ox, ophic in e ac ions. Co espondence Ma i S. Lyly, Sec ion o Ecology, Depa men o Biology, Uni e si y o Tu ku, FI-20014 Tu ku, Finland. Tel: +358 2333 5813; Fax: +358 2333 6568; E-mail: [email p o ec ed] Funding In o ma ion The s udy was inancially suppo ed by Jenny and An i Wihu i Founda ion (g an s o MSL), and he Academy o Finland (g an s no. 123379, 136717 and 250709 o EKo ). Recei ed: 21 May 2014; Re ised: 23 No embe 2014; Accep ed: 25 No embe 2014 Ecology and E olu ion 2015, 5(2): 503–514 doi: 10.1002/ece3.1370 Abs ac Top p eda o s may induce ex ensi e cascading e ec s on lowe ophic le els, o example, h ough in aguild p eda ion (IGP). The impac s o bo h mamma- lian and a ian op p eda o s on species o he same class ha e been ex ensi ely s udied, bu he e ec s o he la e upon mammalian mesop eda o s a e no ye as well known. We examined he impac o he p eda ion isk imposed by a la ge a ian p eda o , he golden eagle (Aquila ch ysae os, L.), on i s po en ial mammalian mesop eda o p ey, he ed ox (Vulpes ulpes, L.), and he pine ma en (Ma es ma es, L.). The s udy combined 23 yea s o coun ywide da a om nes ing eco ds o eagles and wildli e ack coun s o mesop eda o s in Finland, no he n Eu ope. The p eda ion isk o he golden eagle was modeled as a unc ion o e i o y densi y, densi y o ledglings p oduced, and dis ance o nea es ac i e eagle e i o y, wi h he expec a ion ha a high p eda ion isk would educe he abundances o smalle sized pine ma ens in pa icula . Red oxes appea ed no o su e om eagle p eda ion, being in ac mos nume ous close o eagle nes s and in a eas wi h mo e eagle e i o ies. This is likely due o simila p ey p e e ences o he wo p eda o s and he la ge size o oxes enabling hem o escape eagle p eda ion isk. Somewha con a y o ou p edic- ion, he abundance o pine ma ens inc eased om low o in e media e e i- o y densi y and a close p oximi y o eagle nes s, possibly because o simila habi a p e e ences o ma ens and eagles. We ound a sligh ly dec easing end o ma en abundance a high e i o y densi y, which could indica e ha he esponse in ma en popula ions is dependen on eagle densi y. Howe e , mo e esea ch is needed o be e es ablish whe he mesop eda o s a e in imida ed o p eda ed by golden eagles, and whe he such e ec s could in u n cascade o lowe ophic le els, bene i ing he bi o ous species. In oduc ion The adi ional iew o ophic dynamics has emphasized he impo ance o bo om-up e ec s in ecosys ems, bu ecen s udies ha e p o ided inc easing e idence o show ha op-down p ocesses a e also ex ensi e and in luen ial in shaping communi ies (e.g., Hebblewhi e e al. 2005; Bo all and Ebenman 2006; Te bo gh and Es es 2010). A ypical case o such p ocess is op p eda o s p eying upon and limi ing he numbe s o he bi o es. Indeed, op p e- da o s a e o en key species in hei ecosys ems and may induce ex ensi e cascading e ec s on he lowe ophic le els (Es es e al. 2011; Ripple e al. 2014). Ano he pa hway o op-down e ec s is in aguild p e- da ion (IGP), whe e p eda o y species compe e o e sha ed p ey bu also p ey on each o he (Polis e al. 1989; Lou enc ßo e al. 2014). Among e eb a es, IGP is ypically in lic ed by la ge species on hei smalle compe i o s, he kille species being usually mo e han h ee imes la ge han he ic im species (Paloma es and Ca o 1999; Se gio and Hi aldo 2008). In addi ion o di ec killing and consump ion, IGP o en induces a oidance and ª2015 The Au ho s. Ecology and E olu ion published by John Wiley & Sons L d. This is an open access a icle unde he e ms o he C ea i e Commons A ibu ion License, which pe mi s use, dis ibu ion and ep oduc ion in any medium, p o ided he o iginal wo k is p ope ly ci ed. 503 isk-sensi i e habi a selec ion in he p ey, educing hei b eeding success and su i al (Mi chell and Banks 2005; Se gio and Hi aldo 2008; Mukhe jee e al. 2009). Top p eda o s may hus ac as an impo an mo ali y ac o o smalle p eda o s (Ri chie and Johnson 2009; Pasa- nen-Mo ensen e al. 2013) and gene a e a “landscape o ea ” whe e mesop eda o s expe ience di e ing le els o p eda ion isk (Laund  e e al. 2001; Swanson e al. 2014). Howe e , i op p eda o popula ions decline, hey can cease o limi popula ions o mesop eda o s, as is sug- ges ed by he mesop eda o elease hypo hesis (Soul e e al. 1988; C ooks and Soul e 1999; P ugh e al. 2009). Fo example, in Aus alia, he absence o a op p eda o , he dingo (Canis dingo, Meye ), has esul ed in a meso- p eda o elease, which in u n has caused de as a ing cascading e ec s on he small ma supial auna (Johnson e al. 2007). Mos o he p e ious s udies on e eb a e IGP ha e ocused on wi hin-class p eda o guilds (mammal–mam- mal IGP: Paloma es e al. 1995; Cou champ e al. 1999; Helldin e al. 2006; Le nic e al. 2011; bi d–bi d IGP: Hakka ainen and Ko pim€ aki 1996; Fielding e al. 2003; Se gio and Hi aldo 2008; Lou enҫo e al. 2011). Fo example, s udies om No h Ame ica show ha he wol (Canis lupus, L.) has a nega i e impac on he coyo e (Canis la ans Say) no jus ia p eda ion bu also h ough in imida ion and ood compe i ion (Be ge and Gese 2007; Mille e al. 2012). Mo eo e , he e is a clea in e ac ion ex ending o he ed ox (Vulpes ulpes, L.): when wol es a e p esen , oxes a e mo e nume ous han coyo es, whe eas when wol es a e absen , coyo es domi- na e in numbe s (Newsome and Ripple 2014). None he- less, obse a ions om he p ey emains o la ge ap o s show ha mammalian mesop eda o s succumb o p eda- ion by bi ds, oo (Sulka a e al. 1997, 2008; Wa son 2010). This sugges s ha la ge bi ds o p ey could also a ec he abundance and dis ibu ion o mammalian mes- op eda o s. Ye , such p eda o y in e ac ions ha e been epo ed only in a ew a icles (Ko pim€ aki and No dahl 1989; Roeme e al. 2002; Moeh enschlage e al. 2007; Salo e al. 2008). One o he la ges ap o species in he No he n Hemisphe e is he golden eagle (Aquila ch ysae os, L.). Many o i s popula ions c ashed du ing he ea ly 20 h cen u y, la gely due o pe secu ion (Whi ield e al. 2004; Ollila and Koskimies 2006; Wa son 2010). Cu en ly golden eagle popula ions a e mainly s able o inc easing, bu hei eco e y has been slowed down by habi a ag- men a ion and inc eased human dis u bance (Wa son 2010). The golden eagle is a op p eda o , which can p ey o e a wide ange o species, om small bi ds, and oden s e en o ungula es as la ge as dee . The main p ey i ems come om he g oups o ha es (Lepo idae), g ouse (Te aonidae and Phasianidae), and squi els (Sciu idae) (Wa son 2010). Wha di e en ia es golden eagles om many o he ap o s is he ela i ely high pe cen age o mammalian p eda o s in hei die (Valkama e al. 2005; Lou enҫo e al. 2011), ypically anging om 2% o as much as 10–20% (Wa son 2010). The e o e, he pauci y o absence o he golden eagle may ha e con ibu ed o he inc ease o mesop eda o s in many ecosys ems du ing he pas decades (Ko pim€ aki and No ds € om 2004). The aim o his a icle was o expand he unde s and- ing o in aguild ela ionships be ween a ian op p eda- o s and mammalian mesop eda o p ey. We do so by examining whe he he Finnish golden eagle popula ion impac s abundances o wo ca ni o ous mesop eda o s, he ed ox, and he pine ma en (Ma es ma es, L.). Bo h species a e common in Finland (Wikman 2010) and known o be p eyed upon by eagles (ca. 1% o esh p ey emains a nes si es each, Sulka a e al. 1999). We com- bine wo coun ywide da a se s ga he ed om Finland o e se e al decades: he nes ing in en o y da a o golden eagles and he wildli e snow- ack census da a, which as annual moni o ing scheme o mul iple species is unique in i s ex en and longe i y. We use es ima ed eagle e i- o y and ledgling densi ies oge he wi h dis ance o he nea es eagle nes as p oxies o he p eda ion isk, he impac o which we hen examine on mesop eda o abun- dances. Howe e , he mechanism o such impac can be ei he di ec p eda ion, dispe sal due o in imida ion, o bo h, and he wo phenomena canno be se apa wi h obse a ional da a. Ins ead, we s udy he o e all impac o eagles on mesop eda o abundances. We p edic ha he abundance o he pine ma en, simila in size wi h he main p ey i ems o eagle, will be lowe in a eas whe e he e a e mo e ac i e eagle e i o ies and ledglings p o- duced, and also in he icini y o inhabi ed eagle nes s. In con as , we expec he la ge ed ox, which is a he uppe end o he weigh scale o eagle p ey, no o espond o ou p oxies o p eda ion p essu e. Ma e ials and Me hods Golden eagle da a In Finland, he golden eagle is dispe sed o e an a ea la - ge han 150,000 km²(Fig. 1), mainly wi hin he eindee (Rangi e a andus, L.) husband y a ea in he no he n pa s o he coun y (Ollila and Ilmonen 2009). Du ing he pas decades, he numbe o b eeding eagle pai s has been inc easing (Ollila and Koskimies 2006; Fig. 2), eaching 348 inhabi ed e i o ies in 2014 (Ollila 2014). The nes ing da a o he golden eagle, p o ided by he Finnish Fo es and Pa k Se ice, include in o ma ion abou nes occupancy and b eeding success o eagles in 504 ª2015 The Au ho s. Ecology and E olu ion published by John Wiley & Sons L d. Eagle P eda ion Impac upon Ma en and Fox M. S. Lyly e al. Finnish nes ing si es. The da a a e de i ed om an ongo- ing la ge-scale moni o ing o all known e i o ies and ac i e sea ching o new ones, ca ied ou annually by some 40 olun a y bi d inge s. In his s udy, we included all he known e i o ies ha had eco ds o eagle p esence du ing 1988–2011 (n=477, 6569 indi idual eco ds). Nes s we e annually ca ego ized in o wo classes acco ding o hei b eeding success: (1) unsuccess ully b eeding pai s (p esen bu no b eeding, wi h possible eco ds o a ailed b eeding a emp , such as b oken egg shells o emains o chicks); and (2) success ully b eeding pai s (p esen wi h ≥1 ledglings p oduced). To analyze he impac o golden eagles upon mesop eda o species, we hen used he eagle nes loca ions o o m wo ypes o annual eagle densi y maps. In he e i o y densi y (TD) maps, all ac i e golden eagle e i o ies we e included (classes 1–2), whe eas ledgling densi y (FD) maps we e o med using only he nes s wi h ledglings p oduced (class 2), mul iplied wi h he numbe o ledged chicks (1 o 2). The ob ained coun ywide densi y maps e lec he p e- da ion p essu e –bo h di ec and indi ec –by he golden eagle. The eagle densi y maps we e p oduced by compu ing a smoo hed in ensi y unc ion ( as e size 1 km) om he nes loca ions, using he “densi y.ppp” unc ion om he R package “spa s a ” (Baddeley and Tu ne 2005). The densi y.ppp compu es a ke nel es ima e (Diggle 1985) o he in ensi y unc ion o a poin p ocess, which gene a ed he pa e n o nes s (u). I compu es he con o- lu ion o he iso opic Gaussian ke nel wi h poin masses a each da a poin in u: kðuÞ¼ X n i kðxiuÞwi () eðuÞ(1) In he in ensi y unc ion k(u), he poin s in he neigh- bo hood o a single poin ua e gi en by x i and hei weigh s by w i , equaling o 1 in ou case. The edge co ec- ion ac o o he densi y unc ion is gi en by e(u). The amoun o smoo hing (i.e., s anda d de ia ion) in Gauss- ian smoo hing ke nel kis de ined by as ollows: 0 50 100 150 200 km N Figu e 1. Map o Finland showing he loca ions o wildli e iangles ( iangle symbols) and golden eagle nes ing si es (g ey a ea) included in he s udy. The nes ing si es a e p esen ed wi h a andomized ≤10 km o se and a 10 km bu e zone. The sou he n bo de o eindee husband y a ea is p esen ed wi h a da k g ey line. 0 100 200 300 400 Yea Numbe o e i o ies/ ledglings 88 90 92 94 96 98 00 02 04 06 08 10 Figu e 2. Popula ion ends o he golden eagle du ing 1988–2011 in Finland. The numbe o ac i e e i o ies (solid line) and ledglings p oduced (dashed line) a e shown annually. ª2015 The Au ho s. Ecology and E olu ion published by John Wiley & Sons L d. 505 M. S. Lyly e al. Eagle P eda ion Impac upon Ma en and Fox kðx;yÞ¼ 1 2p 2ex2þy2 2 2(2) The co e o a golden eagle e i o y does no ypically ex end mo e han 3 km om he nes , bu eagles egula ly isi si es u he om hei nes (McG ady e al. 2002), especially a e he chicks ha e ledged. The e o e, he spa- ial ex en wi hin which he eagle p eda ion migh be s onges was es ed by calcula ing he densi y maps a i e di e en le els o smoo hing, a 1–5 km. An op imal le el o was hen selec ed h ough model compa ison. Game animal abundance da a The abundance da a o game animals om he wildli e iangle scheme comp ised o snow- ack coun s o se e al mammalian species om 1989 o 2011, de i ed om 1610 indi idual iangles (17,808 indi idual eco ds) (Fig. 1). Da a we e p o ided by he Finnish Game Resea ch Ins i u e, which coo dina es he annual na ion- wide census pe o med by hun e s. T iangle loca ions a e ixed, and mos , bu no all, iangles a e censused e e y yea . The ack coun s a e ca ied ou mainly du ing Feb- ua y by moni o ing he iangula 3 94 km ansec lines, es ablished h oughou he coun y (Lind en e al. 1996; Pellikka e al. 2005). Old acks a e co e ed by snow all o ack coun e s, and a e a su icien ack accumula ion ime, new acks a e coun ed. Typically, he esul s o he ack coun s a e exp essed as ack densi y (c ossings pe 24 h pe 10 km) (Pellikka e al. 2005), bu in his s udy he o iginal ack coun obse a ions we e used, complemen ed by a a iable epo ing he ack accumula ion ime in days. Obse a ions wi h an espe- cially long accumula ion ime (>10 days) we e disca ded. Addi ionally, isual sigh ings o g ouse a e eco ded du - ing he census, and hese abundance da a we e also employed in his s udy. Abundance index da a o wo common mesop eda o species, ed ox and pine ma en, we e used as esponse a iables. These wo mesop eda o s a e ai ly ac i e du - ing win e and a e hus well p esen ed in he snow- ack coun s. The ed ox, as a la ge species, may also p ey on he smalle pine ma en (Linds € om e al. 1995). The indices o moun ain ha e (Lepus imidus, L.), cape caillie (Te ao u ogallus, L.), black g ouse (Te ao e ix, L.), hazel g ouse (Te as es bonasia, L.), and willow g ouse (Lagopus lagopus, L.) we e also employed as explana o y p ey co a ia es. Obse a ions o all he bi d species we e summed oge he as he a iable “g ouse”. T ack coun s, when used as a co a ia e, we e scaled wi h ack accumu- la ion ime. Using he midpoin coo dina es o he wildli e ian- gles, he obse ed game abundances o an indi idual iangle we e linked wi h he eagle densi ies ex ac ed om he TD and FD as e maps a he same loca ion. In a gi en yea , he game obse a ions we e linked wi h TD o he same yea and wi h FD om he p e ious yea . The eason o his was ha TD co esponds well o eagle densi y du ing he win e since adul eagles keep o hei e i o ies yea - ound, apa om he no he nmos pop- ula ions ha a e o ced o mig a e because o se e e win- e condi ions (McG ady e al. 2002; Wa son 2010). Meanwhile, FD is a good p oxy o he hun ing p essu e posed by eagles du ing he summe and au umn, when he ledglings mo e abou he e i o y and hei pa en s hun la ge amoun s o ood o hei o sp ing and hem- sel es (Wa son 2010). Thus, FD would mos likely a ec he obse ed mesop eda o abundances in he ollowing win e . In addi ion o he densi y a iables, he dis ance o he closes ac i e eagle nes om each wildli e iangle was calcula ed o each yea . The a iable “dis ance o nea es nes ” (DNN) was hen used as a compe ing explana o y eagle a iable and compa ed wi h TD and FD. DNN inc eases linea ly when mo ing u he om he nes , and he e o e, i may be e desc ibe he p eda ion e ec o eagle a he pe iphe y o he e i o ies. Howe e , he e ec o DNN is o in e es only wi hin dis ances wi hin which eagles and mesop eda o s mo e abou , as i is no plausible ha eagles would educe mesop eda o numbe s a ou side hei e i o y limi s. Densi y a iables TD and FD app oach ze o qui e quickly when mo ing u he away om he e i o y co e a ea, bu hei ad an age is ha hey a e able o accoun o wo o mo e e i o ies close o each o he , whe eas DNN does no e lec he o e all local eagle abundance well. Habi a da a As mesop eda o abundances depend on landscape ea- u es, a iables desc ibing habi a composi ion we e inco po a ed in o ou analyses. Habi a in o ma ion was acqui ed om CORILIS da a ( as e size 1 91 km), which gi es he p opo ion o di e en Co ine Land Co e (CLC) classes wi hin a smoo hing adius o 5 km (Eu opean En i onmen Agency 2009). The da a a e de i ed om Landsa sa elli e image y om 2000. The p opo ions o a mland and o es in he landscape we e calcula ed by summing he p opo ions o CLC classes 12–22 and 23–29, espec i ely. Habi a p opo ions we e hen assigned o each wildli e iangle acco ding o he as e cell in which he iangle midpoin was loca ed, so ha he a iables gi e pe cen ages o a mland and o es in he landscape wi hin a 5 km adius om he iangle cen e s. The amoun o a mland and o es habi a wi hin landscape has p e iously been ound o in luence 506 ª2015 The Au ho s. Ecology and E olu ion published by John Wiley & Sons L d. Eagle P eda ion Impac upon Ma en and Fox M. S. Lyly e al. he abundances o ed oxes and pine ma ens (Ku ki e al. 1998). La i ude and longi ude a iables we e included in he models as hey we e expec ed o explain he impac o egionally a ying en i onmen al condi ions upon he mesop eda o s. The p oduc i i y o he o es s in no h- e n Finland is dis inc ly lowe han in he sou he n pa s o he coun y (Pel ola 2009). Rela ed o his, condi ions in no he n Finland a e ha she han in he sou h, as empe a u es a e lowe and he e is ypically mo e snow- all du ing he da k pola win e s. On he o he hand, eindee husband y occu s only in he no he n pa s o Finland (Fig. 1), and in his a ea, he die o he golden eagle also includes eindee s, mainly in he o m o cal es and ca casses (Nybakk e al. 1999; Sulka a e al. 1999; No be g e al. 2006). S a is ical analyses Fo da a handling and analyses, we used R-so wa e, e - sion 3.0.2 (R Co e Team 2013). S a is ical analyses we e conduc ed wi h gene alized addi i e mixed model (GAMM), using he “gamm” unc ion om he package “mgc ” (Wood 2011). To analyze he impac o golden eagle upon he abun- dance indices o ed ox and pine ma en, we buil mod- els wi h quasi-Poisson dis ibu ion. DNN, and TD and FD wi h 1–5 km we e all se as explana o y a iables in sepa a e models (11 models in o al). The in e ac ion o longi ude and la i ude was included in he models o accoun o la ge-scale spa ial au oco ela ion. Also, he ollowing co a ia es we e cen e ed and scaled, ollowing Schielze h (2010), and included in he model s uc u e: he p opo ion o a mland, he p opo ion o o es , moun ain ha e abundance, g ouse abundance, and he ime o mesop eda o ack accumula ion in days. In he pine ma en models, he abundance o he la ge ed ox was also included as a p eda o co a ia e. To accoun o epea ed samples om he same iangles, ha is, o han- dle pseudo- eplica ion issues (Hu lbe 1984), a andom s uc u e o indi idual iangles and an AR1 co ela ion s uc u e o iangle ID wi hin yea we e included in all models. Model selec ion was s a ed by i ing a smoo he o all co a ia es excep he ack accumula ion day, which was expec ed o ha e a linea ela ion wi h he esponse a i- able. Fo o he a iables han eagle and coo dina es, he k (uppe limi on he deg ees o eedom associa ed wi h a smoo h) was cons ic ed o imp o e con e gence. F om linea es ima es, he unnecessa y smoo he s we e emo ed and hen he model was e i ed. A e his, all nonsigni i- can a iables we e emo ed –only eagle a iables we e always e ained as a minimal model s uc u e in o de o imp o e he compa abili y o he models. The bes o TD and FD was selec ed based on he amoun o de iance explained in he inal models (in TD models 4 km o ed ox, 5 km o pine ma en; in FD models 5 km o bo h species, see Table 1). Resul s a e epo ed only o models TD ,FD ,and DNN o ed ox and pine ma en ( o a iables e ained in inal models, see Table 2). Resul s Red ox The ela ionship be ween he e i o y densi y (TD, 4 km) and ed ox abundance was linea : he mo e eagles he e we e, he highe ox numbe s we e ( =2.305, P=0.021, Fig. 3A). In con as , ledgling densi y o eagles in he p e ious yea (FD, 5 km) had no ob ious associa ion wi h ox abundance ( =0.777, P=0.437, Fig. 3B). The in luence o dis ance o nea es eagle nes (DNN) was nonlinea , bu when examined only a close dis ances (up o 30 km) i was ai ly linea : he sho e he dis ance o ac i e eagle nes s was, he mo e oxes he e we e (F=6.337, P<0.001, Fig. 3C). Table 1. The amoun o de iance explained (DE) in all ed ox and pine ma en models. Models selec ed based on hei DE ha e bolded alues. Red ox models DE Pine ma en models DE Dis ance o nea es nes 95145.24 Dis ance o nea es nes 19079.21 Te i o y densi y, 4km 94824.11 Te i o y densi y, 5km 18762.96 Te i o y densi y, 3 km 94823.30 Te i o y densi y, 4 km 18750.45 Te i o y densi y, 5 km 94811.90 Te i o y densi y, 3 km 18668.46 Te i o y densi y, 2 km 94786.98 Te i o y densi y, 2 km 18639.04 Fledgling densi y, 5km 94778.21 Fledgling densi y, 5km 18630.63 Fledgling densi y, 4 km 94773.42 Fledgling densi y, 4 km 18607.4 Fledgling densi y, 3 km 94759.72 Fledgling densi y, 3 km 18597.67 Fledgling densi y, 1 km 94753.89 Fledgling densi y, 2 km 18581.58 Fledgling densi y, 2 km 94751.36 Fledgling densi y, 1 km 18568.67 Te i o y densi y, 1 km 94751.32 Te i o y densi y, 1 km 18567.63 ª2015 The Au ho s. Ecology and E olu ion published by John Wiley & Sons L d. 507 M. S. Lyly e al. Eagle P eda ion Impac upon Ma en and Fox The impac o habi a a iables on ed oxes was he same in all h ee models (see Table S1 o all esul s). In e media e p opo ions o a mland and low o in e - media e p opo ions o o es in he landscape we e mos bene icial o oxes. As expec ed, he abundance o oxes inc eased wi h inc easing moun ain ha e numbe s, al hough his associa ion le eled o a e y high ha e abundances. In spa ial e ms, he model p edic ed highes ox abundances in ag icul u e- and o es -domina ed sou hwes Finland, g adually dec easing owa d no heas , wi h a local dec ease in he eas e n pa s o Middle Finland. Pine ma en Te i o y densi y (TD, 5 km) had a nonlinea ela ion- ship wi h pine ma en abundance: om low o in e medi- a e TD, he amoun o ma ens sligh ly inc eased, bu a high TD, i appea ed o dec ease (F=5.300, P=0.004, Fig. 4A). Howe e , wide con idence in e als esul ing om ela i ely ew da a poin s a high eagle densi ies necessi a e cau ion when in e p e ing he associa ion. Fledgling densi y (FD, 5 km) had no ob ious associa- ion wi h he ma en abundances ( =1.412, P=0.158, Fig. 4B). Dis ance o nea es nes (DNN) had a nonlinea impac on pine ma en abundance, which a close ange o ac i e eagle nes s ansla ed o he sho e he dis ance, he mo e ma ens he e we e (F=5.008, P<0.001, Fig. 4C). The esponse o ed ox abundance was signi ican and iden ical in all o he h ee pine ma en models: he num- be o ma ens obse ed inc eased sligh ly when he ox abundance inc eased om low o in e media e ox abun- dance, bu a high ox abundances he numbe o ma ens dec eased (F=23.695, P<0.001, Fig. 5). To be conse a- i e, he esul s a e epo ed om he DNN-model, which had he lowes F- alue. Mo eo e , all h ee models p e- dic ed pine ma en abundances o be highes when he p opo ion o o es in he landscape was high (see Table S1 o all esul s). The esponse o p opo ion o a m- land was nonlinea , wi h lowes ma en abundances occu ing a landscapes wi h in e media e amoun s o a mland. Spa ially examined he highes ma en abun- dances occu ed in sou he n and cen al Finland and he lowes in he wes e n pa s o Middle Finland. Discussion We examined he abundances o wo mammalian meso- p eda o s, he ed ox, and he pine ma en, in ela ion o he p eda ion isk imposed by he golden eagle by com- bining long- e m da a om moni o ing o game animal abundances and eagle nes ing. We p edic ed ha ma en abundance indices would be lowe in a eas whe e he e a e mo e ac i e eagle e i o ies and ledglings p oduced, and also when close o inhabi ed eagle nes s, whe eas ed ox abundances would no espond nega i ely o hese ac o s. Ou esul s show ha bo h mesop eda o s we e mos abundan in a eas o high e i o y densi y and e y close o ac i e eagle nes s. In ega ds o he pine ma en, his was he opposi e o ou p edic ion. Howe e , he esul s ga e some indica ion ha eagles could be ha m ul o pine ma ens a e y high e i o y densi ies. Acco ding o ou esul s, ed oxes seem o be abun- dan in he icini y o ac i e golden eagle nes s. Al hough he amoun o o es and a mland in he landscape was accoun ed o , we suspec ha he e we e s ill some unex- plained ac o s ha bo h eagles and oxes bene i om, esul ing in inc eased densi ies o bo h species in he same a eas. The golden eagle is sensi i e o any human dis u - bance (Ollila and Koskimies 2006; Kaisanlah i-Jokim€ aki e al. 2008), and he ed ox can manage well in emo e a eas oo, al hough i bene i s om agmen ed landscapes occu ing close o human inhabi a ion (Ku ki e al. 1998). Foxes and eagles p ey pa ly on same p ey species, such as ha es, g ouse, and o he bi ds (Kauhala e al. 1998; Sulka a e al. 1999; Dell’A e e al. 2007). The amoun o ha e and g ouse was aken in o accoun , bu su icien a ailabili y o o he sha ed p ey could enable he wo species o h i e in same a eas. Fo example, he local a ailabili y o eindee ca casses would p o i bo h species. All in all, i seems ha a he popula ion le el Table 2. Re ained a iables in he selec ed GAMM models o ed ox and pine ma en. The a iables applied wi h smoo he s a e in i alics. In e - ac ions a e ma ked wi h a symbol 9. E and N e e o la i udinal and longi udinal coo dina es and abb e ia ion acc.days o he snow- ack accu- mula ion ime in days. Response P eda o s Loca ion, habi a P ey, census ime Red ox Te i o y densi y E9N, a mland, o es Moun ain ha e, acc. days Red ox Fledgling densi y E9N, a mland, o es Moun ain ha e, acc. days Red ox Dis ance o nea es nes E 9N, a mland, o es Moun ain ha e, acc. days Pine ma en Te i o y densi y, ed ox E 9N, a mland, o es Moun ain ha e, acc. days Pine ma en Fledgling densi y, ed ox E 9N, a mland, o es Moun ain ha e, acc. days Pine ma en Dis ance o nea es nes , ed ox E 9N, a mland, o es Moun ain ha e, acc. days 508 ª2015 The Au ho s. Ecology and E olu ion published by John Wiley & Sons L d. Eagle P eda ion Impac upon Ma en and Fox M. S. Lyly e al. oxes a e able o a oid he ha m ul impac s o eagle p e- da ion o in imida ion, as hey did no espond o he landscape o ea modeled by eagle p esence. This can esul om he ac ha ed oxes a e usually o he same weigh o e en la ge han golden eagles, whe eas in IGP, he kille is ypically clea ly la ge han he p ey (Palo- ma es and Ca o 1999; Se gio and Hi aldo 2008). Te i o y densi y ( e i o ies/100 km²) Red ox abundance ( acks obse ed/day) (A) 01234 23456789 0.0 0.5 1.0 1.5 2.5 3.0 3.5 4.0 4.5 5.0 0 5 10 15 20 25 30 3.0 3.5 4.0 4.5 5.0 Fledgling densi y ( ledglings/100 km²) Red ox abundance ( acks obse ed/day) (B) Dis ance o nea es nes (km) Red ox abundance ( acks obse ed/day) (C) Figu e 3. Red ox snow- ack abundance es ima es a a ying golden eagle e i o y densi y (TD, panel A), ledgling densi y (FD, panel B), and dis ance o nea es nes (DNN, panel C), p esen ed wi h 95% CI (dashed lines). Te i o y densi y ( e i o ies/100 km²) Pine ma en abundance ( acks obse ed/day) (A) Fledgling densi y ( ledglings/100 km²) Pine ma en abundance ( acks obse ed/day) (B) 0.0 0.5 1.0 1.5 2.0 2.5 0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5 0 5 10 15 20 25 30 0.0 0.5 1.0 1.5 Dis ance o nea es nes (km) Pine ma en abundance ( acks obse ed/day) (C) Figu e 4. Pine ma en snow- ack abundance es ima es a a ying golden eagle e i o y densi y (TD, panel A), ledgling densi y (FD, panel B), and dis ance o nea es nes (DNN, panel C), p esen ed wi h 95% CI (dashed lines). ª2015 The Au ho s. Ecology and E olu ion published by John Wiley & Sons L d. 509 M. S. Lyly e al. Eagle P eda ion Impac upon Ma en and Fox Simila ly o he ed ox, he abundance o pine ma en was ound o co ela e posi i ely wi h eagle e i o y den- si y, speci ically when mo ing om low o in e media e densi ies. This likely de i es om he ac ha he golden eagle and he pine ma en bene i om simila habi a s. Al hough he amoun o o es co e age was con olled o in he analyses, he e is also a ia ion be ween o es s ands in e ms o wood olume, ee densi y, and s and age. Ma ens eside mainly in o es ed a eas and p e e o mo e in old sp uce- and pine-domina ed s ands (B ain- e d and Rols ad 2002). Golden eagles ypically nes in la ge pines (Ollila and Koskimies 2006), bu habi a s p o- iding hese ees a e limi ed in Finnish o es s. Thus, excluding e y dense o es s ands ha a e di icul o eagles o mo e in, he wo species likely eside in simila ype o habi a s. This was also suppo ed by he ac ha we obse ed highe ma en abundances wi hin sho e dis ances om ac i e eagle nes s. A densi y-dependen p eda ion impac was ound wi h ed oxes on pine ma ens: a high abundances, oxes could ac as an IG p eda o o ma en. Red oxes ha e been obse ed o p ey on pine ma ens, o example, in Sweden, whe e he dec ease o oxes in 1980s due o sa c- op ic mange, caused by a pa asi ic mi e, was ollowed by an inc ease in he numbe o ma ens (Linds € om e al. 1995). Howe e , landscape-le el s udies om Fennoscan- dia ha e epo ed con a y esul s (Ku ki e al. 1998). Conside ing ed oxes and pine ma ens, i is impo an o no e ha hese gene alis p eda o s may also compe e o e he same ood esou ces, which likely in ensi ies hei in e ac ion a high abundances. Die s udies ha e shown ha he o e all p ey selec ion o he wo species is o e - lapping, al hough ma ens ha e less la ge-sized p ey in hei die han oxes do (Pulliainen and Ollinm€ aki 1996; Kauhala e al. 1998). Ou esul s emphasize he impo ance o he iming o obse a ions; in con as o e i o y densi ies om he same ime pe iod, mesop eda o abundances did no espond o he ledgling densi y o eagles in he p e ious summe . One eason o his could be ha a e ledging, ju enile eagles mo e ou side he home ange o such an ex en ha a clea p eda ion impac wi hin he e i o y canno be obse ed. In addi ion, young eagles ypically s a dispe sing al eady du ing la e au umn (Wa son 2010), a e which he p eda ion impac is gene a ed solely by he adul s emaining a he e i o y. Fu he - mo e, mo emen s o he mesop eda o p ey can dilu e he e ec o eagle p eda ion. The e o e, a ime lag o o e 6 mon hs may be oo long o any p eda ion impac o show on mesop eda o abundances. Ins ead, i could be be e o s udy he impac o b eeding eagles and hei ledglings on mesop eda o abundances in au umn. Es i- ma ed eagle e i o y densi y, which is able o accoun o mul iple close e i o ies, was pe haps he bes p oxy o p eda ion isk, al hough he dis ance o eagle nes explained he la ges amoun o a ia ion in he da a. We expec ed o de ec a p eda ion impac o golden eagle upon pine ma en abundance, bu he analyses did no p o ide subs an ial suppo o his. Howe e , we wish o no e ha he da a showed a dec easing end in ma en abundance a high e i o y densi ies o eagles, ha is, when he e is one o mo e esiden eagle pai wi hin he a ea. As he impac o oxes was sepa a ely con olled o in ou analyses, we do no belie e ha his end was a esul o combined p eda ion p essu e by eagle and ox. Ins ead, a esponse o he landscape o ea shaped solely by he eagle emains a mo e plausible expla- na ion. A e all, ma ens weigh app oxima ely only one- hi d o he weigh o he eagle and a e hus an easie p ey o he eagle o hun han oxes a e. I has p e iously been no ed ha pine ma ens a oid e y open a eas and clea cu s (B aine d and Rols ad 2002), and his local- scale habi a p e e ence o ma ens could be ela ed o he h ea o golden eagle p eda ion (Ko pim€ aki and No d- s € om 2004). In gene al, medium-size p eda o s ac i ely a oid habi a s u ilized by apex p eda o s (Fed iani e al. 1999; Mukhe jee e al. 2009). Simila ly, he eagle could, in addi ion o di ec killing, educe ma en numbe s by in imida ion. Also, eagles may dis u b ma ens and ham- pe hei o aging, hence dec easing hei i ness. In an ea lie s udy, p eda ion isk imposed by he whi e- ailed sea eagle (Haliaee us albicilla, L.) has been shown o educe he swimming ips o ano he mus elid species, he Ame ican mink (Neo ison ison, Sch ebe ) (Salo e al. 2008). In con as , s ong ood compe i ion among golden eagles and pine ma ens is unlikely, as he die o 0 50 100 150 200 0.0 0.5 1.0 1.5 2.0 Red ox abundance ( acks obse ed/day) Pine ma en abundance ( acks obse ed/day) Figu e 5. Pine ma en snow- ack abundance es ima es a a ying ed ox abundances, e ie ed om he dis ance o nea es nes (DNN) model. 95% CI a e p esen ed wi h dashed lines. 510 ª2015 The Au ho s. Ecology and E olu ion published by John Wiley & Sons L d. Eagle P eda ion Impac upon Ma en and Fox M. S. Lyly e al. hese species is ai ly b oad ye no o e lapping much (Pulliainen and Ollinm€ aki 1996; Wa son 2010). Resea ch conduc ed a ine spa ial scales could help o es ablish whe he an abundan eagle popula ion causes any educ- ion in pine ma en numbe s, o example, ia beha io al changes ha esul in lowe i ness. New in o ma ion on he impac o he golden eagle could be acqui ed, o example, by adio- acking habi a use o ma ens and obse ing he p eda ion beha io o eagles. When abundan , mesop eda o s hold he po en ial o induce bo h ecological and economic cos s wi h hei ou - b eaks (P ugh e al. 2009), and he e o e, unde s anding p ocesses ela ed o hei abundance is impo an . Species such as he ed ox, pine ma en, and Ame ican mink a e sugges ed o limi o be he cause o decline o ha e and bi d popula ions (Ma cs € om e al. 1988; Linds € om e al. 1994; Kauhala and Helle 2000; No ds € om e al. 2002). No su p isingly, IGP om a op p eda o upon mesop e- da o s may also cause cascading impac s on he bi o es (e.g., Henke and B yan 1999; Helldin e al. 2006; P ugh e al. 2009). Elmhagen e al. (2010) showed ha when he ecolonizing lynxes (Lynx lynx, L.) limi ed ox popula- ions, his has an indi ec posi i e impac on moun ain ha e abundance. Fielding e al. (2003) sugges ed ha h ough supp essing medium-sized ap o s he golden eagle could educe he o e all p eda ion p essu e on game species. Howe e , such cascading e ec s wi h a ian–mam- malian IGP ha e no eally been examined, e en hough hey could ha e ex ensi e in luence on ecosys ems. Fo example, on he Cali o nia Channel Islands (USA), he in oduc ion o an exo ic p ey has enabled he golden eagle o s ongly eshape he local mesop eda o commu- ni y and indi ec ly a ec he g ani o e p ey (Roeme e al. 2009). In no he n Eu ope, he eagle owl (Bubo bubo) could also be ac ing as p ominen a ian op p eda- o (Ko pim€ aki and No ds € om 2004). Based on he esul s p esen ed he e, i is no clea whe he he bi o es (e.g., ha es and g ouse) would bene i om eagles p eying upon and in imida ing ed ox and pine ma en. None heless, i is impo an o keep in mind ha he Finnish mesop eda o guild en ails se e al o he species ha may be in luenced by eagles. Mo eo e , by expanding hei ange in he u u e, golden eagles (Ollila and Ilmonen 2009; Wa son 2010) as well as o he ap o s may inc ease p eda ion on mesop eda o s in new a eas. This could, in u n, imp o e he p o ec ion s a us o golden eagles and ap o s in gene al i ap o -induced mesop eda- o supp ession we e o bene i he bi o ous species. In addi ion, he p eda ion impac s o ap o s could be com- pa ed wi h hose caused by mammalian op p eda o s, which in Finland include he inc easingly nume ous lynx and bea (U sus a c os), as well as he less common and p obably only locally ele an wol and wol e ine (Gulo gulo, L.) (Wikman 2010). This ype o compa ison o he s eng h o op-down supp ession would help o assess he impo ance o ap o s in bo eal ecosys ems. Conclusions This s udy p o ides no el in o ma ion on a ian–mamma- lian IGP by examining he ela ionship be ween he golden eagle and i s po en ial p ey, he pine ma en and he ed ox. We ound bo h mesop eda o species o be mos abundan in p oximi y o he golden eagle sugges - ing ha killing and in imida ion by eagles may no be a ele an cause o dec ease o popula ions o ma ens and oxes. Howe e , we su mise ha pine ma ens could su - e om p eda ion a high densi ies o eagle e i o ies, pa icula ly due o hei smalle size. The in o ma ion epo ed in his a icle imp o es ou unde s anding o he ole o a ian op p eda o s in e es ial communi ies, bu u he long- e m s udies a e equi ed o o m a clea e pic u e o he impac s ha eco e ing a ian op p eda- o s, such as he golden eagle, ha e on p eda o guilds. Ga he ing beha io al da a would also help us o unde - s and in mo e de ail how mesop eda o s cope wi hin he p esence o la ge a ian p eda o s. Po en ial cascading e ec s o mesop eda o supp ession on lowe ophic le els would be o g ea in e es , oo. Acknowledgmen s We hank all he olun ee s who ha e helped o collec he golden eagle and game abundance da a. We also hank E ic Le To o ec o e ising he language and he edi o s and an anonymous e iewe o help ul commen s on he d a manusc ip . The s udy was inancially sup- po ed by Jenny and An i Wihu i Founda ion (g an s o MSL), and he Academy o Finland (G an Nos. 123379, 136717 and 250709 o EKo ). Con lic o In e es None decla ed. Da a Accessibili y Da a used o his s udy can be ound in D yad da a a chi e (ASCII ex ile, doi: 10.5061/d yad.qc6n ). The loca ion da a a e no included o a oid endange ing nes - ing si es o he golden eagle. Habi a da a come om he Eu opean En i onmen Agency (2009). Re e ences Baddeley, A., and R. Tu ne . 2005. spa s a : an R package o analyzing spa ial poin pa e ns. J. S a . So w. 12:1–42. ª2015 The Au ho s. Ecology and E olu ion published by John Wiley & Sons L d. 511 M. S. Lyly e al. Eagle P eda ion Impac upon Ma en and Fox