Review The Effect of Edge on Avian Nest Success: How Strong Is the Evidence? PETER W. C. PATON Department of Fisheries and Wildlife and Utah Cooperative Fish and Wildlife Research Unit Utah State University Logan, UT 84322, U.S.A. Abstract: Wildlife biologists historically considered the edge El efecto de borde sobre el (~dto en la nidiflcaci6n: ~Cuan between adjacent habitat types highly productive and bene- firme es la evidencia? ficial to wildlife A currant dogma is that edges adversely Resumen: His~t~ los bi6logos de fauna silvestre affect a wide range of avian species by increasing depreda- conslderaron a los hordes entre tipos de habitats adyacentes tion and parasitism rates of nest~ I critically evaluated ex- isting empirical evidence to test whether there was a grada- como altamente productivos y beneficiosos para la fauna silvestre Un dogngt actual sostiene que los hordes afectan tion in nest success as a function of distance from an edge Researchers investigating this question have been inconsis- adversamente a un ampllo rango de especies de ayes al in- tent in their experimental design~ making generalizations crementar las tasas de predact6n y parasittsmo de los nldo~ Yo evalud criticamente la evldencla emptrtca existente para about edge-effect patterns difficul£ The majority of studies I examined found nest success varied near edge~ with both establecer st existla un gradtente en el dxito de los nldos en funci6n de la distancla a un bord~ Los investigadores que depredation rates (10 of 14 artificial nest studie~ and 4 of 7 natural nest studies) and parastttsm rotes (3 of 5 studies) anal~ esta cuestidn hart sldo inconsistentes en el di- increasing near edger In addittor~ there was a positive rela. sefJo de sus experimento~ haciendo diftcil las generaliza- tlonship between nest success and patch size (8 of 8 studies). clones acerca de los patrones del efecto de borde& La mayotqa de los estudios que examind establecieron que el The most conclusive studies suggest that edge effects usually occur within 50 m of an edgg whereas studiesproposing that ~xtto de los nldos varl6 cerca de los borde~ obsenmndose un increased depredation rates extend farther than 50 m from i~to tanto en las tasas de predaci6n (10 de 14 estu- an edge are less convincing. Prior research has probably fo- dios de nldos artiflciales y 4 de 7 estudios de nidos natu- cused on distances too far from an edge to detect threshold rales) como de parusittsmo (3 de 5 estudlos) cerca de los value~ and future research should emphasize smaller scale~. borde~ Ademds existi6 una relaci6n posttiva entre el dxtto 100-200 m from an edge at 20-25 m inotnnent~ Research- de los nldos y el tamatio de los parches (8 de los 8 estudios). ers often use relatively arbitrary habitat characteristics to Los estudios mds decisivos sugieren que los efectos de horde define an edge Therefore, Ipropose that only openings in the ocurren usualmente a menos de 50 m del bordg mientras forest cxmopy with a diameter three times or more the height que los estudios quc proponlan que los incrementos en las of the adjacent trees should be included in edge analyse,~ tasas de predacidn se extendlan rods alld de 50 m del horde This review suggests that fragmentation of eastern North fueron poco convincente~ Investigaclones anteriores se con- Ame~can temperate forests could lead to increased nest pre- centraro~ posiblementg en distancias mwy aleJadas del dation and parasitisnt and there is need to determine if horde como para detectar valores de umbral y la investtga- similar processes occur in other forested regions of North ct(m futura deberla enfattzar menores escalas; 100-200 m America dei horde con incrementos de 20-25 nt Los investigadores usan frecuentemente caracter~ticas del hdbitat relativa. mente awbitrarias para definir un horde. Por consiguiente Paper submitted August 3, 1992; retdsed manuscript accepted July propon8o que sdlo se incluyan en los andlisis de horde a- 18, 1992t berttwas en el canopeo forestal q~ posean un didmetro su- 17 Conservation Biology,Pages 17-26 Volume 8, No. 1, March 1994 18 Edge Effects on Avian Nests P~ton perlor a 3 veces la altura de los drboles adJacente~ Esta revisi6n sugiere que la fragmentaci6n de las forestas del este de Norte Aradrica podria conducir a un i~to en la predact6n y parasitismo de nido~ y que es necesarto deter- minar si procesos similares ocurt~ en otras regiones fores- tales de Notre Amdrica Introduction Methods Many biologists consider the edge between adjacent I have divided this review into two sections. The first habitat types a positive feature of the landscape for wild- section, on distance as an independent variable, deals life (see Kremsater & Bunnell 1992). This belief is based with studies that used artificial or natural nests to quart. in part on Leopold's (1933) "law of interspersion," tiff/the relationship between nest success and distance which postulated that increases in the amount of edge from an edge. I attempted to include in this review all habitat resulted in higher population densities (Guthery studies that have directly measured nest success as a & Bingham 1992). Leopold's edge-effects hypothesis led function of distance from an edge. The second section, many wildlife biologists to create edges to enhance hab- on patch size as an independent variable, analyzes stud- itat for game species (for example, Yoakum et al. 1980: ies that indirectly investigated edge effects by quantify- 402). Within the past decade, however, wildlife biolo- ing the relationship between nest success and patch gists have pointed out that Leopold's hypothesis size. There is a large body of literature focusing on the pertained only to certain species and have challenged relationship between patch size and avian species rich- the idea that edges benefit most wildlife (see Reese & ness (for example, Lovejoy et al. 1986; Hagan & Ratti 1988; Yahner 1988). Johnston 1992), but there are fewer published papers Much of the recent interest in edge is due to avian that have quantified the relationship between patch size population declines in fragmented landscapes, as exem- and nest success. plified by neotropical migrants in eastern North Amer- Study-site habitat types ranged from tropical forests in ica (Askins et al. 1990; Hagan & Johnston 1992). In a Central America (one study), coniferous forests in west- recent summary of breeding season mortality factors for ern North America (two studies), deciduous forests in 32 species of neotroptcal migrants, Martin (1992) iden- eastern North America (five studies), and a mixture of tiffed nest predation as the most important cause of nest habitat types in Europe (six studies; Table 1). Because failure. Therefore, determining what factors influence nests were placed across a broad spectrum of different nest predation is imperative if biologists hope to suc- habitat types, they were exposed to a variety of poten- cessfully manage many avian populations. tial predators (Table 1). Five of the studies (Yahner & One of the most frequently cited explanations for Wright 1985; Angelstam 1986; Andr~n & Angelstam population declines in fragmented landscapes is higher 1988; Moiler 1989; Yahner 1991) found that corvids nest depredation rates near edges, hereafter referred to were the primary predators at their study sites, whereas as edge effects. Gates and Gysel (1978) developed the red squirrels (Tamiasciurus hudsonicus) were in Al- "ecological trap" hypothesis, which postulated that nest berta (Boag et al. 1984), skunks (Mephitis mephitis) predation rates were density-dependent, with greater were in Maine (Vickery et al. 1992), and snakes were in nest densities and a concomitant increase in depreda- grasslands of Illinois (Best 1978). The other studies did tion rates near edges. Subsequent studies have refuted not identify a principal predator. this hypothesis, however, based on data with artificial I tested whether nest success was independent of ei- nests showing that nest success did not decline near ther distance from an edge or of patch size using a like- edges (Angelstam 1986; Ratti & Reese 1988). lihood-ratio chi square (G) test (Agresti 1990 ). Many of Because the data are equivocal, I was interested in the the studies presented here originally used a likelihood- circumstances when edge effects appear to be an im- ratio test to analyze their data (Yahner & Wright 1985; portant biological phenomenon. My purpose is to (1) Moiler 1988, 1989; Berg et al. 1992). Each nest was summarize existing literature that quantified the rela- considered an independent observation (although see tionship between nest loss (either by predation or par- below for a discussion of Gates & Gysel [1978] and asitism) and distance from an edge (or patch size ), (2) Chasko & Gates [1982]). I considered p values <0.1 re-analyze these data where possible, (3) snmmarize the significant, as I wanted to be statistically conservative to conditions when predation and parasitism rates change minimize Type II errors (concluding there was no evi- near edges (for example, degree of contrast between dence for edge effects, when in fact they existed) and adjoining habitats, forested versus unforested habitats, was not as concerned with Type I errors (concluding type of predator ), and, finally, (4) discuss what conclu- edge effects exist when they do not ). sions I believe realistically can be drawn from existing Most artificial nests studies reviewed here used either empirical evidence. real chicken or quail (Coturntx coturntx) eggs, with cmz~zvationBiology Volume8, No. 1, March 1994 Paton Edge/~cts on A~ Nests 19 Tabk 1. Smmry og pm~ pmmm ~Rd mtt~ ~ ta nm ta~.e~ mdte~ I Prlma,7 Type of Type of Exposure Refet~c~e Location Pmdato~ Nes# Eggs (n) c Days Artificial Nests Andr~n & Angelstam 1988 Sweden C t RF PM OG C (2) 8 Angelstam 1986 Sweden C t RFPM OG C(2) 14 Avery et aL 1989 Scotland STRFC 2 OG C(1) 8 Berg et aL 1992 Sweden ND OG Q(2) 28 Boag et al.