j RaptorRes. 31 (4):370-375 ¸ 1997 The Raptor ResearchFoundation, Inc. SELECTIVE PREDATION ON SCANSORIAL AND ARBOREAL MAMMALS BY RUFOUS-LEGGED OWLS (STRIX RUFIPES) IN SOUTHERN CHILEAN RAINFOREST DAVID R. MARTINEZ Laboratoriode Ecologia, Departamento de CiendasBtisicas, Universidad de LosLagos, Casilla 933, Osorno, Chile F• M. JAKSIC Departamentode Ecologia, Pontificia Universidad Catdlica de Chile,Casilla 114-D, Santiago,Chile ABSTRACT.--Weestimated the annual mammalian diet of Rufous-leggedOwls (Strix rufipes)in three temperate rainforest sitesin southern Chile (Anticura, Rupanco and San Martin), and mammal abun- dance at San Martin and Anticura. Diets were estimatedby analysisof 894 pelletscollected from 1988- 95. In all sites,arboreal (arborealmouse, Irenomys tarsalis and colocoloopossum, Dromiciops australis) and scansorial(long-tailed mouse, Oryzomyslongicaudatus) mammals occurred significantlymore fre- quently than those mammals with cursorial habits, such as the olivaceousfield mouse (Akodonolivaceus) and long-haired field mouse (A. longipilis).Although the body size and abundance of the olivaceous field mousewere similar to thoseof the long-tailedmouse, the former specieswas seldomtaken by Rufous-leggedOwls, likely becauseit wasassociated with vegetationalfeatures that offer overheadpro- tection. KEYWORDS: Strix rufipes;Rufous-legged Owl; prey selection; Patagonian Wood Owl; Chile,,temperate forest. Depradaci6nselectiva sobre mamlferos escansoriales y arbor/colaspor concones(Strix rufipes) en la pluviselvadel sur de Chile RESUMEN.--Estimamosla diem anual del bfiho de bosqueo conc6n (Strixrufipes) en tres localidadesde pluviselvatemplada (Anticura, Rupanco y San Martin) y la relacionamoscon la abundanciaanual de micromamlferospresentes en San Martin y Anticura. La dieta se determin6 mediante el anfilisisde 894 egagr6pilascolectadas entre 1988-95. En los tres sitios,los micromamlferosde hfibitosarbor/colas (Ir- enomystarsalis y Dromidopsaustralis) y escansoriales( Oryzomyslongicaudatus) fueron significativamente milsdepredados que los de hfibitoscursoriales tales como Akodon olivaceus y A. longipilis.A pesarque A. olivaceusexhibi6 abundanciaen terreno y tamafiocorporal similares a los de O. longicaudatus,casi no fue depredadopor losbfihos. Esto puede atribuirsea su asociaci6ncon variablesvegetacionales que le ofrecerian protecci6n desde un plano vertical. [Traducci6n de autores] The Rufous-legged Owl or Patagonian Wood sidered to be declining in southern Chile because Owl (Strix rufipes)is one of the least known noc- of decreasinghabitat brought about by logging turnal raptors of South America, inhabiting the (Jaksi• and Jimfnez 1986). This was confirmed temperate rainforest region shared by Chile and when Martinez andJaksi• (1996) reportedthe first Argentina (Jaksi•1997). What information there is quantitative information on the relative abundance about this species is either anecdotal (Housse of Rufous-leggedOwls in forest remnantsof south- 1945, Johnson 1967) or consistsof brief accounts ern Chile, and described forest stands used by on taxonomy and distribution (e.g., Humphrey et these owls. al. 1970, Vuilleumier 1985,Jaksi• and Feinsinger The first quantitativereport on the diet of the 1991). Rufous-leggedOwl wasbased on an analysisof 161 From a conservation viewpoint, Glade (1988) pellets (Martinez 1993). Although the study con- listed this speciesas inadequately known. While the cludedthat Rufous-leggedOwls are generalistfeed- Rufous-leggedOwl is relatively common it is con- ers, the most frequent vertebrateprey were arbo- 37O DECEMBER 1997 SELECTrv• PREDATION BY RUFOUS-LEGGED OWLS 371 real and scansorialsmall mammals. Cursorial spe- equipped with one trap for a total of 144 traps covering cies, despite their similar or higher diversity(Mes- an area measuring132 X 132 m (1.7 ha). Each census lastedsix consecutivenights every month for a full year erve and Jaksi• 1991), were poorly represented. The mean annual abundance (recaptures were not con- Here, we report additional data on the diet of Ru- sidered) of each small mammal speciestrapped wasused fous-legged Owls that inhabit three rainforest to estimate the expected frequency of prey consumpuon stands in southern Chile and test the assumed se- by Rufous-leggedOwls (e.g., proportion of mammal spe- cies i in the field with regard to the abundance of all lective predation noted by Martinez (1993). speciestrapped). The actual prey identified in the d•et STUDY AREAS AND METHODS constituted the observed frequency. For Anticura, we used data reported by Rau et al. (1995), who assessedthe We studied Rufous-leggedOwls in three sites in the abundance of ground-dwelling small mammals with two Valdivian Rainforestregion of southern Chile (sensuVeb- trap lines 2 m apart with 30 stationsset at 12-m intervals len et al. 1983). The first site was the San Martin Exper- and equipped with one medium-sizedSherman trap per imental Forest (39ø38'S, 73ø11'W; 20 m elevation) in the station for a total of 60 traps in an area measuring 348 coastal ranges near the city of Valdivia. This 150-ha site X 4 m (0.14 ha). Rau et al. (1995) simultaneouslyesu- was covered by a second-growthmultistoried forest rem- mated the abundance of tree-dwelling small mammals nant dominated by Aextoxiconpunctatum and Podocarpus with 60 hair-samplingtubes nailed to trees along the trap salignawith substantial Gevuinaavellana and some scat- lines. An empirical estimate of abundance was obtained tered old individualsof emergent Nothofagusdombeyi. The by pooling these data, given that the trapping effort was understory was sparse,and the soil was covered with a the same for ground- and tree-dwelling small mammals. thick layer of mossesand litter. A Chi-squaretest was performed to test the goodness- The second site wasa 180-ha second-growthforest near of-fit of the frequency distributionsof prey in the d•et Lake Rupanco(40ø45'S, 72ø38'W; 150 m elevation)that and in the field (Zar 1984). Nonsignificant values were consistedof an almost pure stand of A. punctatumwith interpreted as if Rufous-leggedOwls used prey in pro- some Laurelia sempervirens,G. avellanaand old N. dombeyi portion to their field abundance, whereas significant de- trees. viations suggestedthat Rufous-leggedOwls "preferred" The third site, Anticura (40ø40'S, 72ø10'W; 600 m el- or "avoided" some small mammal species,thus appear- evation), was an extensive preAndean forest located in ing to select prey. Puyehue National Park, near the city of Osorno. It wasa Becausewith the Chi-square test it is not possible to multistoried old-growth stand with emergent trees (N. determine which prey specieswas individually avoided or dombeyi)35-40 m in height; also present were L. phillipi- preferred, we constructed confidence intervals for each ana, Eucryphiacordifolia and shade-toleranttrees such as prey using the Bonferroni adjustment(Neu et al. 1974, Weinmania trichospermaand A. punctatum.Shrubs were Byerset al. 1984): scarce and the understory was comprised of saplingsof Pi-- Za/2k%/Pi(1 -- Pi)/n --•Pi --• Pi + Za/2k%/Pi( 1 -- Pi)/n, shade-tolerant trees, ferns, a few bromeliads and both bamboo (Chusqueaspp.) and Ribesmagellanicum clumps. where Pi is the proportion of consumptionof prey spe- The ground was covered by mosses(Sphagnum sp.), and cies i, k is the number of prey species,n is the total num- •n damp areas ferns reached tree-like sizes. ber of prey and Za/, is the upper standardnormal vanate Owl pairs were resident in all three sitesyear-round corresponding to a probability tail area of a/2k. The 2k and their pellets were collected under roosting trees. In denominator under a is used because multiple confi- SanMartin, we collected72 pelletsfrom April 1988 (early dence intervals were computed simultaneously.If the autumn)-February 1990 (late summer). In Rupanco, we confidenceinterval included the expectedproportion of collected 347 pellets from February 1993-September consumption(Pio), then we did not reject the hypothes•s 1995 (early spring) and, in Anticura,we collected475 that prey specieswere preferred or avoided. If the ex- pellets from September 1994-October 1995 (spring). We pected proportion of consumption was not included •n •dentified and quantified most vertebratesin the pellets the interval, we concluded that observedand expected on the basisof skulls and dentary pairs, whichever gave consumption differed significantly.Statistical significance was set at P -• 0.05 for all tests unless otherwise stated. the highest count. For other remains, such as hair, we used reference collectionsand quantified these prey as- RESULTS suming the smallestpossible number of individuals (e.g., presence of hair of a given specieswas deemed as rep- In the three sites,both by number and biomass, resenting only one individual). We estimatedthe biomass the scansoriallong-tailed mouse (Oryzomyslongicau- contribution of each prey type in the diet by multiplying the number of individualsin the pellet by the mean body datus)was the staple prey for Rufous-leggedOwls mass of that prey item. We assumedthat massesof un- (Table 1). Although the biomass contribution identified prey were similar to the average mass of the made by the arboreal mouse (Irenomystarsalis) was most closelyrelated identified taxa. We did not consider lower, it was consistentlyeaten at all sites.This was nonmammalian prey items, becausetheir contribution to also true of the scansorial black rat (Rattus rattus). the biomassconsumed was minimal in all sites (•10%). In San Martin, small mammal