Relationships of surrounding riparian habitat to nest-box use and

0 downloads 0 Views 2MB Size Report
open habitats were actively selected based on their higher probability of success. Predation ... Habitat features at or surrounding a nest site can ... In studies of secondary cavity-nesting birds ..... SMale use was counted when a box was filled with twigs but no eggs were laid. ..... In this DA, a total of 75% of 44 outcomes was.
This file was created by scanning the printed publication. Errors identified by the software have been corrected; however, some errors may remain.

The Condor91:848-859 Society 1989 ? The CooperOrnithological

RELATIONSHIPSOF SURROUNDING RIPARIAN HABITAT TO NEST-BOX USE AND REPRODUCTIVEOUTCOMEIN HOUSE WRENS' DEBORAHM. FINCH RockyMountainForestand Range ExperimentStation, 222 South 22nd Street, Laramie, WY82070

Abstract.I assessedrelationships nest-siteselection,and reamonghabitatstructure, threenest-box productiveoutcomeof HouseWrens(Troglodytes aedon)by establishing gridsin riparianwoodlandsin southeastern Wyoming.Overa 3-yearperiod,37%of the boxescontained HouseWrennests;20%contained unusednestsbuiltbymaleHouseWrens; and42%wereneverusedby wrens.Directdiscriminant threebox-use analysesseparated groups(unused,1-yearuse,2- to 3-yearuse)andthreenestingoutcomepossibilities (failure all years,1-yearsuccess,2- to 3-yearsuccess)alonggradientsof habitatcoverandfoliage density.Comparedwith unusedboxes,thoseselectedrepeatedly by HouseWrenswere locatedin habitatswithsparserunderstories. Similarly,wrensweremorelikelyto fledge fromboxeschosenin sparselyfoliatedhabitats.I concludethatcharacteristics of offspring the surrounding wrenreproductive habitatinfluenced outcomeand suggestthatboxesin of success.Predation openhabitatswereactivelyselectedbasedon theirhigherprobability wasthe majorcauseof nestingfailure.By nestingin openhabitats,HouseWrensmaybe moreadeptat detecting anddeflecting intruders beforenestsarediscovered and cavity-nest destroyed. Key words: House Wren;dummy nests;foliage density;habitatstructure;nest-box use; nestingoutcome;predationrisk;riparianwoodlands;secondarycavity-nestingbirds.

INTRODUCTION Habitatfeaturesat or surroundinga nest site can significantlyinfluencethe rate of nestingsuccess through effects on probability of predation (Westmorelandand Best 1985, Martinand Roper 1988),energyexpensesrelatedto nest or holeentrance orientation (Inouye 1976, Walsberg 1981), and food supply (Martin 1987). If reproductivesuccessis greaterin certainhabitats,then birdsmay use habitatfeaturesto evaluatequality of nest sites(Caccamise1977, Montevecchi1978, Cody 1981). In addition, females may choose mates on the basis of nest-sitequality(Kendeigh 1941, Askenmo 1984, Slagsvold 1986) and, therefore,males may construct multiple nests, which offer females an assortment of nest-site alternatives(Kendeigh 1941). In studies of secondary cavity-nesting birds usingartificialnest sites, disparitiesin surrounding habitat structurehave been used to explain patternsof nest-boxoccupancy,but the relationship between habitat variation and nesting success has rarelybeen addressed(Nilsson 1984). When nest boxes are identical in size, interior

color, and hole entrance diameter (Lumsden 1986), nest sites are presumablyselected on the basis of criteriaexternalto the box itself (but see Brawn 1988). Such factors may include proximity to nestingconspecifics(Muldalet al. 1985) or environmentaldisturbance(Zachand Mayoh 1982),food supplyor foragingsubstrate(Franzreb 1977, Hussel and Quinney 1987), probabilityof predation or competition (Nilsson 1984), features of the nest-box tree (Yahner 1983/1984), or adjacenthabitatstructure(Willneret al. 1983, Munro and Rounds 1985, Belles-Islesand Picman 1986a).Habitatcharacteristicssurrounding the cavity site may be used as cues by nesting pairsto predictthe likelihood of nesting success (Nilsson 1984). Under conditionsof nest-holelimitation,cavity-nesting birds have been shown experimentally to compete for nest boxes and naturaltree holes (van Balen et al. 1982, Gustafsson 1988). House Wrens (Troglodytes aedon) are obligate

cavity-nesterswhose abundancecan be limited by availabilityof nest sites (Yahner 1983/1984, Belles-Isles and Picman 1986a). Intense competition for nest sites and mates in House Wrens can lead to increasedaggression,nest usurpation, bigamy, killing of adult conspecifics,or infanti24 1989. Final 23 acceptance 'Received January cide (Belles-Islesand Picman 1986b, 1987; PicJuly1989. [848]

USEAND NESTINGOUTCOME 849 NEST-BOX man and Belles-Isles1988;Freed 1986a, 1986b). If nest sites are scarce,some individualsmay be forcedto select low-qualitysites, and thus, nestsite preferencesmay be difficultto detect. In an effortto reduce the confoundingeffects of interferencecompetition on nest-site selection, I increasedthe availabilityof nest sites to House Wrens by establishingnest boxes in riparianwoodlands(see also Drillingand Thompson 1984, Munro and Rounds 1985). I then assessedwhetherhabitatfeaturesat nest boxes used by wrens differedfrom those at unused boxes. Further,I determinedwhether the same boxes selected in multiple years were more likely to contain successfulnests, and if so, whethersuch boxes were associated with specific habitat attributes.In contrastto other studies which have examinedwithin-yearassociationsbetweenhabitat and nest placement (or nesting success), I definedhabitatdifferencesamongnest sites based on numbersof yearseach box was used. Because the nest site ratherthan a site-shiftingbird was the experimentalunit, this approachproved to be effective in discerninghigh-qualityhabitats, i.e., habitatswhere nesting success is favored. STUDY AREAS Three study plots were established as grids in May 1982 in streamside habitats in Carbon County, southeasternWyoming, at elevations rangingbetween2,050 m and 2,250 m. Two plots were alongthe North PlatteRiver, 13 km northwest and 21 km southeastof Saratoga,and one plot was at Rock Creek, 5 km northeastof Arlington. Narrowleaf cottonwood (Populus angustifolia) dominated the overstory along with scatteredplainscottonwood(P. deltoides),quaking aspen(P. tremuloides),peachleafwillow (Salix amygdaloides),and Rocky Mountainjuniper (Juniperusscopulorum).Midstoryvegetationincludedbush willow species(S. exigua, S. lasiandra, S. bebbiana,S. moniticola, S. ligulifolia), thin-leaf alder (Alnustenuifolia),mountain maple (Acerglabrum), river hawthorn (Crataegus rivularis),common chokecherry(Prunus virginiana), and western serviceberry(Amelanchier alnifolia).Understoriesweredominatedby western snowberry (Symphoricarposoccidentalis), goldencurrant(Ribesaureum),gooseberry(Ribes spp.), cinquefoil (Potentilla grqcilis, P. fructicosa), wild rose (Rosa woodsii), red raspberry (Rubusidaeus),red-osierdogwood (Cornusstolonifera), and a variety of grasses, sedges, and

forbs. Shortgrassprairieinterspersedwith sagebrush (Artemisiaspp.) borderedriparianwoodlands. METHODS On each of the three plots, 21-22 nest boxes (65 boxes total) were mounted 2 m high on live deciduous trees >10 cm dbh. Boxes were placed 30-35 m apartin gridsthat varied in lengthand width owingto size of the ripariancorridor.Nest boxes were built of 1.7-cm thick cedar, 14 x 14 x 28 cm in dimension, with 3.8-cm-diameter entrancesand latchabletop doors, and were labeled with grid coordinates.To standardizeinfluencesof solar radiationand prevailingwinds on nest-box choice, I placed all boxes with their entrancesfacingtowardthe south or southeast --orientations preferredby some cavity-nesting species (Pinkowski 1976, Lumsden 1986). I ascertainedratesof box occupancyand nesting success by checking nest boxes early in the afternoonevery 2-4 days from mid-May to late Julyof 1983, 1984, and 1985. Becauseindividual male House Wrens frequentlyfill multiple cavities with twigs, the appearanceof an egg was used as evidence of site selection by a nesting female. Box use by male wrens was recorded when incomplete twig nests were found but no eggswere laid. Due to the shortbreedingseason (May throughJuly)in these high altitudewoodlands,only a few boxes werenestedin twice during a single breeding season. Box reuse by the same pair of House Wrens could not be distinguishedfrom late box settlementby a new pair. Such "second" nesting attempts were excluded from analyses to control for effects of seasonal variation on reproductivesuccess. Boxes were cleaned out each Septemberso that new twigs and nests were not confused with box contents from the year before. Vegetationstructurewas sampledin June and July 1984 at all nest-box sites to determine if habitat variation affected nest-box choice or nesting success. The period of vegetation sampling was arbitrarilytimed so that samples reflectedhabitatencounteredbetweenthe periods of nest construction(Mayand June)and fledging of young (July and August). At each nest-box tree, I measured 34 habitat variables using a point-centered quarter sampling procedure (Finch 1989). Habitat features were sampled by dividing each location into four quadrants orientedby cardinalcompass directionsaround

850

DEBORAHM. FINCH

TABLE 1. Structuralvariablesand transformationsused in analyses. Mnemonic acronym

Transformation*

Variable

DBH

SQRT

Tree diameter

CANHT

RCPL

Canopyheight

SHHT

LN

Shrubheight

SHCD

LN

Shrubcrowndiameter

SHDIS VFDI

SQRT RCPL

VFD2

SQRT

VFD3

SQRT

VFD4

SQRT

VFD5

LN

EVH

LN

Shrubdispersion Verticalfoliagedensity in grass-forblayer Verticalfoliagedensity in small shrublayer Verticalfoliagedensity in mid-canopylayer Verticalfoliagedensity in lower overstory Verticalfoliagedensity in upperoverstory Effectivevegetationheight

WILL

SQRT

Percentwillow

SAP

LN

Percentsaplings

BARE

SQRT

Percentbareground

HERB

SQRT

Grass-forbgroundcover

LOGS

SQRT

Log cover

SHRUB

LN

Shrubcover

COVER

SQRT

Live and dead woody cover

Samplingmethod

Diameter(cm) at breastheight of nearesttrees (>3 cm) in each quadrant. Mean height(m) of nearesttrees (or shrubsif no trees in sample)in each quadrant. Mean height (m) of nearestshrubs(>1 m tall) in each quadrant. Mean diameter(cm) at widest crown of nearestshrubs (>1 m tall) in each quadrant. Mean distance(m) to nearestshrub(> Im tall). Mean numberof vegetationcontactsfallingagainst verticalrod in 9-m interval. Height at which a 20-cm-wideboardis >90% obscuredby vegetationat a distanceof 5 m (Wiens 1969). Proportionof shrubspecies in distancesample that are willows. Proportionof plant species in shrubdistancesample that are saplings(3 cm diameter)measuredwith oculartube (Jamesand Shugart1970). Percentcover of small shrubs(< Im tall) measured with oculartube (Jamesand Shugart1970). Percentcover of woody plants, saplings,and downed logs measuredwith oculartube (Jamesand Shugart 1970).

SAcronyms for variable transformations are defined as: SQRT = square root; RCPL = reciprocal; LN = natural log.

the nest-box tree. Pearson's product-moment correlations were used to assess relationships among habitat variables. Sixteen of the original variables were deleted from the analyses because they were either invariant or highly correlated with other variables. Within highly correlated variables, those having a sampling distribution most closely approaching normality were retained (18 variables, Table 1). Degree of normality was determined by examining normal probability plots (Afifi and Clark 1984). DATA ANALYSES Chi-square analysis was used to test whether the frequency of nest-box use and reproductive suc-

cess differed among years. Four levels of box use were identified: (1) nest boxes never used by wrens in all three study years (NO); (2) boxes used only by male wrens for stuffing twigs, i.e., dummy nests (M); (3) boxes selected by nesting pairs during one study year (1YR); and (4) boxes used by nesting pairs for 2 or 3 years (23YR). Reproductive outcomes were classified as: (1) boxes with nesting wrens that failed to fledge offspring all years of use (F); (2) boxes that successfully fledged at least one wren in one study year (S1); and (3) boxes that fledged wrens in 2-3 years (S23). Predation was assumed if nest material was disturbed, eggs were broken, nestlings were partially eaten, predator feces were found in the

NEST-BOXUSE AND NESTING OUTCOME

nest, eggsor nestlingswere on the ground,boxes were damaged or unlatched (probablyby raccoons, Procyonlotor),or nestswereemptybefore nestlingswere due to fledge.Nests were considered abandonedif nest contents failed to hatch or fledge,and adultswere no longeractive at the nest. Partiallosses of clutches and broods were common, resultingfrom egg puncturingby conspecifics, hatching failure, and nestling starvation. Nests in reduced broods were considered successfulif the nest remainedactive to the time of fledging. Multivariatetechniques were applied in this studybecausehabitatgradientsand relationships among multipletreatmentsand multiple habitat factorsare not readily detected or interpretable using univariatestatistics (Green 1979, Klecka 1980). To counteractthe criticism that multivariate tests are prone to producesignificantresults even when data contain no relationships (Rextadet al. 1988),I firstdeterminedthat,within a set of univariatetests, the number of variables that differedsignificantlywas greaterthan thatexpectedby chance(P < 0.05). Forexample, for a set of 18 habitat variables,the number of variables expected to differ by chance alone is 18 x 0.05 < 1.0. Individualhabitatfeatureswere comparedamong use groups using randomized complete block analysis of variance (ANOVA), with the three study plots servingas blocks and the box site representingthe experimentalunit (n = 65). This design improves the accuracyof the comparisonsbetween use classes by eliminatingthe variabilityamong plots (Montgomery 1984) that may arise due to differencesin wren or predatordensities, and habitats. I applied multivariate analysis of variance (MANOVA)to assess whetherhabitatcentroids differedamong box-use groups or nesting outcomes (Marascuiloand Levin 1983). Direct discriminant analysis (DA) with orthogonal rotation (VARIMAX solution, SPSS, Klecka 1975) was used to separateand classifynest-boxuse or fledgingsuccess along axes of habitat structure. To compensatefor unequalsample sizes among groups, I based prior probabilitiesfor classificationon proportionsof caseswithineachgroup. Habitat variables were selected for direct DAs based on significancelevels (P < 0.1) of the 18 transformedvariablesusedin theunivariateblock ANOVAs of box use. Additionalvariablesused in direct DAs were those identifiedby stepwise DA usingthe Mahalanobisdistancecriterionfor

851

maximizingseparationof groups(Klecka 1975). Thirteen habitat featureswere selected for the multivariateanalysesof box-usegroupsand nine variableswere selectedfor the analysesof reproductive outcomes. Based on univariatetests indicatinghabitatsimilaritybetweenunusedboxes and boxes with dummy nests, I created a new use class (NM) by addingthe sampleof dummynest boxes to the unused group;this consolidation improvedsamplesize. Sufficientsamplesize was defined using Klecka's (1980) criteria that each group n must be at least two times greater than the numberof variables. Box's modificationof Bartlett'stest was used to evaluatethe homogeneityof variance-covariance matrices (Williams 1983). Because Box's test is conservative(Green1979),I used an alpha level of 0.01 to determineif covariancematrices differed.Discriminantfunctionswere tested for statisticalsignificance(P < 0.05) by transforming Wilk'slambdainto a Chi-squarestatistic(Marascuilo and Levin 1983). Tukey's procedurefor hypothesis testing (Dunnett 1980) was used to determinethe significance(P < 0.05) of group separationof meandiscriminantscoresuponeach individual discriminantaxis. The originalcases were classifiedinto membershipgroupsto check the adequacyof the discriminantfunctions(Hand 1981, Williams 1983). Three sets of data using raw, log-transformed (LN), and a combinationof squareroot (SQRT), reciprocal(RCPL),and log-transformedvariates (Kleinbaumand Kupper 1978) were analyzed. Resultsusinga combinationof transformedvariates were reportedfor subsequentanalyses because they most closely adheredto statisticalassumptions of normality and equality of covariance matrices. For ease in interpreting variables,mean values of raw data were given. RESULTS House Wrensnested in 21 boxes (32%of 65) in 1983, 29 (45%)in 1984, and 23 (35%)in 1985 (Table 2). Male wrens deposited twigs into an additional 15 boxes (23%)in 1983, 13 (20%)in 1984, and 11 (17%) in 1985. Over the 3-year period,43%of all boxes werenot used by wrens, 20%had male dummynests, and 37%contained true wren nests. Occupancyratesof boxes on each plot did not differamong years for nesting wrens (x2 = 1.27, df = 4, P > 0.75), male wrens (x2 = 6.29, df =

852

DEBORAH M. FINCH

TABLE2. Frequency of nest-boxuse andnestingsuccess,andratesof box occupancy (%use by plot)and success(%neststhatfledged onenestling)of HouseWrenson threestudyplotsin 1983,1984,and1985. _ Yearand plot

Unusedboxes(%)

Male(%)'

HouseWrenuse Nest (%)

Success(%)b

1983 FooteCamp Treasure Island RockCreek 1984 FooteCamp Treasure Island RockCreek

9 (40.9) 14(66.7) 6 (27.3)

6 (27.3) 2 (9.5) 7 (31.8)

7 (31.8) 5 (23.8) 9 (40.9)

7 (100) 5 (100) 8 (88)

10(45.5) 7 (33.3) 6 (27.3)

3 (13.6) 5 (23.8) 5 (22.7)

9 (40.9) 9 (42.9) 11(50.0)

1 (11) 8 (89) 9 (82)

1985 Foote Camp TreasureIsland

7 (31.8) 11(52.4)

5 (22.7) 5 (23.8)

10 (45.5) 5 (23.8)

6 (60) 4 (80)

RockCreek Grandtotal

13(59.1) 62

1 (4.5) 39

8 (36.4) 73

5 (63) 53 (73)

use was countedwhena box was filledwith twigsbut no eggswerelaid. bSMale Numberof successesis the numberof usedboxes that fledgedat least one offspring.

4, P > 0.25), and nestingwrensand male wrens combined (x2 = 4.26, df = 4, P > 0.50) (Table 2). Box occupancywas dependent on whether the same boxes wereoccupiedthe precedingyear (1984: x2 = 14.81, df= 4, P= 0.005, and 1985: 2 = 16.06, df = 4, P = 0.003). House Wrens fledgedoffspringfrom 53 of 73 nests (73%)over the 3-year period. The rate of nesting success decreasedfrom a plot average(i + SD) of 96.0 ? 6.9% in 1983 to 60.7 ? 43.2% in 1984 and 67.7 ? 10.8%in 1985 (Welch's F = 6.6, P = 0.08). Boxes containingsuccessfulnests in 1983 weremore likelyto be used in 1984 by eitherthe same or differentbirds,whereasunusedor failed boxeswerenot likelyto be occupiedthe following year (x2 = 9.41, df = 2, P = 0.009). Box use in 1985 was similarly related to previous reproductive success(x2 = 10.98, df= 2, P = 0.004). UNIVARIATE TESTSOFBOXUSE

(Table 3). Because the number of features expected to differby chance alone was less than 1 at the alpha level of 0.05, I concludedthat patternsof box use weresignificantlyrelatedto habitat differencesand that the use of multivariate tests was appropriatein subsequentanalyses. Ingeneral,trendsdetectedusingunivariatetests weresimilarto those found in multivariateanalyses so univariate results are only briefly described.In sum, nestingwrens selectedboxes in one or more years in areas with smaller trees, fewerand largershrubs,fewersaplings,and more open (herbaceousratherthanwoody)groundand understorycover than areas containingunused boxes. ANALYSIS OFBOXUSE DISCRIMINANT

An overallMANOVA for 13 variablesand three box-use levels (NM, 1YR, 23YR) indicatedthat Thirteenboxes were never used by wrens;eight the habitatcentroidssignificantlydifferedamong boxesheldonlydummynestsbuiltby malewrens; use groups (Hotelling'strace = 1.54, F = 2.54, 21 boxes were occupied by nesting wrens in 1 P = 0.0007). Box's test indicated that the coyearonly;and 23 boxes had truewren nests dur- variance matrices computed using transformed ing 2 or 3 years.Boxes containingmale dummy variablesdid not differin the analysesof box use nests were locatedin habitatssimilarto those in (P = 0.056) andreproductivesuccess(P = 0.250). which boxes were never used (17 of 18 habitat Discriminant analysis of 13 habitat variables features, P > 0.05); therefore,in multivariate producedtwo significantdiscriminantfunctions analyses,thesetwo classeswerecombined(NM). that distinguishedamong three levels of box use In pairwisecomparisons,13 featuresdifferedbe- by House Wrens(Table4). The firstdiscriminant tween various combinationsof wren-useclasses functionaccountedfor 48%of the varianceand (NM vs. lYR, NM vs. 23YR, lYR vs. 23YR) was most highly correlatedwith SAP, SHDIS,

NEST-BOXUSE AND NESTING OUTCOME

853

TABLE3. Means?-CL of 18 habitatfeaturesmeasuredat nest boxes never used by House Wrensin all three study years(NO);used only by male wrens(M);used as a wren nest site in 1 year(1YR);or used as a wrennest site in 2-3 years (23YR). Total sample size = 65.a No use all years(NO)

Maleuse (M)

(n = 13)

(n = 8)

Habitat feature

x

CL

2

DBH (cm) CANHT (m) SHHT (m) SHCD (cm) SHDIS (m) VFD1 (no. hits) VFD2 (no. hits) VFD3 (no. hits) VFD4 (no. hits) VFD5 (no. hits) EVH (m) WILL(%) SAP (%) BARE(%) HERB (%) LOGS(%) SHRUB (%) COVER(%)

27.6 11.4 1.9 120.3 2.9 1.5 0.8 1.0 3.3 0.8 1.0 9.6 4.8 28.8 44.2 3.8 18.2 26.8

7.2 3.2 0.5 38.4 1.6 0.6 0.7 0.6 1.0 0.6 0.5 9.8 5.0 15.6 18.0 3.7 13.5 13.9

34.3 13.1 1.9 107.2 4.0 2.2 0.3 0.7 3.8 0.7 0.6 15.6 6.3 21.9 56.4 6.3 9.3 21.8

Box-useclasses Nest 1 year(lYR) (n = 21)

CL

x

14.6 8.1 0.4 54.5 5.6 0.6 0.3 0.4 1.4 0.7 0.5 19.2 7.9 16.8 19.7 7.9 14.5 16.7

23.7 9.7 2.4 149.7 5.8 1.8 0.4 1.0 5.1 1.3 0.4 3.6 1.8 26.7 61.6 5.8 4.2 11.8

Nest 2-3 years(23YR) (n = 23)

CL

x

5.2 2.7 0.7 28.6 1.9 0.5 0.2 0.6 1.4 0.9 0.3 4.1 3.7 12.0 10.8 4.2 3.3 6.1

24.4 10.3 2.1 148.7 5.3 2.5 0.5 1.1 3.5 0.5 0.4 10.9 0.5 23.3 64.8 2.8 8.6 11.9

CL

Comparisonsb

4.6 3.8 0.4 23.2 1.8 0.7 0.3 0.4 0.8 0.4 0.2 8.5 1.1 10.9 11.2 2.3 5.3 5.3

C C NS D CD ABD NS NS BC BD D B CD NS CD NS C CD

a See Table 1 for descriptions of habitat acronyms. Data were backtransformed for ease of interpretation. b Habitat features were statistically compared among use classes using randomized block ANOVA. The classes NO and M differed in only one feature so a new class NM was formed. Habitat features that differed significantly (P < 0.1) between classes are indicated by: A = NO vs. M, B = 1YR vs. 23YR, C = NM vs. 1YR, and D = NM vs. 23YR. NS = not significant.

SHCD, and HERB (Fig. 1). This function, which described a situation where percentage saplings decreased as shrub dispersion, shrub size, and herbaceous cover increased, distinguished unused boxes from those occupied by nesting wrens. Wrens avoided boxes positioned in habitats with dense sapling-small shrub understories but frequently nested in boxes at sites with open, herbaceous ground cover and dispersed large shrubs. Pairwise comparisons using Tukey's test indicated that the means of discriminant scores of used and unused box sites were significantly separated on the first discriminant function (Table

DISCRIMINANT ANALYSIS OF NESTINGSUCCESS An overallMANOVA for nine habitatvariables and three nesting outcome levels indicatedthat the habitatcentroidswere significantlydifferent among outcomes (Hotelling'strace = 1.55, F = 2.33, P = 0.009) in House Wrens.Two significant discriminantfunctionsdistinguishedamongthree outcome levels in DAs of nine variables(Table 4). The first function accountedfor 44% of the total variance and was most highly correlated with VFD5 and VFDI (Fig. 2). This function

5). The second discriminant function explained 34% of the variance among groups and represented a gradient of decreasing foliage density at the ground level with increasing foliage density in the overstory. Means of boxes used 2-3 years by wrens were significantly separated on the second function from means of boxes used 1 year as well as from means of boxes that failed (Table 5). Apparently, boxes placed in locations with denser overstory cover and thinner ground foliage were more often used as nest sites for 1 year rather than for 2 or 3 years. DA correctly classified 78.5% of all boxes.

TABLE4. Two discriminant functionanalysesdisandreprotinguishing amongnest-boxusecategories ductiveoutcomesin HouseWrens.a

Characteristic

% variance x2 df P

Box use Function Function 1 2

Nesting outcome Function Function 1 2

48.133 60.123 26 0.000

44.259 37.370 18 0.005

34.087 23.353 12 0.025

34.640 15.742 8 0.046

Use categories in House Wrens are (1) box empty or used by male wrens in all study years, (2) box used by females in 1 year, and (3) box used by females in 2-3 years. Outcome groups are (1) ailed in all years, (2) fledged at least one young in 1 year, and (3) fledged at least one young/ year in 2-3 years.

854

DEBORAHM. FINCH

4

Unused

IL

I-.-0 ..o.

I L o

1-year

A

> r ?c•o AO CCt ,

2 *

.

use A

A

2-3-year use

* * ?-- La -

*



A

A•A•

oo o o

o

....J

.. J-"

I*

I 4

IL.-,* *

2.4

0*

u

-?urn

o00

-4

-2

.

.-

0

DF1

2

4 Low High High High

Percent saplings Shrub dispersion Shrub crown diameter Percent herbaceous cover

High Low Low Low

FIGURE 1. Discriminantscores on two axes derivedfrom discriminantanalysisof box-usegroupsin House Wrens.Mean vectors for each groupare labeledas 0 = unused, 1 = 1-yearuse, 2 = 2- to 3-yearuse. Habitat variables,used to interpretgrouppositions, are describedin Table 1. significantly separated boxes with nesting attempts that always failed from those with successful outcomes one or more times (Table 5). Compared with successful boxes, failures were

associatedwith denseroverstoryfoliage, sparser herbaceousfoliagein the surfacelayer,and higher cover of logs >3 cm diameter. The second discriminantfunction accounted

TABLE5. Groupmeans and pairwisecomparisonsof discriminantfunctions,priorprobabilities,and classificationratesfrom two sets of discriminantanalysesthat separatedpatternsof nest-boxuse or fledgingsuccess in House Wrens.

Group

Box use Unused all years Nest 1 year Nest 2-3 years Nesting outcome Failureall years Success 1 year Success2-3 years

n

Prior probability Tukey'

Function1 prob

Function2 SD

SD

Tukey"

correctly classifiedb

21 21 23

0.32 0.32 0.35

A B B

-1.39 0.87 0.38

1.21 0.70 0.88

A B C

0.03 0.93 -0.87

1.10 0.90 0.99

76.2 85.7 73.9

11 19 14

0.25 0.43 0.32

A B C

1.14 0.15 -1.12

1.04 1.05 0.89

A B A

0.75 -0.80 0.50

0.68 1.26 0.80

81.8 68.4 78.6

SPairwisecomparisonsof mean discriminantscoresof use and outcomeclasseswereassessedusingTukey'smultiplecomparisontest (Dunnett 1980). Significant(P < 0.05) differencesbetweenmeansare indicatedby differentletters(e.g.,A vs. B vs. C). b Percentage of cases that were correctly classified by the discriminant analyses.

NEST-BOX USEAND NESTINGOUTCOME 855

4 .tw

r-t CC

Failure

E

1-year success

2 -0 S_

c

o

o

>

ccess

0

0

* 2

0*

--o drr

Ho

1u uor -'-'f;; c,

o

2-3-year su

,*

e

??h

Low

A

u Y-42:

>f

T....

*

tc

0

K

'-

2

4

OAF

-4

-

Low-High--Low

-

-

-2 0 2 High-overstory foliage density-Foliage density grass-forb layer--Percent cover of logs

4 High Low High

of discriminant FIGURE2. Distribution scoreson twofunctions,basedontheanalysisof nestingoutcomes in boxesselectedby HouseWrens.Meanvectorsforeachgrouparelabeledas 0 = failure,1 = 1-yearsuccess, and2 = 2- to 3-yearsuccess.Habitatvariables,usedto interpret grouppositions,aredescribedin Table1. for 35%of the total variancein the House Wren DA and significantlydistinguishedbetweenboxes with 1 year successes and boxes with either failed attempts or 2-3 year successes (Tables 4 and 5). This resultsuggestedthat successfuloutcomes in reused boxes were more likely to be repeatedin habitatswheregroundsurfaceswere bare,shrubslarge,and loweroverstories(VFD4) sparse (Fig. 2). When both functions are taken into account, wrens were most reproductively successfulif they selectedboxes in habitatswith open canopies and open surfaceswith few decaying logs or dwarf shrubs.The probabilityof failurewas greatestin habitatswith thick canopy foliage and heavy nonherbaceousgroundcover. In this DA, a total of 75% of 44 outcomes was correctlyclassified(Table 5).

DISCUSSION NEST-BOX BYHOUSEWRENS SELECTION Boxes used by House Wrensfor nestingdiffered from unused boxes in that they were located in open habitats. In experimental woodlands of Pennsylvaniaand naturalwoodlandsof the Midwest and centralRockyMountains,House Wrens were reportedto be abundantin areas with low shrubcover and foliage density (DeGraaf 1987, Sedgwick and Knopf 1987), selecting natural cavities in smaller trees and snags than those used by largercavity-nestingspecies(Staufferand Best 1982, Gutzwillerand Anderson 1987). In othernest-boxstudies,House Wrensshowedthe same tendency to select boxes in sparservegetation(Belles-Islesand Picman 1986a)and, when

856

M. FINCH DEBORAH

given a choice between boxes in logged or unloggedstands,selectedmoreboxesin loggedareas with greaterherbaceouscover and fewer shrubs and saplings(Drillingand Thompson 1984). In areaswith moreopen fieldand pasture,however, wrens chose boxes closer to trees, saplings,and shrubs(Willneret al. 1983, Munroand Rounds 1985). Thus, over a gradientof habitatsranging from open fields to dense, deciduous forests, HouseWrensapparentlyselectintermediatesites. House Wrensforageon the groundand in the grass-forb-shrublayer (DeGraaf et al. 1985, Sedgwickand Knopf 1987)and may be attracted to nesting habitats with accessible surfacesfor foraging, particularlyif litter and herbaceous vegetationhave moreinsectspreferredby wrens. Increasedforagingby House Wrensin loggedand burnedareas (Franzreb1977) suggestsselection for open foraging substrate with herbaceous groundcover. Boxes placedin sparsevegetation receive greater solar radiation (McComb and Noble 1981), so microclimatemay play an additional role in nest-box choice, particularlyin the unpredictable,high altitude climate of Carbon County,Wyoming.Alternatively,box selection may be frequencydependent(Brawn1988). For instance,if boxes in open habitatsare more easily found, then such highlyvisible boxes may be selected more often by wrens. Male wrensconstructeddummy nests in more varied sites than those ultimatelyselectedby females. Boxes containing unused dummy nests were located in habitats similar to those where boxes remainedempty all 3 years. Indeed, habitat structuresampledat centersof male-defended territorieswas similar to that at randomly sampled sites (Finch 1989), suggestingthat territoryselectionby male wrensis nonspecialized. Male House Wrens are aggressivelyterritorial (Kendeigh1941)and could easily defendseveral boxes given the box-spacingpatternin this study (pers. observ.). Nonspecializeduse of boxes by multiple-nestmales may be adaptive under the following circumstances.By building nests in multiple boxes within its territory,a male may improve its chances of attractinga mate (Kendeigh 1941), even if some boxes are positioned in low-qualitysites. By matingwith a multiplenest male, a female House Wrencan readilyrenest in a surplusnest hole if the firstattemptfails because of nest usurpationor predation. High densities of dummy nests may also protect the actual nest site from search-strategypredators,

as was demonstratedfor the Marsh Wren, Cistothoruspalustris(Leonardand Picman 1987). IN HOUSEWRENS NESTINGSUCCESS Factors affectingnesting success, such as food supply, microclimate,and predation,can be influencedby habitatfeaturesat the nest site (Walsberg 1981, Westmorelandand Best 1985, Martin 1987). In this study, I focusedon habitatfactors that may limit predationrisk because nest predation was the major source of nesting failure and failurewas associatedwith habitat characteristics.Previousresearchon open-nestingbirds has shown that increasedfoliagedensitymay reduceriskof nest discoveryby concealingthe nest (Nolan 1978, Murphy 1983, Westmorelandand Best 1985), impeding predatortravel or transmission of cues (Bowmanand Harris 1980), or increasingthe number of possible nest sites a predatormust inspect(Martinand Roper 1988). Reproductivesuccessof House Wrensin Wyoming was associatedwith nest sites surrounded by sparse vegetation and little downed wood, ratherthan dense foliage and groundcover. Although few studies of cavity-nestingbirds have considered habitat influences on reproductive outcome(see Nilsson 1984),Belles-Islesand Picman (1986a) reportedthat productivityof boxnestingwrensin Ontario,Canada,was greaterin open habitats because nest losses to predators were minimized. Over the 3-yearperiod of this study, 18 of 20 unsuccessfulnests were destroyedby predators, and two were desertedafterpartialclutchesdisappeared.At least four of the depredatednests showed signs of destructionby House Wrens. Evidenceof conspecificintrusionincludedpuncturedeggs,nest materialpulledthroughentrance holes (see Belles-Islesand Picman 1986b), and observationsof box entryby morethantwo wrens (all box usersviewed simultaneouslyin vicinity). Conspecificsarereportedlymajornest destroyers in House Wrens, pecking eggs and nestlings to death to confiscatenest sites or mates, or to enhance sexual receptivity in females, allowing forced extra-paircopulations by males (BellesIsles and Picman 1986a; Freed 1986a, 1986b; Quinnand Holroyd1989).Predatorsin this study were undoubtedlysmall (limited by the 3.8-cm hole entrance)and possibly accustomed to enteringdarkcavities becauseof their own nesting habits. Small cavity-nestpredatorstypicallyincludewoodpeckersand mice (Nilsson 1984),sec-

USEANDNESTINGOUTCOME 857 NEST-BOX ondarycavity-nestingbirds (Freed 1987, Butler and Campbell 1987), weasels (Dunn 1977), and snakes (Nolan 1959). In riparian habitats of northcentral Wyoming, bullsnakes (Pituophis melanoleucussayi) andlong-tailedweasels(Mustelafrenata)were detectedenteringand destroying House Wren nests (L. Scott Johnson and HenryKermott,pers.comm.). Both of these vertebratespecieswere observedin my study areas. Becausethe contents of some wren nests in this studywereremovedwithoutdisturbanceto nesting material,the tree-climbingbullsnakeis a suspected predator.Largepredatorsinclude mammals able to open a box (e.g., unlatchedboxes and tracksin the vicinity implicate raccoonsin this study) or poke a sweepingpaw throughan entrance (e.g., red squirrel, Tamiasciurushudsonicus, filmed by the author with time-lapse camera). The restlessactivityand vociferouschatterand song of House Wrens may attractpredatorsto nest sites (see also Skutch 1949, Willis 1973). Cavity nests may be easily approachedby predatorsandconspecificsif densefoliageanddowned logs conceal their approachfrom nesting adults (Belles-Islesand Picman 1986a). For the nonsedentaryHouse Wren,the abilityto detectpredators first may be an important aspect of nest if predatorvisibilityis greater defense,particularly around nest sites with little obstructivefoliage. When a predatoror conspecific is sighted, the nestingadultscan modify theirbehavioraccordingly, by minimizing activity at the nest, distracting the animal's attention away from the cavity, or by attacking the intruder. Direct aggressionis usuallyeffectiveif the intruderis a conspecific(Kendeigh 1941, Grove 1981). High visibility may actually be advantageousin territorialdefenseof nest sites againstconspecifics. In addition, vegetation immediatelyadjacent to the nest site may facilitatepredationby supplying a bridgeto the nest. Bullsnakeswere observedenteringthreeHouse Wrennestsin northcentral Wyoming by descending from foliage directlyabove each nest box (L. Scott Johnson, pers.comm.). House Wrensmay avoid nest sites where foliage is profuse if close foliage permits easier nest access by climbing snakes. Increasesin both nest-boxuse and nestingsuccess of wrenswere associatedwith open surfaces and open canopies. Thus, boxes that were occupied for multiple years were in habitats associated with greater reproductive success, sug-

gestingthat boxes in open habitatswere actively selectedbased on theirhigherprobabilityof success. My resultsshowedthatbox use in 1984 and 1985 was relatedto box occupancyand nesting outcome from the precedingyear. In a study of marked birds in central Illinois, Drilling and Thompson (1988) demonstrated that female House Wrensthat returnedto breedingsites had producedmore offspringin the previous breeding season than had nonreturningfemales. If nesting outcome the year before influencedrate of box reuseby returningbirdsin my study,then the association between multiple-yearuse and sparsevegetationmay be an incidentalresult of the underlyingassociationbetween nesting outcome and habitat.Hence, placementof nests by House Wrens and other bird species may ultimately depend on age and prior nesting experience (Sonerud 1985, Marzluff1988). ACKNOWLEDGMENTS G. J. Sherman, K. A. Conine, I thankP. F. Gutzwiller, C.L.Canaday, andR.D. Greerforassistance in checkhabitatcharacteristics, ing nestboxesandmeasuring andA. L.Wardforhisencouragement duringthestudy. to M. G. Raphaelforadviceon statistical I amgrateful design,andJ. D. Brawn,F. L.Knopf,K.J. Gutzwiller, T. E. Martin,B. A. Maurer, J. Picman,R. C. Rounds, J. A. Sedgwick,and R. H. Yahnerfor reviewof the manuscript. LITERATURECITED A. A., ANDV. CLARK.1984. Computer-aided AFnII, multivariateanalysis.LifetimeLearningPublications,Belmont,CA. C.E.H. 1984. Polygynyand nest site seASKENMO,

lectionin the PiedFlycatcher. Anim.Behav.32: 972-980. BALEN, J. H. VAN,C.J.H.BooY, J. A. FRANEKER,AND E. R. OSIECK. 1982. Studieson hole nestingbirds

innatural nestsites.I. Availability andoccupation of naturalnestsites.Ardea70:1-24. BELLES-ISLES,J.

C., AND J.

PICMAN.

1986a. Nesting

lossesandnestsite preferences in HouseWrens. Condor88:483-486.

BELLES-ISLES,J. C., AND

J. PICMAN. 1986b. House

Wrennest-destroying behavior.Condor88:190-

193. J. C., ANDJ. PICMAN.1987. Suspected BELLES-ISLES, adultintraspecifickillingby House Wrens.Wilson

Bull.99:497-498.

G. B., ANDL. D. HARRIS.1980. Effectsof BOWMAN, spatialheterogeneityon ground-nestdepredation. J. Wildl. Manage.44:806-813. J. D. 1988. Selectivityand ecologicalconBRAWN, sequences of cavity-nestersusing naturalvs. artificialnest sites. Auk 105:789-791. BUTLER, R. W., AND C. A. CAMPBELL.1987.

Nest

858

DEBORAHM. FINCH

appropriationand interspecificfeeding between HUSSELL, D.J.T., AND T. E. QUINNEY. 1987. Food abundance and clutch size of Tree Swallows Tree Swallows, Tachycinetabicolor, and Barn Swallows,Hirundorustica.Can. Field-Nat. 101: Tachycinetabicolor.Ibis 129:243-258. 433-434. D. W. 1976. Nonrandomorientationof enINOUYE, D. F. 1977. Breeding success and nest tranceholes to woodpeckernests in aspen trees. CACCAMISE, Condor78:101-102. site characteristicsof the Red-wingedBlackbird. Wilson Bull. 89:396-408. 1970. A quantiJAMES, F. C., AND H. H. SHUGART. tative method of habitat description. Audubon CODY,M. L. 1981. Habitat selection in birds: the Field-Notes 24:727-736. rolesof vegetationstructure,competitors,andproductivity.Bioscience31:107-113. KENDEIGH, S. C. 1941. Territorial and mating beR. M. 1987. Breeding birds and gypsy moth havior of the House Wren.Ill. Biol. Monogr. 18: DEGRAAF, 1-20. defoliation:Short-termresponsesof species and guilds. Wildl. Soc. Bull. 15:217-221. KLECKA, W. R. 1975. Discriminant analysis, p. 434467. In N. H. Nie, C. H. Hull, J. G. Jenkins,K. DEGRAAF,R. M., N. G. TILGHMAN,AND S. H. ANDERSON.1985. Foraging guilds of North Steinbrenner,and D. H. Bent [eds.],SPSS,StatisAmericanbirds. Environ.Manage.9:493-536. tical package for the social sciences. 2nd ed. N.E., ANDC. F. THOMPSON.1984. The use DRILLING, McGraw-Hill,New York. of nestboxesto assessthe effectof selectivelogging KLECKA, W. R. 1980. Discriminantanalysis. Sage on House Wren populations,p. 188-196. In W. Publications,BeverlyHills, CA. C. McComb [ed.], Proceedingsof the workshop KLEINBAUM,D. G., AND L. L. KUPPER. 1978. Applied onmanagement speciesandecological ofnongame regressionanalysisand other multivariatemethods. DuxburyPress, North Scituate,MA. communities.Univ. of Kentucky,Lexington. DRILLING,

N. E., AND C. F. THOMPSON. 1988.

Natal

and breedingdispersalin House Wrens (Troglodytes aedon).Auk 105:480-491. DUNN, E. 1977. Predationby weasels (Mustela nivalis) on breedingtits (Parusspp.) in relationto the densityof rodents.J. Anim. Ecol.46:633-657. C. W. 1980. Pairwisemultiple compariDUNNETT, sons in the homogeneousvariance,unequalsample size case. J. Am. Stat. Assoc. 75:789-795. D. M. 1989. Habitatuse and habitatoverlap FINCH, of riparianbirds in three elevationalzones. Ecology 70:866-880. K. E. 1977. Bird population changes after FRANZREB,

timberharvestingof a mixed coniferforestin Arizona. USDA For. Ser. Res. PaperRM-184. Fort Collins, CO. FREED, L. A.

1986a. Usurpatory and opportunistic

bigamy in tropical House Wrens. Anim. Behav. 34:1894-1896. L. A. 1986b. Territorytakeoverand sexually FREED, selectedinfanticidein tropicalHouse Wrens.Behav. Ecol. Sociobiol. 19:197-206. FREED,L. A.

1987.

Rufous-and-white Wrens kill

LEONARD,M. L., AND J. PICMAN. 1987. The adaptive

significanceof multiple nest building by male MarshWrens.Anim. Behav. 35:271-277. H. G. 1986. Choice of nest boxes by Tree LUMSDEN,

Swallows, Tachycineta bicolor, House Wrens, Troglodytesaedon,EasternBluebirds,Sialiasialis, and EuropeanStarlings,Sturnus vulgaris.Can. Field-Nat. 100:343-349. MARASCUILO, L. A., AND J. R. LEVIN.

MARTIN, T. E.

1987. Food as a limit on breeding

birds:a life-historyperspective.Annu. Rev. Ecol. Syst. 18:453-487. MARTIN, T. E., AND J. J. ROPER. 1988. Nest predation and nest-site selectionof a westernpopulationof the HermitThrush.Condor90:51-57. MARZLUFF, J. M.

1988. Do pinyon jays learn from

priorexperiencewhereto placetheirnests?Anim. Behav. 36:1-10. MCCOMB, W. C., AND R. E. NOBLE. 1981. Microclimate of nest boxes and naturalcavitiesin bottomland hardwoods.J. Wildl. Manage.45:284-289.

House Wren nestlings during a food shortage. Condor89:195-197. R. H. 1979. Samplingdesign and statistical GREEN, methodsforenvironmentalbiologists.JohnWiley and Sons, New York.

MONTEVECCHI,W. A.

GROVE,P. E. 1981. The effect of location and stage

MULDAL,

discriminationin of nestingon neighbor/stranger the House Wren. Ph.D.diss. City Univ. of New York.

1983. Multi-

variate statistics in the social sciences. Brooks/ Cole PublishingCompany,Monterey,CA.

1978. Nest site selection and

its survivalvalue among LaughingGulls. Behav. Ecol. Sociobiol. 4:143-161. MONTGOMERY,D. C.

1984. Design and analysis of

experiments.John Wiley and Sons, New York. A., H. L. GIBBS,

AND

R. J. ROBERTSON.

1985.

Preferrednestspacingof an obligatecavity-nesting bird, the Tree Swallow.Condor87:356-363. MUNRO,H. L., ANDR. C. ROUNDS. 1985. Selection

of artificialnest sites by five sympatricpasserines. J. Wildl. Manage.49:264-276. M. T. 1983. Nest successandnestinghabits MURPHY, of EasternKingbirdsand other flycatchers.Condor 85:208-219. 1987. MulGUTZWILLER, K. J., AND S. H. ANDERSON. tiscale associationsbetween cavity-nestingbirds NILSSON,S. G. 1984. The evolution of nest-site selection amonghole-nestingbirds:The importance and featuresof Wyomingstreamsidewoodlands. Condor89:534-548. of nest predationand competition. Ornis Scand. 15:167-175. HAND,D. J. 1981. Discriminationand classification. John Wiley and Sons, New York. NOLAN, V. 1959. Pileated Woodpecker attacks pilot GUSTAFSSON, L. 1988. Inter- and intraspecific com-

petition for nest holes in a populationof the Collared FlycatcherFicedulaalbicollis.Ibis 130:1116.

NEST-BOXUSE AND NESTING OUTCOME black snake at tree cavity. Wilson Bull. 71:381382. NOLAN,V. 1978. The ecology and behavior of the Prairie Warbler Dendroica discolor. Ornithol. Monogr.No. 26. AmericanOrnithologists'Union, Washington,DC. 1988. Evidence PICMAN, J., ANDJ. C. BELLES-ISLES.

for intraspecificbrood parasitism in the House Wren.Condor90:513-514. PrNKowSKI, B.C. 1976. Use of treecavitiesby nesting EasternBluebirds.J. Wildl. Manage.40:556-563. QuINN,M. S., ANDG. L. HOLROYD.1989. Nestling and egg destructionby House Wrens.Condor91: 206-207. E. A., D. D. MILLER, E. M. C. H. FLATHER, REXTAD, ANDERSON,J. W. HuPP, D. ANDERSON. 1988.

859

dation involving long term memory of the predator. J. Anim. Ecol. 54:179-192. D. E., AND L. B. BEST. 1982. Nest-site STAUFFER,

selection by cavity-nestingbirds of riparianhabitats in Iowa. Wilson Bull. 94:329-337. WALSBERG, G. E. 1981. Nest-site selection and the radiative environment of the Warbling Vireo. Condor83:86-88. WESTMORELAND, D., ANDL. B. BEST. 1985. The effect

of disturbanceon MourningDove nestingsuccess. Auk 102:774-780. WIENS,J. A. 1969. An approachto the study of ecological relationshipsamong grasslandbirds. Ornithol. Monogr.No. 8. AmericanOrnithologists' Union, Washington,DC. B. K. 1983. Some observationson the use WILLIAMS, multivariate statisticalinferencein of discriminantanalysis in ecology. Ecology 64: Questionable 1283-1291. wildlifehabitatandcommunitystudies.J. Wildl. WLLIS, E. O. 1973. Survival rates for visited and Manage.52:794-798. unvisited nests of Bicolored Antbirds. Auk 90: 1987. Breeding SEDGWICK,J. A., AND F. L. KNOPF. 263-267. birdresponseto cattlegrazingof a cottonwood

bottomland. J. Wildl.Manage.51:230-237. A. F. 1949. Do tropicalbirdsrearas many SKUTCH,

WILLNER,

G. R., J. E. GATES, AND W. J. DEVLIN. 1983.

Nest box use by cavity-nestingbirds. Am. Midl.

Nat. 109:194-201. young as they can nourish?Ibis 91:430-455. R. H. 1983/1984.Avianuseof nestboxes T. 1986. Nest site settlementby the Pied YAHNER, SLAGSVOLD, inMinnesota shelterbelts. farmstead J.Minn.Acad. Flycatcher:Does the female choose her mate for Sci.49:18-20. qualityof his house or himself?Ornis Scand. 17: 210-220. K.R. MAYOH. 1982. Breeding R.,AND ZACH, biology of TreeSwallowsandHouseWrensin a gradient SONERUD, G. A. 1985. Nestholeshiftin Tengmalm's as defenseagainstnestpreOwlAegoliusfunereus

of gammaradiation.Ecology63:1720-1728.