The Auk 114(4):708-716, 1997

EASTERN NUMBERS INCREASE FOLLOWING DEFOLIATION BY GYPSY MOTHS

JENNIFERL. BELL• AND ROBERTC. WHITMORE WestVirginia University, Division of Forestry,P.O. Box 6125, Morgantown, West Virginia 26506, USA

ABSTRACT.--Birdpopulations and habitat were monitored each year before (1984 to 1986), during(1987 to 1988),and after (1989to 1996)a majorgypsy moth (Lymantria dispar) out- breakin the easternpanhandle of WestVirginia. Extensive tree mortality caused by repeated defoliationsby gypsymoths resulted in the releaseof understoryvegetation. Densities of EasternTowhees ( erythrophthalmus) were significantlyhigher in the periodfollowing the gypsymoth outbreak. Both before and after the outbreak, Eastern Towhee densities were higherin areasof the forestwith lessoverstory (particularly high canopy)and lowerden- sitiesof live trees(particularly small trees). This indicates that not all formsof early suc- cessionalhabitat, specifically areas with a high densityof small-diametertrees, are suitable for speciesdeemed as "early successional." Although the gypsymoth outbreak resulted in an increasein the numberof saplings,it alsoopened up the canopyand createda dense layerof shrubsin manyareas. Because Eastern forage and nest on the groundand in shrubs,the outbreakincreased the amountof suitablehabitat for this species.Given that densitiesof EasternTowhee are decliningin the state,it is usefulto documenthabitat fea- turesthat areimportant for sustainingtowhee populations. Received 13 January 1997, accepted 13 May 1997.

SINCE ITS INTRODUCTIONaround 1869, the 1987,Hix et al. 1991,Twery 1991).Thus, tree gypsymoth (Lymantriadispar) has become one mortality causedby repeated defoliationsby of the most economicallyand ecologicallyim- gypsymoths can create suitable habitat for ear- portantforest pests in the UnitedStates. It has ly successionalspecies, such as the EasternTo- infested approximately25% of the nation's whee(Pipilo erythrophthalmus). TheEastern To- hardwoodforests, and the rangeof the gypsy wheeis characterizedas a forestgeneralist typ- motheventually is expectedto includemost of icallyfound in earlysuccessional habitat, edge theeastern (Gottschalk 1991). De- or openareas, or in forestswith a well-devel- spite the presenceof gypsy moths in North oped understory(James 1971, Maurer et al. America for more than 100 years, few studies 1980, Hagan 1993). Our objectivewas to de- haveexamined their effectson populations scribethe types of habitatsEastern Towbees and habitats(e.g. Cooper et al. 1987,DeGraaf were associated with both before and after a 1987,Thurber et al. 1994).Repeated defoliation majoroutbreak of gypsymoths. We wantedto by gypsymoths can result in widespreadmor- determine if more suitable habitat was created tality of trees,particularly oaks (Quercus spp.), for towheesas a resultof gypsymoth defolia- whichare a preferredfood item of the larvae. tion,and to assesshow vegetationchanges may Thesefactors can result in majorchanges in for- affect towheepopulations in our study area. est structureand speciescomposition. Thesefindings are of particularinterest in light There has been concernthat speciesof ver- of recentwork showingthat numbersof East- tebratesdependent on early successionalhabi- ern Towheeare decliningin easternNorth tats have been declining,particularly in the America(Hagan 1993). Determining the habi- (Hagan 1993, Litvaitis tat associationsof EasternTowbees, particular- 1993,Yahner 1995), and that thesedeclines may ly wherethey are foundin high numbers,may be linkedto forestmaturation. Gypsy moths set helpto shedlight on the causesof theirdecline back successionby decreasingthe numberof in other areas. large overstory trees and increasing the amountof understoryvegetation and the num- STUDY AREA AND METHODS ber of smallertrees via regeneration(Quimby Thisstudy was conducted within theSleepy Creek • [email protected] PublicHunting and FishingArea (hereafterSleepy

7O8 October1997] Towheesand Gypsy Moth Defoliation 709

TABLE1. Averagedefoliation values (2 ñ SE) from mothnumbers were increasing (Cooper et al. 1987). 1987to 1996for all trees>15.2 cm dbh.Averages Two consecutiveyears of heavydefoliation occurred basedon defoliationestimates of 0 to 3 (0 = no de- in 1987 and 1988 as gypsy moth numbersreached foliation, 1 = 1 to 33% defoliation, 2 = 34 to 67%, 3 = >67%). outbreaklevels (Cooperet al. 1993;Table 1). Defoli- ation returned to low levels until another outbreak occurred in 1992, and then were low from 1992 Average Year defoliation through1996 (Table 1). Six transects,each approximately 2 km in length, 1987 1.0 +_ 0.1 1988 1.2 +_ 0.1 wereestablished on SleepyCreek Mountain in 1984. 1989 0.3 _+ 0.0 (Fig. 1). Two transectswere placedalong each ele- 1990 0.7 ñ 0.1 vationalgradient (ridgetop, midslope, and valley)to 1991 0.4 ñ 0.0 represent a uniform sample of locationson the 1992 1.0 +- 0.1 mountain. In 1985, a seventh transect was added 1993 0.7 ñ 0.1 alongthe midslopeof Third Hill Mountain,which 1994 0.0 +_ 0.0 lies parallelto SleepyCreek Mountain (Fig. 1). Six 1995 0.0 ñ 0.0 circular(125-m radius) plots were placed along each 1996 0.0 +_ 0.0 transect,for a total of 42 plots. The centersof these plotswere used as bird samplingpoints and wereat least 325 m from the nearestadjacent plot center. Creek),an 8,000-haarea located on Sleepy Creek and This placementmaximized independence among Third Hill mountains(elevation 240 to 662 m) in plotswhile minimizinggeographic variation in veg- Morgan and Berkeleycounties, West Virginia. This etation structure and composition. Five 0.04-ha arealies within the Ridgeand Valleyprovince and is (1/10 acre)subplots were placedwithin eachplot, typical of the oak-hickoryforests of the region. for a total of 210 subplots.One subplotwas always SleepyCreek is maintainedby the WestVirginia Di- locatedat the plot center,and onewas randomly lo- vision of Natural Resources, and no efforts were catedwithin eachof the four quadrantsof the 125-m taken to controlgypsy mothsduring the courseof plot (seeFig. 1). Thesesubplots were usedas vege- our study.Bird populationsand habitatwere moni- tationsampling plots to be associatedwith the bird tored before(1984 to 1986),during (1987to 1988), countsdone at the plot centers.Values for eachveg- and after (1989to 1996) a severegypsy moth out- etation variableat eachplot were basedon the mean break.Some of the studyplots on the low valleyand of the five subplots.A tree in the centerof eachplot midslopeareas received defoliation in 1986as gypsy and subplotwas permanentlymarked with paint

Ftc. 1. Plot layouton the ridgetop,midslope, and low valleyof SleepyCreek and Third Hill mountains. Dotsare 125-m radius plots, each of whichcontained five 0.04-ha vegetation sampling subplots (inset). 710 BELLAND WHITMORE [Auk, Vol. 114 and aluminumtags to facilitaterelocation. The same rangedfrom 0 to 100% at eachsubplot. If foliage plotsand subplotswere maintainedthroughout the from any woody-stemmedplant less than 3 m in duration of the study. heightwas interceptedby the crosshairs,it was re- Bird countswere conductedusing the variablecir- corded by speciesor speciesgroup and countedas cularplot technique(Reynolds et al. 1980).Eachtran- shrubcover. Foliage greater than 3 m in heightwas sect (and thus each plot) was surveyedfive times classifiedas canopy cover.Presence (as opposed to overthe course of thebreeding season, from late May absence)of canopycover was recordedas total can- to mid-Julyeach year Approximatelyfive transects opy cover.If the highestlayer of canopypresent was were surveyedeach week over a period of roughly between 3 and 12 m, it was recorded as maximum sevenweeks. All transectswere surveyedonce be- low canopycover. If the highestlayer of canopypres- fore any transect was sampled a second time, all ent was greaterthan 12 m, it was recordedas maxi- transectswere surveyed a second time before any mum high canopycover. Observers were trainedto transectwas sampleda third time,and soon until all estimatecanopy heights with theuse of a clinometer, transectswere surveyedfive times. The order the and premeasuredreference trees were recorded at plotswere surveyedwithin a transectand the order each plot. Cover and canopy variableswere mea- of the transectswithin eachsampling bout were de- suredeach year during late May and Juneto assess cidedby randomdrawing. Observers were switched the effectsof defoliationprior to refoliation,which amongtransects to avoidan observer/ transect bias, sometimesoccurs after heavy (i.e. >60%) defoliation. i.e. having the same observersampling the same The subplotswere revisited and thenumber, and di- transect each time. ameterat breastheight (dbh) of eachspecies of tree Bird countsbegan just after sunriseand were com- (any woody plant >3 m in height) within the sub- pleted by approximately0830 EDT. To reduceob- plotswere measuredusing a modifiedBiltmore stick servervariability, all fieldtechnicians were trained in (Jamesand Shugart 1970, Husch et al. 1972). Each bird identification(by sightand sound)and distance tree >15.2 cmdbh was assigned to a defoliationlevel estimationprior to actualdata collection(Kepler and (0 = no defoliation, 1 = 1 to 33% defoliation, 2 = 34 Scott 1981). Practiceof distance estimation was aided to 67%, 3 = >67%). A defoliation value was calcu- with the use of a measuringtape so that observers lated for eachyear based on the averageof all defo- becamefamiliar with birdsongat differentdistances liation classesfor all trees assigneda value (Table 1). at samplingplots. An eight-minutesurvey was con- Eighteenhabitat variables were includedin the ductedat eachplot, during which all seenor analyses(see Table 4): number of live trees (all spe- heard were recorded. Distance was estimated from cies)in six size classes(0 to 7.6 cm dbh, >7.6 to 15.2 where the observerwas standingto a point on the cm, >15.2 to 22.8 cm, >22.8 to 30.4 cm, >30.4 to 38.1 grounddirectly below the bird. Detectioncutoff val- cm, and >38.1 cm); numberof dead treesin the same ues were calculated according to Reynolds et al. six size classes;total number of live trees; total num- (1980) with all observationsfrom all years pooled ber of dead trees;percent cover of woody-stemmed (ca. 2,400 observationsof towhees made from 42 vegetation<3 m in height(i.e. shrubcover); percent plotsover 13 years).The calculateddetection cutoff total canopycover (percent of canopypresent vs. ab- valuefor the EasternTowhee on our studyplots was sent);percent cover of maximumhigh canopy (when 70 m, and this was used as the cutoff value for all theuppermost canopy was > 12m high);and percent plotswithin the study area.The numberof towhees coverof maximumlow canopy(when the uppermost countedat eachplot was averagedover the fivevisits canopywas 3 to 12 m high). Habitat variableswere and is expressedas number of towheesper 70-mplot measuredat eachsubplot, and the valueswere av- (i.e. 15,394m2). Each plot was consideredseparately eragedover the five subplotsto get a plot value(n = for statisticalpurposes. 42) to be testedagainst the numberof birds counted Vegetationwas sampledat eachsubplot following at eachplot. a modificationof the methodsof Jamesand Shugart Data analysis.--Pre-outbreakvalues were the av- (1970).There were 5 samplingpoints in eachcardi- erageof eachof the 18variables plus Eastern Towhee nal compassdirection spaced2.26 m apart, for a total densityestimates for eachplot overthe periodfrom of 20 samplingpoints per 0.04-hasubplot. At each 1984to 1986(n = 42 forthe pre-outbreakperiod). Be- sampling point, ground cover and vegetationwere cause the seventh transect was established in 1985, recordedusing a sightingtube with a wire crosshair thevalues for this transect in thepre-outbreak period at the end. The observersighted vertically through representthe averageof 1985and 1986only. Post- the tube (first down, then up), and vegetationthat outbreakvalues were the averagefor eachplot over intersectedthe crosshairswas recordedby height the period from 1989 to 1996. The years 1987 and classfrom the groundup throughthe canopy.Ob- 1988were left out of the analysesbecause we were serverssometimes had to stepback to observecan- interestedprimarily in the periodsbefore and after opy coverthat was obscuredby lowercanopy. Each this major disturbance.The disturbanceyears were "hit" or intersectionof the vegetationwith thecross- left in for repeated-measurestrend analysis only. An hair counted as 5%, so values for each variable c• of 0.05 was used for all tests. October1997] Towheesand Gypsy Moth Defoliation 711

To accountfor correlationamong years due to re- TABLE2. Abundanceof EasternTowhees (no. per peatedsampling of the sameplots, we usedrepeat- 70-m radius plot) in Berkeleyand Morgan coun- ed-measuresanalysis of variance(Hatcher and Ste- ties, West Virginia, 1984 to 1996.Density estimates panski1994) to determineif EasternTowhee density were calculatedusing variable circular plots with estimates increased over time. This was done for the a cutoffdistance of 70 m at eachplot.

36 plots on SleepyCreek Mountain.Due to an un- Mini- Maxi- balanceddesign (measured for 12 and not 13 years), No. % plots Mean mum mum a separateANOVA wasperformed on the sixplots on plots with no. no. no. Third Hill Mountain. sur- tow- tow- tow- tow- A matched-pairst-test on before and after values at Year veyed hees hees• hees hees SE eachplot was used to determineif EasternTowhee 1984 36 52.8 0.3 0 1.2 0.06 numberswere higher in our studyarea after the gypsy 1985 42 47.6 0.3 0 1.2 0.05 moth outbreak. Paired t-tests also were used to deter- 1986 42 50.0 0.2 0 1.0 0.04 mine if valuesfor the vegetationvariables were sig- 1987 42 52.3 0.4 0 1.8 0.08 nificantly different before and after the outbreakfor 1988 42 88.1 0.9 0 2.8 0.10 the studyarea as a whole.Principal components anal- 1989 42 83.3 1.2 0 2.8 0.12 ysis (Isebrandsand Crow 1975, Johnson1981) was 1990 42 95.2 1.5 0 3.4 0.12 usedto describethe vegetational gradients in thehab- 1991 42 95.2 1.7 0 4.0 0.16 1992 42 92.9 1.6 0 4.0 0.16 itat data set. Pearson correlation was used to deter- 1993 42 97.6 1.7 0 4.0 0.14 mine the degreeof associationbetween density esti- 1994 42 95.2 1.8 0 4.4 0.17 matesof towheesand the majorvegetational gradi- 1995 42 95.2 1.7 0 4.4 0.17 entsin the studyarea as determined by PCA.Because 1996 42 85.7 1.2 0 3.4 0.14 our objectiveswere to explain habitat associationsof aValues for densityestimates were madefor eachplot/year com- EastemTowhees, in this paperwe describeonly the binationand then averagedfor eachyear. Maximum and minimum vegetationalgradients that were significantlycorre- valuesare combinedacross plots within years. latedwith towbeedensities. Eastern Towhee density estimatesalso were correlatedwith percentcover of 11 commonspecies of plantswithin the shrublayer, analysistime accountedfor 73% of the total which were indicative of different habitat conditions variance in towhee numbers. Towhee numbers within the studyarea. All variableswere transformed increasedsignificantly on Third Hill Mountain prior to analysis.Percentage data were subjectedto an (F = 18.85, P = 0.0073). Again, a linear trend arcsinesquare-root transformation (Krebs 1989); the transformationlog•0 (variable + 1) wasused for all oth- accountedfor most of the total variance(53%). er variables(Dowdy and Weardon1991). Densityestimates were significantly higher af- To examinethe degree of predationand Brown- ter the outbreakthan before over all 42 plots headedCowbird (Molothrus ater) parasitism at Sleepy (paired t = 14.99,P = 0.0001). Creek, we searched for nests in and around the 42 Prior to the outbreak,the first threeprincipal samplingplots in 1995and 1996.Nests were moni- componentsaccounted for 61.1%of thevariation tored following guidelinessuggested by the Breed- in the vegetationdata set (Table3). During this ing BiologyResearch and MonitoringDatabase (i.e. period,Eastern Towhee density estimates were BBIRD) samplingprotocol (Martin et al. 1996),and contents of nests were recorded. Care was taken not significantlycorrelated only with PC II (r = -0.34, P = 0.0295),which representeda gradi- to disturbvegetation surrounding the nest.Parasit- ism was indicatedby the presenceof a Brown-head- ent from plotswith a high densityof trees(par- ed Cowbird egg or nestling in a towbee nest. Nests ticularlytrees of smalldiameter) and a highcan- were consideredsuccessful if they fledgedat least opy structureto plots with lower densitiesof onehost young. live trees,i.e. plotsof sparsevegetation. After the outbreak,the first three principal RESULTS componentsaccounted for 73.8% of the varia- tion in the vegetationdata set (Table3). During Estimatesof densityranged from an average this period, EasternTowhee density estimates of 0.2 birds per plot (i.e. 15,394m 2) in 1986 to were significantlycorrelated with PC III (r = 1.8 birds per plot in 1994 (Table2). The abun- -0.51, P < 0.0005),which representeda gra- danceof EasternTowhees increased signifi- dientfrom a high densityof trees(particularly cantly acrossall years (F = 211.1, P = 0.0001). small-diametertrees) and thus a closedcanopy, The estimatedregression slope of number of to plotswith more canopygaps and lower den- towheesagainst time was0.13 _+SE of 0.01.The sitiesof live trees(due to high numbersof me- trend was not linear, although in the linear dium and large dead trees;Table 4). 712 BELLAND WHITMORE [Auk, Vol. 114

TABLE3. Resultsof principalcomponents analysis on Eastern Towhee habitat variables before (1984 to 1986) and after (1989to 1996)an outbreakof gypsymoths.

Pre-outbreak Post-outbreak

PC I PC II PC III PC I PC II PC III Eigenvalue 5.7 2.8 2.3 5.6 4.7 2.8 % Relative total variance 32.0 16.0 12.9 31.4 26.6 15.7 % Cumulative total variance 32.0 48.1 61.1 31.4 58.0 73.8

After the gypsy moth outbreak,the study commonshrub-level plant speciesbefore and area as a whole had greater percentageof after the outbreakare givenin Table6. shrub coverand low canopycover, a greater We found 41 Eastern Towhee nests in 1995 number of live trees0 to 7.6 cm dbh, a greater and 1996.Fourteen nests failed duringthe egg number of total live trees, and a greaternum- stage,12 failed duringthe nestlingstage, and ber of deadtrees in the four largestsize classes 15 fledgedat leastone Eastern Towhee. Three (Table5). The study area had a smallerper- of the 41 nestswere parasitizedby Brown- centageof canopy cover and high canopy cover, headed Cowbirds. One nest that failed in the fewer dead trees in the 0 to 7.6 cm size class, nestlingstage contained one unhatchedcow- and a smaller number of live trees in the four bird egg,whereas two successfulnests fledged largestsize classes.Because the decreasein one cowbirdeach. One of thesenests fledged numbers of small dead trees offset the increase threehost young and onecowbird young, and in numbers of medium and large dead trees, the otherfledged one host young and onecow- the total numberof deadtrees did not change bird young. Successfulnests fledged an aver- significantly.The numberof live trees>7.6 to age of 2.73 young. 15.2 cm and >15.2 to 22.8 cm, and the number of dead trees >7.6 to 15.2 cm also did not DISCUSSION changesignificantly from the pre- to post-out- breakperiods. Results of the correlationanal- The SleepyCreek site as a whole became ysis betweentowhee densityestimates with more attractive to towhees after the influence of

TABLE4. Pearsoncorrelations between statistically significant vegetational gradients from principalcom- ponentsanalysis and EasternTowhee density estimates (n = 42) before(1984 to 1986)and after(1989 to 1996)an outbreakof gypsymoths.

Pre-outbreak (PC II) Post-outbreak(PC III) Variable r P r P Towheedensity estimate -0.34 0.0295 -0.51 0.0005 % vegetation<3 m height 0.54 0.0002 0.15 0.3381 % total canopycover 0.24 0.1230 0.83 0.0001 % max. low canopy cover - 0.51 0.0005 0.30 0.0557 % max. high canopycover 0.60 0.0001 0.45 0.0025 No. live trees(0 to 7.6 cm dbh) 0.84 0.0001 0.40 0.0079 No. live trees (>7.6 to 15.2 cm) 0.42 0.0057 0.68 0.0001 No. live trees (>15.2 to 22.8 cm) -0.15 0.3424 0.37 0.0145 No. live trees (>22.8 to 30.4 cm) -0.16 0.3247 0.01 0.9726 No. live trees (>30.4 to 38.1 cm) -0.05 0.7334 0.14 0.3807 No. live trees (>38.1 cm) 0.09 0.5716 0.15 0.3437 Total no. live trees 0.79 0.0001 0.57 0.0001 No. dead trees 0 to 7.6 cm) 0.44 0.0037 0.01 0.9628 No. dead trees >7.6 to 15.2 cm) - 0.05 0.7543 0.01 0.9757 No. dead trees >15.2 to 22.8 cm) -0.03 0.8578 -0.40 0.0082 No. dead trees >22.8 to 30.4 cm) 0.11 0.4809 -0.55 0.0001 No. dead trees >30.4 to 38.1 cm) 0.23 0.1438 -0.37 0.0147 No. dead trees >38.1 cm) - 0.04 0.7804 - 0.27 0.0797 Total no. dead trees 0.27 0.0851 -0.14 0.3859 October1997] Towheesand Gypsy Moth Defoliation 713

T^I31•E5. Means(_+ SE) for habitatvariables and EasternTowbee density before (1984 to 1986)and after (1989to 1996)an outbreakof gypsymoths. P-values are from pairedt-tests comparing pre- and post- outbreak values.

Variable Pre-outbreak Post-outbreak t P Towheedensity estimate 0.3 _+0.0 1.5 +_0.1 -14.84 0.0001 % total canopycover 88.8 _+0.6 75.8 _+0.6 10.31 0.0001 % shrubcover (<3 m high) 41.5 _+0.6 67.7 _+0.7 -14.82 0.0001 % max. low canopycover 23.7 _+1.6 44.5 _+0.8 -8.23 0.0001 % max.high canopycover 65.7+_ 1.7 31.7 _+1.0 12.57 0.0001 No. live trees (0 to 7.6 cm dbh) 256.1 ñ 15.8 394.4 _+20.7 -7.66 0.0001 No. live trees (>7.6 to 15.2 cm) 101.5 _+5.8 110.0 ñ 4.2 -2.01 0.0522 No. live trees (>15.2 to 22.8 cm) 47.7 +_4.5 39.0 _+3.5 2.73 0.0093 No. live trees(>22.8 to 30.4 cm) 22.7 _+1.2 17.5 +_1.6 4.76 0.0001 No. live trees (>30.4 to 38.1 cm) 17.3 _+1.0 9.5 -+ 0.8 6.59 0.0001 No. live trees (>38.1 cm) 20.9 +_1.6 10.8 _+1.0 6.40 0.0001 Total no. live trees 466.5 _+ 20.1 581.3 _+ 21.9 -5.86 0.0001 No. dead trees(0 to 7.6 cm) 70.9 +_4.8 40.4 _+2.4 9.85 0.0001 No. dead trees (>7.6 to 15.2 cm) 45.9 ñ 3.9 43.2 ñ 4.2 1.77 0.0835 No. dead trees(>15.2 to 22.8 cm) 9.0 _+0.8 18.1 +_1.6 -6.23 0.0001 No. dead trees (>22.8 to 30.4 cm) 2.7 _+0.3 10.2 _+0.8 -9.25 0.0001 No. dead trees(>30.4 to 38.1 cm) 2.0 +_0.3 7.9 _+0.7 -8.78 0.0001 No. dead trees (>38.1 cm) 1.0 +_1.6 9.3 +_0.9 -15.63 0.0001 Total no. dead trees 131.3 _+ 7.3 129.0 _+ 7.1 1.99 0.0522 the gypsymoth outbreak. Mortality of oversto- cessional"may be too vague to properlyde- ry treesand subsequentopening of forestcan- scribe optimal towhee habitat, and that the opy causedincreases in understoryvegetation structureand/or compositionof early succes- due to increasedlight penetrationand nutrient- sionalhabitat must be considered.Among other laden gypsy moth frass reachingthe forest things,"early successional" can indicate a dense floor (seeCollins 1961). Because a gypsymoth standof saplingtrees. Both before and after the outbreak can set back succession(i.e. decrease outbreak, towhee densities consistentlywere the numberof large- and medium-sizedtrees negativelyassociated with highdensities of live and increasethe amount understoryvegeta- trees,particularly when a largenumber of the tion), it is beneficialfor EasternTowhees. The trees were of a small diameter However, al- increasesin understoryvegetation (particular- thoughEastern Towhees nest and forage on the ly shrubs)increased nesting cover and foraging groundand in shrubs,we did notfind a positive opportunitiesfor this species. associationbetween estimates of towheedensity Our resultssuggest that the term"early suc- andthe amount of vegetation<3 m in heightat

TABLE6. Pearsoncorrelations between Eastern Towbee density estimates (no. per 70-m radius plot) and percentcover of 11 commonplant speciesin the shrublayer (<3 m high)before (1984 to 1986)and after (1989to 1996)an outbreakof gypsymoths (n = 42).

Pre-outbreak Post-outbreak

Species r P r P Blackberry(Rubus spp.) 0.37 0.0171 0.68 0.0001 Blackbirch (Betulalenta) -0.10 0.5490 -0.39 0.0097 Blackcherry (Prunus serotina) 0.49 0.0010 0.39 0.0011 Blackgum (Nyssasylvatica) -0.42 0.0059 -0.47 0.0019 Blacklocust (Robiniapseudoacacia) 0.32 0.0396 0.41 0.0064 Grapevine(Vitis spp.) 0.23 0.1503 0.50 0.0007 Hickory (Caryaspp.) 0.16 0.32 0.53 0.0004 Red maple (Acerrubrum) -0.54 0.0003 -0.58 0.0001 Spicebush(Lindera benzoin) -0.27 0.0887 -0.35 0.0242 Witch hazel (Hamamelisvirginiana) -0.37 0.0295 -0.44 0.0036 Yellowpoplar (Liriodendrontulipifera) -0.22 0.1576 -0.34 0.0281 714 BELLAND WHITMORE [Auk, Vol. 114 our site.This result may have occurredbecause numbers of small trees. If towhee numbers are "% shrubcover" encompassed all vegetation<3 lower in areas with more small trees, then de- m in height,regardless of speciesor height with- foliation actually may make the habitat less in the category.For example,tree saplingsand suitable for towhees. However, shrub cover shrubswere lumped as "shrub cover." When to- also increased after defoliation. Thus, defolia- whee densities were correlated with individual tion would be detrimentalto towheesonly if plant species,however, several patterns of as- shrub cover declined as a result of the increase sociationemerged. in small trees. Towhees were positively associatedwith The large increasein EasternTowhee densi- plant speciesthat were most abundanton the ties after the gypsy moth outbreak suggests drier ridgetop and midslope plots at Sleepy that the habitat became more suitable. This con- Creek. The highest densitiesof towheesoc- clusionis basedon the assumptionthat density curred on the ridgetops,and theseplots were of towheesreflects habitat suitability.The as- associatedwith the most opencanopy, lowest sociationbetween density and habitatsuitabil- tree density,and the lowestsmall-tree density, ity has been debated elsewhere(Wray et al. particularly in the post-outbreakperiod. In 1982,VanHorne 1983, Vickery et al. 1992).The canopygaps on the drier uplandsites, vegeta- fact that habitat associations were similar both tion suchas Vitis spp. and particularlyRubus before and after outbreak, and that before the spp. grew in abundance.The ability of shade- outbreak31% of the plotshad no towhees,sup- intolerantseedlings and saplingssuch as black portsthe idea that higherdensities of towhees cherry,black locust, and certain hickory species indicatesuitable habitat in our studyarea. Sim- (see Table 6 for scientificnames of plants) to ilarly, Holmeset al. (1996)found that plotswith grow on these upland sites further indicates high densitiesof Black-throatedBlue Warblers the degree of openness(i.e. density of low (Dendroicacaerulescens) were more suitablethan trees) of the stands. plots with low densitiesof warblers.Eastern Towhee density estimateswere negatively Towheenumbers likely did not result from in- associatedwith many plant speciesthat grow creased food through the temporary abun- primarily in the moistercoves and valleyareas danceof gypsy moths.Although many birds, of the site. Both before and after the outbreak, including Eastern Towhees,eat gypsy moth towheeswere negativelyassociated with red larvae, this food sourcemakes up only a small maple and black gum, and after the outbreak, proportionof total diet (Smith1985, Cooper et they were negatively associatedwith other al. 1987, Smith and Lautenschlager1981). In- plant speciessuch as spicebushand witchha- deed, most speciesof birds (exceptCoccyzus zel. Thesespecies are shadetolerant and were cuckoos)that eat lepidopteranlarvae prefer found more often on wet soilsat SleepyCreek. smooth,hairless species over hairy larvaesuch After the outbreak,towhees also were nega- as gypsy moths (Smith 1985, Whelan et al. tively associatedwith certainshade-intolerant 1989).Although other factors could have influ- species,such as yellow poplar and black birch. enced towhee numbers, we believe that the dra- However,saplings of thesespecies were found matichabitat changes after the gypsymoth out- primarily in canopygaps in the moistvalley of break provide the most plausibleexplanation the study site and were not abundanton the for the increases in towhees that we observed. drier ridgetopand midslopeplots. In order to better understand the influence of Becausetowhees nest and forageon or near gypsymoth-induced tree mortality on Eastern theground, a highdensity of trees may indicate Towhees, further work needs to be done on that thereis lessspace to carry out daily activ- pairing and reproductivesuccess within out- ities, and towheestherefore may prefer areas break areas. Concern has been raised that trees with lowertree densities.Dense stands of sap- killed by defoliationmay make better vantage lings may not providesuitable nesting cover or pointsavian nest predators and Brown-headed foraginghabitat for towhees.Shrubs such as Cowbirds (Robbins1979, Yahner and Wright Rubusand Vitis may providemore appropriate 1985).However, only three of the 41 neststhat dense cover close to the ground and better we monitored were parasitized by cowbirds, structurein whichto placenests. After the out- despite the many standing dead trees in the break, someof the study plotshad increased area.Predation was the majorcause of nestfail- October1997] Towheesand Gypsy Moth Defoliation 715 ure at SleepyCreek, but thisis typicalfor many ized increasesin their populationsin eastern populationsof (Martin 1993). . Using a combinationof sources,Hagan (1993) reported a major decline in EasternTo- ACKNOWLEDGMENTS whees in the northeastern United States from 1966to 1989.Moreover, he reportedan annual Financialsupport was providedby the McIntire- decreasein towheesof 1.2%per year in West StennisProgram, the WestVirginia Division of Nat- ural Resources,and Cooperative Agreement No. Virginia. Although data from our study area 23-043 between the U.S. Forest Service and West Vir- indicatethat EasternTowhees are increasing ginia University.We gratefully acknowledgeK. M. ratherthen declining, they support Hagan's as- Dodge,D. K. Thurber,and R. J.Cooper for their work sertionthat declinesin EasternTowhees prob- on this project;R. T. Holmes,J. M. Hagan, R. M. ably are relatedto forestsuccession. Breeding DeGraaf,and R. L. Hutto for their suggestions;and Bird Surveydata from the regionsurrounding S. Droegeand P. Peterjohnfor providingBBS infor- our study area (Maryland routes6 and 7; Penn- mation.This paper is publishedwith the approvalof the Director, West Virginia University Agriculture sylvaniaroutes 92 and 194;Virginia routes1, 5, and ForestryExperiment Station, as ScientificPaper 7, and 8; and West Virginia routes48, 49, 50, No. 2612. and 51) indicate no significanttrend (overall trend for all routeswas 0.698, P > 0.05) in East- LITERATURE CITED ernTowhee numbers from 1984 to 1995(Droege andPeterjohn pers. comm). Since 1980, much of BIRCH, t. W., D. A. GANSNER, S. L. ARNER, AND R. H. the areasurrounding the SleepyCreek site has WIDMANN. 1992. Cutting activity on West Vir- beenheavily treated with pesticidesto control giniaTimberlands. Northern Journal of Applied gypsymoths (USDA 1994). Thus, the vegetation Forestry 9:146-148. changesin the areaswhere the surveyswere COLLINS,S. 1961. Benefitsto understory from can- conducted would not be the same as those that opy defoliationby gypsy moth larvae.Ecology 42:836-838. occurredat SleepyCreek. Our dataalso support COOPER,R. J., K. M. DODGE, P. J. MARTINAT, H. R. the ideathat Eastern Towhees can respond im- SMITH, AND R.C. WHITMORE.1987. Impacts of mediatelyto newly createdhabitat (e.g. areas the gypsy moth on nongamebird populations. subjectto gypsymoth defoliationand subse- Pages48-58 in Proceedingsof Coping with the quent understoryrelease) that formerly oc- Gypsy Moth on the New Frontier (S. Fosbroke curred only rarely in a heavily forestedstate and R. R. Hicks, Jr.,Eds.). West Virginia Univer- suchas West Virginia. This ability to rapidlycol- sity Books,Morgantown. COOPER,R. J., K. M. DODGE, D. K. THURBER,R. C. onize new areaswould seemto be especially WHITMORE,AND H. R. SMITH. 1993. Responseof adaptivein bird speciesrestricted to ephemeral ground-levelwildlife food plants to canopyde- habitats. In the absenceof marked individuals, foliationby the gypsy moth. Annual Conference it is impossibleto determinewhere the towhees of the Southeastern Association of Fish and from our studyarea originated. Wildlife Agencies47:268-275. Eastern Towhee populationsmay benefit DEGRAAF,R. M. 1987. Breeding birds and gypsy from eventsthat arelikely to occur(particular- moth defoliation:Short-term responsesof spe- ciesand guilds. Wildlife SocietyBulletin 15:217- ly in WestVirginia) overthe nextdecade or so. 221. Muchof theforest that replaced timber cut ear- DOWDY, S., AND S. WEARDON. 1991. Statistics for re- lier this centuryis reachingmaturity and is search.John Wiley and Sons,New York. likely to be harvestedsoon. Approximately GOTTSCHALK,K. W. 1991. Gypsy moth:Impacts and 89%of the forestedarea in WestVirginia is pri- silvicultural options.Pages 159-169 in Proceed- vately owned,and more than half of the pri- ingsof the Oak Resourcein the UpperMidwest. vatelyowned forest is expectedto be cut over Implicationsfor Management(S. B. Laursenand the next10 years (Birch et al. 1992).Populations J. F. DeBoe, Eds.). Minnesota Extension Service, St. Paul. of vertebratesfrequently reflect land-use prac- HAGAN, J. M. Ill. 1993. Decline of the Rufous-sided tices (Litvaitis 1993); prospective land-use Towhee in the eastern United States. Auk 110: changesmay createmore suitablehabitat for 863-874. EasternTowhees and other early successional HATCHER,L., AND E. J. STEPANSKI.1994. A step-by- species,and we may once again see general- stepapproach to usingthe SASsystem for uni- 716 BELLAND WHITMORE [Auk, Vol. 114

variate and multivariate statistics. SAS Institute, pen, Ed.). U.S. ForestService General Technical Inc. Cary,North Carolina. Report RM-87. HIx, D. M., D. E. FOSBROKE,R. R. HICKS, AND K. W. QUIMBY,J. W. 1987.Impact of gypsymoth defolia- GOTTSCHALK.1991. Developmentof regenera- tion on forest stands.Pages 21-38 in Proceed- tion following gypsy moth defoliationof Ap- ingsof copingwith the gypsymoth on the new palachianPlateau and Ridge and Valley hard- frontier (S. Fosbrokeand R. R. Hicks, Jr., Eds.). wood stands.Pages 347-359 in Proceedingsof WestVirginia UniversityBooks, Morgantown. the 8th Central Hardwood ForestConference (L. REYNOLDS,R. T., J. M. SCOTT,AND R. A. NUSSBAUM. H. McCormickand K. W. Gottschalk,Eds.). U.S. 1980. A variablecircular plot methodfor esti- ForestService General Technical Report NE-148. matingbird numbers.Condor 82:309-313. Radnor,Pennsylvania. ROBBINS,C. S. 1979. Effects of forest fragmentation HOLMES, R. t., R. P. MARRA, AND t. W. SHERRY.1996. on bird populations.Pages 198-213 in Manage- Habitat-specificdemography of breedingBlack- ment of northcentral and northeastern forests throated Blue Warblers(Dendroica caerulescens): for nongamebirds (R. M. DeGraaf and K. E. Implicationsfor populationdynamics. Journal of Evans, Eds.). U.S. Forest ServiceGeneral Tech- Ecology65:183-195. nical ReportNC-51. HUSCH, B., C. I. MILLER, AND t. W. BEERS. 1972. For- SMITH,H. R. 1985. Wildlife and the gypsy moth. est mensuration.John Wiley and Sons,New Wildlife SocietyBulletin 13: 166-174. York. SMITH• H. R., AND R. A. LAUTENSCHLAGER. 1981. ]SEBRANDS,J. G., AND t. g. CROW. 1975. Introduction Gypsy moth predators.Pages 96-124 in The to usesand interpretationof principalcompo- gypsy moth: Researchtoward integratedpest nentsanalysis in forestbiology. U.S. ForestSer- management(C. C. Doane and M. L. McManus, viceGeneral Technical Report NC-17. Eds.).U.S. Department of AgricultureTechnical JAMES,E C. 1971. Ordinations of habitat relation- Bulletin No. 1584. shipsamong breeding birds. Wilson Bulletin 83: THURBER,D. K., W. g. MCCLAIN, AND g. C. WHIT- 215-236. MORE.1994. Indirecteffects of gypsymoth de- JAMES,E C., AND H. H. SHUGART.1970. A quanti- foliation on nest predation.Journal of Wildlife tativemethod of habitatdescription. Audubon Management 58:493-500. Field Notes 24:727-736. TWERY,M. J. 1991. Effectsof defoliationby gypsy JOHNSON,D. H. 1981. The use and misuse of statis- moth.Pages 27-39 in Proceedingsof the United tics in wildlife habitatstudies. Pages 11-19 in StatesDepartment of AgricultureInteragency The use of multivariate statistics in studies of GypsyMoth Research Review (K. W. Gottschalk, wildlife habitat(D. E. Capen,Ed.). U.S. Forest M. J. Twery, and S. I. Smith,Eds.). U.S. Forest ServiceGeneral Technical Report RM-87. ServiceGeneral Technical Report NE-146, Rad- KEPLER,C. B., ANDJ. M. SCOTT.1981. Reducingbird nor, Pennsylvania. count variability by training observers.Pages USDA FORESTSERVICE. 1994. Sprayingthe gypsy 366-371 in Estimating numbers of terrestrial moth.Gypsy Moth News 35:2-11. birds (C. J. Ralphand J. M. Scott,Eds.). Studies VANHORNE,B. 1983. Density as a misleadingindi- in Avian Biology No. 6. catorof habitatquality. Journal of Wildlife Man- KREBS,C. J. 1989. Ecologicalmethodology. Harper agement 47:893-901. and Row, New York. VICKERY,P. D, M. L. HUNTER, AND J. V. WELLS. 1992. LITVAITIS,J. g. 1993. Responseof early successional Is densityan indicatorof breedingsuccess? Auk vertebratesto historicchanges in landuse. Con- 109:706-710. servationBiology 7:866-873. WHELAN, C. J., R. t. HOLMES,AND H. R. SMITH. 1989. MARTIN,t. E. 1993. Nest predation among vegeta- Birdpredation on gypsy moth (: Ly- tion layersand habitattypes: Revising the dog- mantriidae) larvae: An aviary study. Environ- mas. American Naturalist 141:897-913. mental Entomology18:43-54. MARTIN, t. E., C. R. PAINE, C. J. CONWAY,W. M. HO- WRAY, t., K. g. STRAIT, AND g. C. WHITMORE. 1982. CHACHKA, P. ALLEN, AND W. JENKINS. 1996. Reproductivesuccess of grasslandsparrows on BreedingBiology Research and MonitoringDa- a reclaimedsurface mine in WestVirginia. Auk tabase(BBIRD) field protocol.Montana Coop- 99:157-164. erative Wildlife ResearchUnit, University of YAHNER,R. H. 1995. Powerline right-of-ways.Con- Montana, Missoula. servationBiology 9:476. MAURER, B. A., L. B. MCARTHUR, AND g. C. WHIT- YAHNER,g. H., AND g. L. WRIGHT.1985. Depreda- MORE.1980. Habitat associationsof breeding tion on artificial ground nests:Effects of edge birds in clearcut deciduous forests in West Vir- and plot age. Journalof Wildlife Management 49:508-513. ginia. Pages167-172 in The use of multivariate statisticsin studiesof wildlife habitat (D. E. Ca- Associate Editor: R. L. Hutto