POPULATION CONSEQUENCES OF RESTRICTED DISPERSAL FOR AN INSECT HERBIVORE IN A SUBDIVIDED HABITAT

1828

Ecology, 81(7), 2000, pp. 1828–1841q 2000 by the Ecological Society of America

POPULATION CONSEQUENCES OF RESTRICTED DISPERSAL FOR ANINSECT HERBIVORE IN A SUBDIVIDED HABITAT

PATRICIA DOAK1

Section of Ecology and Systematics, Corson Hall, Cornell University, Ithaca, New York 14853 USA

Abstract. This paper addresses how dispersal ability of the geometrid moth Itameandersoni interacts with habitat subdivision to impact population structure, importance ofdifferent habitat types (e.g., patch sizes, habitat configurations), and regional distributionpatterns in a subdivided habitat of patches of its host plant Dryas drummondii (Rosaceae).Habitat subdivision occurs both on the patch scale and on the larger spatial scale of sitesdue to patchy successional patterns. Itame distribution is patchy within areas colonized byDryas. This research combines observational and experimental approaches to examineindividual movement behavior, immigration, and emigration rates of both caterpillars andadult females at the scale of discrete Dryas patches. Itame caterpillars will cross bare groundseparating plant patches; however, movement of caterpillars on bare ground was morerestricted than predicted by a correlated random walk due to large turning angles whichled to area-restricted search. Caterpillar immigration into patches previously cleared ofItame is usually low and is negatively affected by patch isolation when local Itame densitiesare high. Caterpillar emigration is also significantly greater from patches with high Itamedensity. Immigration of females, which have reduced wings, was positively influenced bypatch isolation and negatively influenced by the presence of a barrier preventing adultfemales from walking into patches. All females emigrate from their point of emergenceand continue moving as they oviposit. Females lay eggs singly on foliage, in leaf litter,and on rocks under and surrounding Dryas patches. Significantly more eggs are laid off ofDryas patches than within Dryas patches, but this is not a result of concentration of searchingeffort off of Dryas patches. Females move greater distances than predicted by a correlatedrandom walk. Movement parameters combined with oviposition parameters lead to theprediction that the average female will lay all of her eggs within 5–15 m of her startingpoint. Restricted movement appears to demographically isolate sites separated by distancesof as little as 0.5 km. Limited between-site dispersal may link these populations intometapopulations and may explain patchy regional patterns of distribution. Within sites,densities are likely to build up in smaller patches and subdivided habitat due to preferentialoviposition on bare ground. These smaller patches may in turn be disproportionally im-portant for population processes due to their contribution of dispersers through density-dependent emigration within and between sites. Thus, this research demonstrates how lim-ited dispersal can impact patterns of distribution, abundance, and population structure ina natural system.

Key words: correlated random walk; Dryas drummondii; emigration; habitat patchiness; im-migration; insect dispersal; Itame andersoni; metapopulation; oviposition behavior.

INTRODUCTION

Over the past decade an increasing emphasis hasbeen placed on the importance of spatial patterns forpopulation dynamics, particularly the spatial arrange-ment of habitats. To understand population level pro-cesses of species inhabiting fragmented landscapes, itis not only vital to document the spatial scale of habitatsubdivision but also to study the movement of the spe-cies of interest (Kareiva 1990, May and Southwood1990, Reeve 1990, Taylor 1990, Harrison 1991, Doak

Manuscript received 18 September 1998; revised 2 June 1999;accepted 9 June 1999; final version received 12 July 1999.

1 Present address: Institute of Arctic Biology, Universityof Alaska, P.O. Box 757000, Fairbanks, Alaska 99775 USA.

et al. 1992, Kuussaari et al. 1996, Ims and Yoccoz 1997,Stacey et al. 1997).

Although much of this work has concentrated oneasy-to-measure aspects of habitat subdivision, whatappears as subdivision to the observer may or may notbe subdivision on a scale relevant to resident species.Even small scale subdivision may affect populationdynamics (e.g., Gerber and Templeton 1996, Kindvall1996, Kareiva 1987), however in other cases, dramaticsubdivision may have little or no effect on populationdynamics (e.g., Antolin and Strong 1987, Fahrig andPaloheimo 1988, Harrison and Thomas 1991, Staceyand Taper 1992, Neve et al. 1996). Sedentary organismscan be affected by very slight subdivision, whereashighly mobile organisms may not experience any im-pacts from even drastic large-scale fragmentation.

July 2000 1829INSECT DISPERSAL AND POPULATION PROCESSES

Understanding the interaction of dispersal capabilityand behavior on the one hand, and the qualitative andquantitative aspects of habitat subdivision on the other,is of interest for many aspects of basic population ecol-ogy, agroecology, and conservation biology. Althoughthe questions apply to many taxa, most work on habitatsubdivision has been conducted on insects due to theirshort generation time, ease of manipulation, and rela-tive tractability (see Doak et al. 1992 for a review).However, even for insects, relatively few studies haverigorously investigated both habitat pattern and indi-vidual movement, and still fewer have studied naturallyoccurring patterns of habitat arrangement.

My research concerns the moth, Itame andersoni(Geometridae) (henceforth Itame), which inhabitspatches of its host plant, Dryas drummondii (Rosaceae)(henceforth Dryas). Although Dryas patches are ubiq-uitous along river bars and glacial margins and occupyhundreds of miles of riverways in the Wrangell Moun-tains of Alaska, Itame populations are rare and limitedin spatial extent. I examine immigration and emigrationrates at the level of discrete Dryas patches as well asindividual movement of both larval and adult life stagesto address three issues relating to spatial populationprocesses:

1) What can movement behavior and dispersal dis-tances reveal about the connectivity of habitat patchesand thus the spatial population structure of Itame?

2) How does movement behavior impact the influ-ence of different patch types (e.g., small vs. large, iso-lated vs. surrounded) and habitat arrangements (e.g.,highly subdivided vs. continuous) on Itame popula-tions?

3) Can movement behavior help in explaining Ita-me’s patchy pattern of regional distribution?

A central question in the study of subdivided pop-ulations is the scale at which population processes oc-cur (Hanski and Gilpin 1991). By examining both cat-erpillar and adult movement as well as oviposition pat-terns, I am able to address the scale of population pro-cesses and examine the possible impacts of habitatsubdivision at two distinct spatial scales.

A less emphasized but related issue is how importantdifferent habitat types are for population or metapop-ulation maintenance (Root 1973, Shaffer 1981, Souleand Simberloff 1986, Wilcove et al. 1986, Kuussaariet al. 1996, Thomas and Hanski 1997). Itame popu-lations appear to florish in smaller Dryas patches andhighly subdivided habitat (Doak 2000). I use investi-gations of larval and adult immigration and emigration,and of oviposition behavior to address the contributionof different patch types to population mixing and col-onization of new patches.

Itame’s patchy regional distribution might be ex-plained by variation in host-plant quality, abiotic fac-tors, interspecific interactions, limited dispersal andcolonization potential, or any combination of the

above. Here, I examine displacements of both cater-pillars and ovipositing females to shed light on whetherdispersal is limited enough to play an important rolein these regional patterns of distribution.

NATURAL HISTORY AND STUDY SYSTEM

Habitat pattern

I conducted this research on the east margins of theKennicott Glacier in the Wrangell Mountains near Mc-Carthy, Alaska (618 N, 1438 W).

The host plant, Dryas drummondii, is an early suc-cessional, nitrogen-fixing subshrub (height ;5 cm) thatgrows on glacial moraine and river bars. It forms atwo-dimensional habitat of distinct patches on gravel,rock, and silt substrates. At later successional stagesthese patches grow together, forming extensive matsthat are then overgrown by later successional species.Individual Dryas patches range from minute to quitelarge with many discrete patches within the range of0.1–2 m2 (personal observation). In many locationsDryas exists in an almost complete monoculture.

In this system, subdivision not only occurs on thesmall patch-to-patch scale but also on a larger spatialscale due to patchy successional patterns. The glacialretreat at the study site has left a landscape of dry run-off channels interspersed with older vegetation ‘‘is-lands.’’ The dry channels and lower benches tend toharbor younger populations of Dryas than the older andhigher island areas. Thus, there is a mosaic of veryrecent run-off channels with little or no vegetation,older channels with young patchy distributions ofDryas, island areas with older patchy or continuousgrowth of Dryas, and island areas in later successionalstages that are being taken over by shrubs and trees.Additionally, there exists a great deal of seeminglysuitable habitat that is not occupied by Itame. The dis-tribution of Itame along the Kennicott Glacier and Riv-er is far from uniform, with areas of occupied Dryasinterspersed with areas of unoccupied Dryas. This pat-tern of patchy occupation also occurs on a larger scalewithin the Wrangell Mountains with large regions ofsome drainages unoccupied by Itame despite extensivepopulations of Dryas.

Life history

Itame andersoni is univoltine. Caterpillars hatchfrom overwintering eggs in early to mid-May shortlyafter snowmelt. All larval instars feed on the foliageof Dryas drummondii. I never found Itame associatedwith any other host plant.

Fifth instar caterpillars pupate under rocks or in thelitter beneath host plant patches in mid-June to mid-July. Adults emerge about two weeks after pupation.Neither males nor females feed. Males fly readily bothdiurnally and nocturnally. Females have smaller wingsthan males and seldom fly. Eggs are laid singly in leaflitter, on foliage and predominantly in gravel and rocks

1830 PATRICIA DOAK Ecology, Vol. 81, No. 7

TABLE 1. Number of Dryas patches in the caterpillar im-migration experiment.

Year and site

Patch type

1994

IN GL

1995

IN GL OR

SurroundedIsolatedVery isolated

55···

46···

645

645

555

both under and surrounding Dryas patches. Adult fe-males live for one to two weeks in captivity but lifespanin the field is unknown. The total flight period is ap-proximately one month (Doak 1997).

METHODS

I conducted this research during the summers of1994, 1995, and 1998. Related research on the distri-bution, abundance, and population dynamics of Itameoccurred from 1991–1995 (Doak 2000). Itame was theonly abundant herbivore on Dryas at my study sitesfrom 1990–1998. In this paper ‘‘patches’’ refers toDryas individuals or groups of individuals that havegrown together to form a linked and discrete clump,whereas ‘‘sites’’ encompass groups of Dryas patches.Sites were delineated by natural features such as theextent of Dryas growth, gravel banks, forested areas,and boundaries where patchy areas merged into regionsof huge continuous Dryas mats. The three study sitesused for this research were approximately 3200, 5300,and 6400 m2 (OR, IN, and GL sites respectively). Dis-tances between the sites were 460 (GL and OR), 730(GL and IN), and 760 m (OR and IN).

Caterpillar movement

Caterpillar immigration and patch isolation.—Toexamine the colonization of vacant patches by cater-pillars, I removed all caterpillars from experimentalpatches and then censused these extirpated patches todetermine the arrival of immigrants.

In 1994, I marked 10 patches at each of two sites(GL and IN). These patches were small with mean patchsize of 0.30 6 0.02 m2 (mean 6 1 SE, n 5 20) andwere classified as either surrounded or isolated. TheGL site had high caterpillar densities whereas the INsite had low caterpillar densities throughout my surveystudies (Doak 2000). On 31 May, after virtually all eggshad hatched and when the majority of caterpillars werein the first and second instar, I sprayed each patch withMalathion 50 Plus (Ortho, Columbus, Ohio, USA; 3.91mL/L of water). My observations indicated that Mal-athion was effective in killing these early instar cat-erpillars. I counted immigrants eight times from 11June to 6 July, removing them and noting their instars.I also reared a subset of the immigrants to examine sexratios (n 5 47).

In 1995, I marked 15 patches at each of three sitesand classified these patches as surrounded, isolated, orvery isolated. I used the two sites from 1994 and addedthe OR site which in past years had caterpillar densitiesintermediate between the IN and GL sites (Doak 2000).In 1995 mean patch size was 0.35 6 0.02 m2 (mean 61 SE). I sprayed these patches with Pyrenone (AgrEvo,Wilmington, Delaware, USA; 1.29 mL/L water) on 8June 1995 when the majority of caterpillars were thirdinstars (again the insecticide appeared effective). Ichecked each patch seven times from 14 June to 8 July,

again counting and removing immigrants and notingtheir instar.

I switched from Malathion to Pyrenone because Mal-athion appeared to have a fertilizing effect on patches;this effect did not, however, appear to differ betweenpatch types or sites.

The areas of all patches were estimated as ellipsesbased on their major and minor axes. Patch isolation,for surrounded and isolated patches, was quantified af-ter inclusion of patches in the experiment, and this ledto slightly unequal sample sizes (Table 1). To quantifypatch isolation, I measured the areas of all patches withan edge within 2 m of the edge of each experimentalpatch. I calculated the sum of the areas of these neigh-bors. Since a larger patch encompasses a larger areaby this measurement, I divided the summed neighborareas by the area surveyed. This measure of isolationwas used to classify patches into two categories (iso-lated , 0.2 , surrounded). This quantitative measure,developed to distinguish between the original two cat-egories, did not apply well to very isolated patches.Very isolated patches were distinguished by having noor very few neighbors within at least a 5 m radius.

Number of immigrants was analyzed by ANOVAwith effects of patch isolation, site, and the site byisolation interaction. Contrasts were done using se-quential Bonferroni corrections (Rice 1989) at the over-all a 5 0.05 level. Model assumptions were examinedand adequately met in this and all subsequent analyses.

Caterpillar emigration.—To examine rates of cat-erpillar emigration from patches, I conducted an ex-periment in which I monitored the movement of cat-erpillars out of patches (size: 0.28 6 0.01 m2, mean 61 SE, n 5 36) containing three different initial popu-lation densities. The experiment had two phases: aninitial set-up phase to manipulate population densityand hence the amount of feeding damage and an ex-perimental phase in which I followed emigration. Ichose 30 patches in 10 blocks of three patches each.Within each block, patches were randomly assigned toa low, medium, or high density treatment. In addition,six of the blocks also contained a control patch.

During the set-up phase (beginning 10–13 June),when the majority of caterpillars were in the third andfourth instars, I adjusted densities in the patches so thatlow, medium, and high density patches contained 5,15, and 25 caterpillars respectively. This initial treat-


ment yielded different damage levels which may affectemigration. Because most damage to the host plant isdone by late fourth and fifth instars, my adjustmentswere made early enough to impact damage levels.

I performed the experimental phase in two temporalsets, using one half of the blocks for each. On 15 JuneI removed all caterpillars in each patch of the first setof five blocks. I then placed 10, 20, and 30 caterpillarsin low, medium, and high density patches respectively.Caterpillars were collected from nonexperimental ar-eas, pooled, and randomly chosen for placement nearthe center of each patch. Each patch received halffourth and half fifth instars. The caterpillars in eachpatch within a block were dusted with a different colorof dayglo fluorescent powder. There was no detectablemovement between blocks. Caterpillars appeared to be-have normally when marked (personal observation,this experiment, caterpillar movement experiment be-low, and unpublished data). Patches were censusedeach day from 16–20 June. The fluorescent powder waseasily seen during both night and day censuses. At eachcensus the number of marked caterpillars still presentand visible in each patch was recorded. By doing de-structive samples, I have found that my estimates ofdensity correlate well with actual densities even whensearching for unmarked caterpillars (Pearson correla-tion, r 5 0.79; Doak 1997); therefore, I am confidentthat my censuses of these patches accurately reflectdifferences between treatments.

On 25 June I set up the second set of emigrationpatches in the same manner, but using only fifth instars(10, 20, and 30/patch), because the majority of cater-pillars were in that stage by 25 June. The second setof patches was censused each day from 26 June to 1July and all marked caterpillars were counted.

I also followed control patches to monitor the lossof caterpillars from sources other than emigration (e.g.,the loss of marks when fourth instars molted to fifths,pupation of fifth instars). I marked six control patches,each within an experimental block. Control patcheswere surrounded by clear plastic barriers, buried at theirbase and approximately 12 cm high, with a thin stripof Tanglefoot (Tanglefoot Company, Grand Rapids,Michigan, USA) around the top to prevent caterpillarsfrom leaving the patches (Itame caterpillars will notwalk into Tanglefoot so it does not increase mortality).These patches were manipulated in the same manneras medium density patches, receiving 15 caterpillarsduring the set-up phase and 20 marked caterpillars forthe experimental phase. Three of the six control patcheswere followed during each experimental set and cen-sused on the same dates as the experimental patches.

I used survivorship analysis (PROC LIFETEST; SASInstitute 1989) to examine the effect of density treat-ment on emigration times. Only data from the first threetime intervals (days 0–3) were used due to the largenumbers of caterpillars missing from patches by daythree (.50% for many patches) indicating the break-

down of density treatments. Caterpillars still remainingon day three became right-censored observations.

I pooled the data across the two time sets and acrossblocks, leading to a conservative estimate of treatmenteffects. Both Wilcoxon and log-rank x2 statistics wereused to evaluate the overall treatment effect. I thencalculated Z-statistics for multiple pairwise treatmentcomparisons with an experimentwise a 5 0.05 (Fox1993). I also tested the set and block effects for as-sociation with the loss time using Wilcoxon x2 (PROCLIFETEST).

Individual caterpillar movement.—On 8–12 July1994 and 13–14 July 1998, I observed 20 fifth instarcaterpillars (10 each year) to document their movementoff of Dryas patches and thus their possible contri-bution to population mixing and long distance spread.Caterpillars that engaged in pre-pupation behavior(e.g., searching under rocks for pupation sites) werenot included in these 20 observations. Each caterpillarwas marked with dayglo fluorescent powder to makeit easier to track. The caterpillar was placed on bareground in an area with very few Dryas patches and,after a 10 min acclimation period, its location markedevery 2 min until it stopped moving for 10 min, founda patch of Dryas, or, in 1998, for a maximum of 50min. I measured the distance and compass directionbetween successive positions, and these data were usedto calculate a caterpillar’s net displacement as a func-tion of time.

To assess whether caterpillar movement could be de-scribed as a correlated random walk, I examined therelationship between mean squared displacement(MSD) and time. The correlated random walk modeldescribes a situation in which there is neither corre-lation within or among move lengths or turning angles;both distance and direction of moves can be representedas random draws from frequency distributions of movelengths and turning angles (Kareiva and Shigesada1983, McCulloch and Cain 1989, Cain 1990). Observedvalues of MSD vs. time can be compared to expectedvalues under the assumption that movement fits a cor-related random walk. Movement paths can be describedby move lengths (l) for set time intervals and by clock-wise turning angles (u) between moves. Under a cor-related random walk model the expected mean squareddisplacement after n moves is then:

nc 1 2 c2 2E(MSD) 5 nE(l ) 1 2E(l ) n 21 21 2 c 1 2 c

where E(l2) is the expected value of l2, E(l ) is theexpected value of l, and c is the expected value of cosu(Kareiva and Shigesada 1983). Expected values are cal-culated as the means over all moves and all individualsincluded in an analysis. This model is of particularinterest because when movement fits a correlated ran-dom walk, MSD can be extrapolated as a linear functionof n, thus allowing characterization of movement be-


yond the time period examined (Kareiva and Shigesada1983).

Because the length of observations varied, I had datafrom few individuals for the later time periods (.50min). For this reason, my analysis only included datafor the 11 caterpillars followed for a full 50 min.

To further explore distance moved and the persis-tence of movement direction, I examined both clock-wise turning angles and move lengths. For each indi-vidual I calculated the mean move length and the meancosine of the turning angle for 10 min intervals. Thecosine of the turning angle is a useful measure of per-sistence of direction, with values near 1 indicatingsmall turning angles and a high degree of persistence.Values near 21 indicate frequent reversal of direction(Turchin et al. 1991). I used likelihood ratio chi-squared tests to examine these data for independencein the direction of successive turns (e.g., whether thedirection of a turn at one time is independent of theturn direction at the previous time); if right turns aremore often followed by left turns and vice versa, per-sistence of movement direction is increased. Parame-ters were examined for 10 min intervals both to aidein comparisons between the life stages and becausesmaller time intervals make it difficult to capture theoverall pattern of movement.

Adult movement

Adult female immigration.—I could not accuratelycensus unmarked females immigrating into host plantpatches because female moths are inconspicuous, in-active for much of the day, and prone to hiding in thevegetation. Therefore, I measured immigration indi-rectly by censusing caterpillars one year after removingthem from plant patches.

On 28–30 May 1994, I placed 12 cm high plasticbarriers around 20 experimental patches distributedthroughout an area dominated by smallish (,2 m2),discrete Dryas patches. These patches were classified(using the method described for the caterpillar immi-gration experiment) as either isolated (7 patches) orsurrounded (13 patches) and were randomly assignedto either a treatment that allowed immigration of flyingfemales only, or one that allowed immigration by walk-ing and flying females. I hand-removed caterpillarsfrom these patches on 11, 15, and 17 June. On 17 JuneI trimmed patches to an approximately 50 3 50 cmcircle and sprayed them with Malathion 50 Plus (3.91mL/L water). This first spraying resulted in some knockdown without kill (this had not been a problem whenspraying earlier instars as in the caterpillar immigrationexperiments), so I sprayed all patches again on 21 Junewith a higher concentration of Malathion (7.82 mL/L).

On 16 July when caterpillar activity had ended forthe year and adult activity had begun, I removed bar-riers from half of the patches to allow female movementinto these patches by walking as well as flight. I placeda Tanglefoot band around the top of each remaining

barrier. Barriers were removed after all Itame activityended for the season.

I replaced barriers as soon as I arrived at the fieldsite in 1995 (15–16 May). First instar caterpillars werepresent in the field on these dates. Each barrier wastopped with a band of Tanglefoot. I then counted andremoved all caterpillars on the patches on 13 June andagain on 17–18 June when caterpillars were predom-inantly fourth and fifth instars. I summed the countsfrom the two collection dates to obtain my estimate ofcaterpillar numbers.

Number of caterpillars was analyzed by ANOVAwith effects of patch isolation, barriers to walking, andthe interaction of these two effects.

Adult female emigration.—I tested whether femalemoths emigrate from patches by placing 10 femalemoths on each of six patches on the morning of 10 July1995. Unmarked females were not observed on the ex-perimental patches. The experimental moths had beenreared from fifth instar larvae in captivity, were ,5 dold, and many were virgins although some may havemated while in captivity. These moths were markedwith fluorescent powder which did not appear to affecttheir movement. I censused the patches using a UVlamp on the nights of 11–13 July. During the earlyevening of 12 July, I placed eight captive-reared mothsin each of six more emigration patches. These patcheswere then censused the following two nights, 13 and14 July. I counted the numbers of marked individualsremaining in the experimental patches at each census.

Individual female movement and oviposition behav-ior.—In 1995 I followed individual female moths todocument their movement and oviposition behaviorand to examine the demographic spread and mixing ofpopulations through adult female movement. All of theobservations were conducted in areas where there weremany small and medium sized patches (0.10–2 m2)interspersed with bare ground.

I performed two types of observations. First, I didindividual ‘‘long observations’’ on 31 females (bothcaptive-reared and wild) marked with dayglo fluores-cent powder. I tried to follow each of these females fora minimum of 40 min, but some became inactive orwere lost prior to that point; the average length of ob-servation was 51 min. All females were released onhost plant patches, usually in the morning when fe-males were inactive. Observations began on an indi-vidual female after she became active.

I used a Tandy TRS-80 computer (Model 102; TandyCorporation, Fort Worth, Texas, USA) to record be-havioral data and the time of each event for all obser-vations. I recorded when a female was on Dryas vs.the ground and each transition between substrates, be-havior (sitting, walking, flying, oviposition), and herposition at 10 min time intervals. I then measured thedistance and compass direction between successivepoints and the length of each flight. Although eggs aredifficult to see during oviposition, I was able to ac-


curately distinguish actual oviposition from probing byclosely watching the females’ behavior and body move-ments.

I used data from the long observations to examinenet displacement over time, movement parameters, andoviposition rates. My interest was in characterizingmovement of active females. For this reason, I excludedfour moths from the analysis; one appeared injured,and the other three all had long periods of inactivity.I also excluded all terminal time intervals where fe-males were inactive for more than 70% of the interval.

To test whether adult female movement could beadequately described as a correlated random walk, Iexamined the relationship between mean squared dis-placement (MSD) and time and compared it to the pre-dicted relationship under the correlated random walkmodel. As with the caterpillars, the length of obser-vations varied, and I had data from few individuals forthe later time periods. For this reason my analyses ex-amined data from: (1) 16 females followed for a full40 min, and (2) 10 females followed for a full 60 min.As noted earlier, whether the observed and expectedrelationships match is of interest because MSD can beextrapolated as a linear function of time if movementfollows a correlated random walk.

For each individual I calculated the mean movelength and mean cosine of the turning angles. I usedlikelihood ratio chi-squared tests to examine for in-dependence in the direction of successive moves. I alsocompared move lengths for 10 min periods spent com-pletely on Dryas to those spent completely on theground. For all females that spent time both on and offof Dryas patches (n 5 23), I calculated the total ob-servation time spent on vs. off and the percent of timeon and off that was active (inactivity was defined asrests of 3 min or more; shorter rests are common afteroviposition events).

To examine oviposition rates I calculated the numberof eggs laid per unit of time, as well as the number ofeggs laid per move length. I also calculated the numberof eggs laid per unit time on Dryas and on the groundfor all females that moved on and off of Dryas.

The second type of observation was shorter, lastinga maximum of 15 min. I located 32 unmarked activefemales in the wild and followed them to gain moreinformation on the frequency and length of flights.These data were combined with those from the longobservations to characterize flight behavior.

RESULTS

Caterpillar movement

Caterpillar immigration and patch isolation.—In1994 no immigrants were found in extirpated patcheson the first census date suggesting that censusing beganprior to immigration. Immigration had diminished bythe last census date although a few new arrivals werestill found on this date. In 1995 a few individuals were

found on the first census date suggesting that immi-gration had recently begun and one individual wasfound at the last census.

In 1994 I found a total of 116 caterpillars in 20patches at two sites. In 1995 I found 185 caterpillarsin 45 patches at three sites. In both years the majorityof immigrants were fifth instars (83% in 1994, 82% in1995). In 1994 I found one, and in 1995 two, thirdinstar immigrants. The remaining individuals were allfourth instars. This indicates that the majority of cat-erpillar movement is carried out by fifth instars.

The majority of patches received very few immi-grants (Fig. 1). I found no more than 8 immigrants ina patch, except at the high density GL site where a fewpatches had high numbers of immigrants in both 1994and 1995.

In 1994 there was a significant effect of site and asignificant site by isolation interaction (Table 2). Con-trasts revealed a significant effect of isolation at theGL site (Fig. 1b, F1,16 5 14.01) where isolated patchesreceived fewer immigrants, but the isolation effect wasnot significant at the IN site (Fig. 1a, F1,16 5 1.70) withexperimentwise a 5 0.05.

In 1995 there were again significant site and site byisolation effects (Table 2). The patterns in this yearwere less clear than in 1994. At the GL site the veryisolated patches received few immigrants, but at leastone of the isolated patches received a great many col-onists (Fig. 1e). No clear patterns emerged at either theIN or OR sites, although generally patches receivedfew immigrants (Fig. 1 c and d). Contrasts revealedthat the only significant isolation effect was betweensurrounded and very isolated patches at the GL site(F1,36 5 13.17).

In 1994, the sex ratio of emerging Itame was 45%male and 55% female (n 5 47, mostly from GL site).These values are almost identical to those for a muchlarger sample (n 5 206) reared from caterpillars col-lected at the GL and IN sites in 1994 (44% male, 56%female; unpublished data).

Caterpillar emigration.—My manipulations of cat-erpillar numbers for this experiment were successfulin creating three distinct ranges of starting densities(Table 3). These densities were higher than averagedensities at many of my sites under natural conditionsbut within the range of naturally occurring densities(Doak 2000).

Density treatment, which combines effects of feed-ing damage and actual caterpillar density, had a strikingeffect on emigration rate (Wilcoxon x2 5 27.37, df 53, P 5 0.0001; log-rank x2 5 29.62, df 5 3, P 5 0.0001;Fig. 2). Pairwise treatment comparisons revealed thathigh density patches displayed a significantly differentsurvivorship function than control, low, and mediumdensity patches (Z 5 5.40, Z 5 4.05, and Z 5 3.81respectively) with an experimentwise error of a 5 0.05.While high density patches display more rapid loss ofindividuals, differences between the other treatments


FIG. 1. The number of immigrants per ex-perimental patch in the caterpillar immigrationexperiment. The IN, OR, and GL sites have low,medium, and high background densities respec-tively.

TABLE 3. Initial densities of caterpillars in emigration ex-periment patches.

Density treatmentDensity (individuals/m2)

6 1 SE

LowMediumHighMovement barrier controls

37.47 6 2.4168.30 6 3.88

106.79 6 5.6376.96 6 7.56

TABLE 2. ANOVA for number of caterpillar immigrants (logtransformed) vs. site and patch isolation in 1994 (r2 5 0.59)and 1995 (r2 5 0.40).

Source

1994

df F P

1995

df F P

SiteIsolationSite 3 IsolationError

111

16

10.313.10

12.87

0.00540.09750.0025

224

36

3.401.632.99

0.04440.20930.0316

are not significant (Fig. 2). Tests for association be-tween the survivorship function and the temporal andspatial blocking factors indicate that while temporal sethas a significant association (Wilcoxon x2 5 7.26, P5 0.007), spatial block does not (Wilcoxon x2 5 3.21,P 5 0.073). Despite this effect of set, by combiningdata across the two temporal sets, my tests for densityeffects incorporate more noise and thus are conserva-tive.

Individual caterpillar movement.—I observed notendency for caterpillars to orient to Dryas patches (orany other objects) even at a distance of a few centi-meters; it appeared that they had to touch their hostplant before detecting it.

Comparison of observed and expected MSD for the11 individuals observed for a full 50 min (Fig. 3a)shows that the observed values fall beneath the ex-pected for all time periods. This indicates that cater-pillars moved less from their initial positions thanwould be expected by a correlated random walk. Theresults from the analysis with all individuals were qual-itatively indistinguishable. The low values of MSDmost likely resulted from the tendency for many of theindividuals to turn back on their paths.

Caterpillar move lengths for 10 min periods are dis-tributed between 0 and 180 cm (Fig. 4a) with a meanof 74.30 6 7.67 cm (n 5 20). Turning angles for 10min intervals were fairly evenly distributed betweensmall and large turns (Fig. 4c). The mean cosine of theturning angle was negative for 8 of 18 individuals(where adequate data existed) and the overall mean was0.14 6 0.10 (n 5 18) indicating an average turn of 82degrees and a tendency for their paths to spiral backon themselves. When all turns for all individuals arepooled, the directions of successive turns were not in-dependent with turns being autocorrelated (right–right29, right–left 12, left–right 12, left–left 18; G 5 6.76,P 5 0.009), which further indicates a tendency to turnback on their paths.

Adult movement

Adult female immigration.—Immigration of adult fe-males into extirpated patches (quantified as: ln[numberof caterpillars subsequently found in patch]) was sig-nificantly affected by both patch isolation and the pres-ence of a barrier to walking, although the interactionof these two factors was not significant (Table 4).Patches without barriers had significantly more cater-pillars than did patches with barriers, confirming that


FIG. 2. The mean proportion of individuals remaining in emigration patches 0–4 d after the start of the experiment.

movement by walking is an important alternative toflight in females (Fig. 5). Furthermore, isolated patchescontained more caterpillars than surrounded patches,an unexpected result (Fig. 5). Isolated patches were notcongregated in one region so spatial proximity and lo-cal conditions cannot explain this result.

Female emigration.—In the first set of emigrationtrials, all females disappeared from their release patchwithin three days. The second set was only censusedfor two days but at that time the majority of femaleshad disappeared; one patch had two females and twopatches had one female remaining when last censused.Although predation, as well as emigration, could con-tribute to this disappearance, these results strongly sug-gest that females leave their patches of origin.

In the first trial 23 of the 29 females that remainedin patches on the first night were copulating when Isurveyed the patches. None were seen copulating insubsequent surveys. This suggests that females matebefore moving. Often when I released virgin females,swarms of males arrived immediately after release andcopulation began within minutes.

Individual female movement and oviposition.—1. MSD.—Comparison of observed and expected

MSD for both the 16 females observed for 40 min andthe 10 females observed for 60 min, revealed that theobserved values were consistently higher than the ex-pected line at later time intervals (Fig. 3b). This in-dicates more persistence of direction than would beobtained from a correlated random walk and presentsa strikingly different pattern than seen for caterpillars(Fig. 3a).

2. Movement parameters.—Move lengths for 10 minperiods tended to be short (Fig. 4b), and mean movelengths for individuals did not differ significantly be-tween adult females (76.81 6 9.90, n 5 30) and cat-erpillars (74.30 6 7.67, n 5 20) (t test on ln trans-formed data: t45 5 0.51, P 5 0.61). An ANOVA ex-

amining only moves that were completely on or off ofDryas revealed significant effects of both substrate(F1,31 5 21.81, P , 0.001) and moth identity (F20, 31 52.85, P 5 0.004) on move length. The least squaresmean move length per 10 min interval on Dryas wasonly 23.34 6 0.98 cm (mean 61 SE, n 5 30) whilethat on the ground was 70.39 6 1.52 cm (n 5 23).

Turning angles were skewed towards smaller angles(Fig. 4d), and comparison of mean cos(turning angle)for individual female moths (0.65 6 0.06, n 5 26) andindividual caterpillars (0.14 6 0.10, n 5 18) indicatesthat females make significantly smaller turns (t42 54.72, P , 0.0001). This likely contributes to the per-sistence of direction and the greater than expected val-ues for MSD. If females alternated left and right turnstheir movement paths would be even straighter; how-ever, there was no evidence of nonindependence ofturning directions when all data on adjacent turns weregrouped (G 5 0.188, P 5 0.664).

In the patchy landscape where these observationswere made, there was no evidence that females spentmore time on vs. off of Dryas patches (Table 5; pairedt test: t22 5 1.006, P 5 0.325). However, on average96% of the time on the ground was active while 86%of the time on patches was active; this difference issignificant (Mann Whitney U 5 157.50, P 5 0.009).

The contribution of flight to female movement wassmall. When all moths from the short and long obser-vations are considered, 70% (44 out of 63) of femalesdid not fly during my observation periods and thosethat did fly did so infrequently (Fig. 6a). Furthermore,flights were quite short, with 85% of them being nomore than 1 m (Fig. 4d). I never observed femalesflying higher than 25 cm above the ground. Thus flightdid not account for the majority of female movement.Although infrequent longer flights could have signifi-cantly impacted overall movement, I did not observeeven rare long distance flights. I lost no females during


FIG. 3. Mean squared displacement plotted against timefor (a) 11 caterpillars observed moving on bare ground and(b) adult females observed moving on a mixture of Dryasand bare ground. Dotted lines connect observed values; solidlines gives the expected MSD.

TABLE 4. Adult female immigration experiment: ANOVAfor number of caterpillars found per patch vs. patch iso-lation and whether or not the patch had a barrier to adultwalking in the previous season.

Source df F P

IsolationBarrierIsolation 3 BarrierError

111

16

7.549.161.89

0.0140.0080.188

Note: r2 5 0.55.

FIG. 4. Movement parameters: ten-minutemove lengths for (a) all moves by caterpillarsduring the first 50 min of observation and (b)all moves by adult females during the first 60min of observation; turning angles between suc-cessive moves for (c) all turns by caterpillarsduring the first 50 min of observation and (d)all turns by adult females during the first 60 minof observation.

flights, thus there should be no bias towards recordingonly short flights.

3. Oviposition rate.—On average more than twice asmany eggs were laid per unit time on bare ground thanon Dryas (Table 5; paired t test: t22 5 26.568, P ,0.0001). This surprising difference cannot be explainedby the greater inactivity of females on Dryas, becausethe difference in oviposition rate was still significantwhen oviposition rate was calculated only for activeperiods. Calculated over all females and for all 10 mi-nute intervals, eggs were laid at rates of 0.38 eggs/minand 7.72 eggs/m (note that these are meters travelednot displacements from the point of origin).

The dispersion of eggs over space is of great interestbecause it describes the demographic mixing of a pop-ulation. While I cannot directly calculate this, I canestimate lower and upper bounds on egg dispersion bythe average female. It is clear from my analyses, thatfemales move more than what is predicted by a cor-related random walk (Fig. 3b), so extrapolating dis-placement as a linear function of time (correlated ran-dom walk model) will underestimate movement dis-tance and give a lower bound on displacement. A re-lationship between oviposition and distance can thenbe obtained by combining information on the squareddisplacement of moving females (m2/unit of time) andthe number of eggs laid (eggs/unit of time). For this,


TABLE 5. Comparison of time and oviposition on and offof Dryas patches for 23 females that spent time on bothsubstrates.

Substrate Percent timePercent time

active Eggs/min

DryasGround

0.54 6 0.040.46 6 0.04

0.86 6 0.030.96 6 0.02

0.24 6 0.040.60 6 0.06

Note: Values are means 6 1 SE.

FIG. 6. Adult female flight: (a) the frequency of flightsby 63 females that flew during my observations and (b) thefrequency of flight distances for 39 flights made by 19 fe-males.

FIG. 5. Results from the adult female immigration ex-periment showing the mean number of caterpillars found ineach patch type. Patches are either surrounded or isolated andeither open to adult walking or with a barrier. Means are back-transformed least-squares means (61 SE) from an ANOVA.

I used the lowest estimate of MSD (regression forcedthrough the origin for 10 females observed for 50 min:MSD 5 0.082 m2/min) and the average number of eggslaid per minute (0.38 eggs/min) to arrive at 4.60 eggs/m2. Fig. 7 shows this lower bound as the expectedradius within which the average female spreads hereggs, based on an average egg load of 120 eggs perfemale (Doak 1997). I calculated an upper bound onegg dispersion by taking the average number of eggslaid per distance traveled (on a per move basis) andthen assuming that females moved in a straight line(clearly an overestimate of average displacement) togive the relationship 7.72 eggs/m; this relationship isalso shown in Fig. 7. These two types of estimatesassume very different behaviors and thus describe dif-ferent types of egg dispersion. The first assumes thateggs are dispersed within a circular radius of a givensize, whereas the second assumes that they are laidalong a linear path. Despite this difference, both esti-mates provide information about the maximum dis-tance from a starting point over which eggs are ex-pected to be laid.

DISCUSSION

The scale of movement and oviposition

Caterpillar movement.—Individual caterpillars onbare ground move less than pedicted by a correlated

random walk. Given that areas occupied by Dryas aresurrounded by vast areas of bare ground, river, andglacier, wandering away from Dryas could be risky.The large turning angles shown by caterpillars (Fig.4c) are consistent with the behavior of area restrictedsearch (Kareiva and Odell 1987), which will tend tokeep caterpillars from leaving areas of host plants.Thus, it appears that most movement by caterpillarswill lead to relatively small scale mixing between hostplant patches.

Most movement by larvae involves later instars. Thiscould result from an increased capacity for movement,behavioral changes, or increasing resource depletion atlater instars. Movement by first instars dispersing fromhatching sites to host plants, while not formally in-vestigated, must occur in caterpillars hatching fromeggs laid off of patches. I saw no evidence in the fieldor lab that this species balloons, and in the absence ofballooning, it is unlikely that first instar caterpillarsmove long distances in their search for host plants.Fifth instar caterpillars looking for pupation sites willalso contribute to movement, however these individ-uals will take the first suitable pupation site found. Idid not observe any long movements prior to pupation.


FIG. 7. The percentage of eggs remaining within variousdistances from the starting position for a female starting withthe mean egg load of 120 eggs. The solid line shows theexpected numbers based on the most conservative estimateof displacement from the linear regression for 10 individualsfollowed for 60 min (4.6019 egg/m2); the dotted line is aliberal estimate based on the mean number of eggs laid permeter and the mean move length, assuming that females movein a straight line (7.7216 eggs/m).

Given that 55% of the caterpillars reared from im-migration patches were females and that a female cat-erpillar carries her full reproductive potential with her,much of this caterpillar movement will contribute todemographic mixing within sites, especially at highcaterpillar densities. Rare incidents of long distancecaterpillar movement may lead to larger scale mixingand colonization of new sites, but it appears unlikelythat caterpillars move on a large enough scale to joinneighboring sites into a single demographically (as op-posed to genetically) intermixing population.

Adult female movement and oviposition.—Unlikecaterpillars, all adult females move and contribute topopulation mixing through the spreading of eggs. Fe-males mate at or near their site of emergence and thenmove as they lay eggs. Because Itame females do notfeed or even search for places to drink, these behaviorsdo not lead to greater movement within or beyond thelarval habitat.

Females lay eggs over a relatively short time anddistance. Assuming a constant rate of oviposition, afemale’s full egg complement, which ranges from 30–200 eggs (Doak 2000), can be laid in as little as 1.5–9 h (5.3 h for the mean load of 120 eggs). Furthermore,although eggs are clearly spread beyond the immediatearea of emergence, they do not appear to be spreadvery far. Based on my estimates of movement and ovi-position parameters, the mean radius of egg dispersionis between 5 and 15 m. These estimates are based onthe assumptions that the rate of egg laying and the rateof movement do not change as females age and reducetheir egg loads. However, I cannot rule out the possi-bility that behavior changes with decreasing egg num-bers as is suggested for other species (Chew and Rob-bins 1984, Warren 1987, Rudd and McEvoy 1996). Achange in behavior as egg load declines could lead towider spreading of eggs, although it still appears thatthe majority of eggs will be laid near the emergencesite.

Although flights could greatly increase the move-ment radius and thus the dispersion of eggs, I neverwitnessed females making long flights or even repeatedshort flights, nor did I observe females that had dis-persed far from suitable larval habitat. Thus a separatebehavioral phase of long distance dispersal appears un-likely. When in flight, females stayed close to theground and always appeared clumsy and overladen.The fact that female immigration is significantly de-creased by plastic barriers only 12 cm in height (Fig.5) further indicates that flight is limited and that walk-ing is the predominant mode of movement for thesemoths.

In sum, it appears that Itame females may commonlymove on the order of tens of meters over their lifetime.Thus, the movement of Itame females is quite restrictedcompared to that seen in most Lepidoptera capable offlight. Although it would be most informative to com-pare Itame’s dispersal to taxonomically or ecologically

similar species, the relevant data are not available. Oth-er studies of Lepidoptera inhabiting subdivided sys-tems find maximum generational movement distanceson the order of hundreds of meters (Thomas and Har-rison 1992, Warren 1987, Thomas et al. 1992, Hanskiet al. 1994, Peterson 1994, Harrison et al. 1995, Ruddand McEvoy 1996, Sutcliffe et al. 1997) to severalkilometers (Thomas et al. 1992, Harrison 1989, Neveet al. 1996). Dispersal is often deemed limited even forspecies capable of movements of 0.5–2.5 km (e.g.,Thomas and Harrison 1992, Warren 1987, Thomas etal. 1992, Peterson 1994, Kuussaari et al. 1996). Al-though what constitutes limited dispersal will dependheavily on the spatial organization of habitat, it seemsclear that Itame’s dispersal is very limited comparedto many other Lepidoptera.

The scale of population processes and the role ofdifferent habitat types

Site scale.—Movement data suggest that fully inter-mixing populations are confined within sites. Demo-graphic mixing between neighboring patches withinsites is common for Itame and is largely due to theshort distance movement of females. The movement ofcaterpillars also leads to mixing on this scale and willhave a larger contribution at high population densitiesdue to density-dependent emigration. In contrast, mydata indicate that movement between sites separatedby longer distances (e.g., .100 m) and obstacles tomovement (e.g., other vegetation types, wide expansesof bare ground, gravel ridges) is likely rare. Sites prob-ably are occupied by demographically (but not genet-


ically) distinct populations or perhaps subpopulationswithin a metapopulation.

The movement of adult males, although not studiedin detail, is much more widespread and will almostcertainly lead to genetic mixing on the larger, between-site scale (Doak 1997). Male movement will influencepopulation dynamics when males are limiting or ge-netics are important (Gilpin 1991, Hastings and Har-rison 1994, Hedrick 1996). However, when inbreedingdepression is not a concern, investigation of femalemovement alone will likely capture the important pro-cesses influencing spatial dynamics and populationstructure. Given the high mobility of male Itame, it isunlikely that inbreeding or male limitation are concernsin this system.

These findings are concordant with results of mystudies of Itame population dynamics. I found thatshifts in Itame densities on host plant patches withina site are highly synchronous over years while dynam-ics between sites are asynchronous (Doak 2000). Lim-ited dispersal between asynchronously fluctuating sitesmay be important in maintaining metapopulations ona regional scale through colonization and extinction ofsites or the input of colonists into low density or mar-ginal sites.

Patch scale.—Although Itame movement linksDryas patches on a site-wide scale, small scale sub-division in the system still has the potential to influencepatterns of Itame distribution and abundance. My ear-lier work (Doak 2000) suggests that regions of sub-divided habitat may support more individuals than con-tinuous habitats even though bare ground occupies alarge part of subdivided habitats. Furthermore, most ofthe sites harboring Itame populations have at least someregions of subdivided habitat, whereas many areas cov-ered by large carpets of Dryas appear to be totallyunoccupied by Itame (personal observation). Thus thesmaller patches which form subdivided habitats arelikely disproportionally important in sustaining Itamepopulations. The current investigation of movementand oviposition patterns sheds more light on the pos-sible mechanisms underlying these results.

My observations of adult female oviposition behav-ior indicated that more oviposition occurred on theground than in Dryas patches. This pattern may leadto high caterpillar densities in smaller patches becausecaterpillars hatching in the proportionally large sur-rounding area of bare ground will move into thesepatches. Thus population densities may be highest insites with subdivided habitat. These high density areasmay further contribute to population dynamics throughdensity-dependent caterpillar movement as seen in boththe emigration and immigration experiments.

Isolated patches may also be important in providingdispersers through density dependent processes. Re-sults of the female immigration experiment indicatedthat isolated patches had the highest caterpillar den-sities in the season following experimental manipula-

tion (Fig. 5). This suggests that more oviposition tookplace in isolated patches. Although barriers were notreplaced until just after first instars began to hatch, itis unlikely that immigration of first instars was re-sponsible for the patterns seen. Patches with barriersto adult walking had significantly fewer caterpillarsthan patches without barriers which would not be ex-pected if immigration of first instars was driving thepattern. While the result of higher densities in isolatedpatches is in contrast to the findings of my densitysurveys (Doak 2000), it is likely that they result fromthe high background Itame densities at the immigrationexperiment site. My earlier data also suggest that asdensities get very high, isolated patches display dis-proportionally large increases in density (Doak 2000).

In summary, the results from both my observationsof female oviposition and the female immigration ex-periment indicate that smaller patches and subdividedhabitat are likely to have higher population densitiesand may contribute disproportionally to mixing andspread of populations through density-dependent cat-erpillar emigration. Pokki (1981) finds a similar effectin a field vole metapopulation inhabiting an archipel-ago. In the Dryas–Itame system, smaller patches andsubdivided habitat may be especially important in thedynamics of populations and metapopulations and mayserve to maintain and to stabilize populations in lesssuitable sites.

Regional scale.—The absence of Itame from the vastmajority of host plant habitat could be due to limiteddispersal leading to limited colonization or to the factthat Itame is excluded from many sites. If long distancemovement of females was common, I would expect tosee more populations (even if transient ones) in theKennicott River and McCarthy Creek valleys. The factthat only scattered populations exist along an ;3 kmstretch of the east side of the Kennicott Glacier Valleyand that no populations are found along the west sideof the Kennicott Valley or any along the margins ofMcCarthy Creek, strongly suggests that new sites areseldom successfully colonized for even a single season.Occupied sites in the Kennicott Valley cannot be con-sidered anomalous because a population of Itame existsat the foot of the Nizina Glacier, in similar habitat.Limited dispersal is further suggested by the fact thatI have established persistent populations ($5 yr) inpreviously unoccupied sites along McCarthy Creek andthe Root Glacier (unpublished data). Populations maybe excluded from some areas by biotic (e.g., predationby ants in later successional sites, secondary plantchemistry) or abiotic (e.g., periodic flooding) factors.Even so, if long distance dispersal was common, Iwould expect to find at least temporary populations ofItame at many more sites. In sum, my observationssuggest that Itame is failing to colonize suitable siteswhere populations could persist.


Conclusions

Pattern of distribution and the dispersal capabilitiesof Itame are quite different from those of most Lepi-doptera studied in the context of habitat patchiness.Most other studies find routine movement over muchlonger distances as well as more predictable occupationof habitat patches within a species’ dispersal range.Itame’s pattern of patchy occupation even within hostplant patches most closely resembles that described byHarrison (1994) for a tussock moth with totally flight-less females. She has shown that low dispersal is im-portant in limiting the spread of this species (Harrison1994). The resemblance of distribution patterns in mysystem to those described for a fully flightless specieshighlights the likely importance of limited dispersal incontributing to distribution patterns for Itame.

The limited movement of Itame coupled with thepatchy nature of its habitat does appear to impact itspatterns of abundance and distribution. In the Itame–Dryas system, restricted movement demographicallyisolates populations on a site scale. Limited between-site dispersal may link these populations into meta-populations and thereby contribute to regional persis-tence and stability. Preferential oviposition on bareground as well as the accumulation of eggs in isolatedpatches under some conditions likely contributes tohigh Itame densities in subdivided habitats. Due to highpopulation densities and density-dependent emigrationof caterpillars, subdivided regions of Dryas will strong-ly contribute to population mixing within sites and oc-casional movement between sites. Thus areas of sub-divided Dryas may act as sources in maintaining pop-ulations in marginal habitats or contributing coloniststo new sites. Finally, low dispersal will contribute todetermining Itame’s patchy distribution on a regionalscale. Given Itame’s low dispersal abilities, historicaloccupation patterns are predicted to be important indetermining current distribution with areas in closeproximity to past populations being more likely to becolonized than areas isolated by even minor barriers.Thus, although both caterpillars and adult femalesmove frequently and contribute to demographic mix-ing, this movement is not great enough to negate theimpact of habitat patchiness on Itame populations.

ACKNOWLEDGMENTS

This manuscript has benefited from the thoughtful com-ments of M. Geber, D. Doak, A. Power, R. Root, D. Winkler,S. Juliano, and four anonymous reviewers. D. Doak and B.Morris provided invaluable, on-the-spot advice in McCarthy.I especially thank M. Geber for her support and encourage-ment. Both Tully and Dan provided very helpful enthusiasmthroughout this project. I thank C. McCulloch for his statis-tical advice. Funding was provided by a NSF Graduate Fel-lowship to P. Doak, a NSF Dissertation Improvement Grant9311208 to M. Geber and P. Doak, Sigma Xi, and the Ex-plorers Club.

LITERATURE CITED

Antolin, M. F., and D. R. Strong. 1987. Long-distance dis-persal by a parasitoid (Anagrus delicatus, Mymaridae) andits host. Oecologia 82:162–165.

Cain, M. L. 1990. Models of clonal growth in Solidago al-tissima. Journal of Ecology 78:27–46.

Chew, F. S., and R. K. Robbins. 1984. Egg-laying in butter-flies. The Symposium of the Royal Entomological Societyof London 11:65–80.

Doak, D. F., P. C. Marino, and P. M. Kareiva. 1992. Spatialscale mediates the influence of habitat fragmentation ondispersal success: Implications for conservation. Theoret-ical Population Biology 41:315–336.

Doak, P. 1997. Impacts of natural habitat subdivision on aninsect herbivore: population dynamics, dispersal, and host–parasitoid interactions. Dissertation. Cornell University,Ithaca, New York, USA.

Doak, P. 2000. Natural habitat patchiness and the distribution,abundance, and population dynamics of an insect herbivore.Ecology 81:1842–1857.

Fahrig, L., and J. Paloheimo. 1988. Effect of spatial arrange-ment of habitat patches on local population size. Ecology69:468–475.

Fox, G. A. 1993. Failure-time analysis: emergence, flower-ing, survivorship, and other waiting times. Pages 253–289in S. M. Scheiner and J. Gurevitch, editors. Design andanalysis of ecological experiments. Chapman and Hall,New York, New York, USA.

Gerber, A. S., and A. R. Templeton. 1996. Population sizesand within-deme movement of Trimerotropic saxatillis(Acrididae), a grasshopper with a fragmented distribution.Oecologia 105:343–350.

Gilpin, M. 1991. The genetic effective size of a metapopu-lation. Biological Journal of the Linnean Society 42:165–175.

Hanski, I., and M. Gilpin. 1991. Metapopulation dynamics:brief history and conceptual domain. Biological Journal ofthe Linnean Society 42:3–16.

Hanski, I., M. Kuussaari, and M. Nieminen. 1994. Meta-population structure and migration in the butterfly Melitaeacinzia. Ecology 75:747–762.

Harrison, S. 1989. Long-distance dispersal and colonizationin the Bay Checkerspot butterfly, Euphydryas editha bay-ensis. Ecology 70:1236–1243.

Harrison, S. 1991. Local extinction in a metapopulation con-text: an empirical evaluation. Biological Journal of the Lin-nean Society 42:73–88.

Harrison, S. 1994. Resources and dispersal as factors limitingpopulations of the tussock moth (Orgyia vetusta), a flight-less defoliator. Oecologia 99:27–34.

Harrison, S., and C. D. Thomas. 1991. Patchiness and spatialpattern in the insect community on ragwort, Senecio ja-cobaea. Oikos 56:293–298.

Harrison, S., C. D. Thomas, and T. M. Lewinsohn. 1995.Testing a metapopulation model of coexistence in the insectcommunity on ragwort (Senecio jacobaea). American Nat-uralist 145:546–562.

Hastings, A., and S. Harrison. 1994. Metapopulation dynam-ics and genetics. Annual Review of Ecology and System-atics 25:167–188.

Hedrick, P. W. 1996. Genetics of metapopulations: aspectsof a comprehensive perspective. Pages 29–51 in D. R.McCullough, editor. Metapopulations and wildlife conser-vation. Island Press, Washington, D.C., USA.

Ims, R. A., and N. G. Yoccoz. 1997. Studying transfer pro-cesses in metapopulations: emigration, migration, and col-onization. Pages 247–265 in I. Hanski and M. E. Gilpin,editors. Metapopulation biology: ecology, genetics andevolution. Academic Press, London, UK.

Kareiva, P. M. 1987. Habitat fragmentation and the stabilityof predator–prey interactions. Nature 321:388–391.

Kareiva, P. M. 1990. Population dynamics in spatially com-plex environments: theory and data. Philosophical Trans-actions of the Royal Society of London B 330:175–190.


Kareiva, P., and G. Odell. 1987. Swarms of predators exhibit‘‘preytaxis’’ if individual predators use area-restrictedsearch. American Naturalist 130:233–270.

Kareiva, P. M., and N. Shigesada. 1983. Analyzing insectmovement as a correlated random walk. Oecologia 56:234–238.

Kindvall, O. 1996. Habitat heterogeneity and survival in abush cricket metapopulation. Ecology 77:207–214.

Kuussaari, M., M. Nieminen, and I. Hanski. 1996. An ex-perimental study of migration in the Glanville fritillarybutterfly Melitaea cinxia. Journal of Animal Ecology 65:791–801.

May, R. M., and T. R. E. Southwood. 1990. Introduction.Pages 1–22 in R. M. May and T. R. E. Southwood, editors.Living in a patchy environment. Oxford University Press,Oxford, UK.

McCulloch, C. E., and M. L. Cain. 1989. Analyzing discretemovement data as a correlated random walk. Ecology 70:383–388.

Neve, G., B. Barascud, R. Hughes, J. Aubert, H. Descimon,P. Lebrun, and M. Baguette. 1996. Dispersal, colonizationpower and metapopulation structure in the vulnerable but-terfly Proclossiana eunomia (Lepidoptera: Nymphalidae).Journal of Animal Ecology 33:14–22.

Peterson, M. 1994. The role of topography and distance inisolating populations of the Lycaenid butterfly, Euphilotesenoptes. Dissertation, Cornell University, Ithaca, NewYork, USA.

Pokki, I. 1981. Distribution, demography and dispersal ofthe field vole, Microtus agrestis (L.) , in the Tvarminnearchipelago, Finland. Acta Zoologica Fennica 164:1–48.

Reeve, J. D. 1990. Stability, variability, and persistence inhost–parasitoid systems. Ecology 71:422–426.

Rice, W. R. 1989. Analyzing tables of statistical tests. Evo-lution 43:223–225.

Root, R. B. 1973. Organization of a plant–arthropod asso-ciation in simple and diverse habitats: the fauna of collards(Brassica oleracea). Ecological Monographs 43:95–120.

Rudd, N. T., and P. B. McEvoy. 1996. Local dispersal by theCinnabar Moth, Tyria jacobaeae. Ecological Applications6:285–297.

SAS Institute 1989. SAS/STAT User’s Guide, Volume I, Ver-

sion 6, Fourth Edition. SAS Institute, Cary, North Carolina,USA.

Shaffer, M. L. 1981. Minimum population sizes for speciesconservation. BioScience 31:131–134.

Soule, M. E., and D. Simberloff. 1986. What do genetics andecology tell us about the design of nature reserves? Bio-logical Conservation 35:19–40.

Stacey, P. B., V. A. Johnson, and M. L. Taper. 1997. Migrationwithin metapopulations: the impact upon local populationdynamics. Pages 267–291 in I. Hanski and M. E. Gilpin,editors. Metapopulation biology: ecology, genetics andevolution. Academic Press, London, UK.

Stacey, P. B., and M. Taper. 1992. Environmental variationand the persistence of small populations. Ecological Ap-plications 2:18–29.

Sutcliffe, O. L., C. D. Thomas, and D. Peggie. 1997. Area-dependent migration by ringlet butterflies generates a mix-ture of patchy population and metapopulation attributes.Oecologia 109:229–234.

Taylor, A. D. 1990. Metapopulations, dispersal, and preda-tor–prey dynamics: an overview. Ecology 71:429–433.

Thomas, C. D., and I. Hanski. 1997. Butterfly metapopula-tions. Pages 359–386 in I. Hanski and M. E. Gilpin, editors.Metapopulation biology: ecology, genetics and evolution.Academic Press, London, UK.

Thomas, C. D., and S. Harrison. 1992. Spatial dynamics ofa patchily distributed butterfly species. Journal of AnimalEcology 61:437–446.

Thomas, C. D., J. A. Thomas, and M. S. Warren. 1992. Dis-tribution of occupied and vacant butterfly habitats in frag-mented landscapes. Oecologia 92:563–567.

Turchin, P., F. J. Odendaal, and M. D. Rausher. 1991. Quan-tifying insect movement in the field. Environmental En-tomology 20:955–963.

Warren, M. S. 1987. The ecology and conservation of theheath fritillary butterfly Mellicta athalia, I. Adult popula-tion structure and mobility. Journal of Applied Ecology 24:483–498.

Wilcove, D. S., C. H. McLellan, and A. P. Dobson. 1986.Habitat fragmentation in the temperate zone. Pages 237–256 in M. E. Soule, editor. Conservation biology, the sci-ence of scarcity and diversity. Sinauer Associates, Sun-derland, Massachusetts, USA.

Documents

POPULATION CONSEQUENCES OF RESTRICTED DISPERSAL FOR AN INSECT HERBIVORE IN A SUBDIVIDED HABITAT