nonrandom movement behavior at habitat boundaries in two...

Nonrandom Movement Behavior at Habitat Boundaries in Two Butterfly Species: Implicationsfor DispersalAuthor(s): L. Conradt and T. J. RoperReviewed work(s):Source: Ecology, Vol. 87, No. 1 (Jan., 2006), pp. 125-132Published by: Ecological Society of AmericaStable URL: http://www.jstor.org/stable/20068916 .

Accessed: 12/12/2012 04:08

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

.

Ecological Society of America is collaborating with JSTOR to digitize, preserve and extend access to Ecology.

http://www.jstor.org

This content downloaded by the authorized user from 192.168.52.61 on Wed, 12 Dec 2012 04:08:23 AMAll use subject to JSTOR Terms and Conditions

http://www.jstor.org/action/showPublisher?publisherCode=esa

http://www.jstor.org/stable/20068916?origin=JSTOR-pdf

http://www.jstor.org/page/info/about/policies/terms.jsp


Ecology, 87(1), 2006, pp. 125-132 ? 2006 by the Ecological Society of America

NONRANDOM MOVEMENT BEHAVIOR AT HABITAT BOUNDARIES IN TWO BUTTERFLY SPECIES: IMPLICATIONS FOR DISPERSAL

L. CONRADT1 AND T. J. ROPER

School of Life Sciences, John Maynard Smith Building, University of Sussex, Brighton, UK

Abstract. We observed meadow brown (Mani?la jurtina) and gatekeeper (Pyronia

tithonus) butterflies at habitat boundaries and observed spontaneous movements out of

suitable habitat in order to investigate such movements in relation to dispersal. We found

that butterflies of both species were aware of the position of a highly permeable habitat

boundary without needing to cross it. Nevertheless, a considerable proportion of butterflies

close to the boundary left their habitat (25-43%). Butterflies that crossed the boundary, and moved substantial distances into unsuitable habitat (up to 350 m in M. jurtina and 70

m in P. tithonus), usually returned to their original habitat patch (98-100%). Movement

trajectories, at least in M. jurtina, were significantly different from, and more directed and

systematic than, a correlated random walk. Approximately 70-80% of spontaneous move

ments into unsuitable habitat in both species were "foray" loops comparable to those

described in mammals and birds. We conclude that, since migrants seemed to have con

siderable control over leaving their patch and over their subsequent movement trajectories, chance encounter rates with habitat boundaries, and indeed habitat leaving rates, might be

less crucial in determining dispersal rates than is usually assumed. In addition, random

dispersal trajectories should not be taken for granted in population or evolution models.

Key words: correlated random walks; dispersal trajectories; emigration; habitat edges; habitat

fragmentation; movement behavior; navigation; orientation; search strategies; search trajectories.

Introduction

Dispersal has crucial implications for the dynamics

and persistence of populations, the distribution and

abundance of species, community structure, gene flow

between populations, local adaptation, speciation, and

the evolution of life-history traits, especially in frag mented landscapes (see Hanski [1998], Travis and Dy

tham [1998], and Dieckmann et al. [1999] for reviews). In fragmented landscapes, animal dispersal consists of

two component behaviors (e.g., Matthysen 2002): (i)

emigration out of an original habitat patch and (ii) sub

sequent search for a new habitat patch. With respect to the first, it has been assumed that emigration rates

out of habitat patches depend on the chance rate of

encounter with habitat boundaries (Haddad 1999,

Golden and Christ 2000). This assumption is important from a theoretical point of view because it has been

used to predict emigration rates based on circumfer

ence: area ratios of habitat patches (see Hanski 1998

for a review). It is also practically important because

it implies that modification of the ratio of habitat edge to area could be used as a conservation measure to

influence emigration rates (Stamps et al. 1987, Ries

and Debinski 2001). With respect to the second com

ponent of dispersal, it is usually assumed that dispers es search for new habitat in the manner of a correlated

random walk (e.g., Hanski 1998, Travis and Dytham

1998, Dieckmann et al. 1999, Armsworth et al. 2001,

Byers 2001). Again this assumption is crucial, since it

leaves individual dispersers only very limited scope

for controlling dispersal rates and directions (Zollner

and Lima 1999). Additionally, dispersal distances, dis

persal success, and dispersal mortality depend largely on the search strategy that individuals employ once

they have left their original habitat patch (Conradt et al. 2003, Leon-Cortes et al. 2003, Doerr and Doerr

2005).

Notwithstanding the widespread use of assumptions such as these in theoretical ecology, empirical infor

mation about dispersal is scarce and often only indirect,

relying usually on behavioral observations conducted

solely at habitat boundaries or on mark-release-recap ture (MRR) data (see Schultz and Crone [2001] for a

comparison between emigration rates predicted by

boundary responses and MRR data and Kindlmann et

al. [2004] for a comparison of empirical MRR data and simulation results based on different assumptions of

movement rules; see also Haddad [1999], Matthysen

[2002], Schooley and Wiens [2004], Doerr and Doerr

[2005] for the importance of detailed movement paths). As regards emigration, for example, it is seldom known

whether individuals that leave habitat patches are gen

uine emigrants, since their subsequent destination is

not recorded. Hence, they might return to their original habitat patch rather than settling in a new one (e.g.,

Roper et al. 2003). If individuals recognize habitat

Manuscript received 14 April 2005; revised 11 July 2005;

accepted 18 July 2005; final version received 26 July 2005. Cor

responding Editor: J. T. Cronin. 1 Corresponding author. E-mail: [email protected]

125



126 L. CONRADT AND T. J. ROPER Ecology, Vol. 87, No. 1

boundaries easily and/or return at a high rate once they have left their original habitat patch, they would have

a substantial degree of control over the decision to

emigrate, so that chance encounter rates with habitat

edges and circumference:area ratios could be poor pre

dictors of emigration rates (e.g., Matthysen 2002). As

regards the search strategy of dispersers, several alter

natives to correlated random walk are possible (Bell

1983, Zollner and Lima 1999, Doerr and Doerr 2005) and empirical evidence suggests systematic rather than

random search in at least some species (Conradt et al.

2000, 2001, Doncaster et al. 2001, Bailey et al. 2003,

Roper et al. 2003, Seymour et al. 2003, Kindlmann et

al. 2004, Doerr and Doerr 2005).

Here, we report the behavior of meadow brown but

terflies, Mani?la jurtina, and gatekeeper butterflies, Py

ronia tithonus, at habitat boundaries that posed no

physical obstacles to emigration (i.e., highly permeable

boundaries: Collinge and Palmer [2002]). We used two butterflies as model species because theoretical devel

opment in metapopulation ecology is based to a con

siderable degree on butterfly systems (see Hanski

[1998] for a review). We followed every spontaneous

movement from suitable habitat across unsuitable hab

itat and back to suitable habitat. We aimed to determine

(i) whether butterflies recognize habitat boundaries; (ii) whether butterflies that cross habitat boundaries and

leave a habitat patch are always genuine emigrants/

dispersers, or whether they sometimes return to their

habitat patch; and (iii) whether butterflies fly in the manner of a correlated random walk across unsuitable

habitat or use forms of systematic search.

Methods

Study site

The study site was in Stanmer Park, Brighton, UK.

Butterflies were observed at the boundary of a 190 m

long, 10-40 m wide patch of long grassland (Fig. 1).

This patch was L shaped, consisting of a "short leg"

(80 m, facing east) and a "long leg" (110 m, facing

south). Both legs were bordered along one side by a

high hedge or woodland and on the other side by a

large area of regularly mown, short grassland. The lin

ear extension of both legs consisted of hedges or wood

land. Other patches of long grassland occurred at dis

tances of at least 120 m from the observation patch.

Study species and habitat

Mani?la jurtina and Pyronia tithonus are both com

mon British butterflies. Their larvae feed on grassland

species, and long grassland verges are their natural hab

itat (e.g., Brakefield 1982). Therefore, we refer to long

grassland as "habitat," whereas habitats unsuitable for

the species (e.g., shrubs, hedges, woodland and short

grassland; Conradt et al. [2000]) are termed "nonha

bitat." The edge between habitat and nonhabitat is

termed the "boundary." Since the boundary between

^ ! i ; Study habitat patch

? ̂ c ; : ; ^ foray loop

Y////SS/ I maximum distance moved from habitat

Short grassland

Fig. 1. Schematic drawing of the study site (not to scale),

including a schematic example of a typical foray search loop (Conradt et al. 2000, 2001) and its related maximum distance

moved from habitat.

long and short grassland in the present study did not

pose physical obstacles for butterflies with respect to

leaving the habitat patch, it was considered as poten

tially highly permeable (relative to other typical habitat boundaries which the species encounters naturally,

such as hedges and forest boundaries; Brakefield

[1982], Buechner [1989], Fagan et al. [1999], Ries and Debinski [2001]; see Fig. 1).

Data collection

Data were collected in July and August 2003 (M.

jurtina and P. tithonus) and 2004 (M. jurtina). Obser

vations consisted of patrolling the habitat boundary from the side where it was bordered by short grassland.

Every butterfly seen within 5 m of either side of the

boundary was followed until it either moved further

than 5 m into habitat (in the same patch or in another

patch), was prematurely lost from sight, or moved fur

ther than 350 m (since the logistics to follow further

proved insurmountable). Altogether, 396 M. jurtina and

175 P. tithonus were observed of which only one M.

jurtina was lost from sight and one M. jurtina moved

further than 350 m (and was assumed to have dis

persed). Since data collection had to be opportunistic,

butterflies could not be marked. However, large pop

ulation ?sizes (estimates according to typical densities:

about 430 M. jurtina and 210 P. tithonus; Conradt et

al. [2001]), combined with the short lifespan/residence times of individual butterflies (e.g., median life span

is 7 d in M. jurtina; Conradt et al. [2001]) relative to the observation periods, made it unlikely that pseudo

replications would be problematic. For example, we



January 2006 NONRANDOM MOVEMENTS IN BUTTERFLIES 127

estimated that the probability not to have a single in

dividual replication in the sample of flight paths for M.

jurtina was 95% in 2003 and 99% in 2004; and 97% for P. tithonus in 2003 (assuming a similar life span

for P. tithonus as that reported for M. jurtina).

Data on most flight paths were collected in a rela

tively simple manner for three reasons: first, because

we were largely interested in qualitative rather than

quantitative aspects of behavior; second, in order to

achieve sufficiently large sample sizes; and third, so as

not to bias observations toward shorter flights that were

likely to be mapped in detail more completely than

longer flights. Therefore, the majority of flight paths were recorded in the following manner. Markers were

dropped at the place where a butterfly (i) was first seen

in habitat; (ii) left its original habitat patch; (iii) re

turned to habitat (whether into the same patch or an

other); or (iv) where the observation was terminated.

Where applicable, markers were also dropped at the

furthest distance that the butterfly reached from its orig

inal habitat patch (for example, for a loop-like trajec

tory, this would be the turning point of the loop). Where

butterflies Were lost, a marker was dropped at the point

of loss.

Additionally, 34 flights of M. jurtina were mapped in more detail. Since butterflies flew in stretches of

almost straight movements with distinct recurrent turn

ings, each straight stretch was defined as a move (as

recommended by Turchin [1998:128], in order to avoid the problems of over- or undersampling of paths). We

dropped markers at the end of every move and mea

sured move lengths and turning angles. Observed flight

lengths ranged from three to 30 moves. Altogether, a

pooled sample of 461 turning angles and 495 move

lengths was collected. Only flights which started in

habitat and went into nonhabitat were used, before the

start of the flight it was decided whether the flight would be mapped in more detail and such detailed flight data were collected throughout the field season, so that

these 34 flights represent a random sample of flights

into nonhabitat.

Habitat use and behavior at habitat boundaries

We noted (i) whether a butterfly was first seen in

habitat or nonhabitat; and (ii) whether it crossed the

boundary.

Movement behavior in nonhabitat

Distances moved from habitat and return to habi

tat.?For flights marked using the "simple" method,

we measured by pacing the maximum orthogonal dis

tance which a butterfly moved away from its original

habitat boundary before it returned, got lost, or mi

grated into a new patch (maximum distance moved

from habitat; Fig. 1). Note that this is not necessarily the distance between furthest turning point/point at loss

and habitat leaving point, but the most direct distance

to the habitat boundary. Pacing was calibrated in meters

and always done by the same observer (L. Conradt).

Where the analysis was based on the number of but

terflies that flew at least to a given distance, lost but

terflies could be included if their point of loss was

above the maximum distance used in the analysis (loss

es were small). We also noted, if a butterfly moved into

nonhabitat, whether it returned to the observation patch or a new patch.

Movement trajectories.?

1. Mean net squared displacement.?For 34 flight

paths of M. jurtina that were mapped in detail, the

observed mean net squared displacements were cal

culated for numbers of moves from one through 30

(i.e., the mean over all flights of the squared distance

between the butterfly's starting point and position after

a given number of moves; Turchin [1998]). Those mean

net squared displacements (for given numbers of

moves) were compared to the respective predicted

mean net squared displacements for a correlated ran

dom walk derived by a bootstrapping method (Turchin 1998), as follows. First, 15 000 pseudopaths (based on

a correlated random walk; Turchin [1998]) were sim

ulated by drawing turning angles and move lengths

randomly from our pooled sample of observed turning

angles and move lengths (/ = 461 observed number of

turning angles, m ? 494 observed number of move

lengths). For these simulated 15 000 pseudopaths, the

simulated net squared displacement and the variance

in simulated net squared displacement were calculated

for each path length of 1-30 moves. The average sim

ulated net squared displacement of all 15 000 simula

tions is the predicted net squared displacement for each

given path length for a correlated random walk. The

distribution of the predicted mean net squared displace ment over n simulated flights for a given path length can then be approximated by a normal distribution with

predicted net squared displacement as mean and vari

ance in simulated net squared displacement divided by n as variance (Zar 1984). From these normal distri

butions, 95% confidence intervals were calculated for

the predicted mean net squared displacement for each

given path length, using the observed sample sizes at

each path length as n (note that n varies for different

path lengths, because not all observed paths reached

the maximum length of 30 moves). If observed mean

net squared displacements lay outside this 95% con

fidence interval for the predicted mean net squared dis

placement, observed paths differed significantly (with P < 0.05) from the predictions of a correlated random

walk. Observed mean net squared displacements sig

nificantly larger than the predicted mean net squared

displacements indicate that observed paths are more

directed than expected by a correlated random walk.

Observed mean net squared displacements significantly

smaller than the predicted mean net squared displace ments indicate that observed paths are less directed and

more winding than predicted by a correlated random




walk (see also Turchin [1998] for a very detailed de

scription of this method).

In addition, we investigated whether observed turn

ing angles were autocorrelated (Turchin 1998) by ex

amining whether consecutive turning angles tended to

be less often to the same direction (either right or left) than expected by chance (i.e., whether there was a first

order negative autocorrelation), and whether once-re

moved-consecutive turning angles tended to be more

often to the same direction than expected by chance

(i.e., whether there was a second-order positive auto

correlation). For this purpose, we counted the observed

number of turns in the same direction (i.e., number of

occasions left turn follows on left turn plus number of

occasions right turn follows on right turn: "left-left"

turns plus "right-right" turns), and the observed num

ber of turns in opposite direction (i.e., "left-right" plus

"right-left" turns). We calculated the relevant expected numbers assuming independence of turning directions

(approximately 50% of turns were left, 50% right; thus the probability of observing left-left turns was 25%,

right-right turns 25%, left-right turns 25%, and right

left turns 25%) and compared expected and observed numbers using x2 tests. A significantly larger number

of observed than expected turns in the same direction

means positive autocorrelation of turning angles. A sig

nificantly larger number of observed than expected turns in the opposite direction means negative auto

correlation of turning angles (see also Turchin [1998]

for a detailed description of the method). A first-order

negative together with a second-order positive auto

correlation of turning angles indicates (i) that paths are

not correlated random walks; but (ii) that previous di

rectional changes are compensated for nonrandomly by

following a zigzag trajectory.

2. Self-crossing of paths and qualitative categori

zation of movement trajectories.?In addition, the flight

trajectories were categorized qualitatively by sight with

respect to (i) whether they crossed themselves; and (ii) whether they consisted of loop-like movements (in

volving leaving the habitat and then returning in a half

circle or half-elliptic "loop" back to habitat; see Fig.

1). Only movements >5 m into nonhabitat were in

cluded in this analysis. Such loop-like movements

would be expected if butterflies used a foray search

strategy for dispersal (Durier and Rivault 1999, Con

radt et al. 2000, 2001). Both these qualitative assess

ments were easy in the field because of the high visi

bility in short grassland, the manner of observations

(the observer followed the route of the butterfly from a reasonable distance; Conradt et al. [2000]), and not

least the very unambigious nature of the flights them

selves. To calculate the expected frequencies of (i) self

crossing and (ii) foray-looping flights, pseudopaths based on correlated random walks were simulated, as

described above, for M. jurtina, and assessed visually.

These simulated paths were terminated in the same

manner as the paths with which they were compared,

namely when they (i) either returned to habitat; or (ii) exceeded a distance of 150 m from their original habitat

patch (since none of the observed flights got lost until

they reached at least a distance of 150 m).

Results

Habitat use

In both species, significantly more butterflies were

first detected <5 m inside the boundary (i.e., in habitat)

than were detected <5 m outside of the boundary (i.e.,

in nonhabitat). This supports our classification of the

study area into habitat and nonhabitat (M. jurtina, 330

in habitat, 32 in nonhabitat, x? =

245.3, P <C 0.0001; P. tithonus, 167 in habitat, eight in nonhabitat, Xi

=

144.5, P << 0.0001).

Behavior at habitat boundary

In both species, butterflies first observed <5 m inside the habitat boundary were significantly more likely to

move deeper into habitat (without previously crossing the boundary) than out of habitat (M. jurtina, 187

moved deeper into habitat, 143 out of habitat, x2 =

5.9, P = 0.01; P. tithonus, 126 moved deeper into

habitat, 41 out of habitat, x? =

43.3, P <C 0.0001). Thus, butterflies seemed to recognize the boundary be

tween habitat and nonhabitat before crossing it. Nev

ertheless, a considerable proportion of butterflies left

the habitat patch at least temporarily (43.4% in M. jur tina and 24.6% in P. tithonus), and more so in M.

jurtina than in P. tithonus (x2 test for difference be

tween species, xi =

17.0, P < 0.0001).

Movement behavior in nonhabitat

Distances moved from habitat and return to habi

tat.?In M. jurtina 43% of butterflies moved up to 160

m, and in P. tithonus 25% of butterflies up to 70 m,

away from their original habitat patch into nonhabitat

(Fig. 2a and b). In M. jurtina, only 2.1% (i.e., three)

of movements out of habitat resulted in dispersal to

different habitat patches, while 97.9% (i.e., 139) of

movements ended with a return to the same habitat

patch (n =

142). Return rates to the starting habitat

patch were still very high even for butterflies that had moved a considerable distance into nonhabitat (e.g.,

70% of M. jurtina that flew further than 50 m from their habitat patch returned subsequently; Fig. 2c), and

M. jurtina that returned to their original patch had flown

up to a maximum distance of 150 m into nonhabitat

before returning (compare this to an average dispersal

distance for M. jurtina in MRR experiments of 40-70

m in a landscape consisting of a mixture of pastures

and agricultural fields; Brakefield [1982]). In P. ti

thonus, all butterflies that left the habitat patch returned

to it (n =

41), flying up to 70 m into nonhabitat before

returning. Movement trajectories.?

1. Mean net squared displacement.?The mean net

squared displacement of 34 detailed mapped flights by




1000

2 S 100

?

a. 6

6

-Z3 CD

Q. O

10

1

1000

100

10

1

1.0

0.8

0.6

0.4

0.2

-

's,

"**!

kA^AA.

AAAAAAAA

25 50 75 100 125 150

Distance from habitat (m)

Fig. 2. Numbers of (a) Mani?la jurtina and (b) Pyronia tithonus that flew to a given distance or farther into nonhabitat

(for reasons of clarity, scale is logarithmic); and (c) propor tion of M. jurtina that returned to their original patch after

they had flown to a given distance or farther.

M. jurtina (see Fig. 3 for examples) lay outside the

95% confidence interval derived by bootstrap simula

tions of correlated random walks (Turchin 1998) for

flight lengths of up to 27 moves (see Fig. 4; above 27 moves sample sizes were very small, and, thus, the

95% confidence interval too broad to allow for mean

ingful comparisons). Therefore, the observed move

ment trajectories were significantly different from, and

more directed than, correlated random walks. Addi

tionally, turning angles were significantly first-order

negatively and second-order positively autocorrelated,

suggesting that butterflies compensated for directional

changes in subsequent turns in the manner of a zigzag

flight (number of first-order left-left plus right-right

turns, 126 observed, 213.7 predicted; number of first

order left-right plus right-left turns, 301 observed,

213.3 predicted; x? =

72.0, P <C 0.0001; number of second-order left-left plus right-right turns, 230 ob

served, 196.8 predicted; number of second-order left

right plus right-left turns, 163 observed, 196.2 pre

dicted, x? = 11.2, P < 0.001).

2. Self-crossing of paths.?One of 177 observed

flight trajectories by M. jurtina, and none of 41 ob

served flight trajectories by P. tithonus crossed their

own flight path. In contrast, in 250 simulated pseu

dopaths (in the manner of correlated random walks)

for M. jurtina, 93 (37.2%) trajectories crossed their

Fig. 3. Three examples of M. jurtina flight paths that were

mapped in greater detail. Note that the scale differs among

paths in panels (a)-(c).




E, 6000 c

E 5000 CD

g. 4000

? 3000 CD g- 2000 CO CD C

c CD CD

1000

s

0 5 10 15 20 25 30

Flight length (number of moves)

Fig. 4. Mean net squared displacement as a function of

flight length (in number of moves) for detailed mapped flights of M. jurtina. Solid squares represent observed data (sample

sizes, 1-34 paths), the solid line represents simulated means,

and the dotted lines represent 95% confidence intervals de

rived from simulations. Measurement of mean net squared

displacement is explained in Methods: Movement behavior in

nonhabitat: Movement trajectories: Mean net squared dis

placement.

own flight path at least once (and usually numerous

times). The difference between observed and simulated

trajectories of M. jurtina with respect to self-crossing

was significant (binomial test, P < 0.0001), further

suggesting that movement trajectories were more sys

tematic and consequently more efficient than correlated

random walks.

3. Qualitative categorization of movement trajec

tories .?Fifty-one of 72 (70.8%) M. jurtina flight tra

jectories and 16 of 20 (80%) P. tithonus trajectories (that ventured further than 5 m from the boundary into

nonhabitat) consisted of loops which could be classified

as "foray search trajectories" (i.e., simple, half-oval

shaped loops out of, and back to, the habitat patch

[Conradt et al. 2000, 2001]; see Fig. 1 and Fig. 3a and

b). By comparison, of 100 simulated correlated random

walk pseudopaths for M. jurtina, 56 (56%) were foray

search trajectories. Thus, observed M. jurtina paths

consisted significantly more often of foray search tra

jectories than expected from a random correlated walk

(X? = 6.5, P =

0.01).

Discussion

Butterflies of both species seemed to be aware of a

boundary between suitable and unsuitable habitat with

out needing to cross it. This indicates that butterflies

of both species can actively control their rate of bound

ary crossing, and that butterflies which crossed into

unsuitable habitat did so for a purpose, presumably

related to dispersal (Conradt et al. 2000). A large pro

portion of butterflies moved, at least temporarily, out

of their habitat patch, and several butterflies moved

considerable distances from their original habitat patch

relative to their average dispersal distance. These

movements differed clearly from the typical, more fre

quently turning and slower movements within suitable

habitat (L. Conradt, unpublished data), further sug

gesting that the movements into unsuitable habitat were

not related to normal food plant search behavior (Con

radt et al. 2000, 2001). Most of the movements out of

suitable habitat resulted in a return to the same habitat

patch, and only a relatively small proportion of but

terflies dispersed to a new patch, as has been predicted

by "forced dispersal" experiments in M. jurtina (Con

radt et al. 2000, 2001) and subsequent dispersal models

(Conradt et al. 2003). Movement trajectories, at least

in M. jurtina, differed significantly from those pre

dicted by a correlated random walk. Instead, butterflies

moved in a more directed manner across unsuitable

habitat, compensated for directional changes in sub

sequent turns, and were very efficient in the sense that

paths almost never crossed themselves. All this implies

that the movements out of habitat were related to dis

persal rather than to food plant search (Conradt et al.

2000, 2001, 2003), that dispersing butterflies move sys

tematically, and that the rate of habitat return is under

active and effective control.

The notion that emigration rates depend on chance

encounters with habitat boundaries (and that, therefore,

circumference:area ratios are automatically good pre

dictors of emigration rates out of habitat patches) are

based on the assumptions that (i) the probability that an individual leaves a patch is proportional to the prob

ability that it encounters a patch boundary; and (ii) once

it has left a patch, it returns to the patch only by chance.

Both these assumptions do not hold for the observed

movements in M. jurtina and P. tithonus. If individuals,

like the butterflies in our study, can recognize bound

aries and, thus, have control over their decision to

leave, the rate at which they leave a patch will depend

mainly on their incentive to leave (e.g., on patch qual

ity, population density, or similar). The rate at which

they encounter boundaries will only play a role if hab

itat patch size is so large that individuals with the aim

to emigrate have difficulties to find a boundary (Mat

thysen 2002). Similarly, if butterflies can actively find

their way back to their patch of origin (rather than drift

around "randomly" and return only by chance), the

rate of emigration will largely depend on their incentive

to continue emigration (e.g., on the detectibility of new

patches, the hostility of the unsuitable habitat, or sim

ilar). Therefore, we suggest that movement out of suit

able habitat (and subsequent dispersal) is unlikely to

increase in a simple way with the encounter rate of

habitat boundaries (Haddad 1999, Ries and Debinski

2001, Schultz and Crone 2001), at least in our two study

species. Thus, the use of circumference:area ratios of

habitat patches in order to predict relative emigration

and dispersal rates (e.g., Stamps et al. 1987, Buechner

1989, Hanski 1998) might not be valid, at least not




generally. Matthysen (2002) made similar observations

in Blue and Great Tits. He argued that if boundary

encountering rates are crucial, emigration rates should

be higher in birds whose natal territory lay close to a

habitat boundary. However, he found that this predic tion was not supported in his study site, and suggested instead that boundary encounter rates become crucial

only when habitat patches are so large that "willing"

dispersers cannot easily detect them. Our study sug

gests further that estimates of dispersal rates from ob

servation of habitat departures can be unreliable if the

majority of habitat leavers return.

In a majority of cases of movements into unsuitable

habitat, spontaneous foray loops, similar to those de

scribed for foray search dispersal (Conradt et al. 2000,

2001, 2003), were observed in both butterfly species. This supports observations previously made in "forced

dispersal" experiments (Conradt et al. 2000, 2001). It

suggests that M. jurtina and P. tithonus habitually re

connoiter the landscape surrounding their habitat patch,

presumably as part of dispersal attempts (Conradt et

al. 2000). Predispersal explorations have also been de

scribed in mammals (Roper et al. 2003, Haughland and

Larsen 2004) and birds (Walters et al. 1992, Rivera et

al. 1998, Doerr and Doerr 2005). Foray-search dis

persal trajectories have important implications for dis

persal distances, efficiency, and mortality (Conradt et

al. 2003), and need, therefore, to be taken into account

by relevant models (Conradt et al. 2003, Kindlmann et

al. 2004, Doerr and Doerr 2005, Zollner and Lima

2005). In conclusion, we suggest that population-level mod

els which are based on the assumption of systematic search (e.g., foray search) by dispersing individuals,

rather than correlated random walks, are urgently need

ed (Conradt et al. 2000). First approaches have been

made by Zollner and Lima (1999), Conradt et al. (2003) and Heinz et al. (2005) using simulation models. Their

results suggest that foray-search type pre-dispersal ex

plorations lead to higher average dispersal success and

lower average dispersal mortality, but shorter average and maximum dispersal distances in populations (Zoll ner and Lima 1999, Conradt et al. 2003). They can also

result in lower habitat patch connectivity between dis

tant patches, because of increased "competition" for

dispersers by neighbouring patches (Heinz et al. 2005).

Acknowledgments

We thank the Brighton and Hove Council for permission to work in Stanmer Park. Dr. L, Conradt was supported by a

Royal Society University Research Fellowship.

Literature Cited

Armsworth, P. R., M. K. James, and L. Bode. 2001. When to press on or turn back: dispersal strategies for reef fish larvae. American Naturalist 157:434-450.

Bailey, R. I., M. E. Lineham, C. D. Thomas, and R. K. Butlin. 2003. Measuring dispersal and detecting departures from a random walk model in a grasshopper hybrid zone. Eco

logical Entomology 28:129-138.

Bell, W. J. 1983. Searching behavior: the behavioral ecology of finding resources. Chapman and Hall, London, UK.

Brakefield, P. M. 1982. Ecological studies on the butterfly Mani?la jurtina in Britain. 1. Adult behavior, micro-dis

tribution and dispersal. Journal of Animal Ecology 51:713 726.

Buechner, M. 1989. Are small-scale landscape features im

portant factors for field studies of small mammal dispersal sinks? Landscape Ecology 2:191-199.

Byers, J. A. 2001. Correlated random walk equations of an

imal dispersal resolved by simulation. Ecology 82:1680

1690.

Collinge, S. K., and T. Palmer. 2002. The influences of patch

shape and boundary contrast on insect response to frag mentation in California grasslands. Landscape Ecology 17:

647-656.

Conradt, L., E. J. Bodsworth, T. J. Roper, and C. D. Thomas. 2000. Non-random dispersal in the butterfly Mani?la jur tina: implications for metapopulation models. Proceedings

of the Royal Society of London Series B-Biological Sci ences 267:1505-1510.

Conradt, L., T. J. Roper, and C. D. Thomas. 2001. Dispersal behavior of individuals in metapopulations of two British butterflies. Oikos 95:416-424.

Conradt, L., P. Zollner, T. Roper, K. Frank, and C. Thomas. 2003. Foray search: an effective systematic dispersal strat

egy in fragmented landscapes. American Naturalist 161: 905-915.

Dieckmann, U., B. O'Hara, and W. Weisser. 1999. The evo

lutionary ecology of dispersal. Trends in Ecology and Evo

lution 14:88-90.

Doerr, E. D., and V. A. J. Doerr. 2005. Dispersal range anal

ysis: quantifying individual variation in dispersal behavior.

Oecologia 142:1-10.

Doncaster, C. P., C. Rondinini, and P. C. Johnson. 2001. Field test for environmental correlates of dispersal in hedgehogs

Erinaceus europaeus. Journal of Animal Ecology 70:33 46.

Durier, V, and C. Rivault. 1999. Path integration in cock roach larvae, Blattella germ?nica: direction and distance estimation. Animal Learning and Behavior 27:108-118.

Fagan, W. E., R. S. Cantrell, and C. Cosner. 1999. How habitat edges change species interactions. American Nat uralist 153:165-182.

Golden, D. M., and T O. Crist. 2000. Experimental effects of habitat fragmentation on rove beetles and ants: patch area or edge? Oikos 90:525-538.

Haddad, N. M. 1999. Corridor use predicted from behaviors at habitat boundaries. American Naturalist 153:215-227.

Hanski, I. 1998. Metapopulation dynamics. Nature 396:41 49.

Haughland, D. L., and K. W. Larsen. 2004. Exploration cor relates with settlement: red squirrel dispersal in contrasting habitats. Journal of Animal Ecology 73:1024-1034.

Heinz, S., L. Conradt, C. Wissel, and K. Frank. 2005. A measure for inter-patch connectivity in fragmented land

scapes. Landscape Ecology 20:83-99.

Kindlmann, P., S. Aviron, F. Burel, and A. Ouin. 2004. Can the assumption of a non-random search improve our pre diction of butterfly fluxes between resource patches? Eco

logical Entomology 29:447-456.

Leon-Cortes, J. L., J. J. Lennon, and C. D. Thomas. 2003.

Ecological dynamics of extinct species in empty habitat networks. 1. The role of habitat pattern and quantity, sto

chasticity and dispersal. Oikos 102:449-464.

Matthysen, E. 2002. Boundary effects on dispersal between habitat patches by forest birds (Parus major, P. caeruleus). Landscape Ecology 17:509-515.

Ries, L., and D. M. Debinski. 2001. Butterfly responses to habitat edges in the highly fragmented prairies of Central Iowa. Journal of Animal Ecology 70:840-852.




Rivera, J. H. V, J. H. Rappole, W. J. McShea, and C. A.

Haas. 1998. Wood Thrush postfledging movements and

habitat use in northern Virginia. Condor 100:69-78.

Roper, T. J., J. R. Ostler, and L. Conradt. 2003. The process of dispersal in badgers M?les meles. Mammal Review 33:

314-318.

Schooley, R. L., and J. A. Wiens. 2004. Movements of cactus

bugs: patch transfers, matrix resistance, and edge perme

ability. Landscape Ecology 19:801-810.

Schultz, C. B., and E. E. Crone. 2001. Edge-mediated dis

persal behavior in a prairie butterfly. Ecology 82:1879

1892.

Seymour, A. S., D. Gutierrez, and D. Jordano. 2003. Dis

persal of the lycaenid Plebejus argus in response to patches of its mutualist ant Lasius niger. Oikos 103:162-174.

Stamps, J. A., M. Buechner, and V. V. Krishnan. 1987. The

effects of edge permeability and habitat geometry on em

igration from patches of habitat. American Naturalist 129:

533-552.

Travis, J. M. J., and C. Dytham. 1998. The evolution of

dispersal in a metapopulation: a spatially explicit, individ

ual-based model. Proceedings of the Royal Society of Lon

don Series B-Biological Sciences 265:17-23.

Turchin, P. 1998. Quantitative analysis of movement: mea

suring and modeling population redistribution in animals

and plants. Sinauer Associates, Sunderland, Massachusetts, USA.

Walters, J. R., P. D. Doerr, and J. H. Carter. 1992. Delayed

dispersal and reproduction as a life-history tactic in co

operative breeders?fitness calculations from Red-Cock

aded Woodpeckers. American Naturalist 139:623-643.

Zollner, P. A., and S. L. Lima. 1999. Search strategies for

landscape-level interpatch movements. Ecology 80:1019

1030.

Zollner, P. A., and S. L. Lima. 2005. Behavioral tradeoffs

when dispersing across a patchy landscape. Oikos 108:

219-230.



nonrandom movement behavior at habitat boundaries in two...

Documents