The influence of forest fragmentation on the location of overwintering monarch butterflies in Central Mexico

2007·6.7 MB·

90 Journalofthe Lepidopterists' Society
JournaloftheLepidopterists'Society
6] 2 2007,90-104
THE INFLUENCE OF FOREST FRAGMENTATION ON THE LOCATION OF OVERWINTERING
MONARCH BUTTERFLIES IN CENTRAL MEXICO
Jennifer Williams1 Douglas A. Stow2
J. ,
and
Lincoln P. Brower3
ABSTRACT. Theendangeredstatusoftheoverwinteringphenomenonofmonarchbutterflies(DanaiisplexippusL..Lepidoptera,N\mphal-
idae,Danainae)thatmigratefromeasternNorthAmericatoMexicohasresultedfromanthropogenicdegradationoftheforestswherediebut-
terfliesspenddiewinter. Duringtheirfivemonthoverwinteringperiod,themonarchsneedtoremaininactiveandclusteredinsemi-closedor
intactforeststoreduce mortalitiescausedbyfreezing,lipiddepletionandpredation.Weanalyzedforestandlandscape metricstodetermine
dieforestcharacteristicspreferredbymonarchswhencolonizingtheiroverwinteringsites.
Forestmetricsattwodifferentspatialscaleswerederivedfromaforestcovermapgeneratedfrommulti-spectralIKONOSsatelliteimagery.
Landscape metricsofforestareasoccupiedbymonarchcoloniesduringthe2002-2003overwinteringseasonweresignificantlydifferentthan
randomlyselectednon-colonyareas. Colonysites hadgreaterforestcover(> 60%), thougheven diese forestpatcheswere fragmentedand
thinned. Immediatecolonyareas(100m radial fromcolonycenter)exhibitedgreaterforestcoverageinmorecloselyspacedpatchesthandid
extendedcolonyareas(1km radialfromcolonycenter). Forestdegradationwas moreevidentintheextendedcolonylandscapesthantheim-
mediatecolonyareas.Though forestsinmanyoftheimmediatecolonyareasappearedtohavebeenthinnedandselectivelylogged,mosthave
asemi-closedcanopy.
Conservationeffortsshouldfocusonprotectingtheforestcanopy.Continuedforestdegradationislikelytoincreasemortalityfortheeastern
NorthAmericanpopulationofmonarchbutterflies,andmaycauseextinctionbothofitsmigrationandthespectacularoverwinteringphenom-
enonin Mexico.
Additional key words: butterfly site selection, forest quality, habitat degradation, FRAGSTATS, IKONOS satellite imagery,
landscapemetrics
The destination of the fall migration of North Coast states in early to mid-March when their larval
American monarch butterflies that breed east of the iood source, milkweed, is again available (Malcolm,
Rock)' Mountains was discovered in January 1975 by 1993).
cooperatorsofUrquhartandUrquhart (1976, 1978).We All known overwintering sites in Mexico occurin die
now know that up to a billion butterflies migrate states of Michoacan and Mexico, within an area of
southward out oftheir nearly one million square mile approximately 10,000 km2 (Brower et al, 2002;
breedingrange (Brower, 1999a) andformoverwintering Bojorquez-Tapiaetal, 2003). This area,partofwhichis
colonies on approximately twelve separate densely shown in Fig. 1, is in the central part of the Neo-
forested mountain massifs inCentral Mexico (Broweret Transvolcanic belt that crosses Mexicojust south ofdie
al, 2002; Slayback et al, 2007). Recent data indicate Tropic ofCancer (Brower, 1995). Monarchs colonize in
that the colonies are astoundingly dense, with up to 50 dense, protective, semi-closed oyamel fir forests (Abies
millionbutterfliesperhectare (Broweretal, 2004). The religiosa H.B.K., Pinaceae) in orderto conserve energy
overwintering period lasts from November through and avoid freezing and desiccation during the winter
March andis aunique biological phenomenon (Brower mondis (Mastersetal., 1988;Weissetal., 1991; Alonso-
and Malcolm, 1991). Butterflies en route to the Mejia et al., 1997). Presumably, after the last glacial
overwintering grounds can be up to four generations retreat, these oyamel forests retreated to the higher
removedfromtheirmigratingancestors, andfrequently peaks in this area (Slayback et al., 2007) and are now
return to the same areas of trees as did their restrictedto elevations rangingfrom 2700 m to 3400 m
predecessors (Brower, 1986). The individual monarchs andoccurwithindie summercloudbelt(Brower, 1995).
thatsurvivedie overwinteringseason returntothe Gulf Theoyamelforestmicro-climateprotectsthebutterflies
from severewarm and freezingweather, and also from
C1.ampDaenpialretmDreinvte,oSfanGeDoigergaop,hCy,A9S2a1n82D-i4e4g93o,SUtSaAte. UCnuirvreresnittya,ddr5e5s0s0: wind and desiccation. The cool temperature that is
HouseMartins,ForeStreet, Kentisbeare,Cullompton, Devon, EX15 moderated bv the high altitude forest canopy limits
2AD,England. butterflyactivity, therebyconservingtheirlipidreserves.
2. Department of Geography, San Diego State University, 5500 The extreme concentration ofthe monarchs in so few
Campanile Drive, San Diego, CA 92182-4493, USA. E-mail:
[email protected] small areas during the overwintering season makes the
3. Department of Biology, Sweet Briar College, Sweet Briar, VA entire Eastern North American population vulnerable
24595, USA. E-mail: [email protected]; corresponding author to minor perturbations to the forest system (Brower,
(telephone: [phone])
1977; Browerand Malcolm, 1991; Malcolm, 1993).
Volume 61, Number 2 91
Over the last few decades prior to 1986, forest Van Apeldoorn, 1990), while more recent studies have
degradation increased in and near critical monarch focused on the configuration ofhabitat patches (Hargis
colony sites, primarily from logging for domestic or etah, 1999). Manyfragmentation studies are consistent
commercial use and clearance for cattle grazing or in stating that no single metric can be used to describe
agriculture (Snook, 1993). Neither a 1986 presidential the response of a species to alterations in forest
decree nor the 2000 Monarch Butterfly Biosphere configuration (Haines-Young and Chopping, 1996;
Reserve (MBBR) (Fig. 1) has protectedthe colonysites Gustafson, 1998; Hargis et ah, 1999). The specific
fromascendantillegallogging. Destructionoftheforest metrics, or combination of metrics, used to answer a
continued after the 1986 decree due to ignorance and particular research question need to be determined
deliberate exploitation (Malcolm and Zalucki, 1993b; separately for each independent study. For example.
Broweretah, 2002). The current extent and increasing Luotoetah (2002) usedanumberoflandscape metrics
rate offorestdegradation andfragmentationwithin and derived from satellite imagery, including habitat
around the core overwintering sites is ofgreat concern composition and largest patch size. They discovered
(Broweretah, 2002; Ramirezetah, 2003; Honey-Roses that spatialvariationofhabitat (whichwas notoriginallv
and Galindo, 2004; Anon., 2006). These destructive considered a critical factor) is the dominant factor in
practices not only remove sections of die habitat by determining the distribution of Clouded Apollo
clear-cutting;, but thev also degrade the forest through butterflies (Parnassius mnemosyne (Linnaeus),
thinning and selective logging. Reductions in tree Papilionidae) in southwestern Finland.
canopy cover alter the forest micro-climate, increase Forest fragmentation can have a critical impact on
both exposure and access bypredators, and ultimately the survival of the monarch butterflies (Calvert and
leadtoasevereincreaseinbutterflymortality, especially Brower, 1981; Calvert et ah, 1982; Weiss et ah. 1991;
duringstrongwinterstorms (Calvertetah, 1983; Finket Brower, 1999b; Ramirezetah, 2003). However, limited
ah, 1983; Brower and Calvert, 1985; Anderson and research has been conducted on the detailed forest
Brower, 1996; Bebi et ah, 2001; Brower et ah, 2004). requirements ofcolonizing monarchs at the landscape
Monarchs are also sensitive to small perturbations that scale and, therefore, the full extent to which forest
allow more sunlight to penetrate their overwintering degradation affects the overwintering monarchs
environment. This can warm the butterflies and cause remains unknown.
more rapid burning of their lipid reserves, which can We addressed two general research questions in diis
become critically low during the overwintering season paper: (1) What are the forest fragmentation
(Alonso-Mejiaetah, 1997). The loss and fragmentation characteristics of monarch colony overwintering sites
oftheforesthasimpacted manyofthetraditionalcolony and howdo theydifferrelative to non-colonysites? (2)
sites, some ofwhich have moved, while several others How do characteristics of forests utilized by monarch
have disappeared altogedier (L. Brower et ah, colonies vary at two different spatial scales, one diat is
unpublisheddata). Continuedreductionsinforestcover an immediate (100 m) and one diat is an extended (1
may soon negatively affect the entire overwintering km) radial area, as measured from die center of the
North American monarch populations, both east and colony?
west of the Rocky Mountains (Calvert et ah, 1983; The answers to these questions can help explain the
Browerand Malcolm, 1991; Snook, 1993; Broweretah, relationship between forest degradation caused bv
2002; Browerand Pyle, 2004). human practices, i.e. clearcutting and thinning, and
As deforestation throughout the world continues to monarch colony locations. Answering these questions
increase at alarming rates (Wilson, 2002), research into willalsoleadtoabetterunderstandingofthe monarch's
the ecological impacts caused by forest fragmentation overwintering biology, which in turn should lead to
on specific species has also increased. There are two moreeffectiveconservationeffortstopreventdielossof
primary components to fragmentation, total reduction the monarch's migration and overwintering behavior
of specific habitat types, and the reduction of habitat that clearly has become a severely endangered
into small isolated patches (Meffe et ah, 1997). Even biological phenomenon (Brower and Malcolm, 1991:
slight perturbations to the qualityofthe habitat caused Honey-Roses and Galindo, 2004).
by initial fragmentation can affect the availability of
Methods
resources forsensitive species usingorlivingwithin the
habitat area (Hargis et ah, 1999; Ramirez et ah, 2003; The study area. The general studv area is die
Ramirezetah, 2005). Monarch Butterfly Biosphere Reserve (MBBR) as
Earlyhabitat fragmentation studies concentrated on decreed by Mexican President Ernest Zedillo (2002)
remnanthabitatpatches (Diamond, 1975;Verboom and and is located in central Mexico in the states of
92 Journalofthe Lepidopterists' Society
Michoagan and Mexico. Our primary study area was recordedthe locations ofmonarch colonies (L. Brower,
delineated by the extent of coverage of specific unpublished data) between 31 December 2002 and S
IKONOS satelliteimagery(Fig. 1). January 2(303. The team first conversed with local
Colony and non-colony location data. A number peopleto locatepossiblecolonies andthen traversedon
of trained field personnel from the MBBR and foot through the forest to determine whether colonies
Universidad Nacional Autonoma de Mexico (UNAM) were actually present. Colony locations were recorded
100°30'W 100°15'W
20°0,N-I I
-20°0'N
Mexico
T,Cancer
State: Michoacan
/ State:
Mexico
19°45'N-
-19°45'N
ChamangariO;
Kilometers
5 10 15 20
I 1 1 1 1 LosLobos
• Towns
19°30'N-
-19°30'N
A
2002-2003 Colony Sites Zitacuaro
i i Mexico States
<? 2000 Reserve Core
<? 2000 Reserve Buffer
LZ1 IKONOS Image Footprint
January 19, 2003
I
100°30'W 100°15'W
Figure 1. Location of the Monarch Butterfly Biosphere Reserve zones showing 11 overwintering sites located during the
2002-2003season (thethirdblacktrianglefromthetoprepresentstwositesandthefourththree).
Volume 61, Number 2 93
using a Garmin Global Positioning System (GPS) with identify canopy reference data points (randomly
an estimated 5-15 m horizontal positional error. The generated) as eitheropen orclosed forest. Sample data
errorisafunctionoftheinstrumentspositional accuracy (1200 points) were randomly generated to achieve
(Garmin, 2003) and a reduced signal caused by dense approximately one point per 50 hectares. Congalton
canopycover. (1991) suggested that at least 250 reference pixels are
To account for a broader range offorest conditions required to determine the mean accuracy to within +
(Pereira and Itami, 1991), in addition to data from the 5% (assuming an overall accuracy of 85%). Thus, we
11 observed colonylocations, we generated 33 random used three-quarters ofthe canopy reference data (900
sample points, referred to as 'available' or likely non- points) to train the image classification processes, and
colonv points for the MBBR area. These non-colony the remaining one-quarter (300 points) to validate the
locations may have been utilized by monarchs during ouqiut classification image.
the studyperiod, but extensive aerial and field surveys We produced a map of forest and non-forest pixels
during four subsequent overwintering seasons using an expert classifier approach with inputs from
(Slayback et al., 2007; and Slayback and Brower, in supervised and unsupervised classification procedures,
press) indicated that no colonies were located at these as well as by selecting class-boundary thresholds for
sites. spectralandspatialtransformimages (Stowetal.,2003).
Fine spatial resolution satellite data. IKONOS The fine 4 m spatial resolution ofthe IKONOS multi-
imagery has been used in many studies to map and spectral imagery enabled tree canopy cover and
classifylandcovercategories (Franklinetal., 2001; Song occasionally individual trees to be identified. To
andWoodcock, 2002; AsnerandWamer, 2003; Roberts effectively distinguish tree canopy from shadow or
et al, 2003; Thenkabail et al, 2004). The high spatial agricultural vegetation, the expert classifier
resolution (1 m) of the IKONOS panchromatic data incorporated both IKONOS spectral band and image
enables individualplant canopies to be detected (Asner transform data, such as the normalized differential
andWarner, 2003). IKONOS multi-spectralimagedata, vegetation index (Read and Lam, 2002: Staus et al.,
widianominalgroundsamplingdistanceof4 m alsohas 2002; Goetzetal, 2003; Chenetal, 2004), tasseledcap
a very high spatial resolution by satellite image indices (Home, 2003), and texture enhancement
standards (Cablk and Minor, 2003), and provides (Franklin et al, 2000; Franklin andWulder, 2002). We
information that can be utilized to separate subtle integratedtheseimage-basedproductsintothe ERDAS
differenceswithincovertypes, suchastreecover, urban Knowledge Engineer software tool to produce a digital
areas, orriparian zones (Goetzetal., 2003). map consisting of forest and non-forest cover classes.
We acquired IKONOS panchromatic and multi- We generalized this two-category map through die
spectralimagery(11-bitradiometricquantization) ofthe eliminationofsinglepixels ofagivenclass, sodiatforest
Central Mexico study area for 13 January 2003, fragmentation parameters could be extracted more
coinciding with the winter dry season for which the effectively and efficiently. The accuracy of the
colony data were recorded. We ortho-rectified the classification was determined from 300 randomlv
IKONOS imagery to the UTM WGS84 coordinate generated points visually interpreted as forest or non-
system, usinga digital elevation model based on a 12 m forest from the IKONOS panchromatic image anddie
grid. The ortho-rectification procedure accounted for aerial photographs. While image classification
the extreme reliefdisplacement in the study area and procedures were usedto extract forest canopypatches,
transformed the image so that it was planimetrically the canopy/non-canopv classification process and
correctandthereforealignedtothe monarchcolonysite resultant map are referred to as forest and non-forest
locations. Ascene-specific IKONOS rationalpolynomial areas in this study, the more common terminologyused
coefficients (RPC) model was used in the ortho- in forest fragmentation studies.
rectification procedure to define die interior and Fragmentation analysis. Toquantifythepatternof
exteriorgeometry ofthe sensor (Dial etal., 2003). The forest cover, we used FRAGSTATS version 3.3
root mean square error of the ortho-rectification was (McGarigaletal,2002) tocreate250landscape metrics,
approximately 2 pixels (or 8 m), based on 124 such as patch size, edge lengdi, and patch density, from
independentcheckpoints. the forest and non-forest map. For diose unfamiliar
We visually interpreted digital color imagery (2 m with the FRAGSTATSprocedureandlandscape metrics
spatial resolution) produced by scanning color aerial terminology, see Anon. (2007). The landscape metrics
photographs capturedbyArmandoPeraltaofUNAM in were calculated at the IKONOS pixel resolution (4 m)
2003, and IKONOS panchromatic imagery (1 m spatial and extracted at two different scales, immediate colonv
resolution) captured by Space Imaging, Inc. in 2003, to (circle with 100 m radius) and extended colony (circle
94 Journalofthe Lepidopterists' Society
with 1000 m radius) centeredonthe 11 monarchcolony both the class-level (forest only) and landscape-level.
and 33 random non-colony sites (Fig. 2). The radial Class-level metrics refer to the relationship between
distance used to generate the immediate colony forest and non-forest pixels, while landscape-level
landscapeswas determinedfrom the average radial size metrics refer to the composition or configuration of
ofcolonies (100 m) andthe estimatedpenetration range forest and non-forest land within a specified area.
of edge effects, such as changes to microclimate and Similar to the methods ofImbemon and Branthomme
predation rate (Chen and Franklin, 1990; Staus et al., (2001),severalstepswerefollowedtoselectappropriate
2002). The radial distance for the extended colony metrics. We generated a correlation matrix for all the
landscapes was based on the estimated active range of landscape pattern metrics and removed redundant
monarchs within the overwintering area. Research metrics where correlation values were greaterthan 0.8.
shows that most colonies are located within 1 km ofa Descriptive and inferential statistics were used to
water source to and from which monarchs regularlyfly compare landscape pattern metrics forcolonyand non-
(Calvert et al, 1983; Masters et al, 1988; Calvert and colony ('available') sites. Based on interpretation of
Lawton, 1993; Alonso-Mejiaetal., 1998; Broweretal., statisticalboxplots, metricswith similardistributions for
unpublished data). colonyand non-colony forest areas were removed from
FRAGSTATS metrics were generated from the further analysis, i.e., were deemed to have no
classified image representing forest cover (Fig. 3) at explanatory power for colonization requirements. We
Figure2. Concentricradiallandscapesextendfromdiemonarchcolonycentroid.
Figure4. An example of non-colonyareas atthe immediate colonyscale (100 m radius).Volume 61,
v
"i
Colony100mRadius Non-Colony100mRadius
|Forest |Forest
\Non-Forest |Non-Forest
Figure3.Anexampleofcolonyareasattheimmediatecolony Figure 4. An example ofnon-colonyareas at the immediate
scale(100m radius). colonyscale (100mradius)
used the small sample Student's f-tests and one-sided indicatinglandscapeheterogeneity) arelistedinTable 1
unequal variance Welch's tests to compare the mean (foradetailed explanation ofFRAGSTATS metrics, see
values ofthe colony and non-colony landscapes forthe Williams, 2005, Appendix D). Abbreviations for
remaining metrics. Metrics showing no significant landscape metrics listed in Table 1 are provided-within
difference (p > 0.05) between the landscapes were parentheses throughout the Results section.
removed. The final set of selected metrics indicated We alsousedStudent'sf-teststoassess differencesin
specific monarch colony habitat characteristics, and landscape metrics for the immediate and extended
differentiatedbetweencolonyandnon-colonysites. The colony areas. This was done to explore whether or not
complete set of metrics used to quantify landscape differenceswere evidentin die spatial characteristics of
pattern, structure and composition at both the single- forests required by the monarchs to colonize
class (forest) and multiple-class (forest and non-forest, (immediate colony) and to fly to water sources
Table 1. SelectedFRAGSTATSfragmentationmetrics,(modifiedfrom McGarigalandMarks, 1994,p.24).
Patchdensity andsizemetrics Coreareametrics
PD PatchDensity(#/100ha) TCA TotalCoreArea(ha)
PLAND PercentageofLandscape(%) CORE CoreArea
AREA PatchAreaDistribution CAI CoreAreaIndexDistribution
NDCA NumberofDisjunctCoreAreas(#)
Edgemetrics DCAD DisjunctCoreAreaDensih'(#/100ha)
ED EdgeDensity(m/ha)
Shapemetrics Isolationandproximitymetrics
LSI LandscapeShapeIndex ENN EuclideanNearestNeighborDistance
GYRATE RadiusofGyrationDistribution Distribution
SHAPE ShapeIndexDistribution
Contagionmetrics
FRAC FractalDimension Index Distribution AI Aggregation Index(%)
CIRCLE RelatedCircumscribingCircle
Distribution PLADJ PercentageofLikeAdjacencies(%)
CONTIG ContiguityIndexDistribution Diversitymetrics
PRD PatchRichnessDensity-(#/100ha)
Connectivity metrics
COHESION PatchCohesionIndex
96 Journalofthe Lepidopterists' Society
(extended colony) during the 2002-2003 overwintering (GYRATE 0.0374), and irregular in shape (SHAPE
season. 0.0417). However, they were also less contiguous
The comparison of colony to non-colony areas was (CONTIG 0.0396) and aggregated (LSI, AI) than the
conducted in order to determine (1) whether specific non-colony sites. The aggregation index (AI)
forestcharacteristics affectwhere monarchcolonies are (approximately 73%) indicated that colony sites
located, or (2) whethercolonylocations are determined containedcloselyspacedpatches, butingeneral, colony
independently of forest characteristics, i.e. forest sites were less aggregated than non-colony sites. This
characteristics in colony and non-colony areas are not maybedue, inpart, toaggregated non-forestpatches at
significantly different. The comparison of immediate the non-colony sites. Despite high edge densities (ED)
and extended colony areas characterized the whole within colony sites, the high percentage of forest in
environmentinwhichthe monarchs reside andinteract combination with the close proximity offorest patches
whilst overwintering, rather thanjust their colony site. (AI) appears to provide habitat conditions suitable for
This may indicate that for an area to be suitable for monarch colonies (e.g., good canopy cover and
monarchs to colonize, the forest composition and protection). In some cases the negative effects ofedge
configuration at both the immediate and extended (orhigh edge density, ED), such ashighpredation, may
scales mustmatchthe monarchs' forest requirements. be reduced or counter-acted by positive edge effects,
such as slight increases in insolation, when patches are
Results
locatedin close proximity (Weissetat., 1991).
Classification accuracy. Overall classification Examples of the more discriminatory landscape
accuracy values for the forest/non-forest map derived metrics at the 1 km radial scale are shown in Table 3.
from IKONOS multi-spectral data were 92.3% and Colony landscapes contained smaller forest patches
88.6%, with kappa values of 0.847 and 0.766, when (AREA, ED 0.0042), which were more complex
compared against reference data generated from visual (FRAC), elongated (GYRATE), irregular in shape
interpretation of IKONOS panchromatic imagery and (SHAPE 0.0009), contained less core area (CORE
aerial photography, respectively. The kappa statistic 0.0005), and were higher in density (PD 0.0499), than
incorporates omission and commission errors and the non-colonysites. The forest canopyin the extended
corrects for chance agreement between reference and colonylandscape was more highlyfragmented than the
classifieddata (Jensen 1996, Lillesandand Kiefer2000). non-colony sites, there were fewer core areas (CORE
The average overall classification accuracyestimated 0.0005) and as a result, much of the forest area
with the two reference sets was 90.4% and the average contained a high proportion of 'edge'. The forest
kappa value was 0.806. For the forest category, user's patches in the colony landscapes were less contiguous
(commission) accuracy values were 92.3% and 87.3%, (CONTIG), connected (COHESION 0.0418), and
and producer's (omission) accuracyvalues were 92.9% circular (CIRCLE 0.0111) than the non-colony
and 92.9%, based on IKONOS panchromatic imagery landscapes, but were also less isolated (ENN). Low
and aerialphotography, respectively. Forthe non-forest patch connectivity (COHESION) may indicate that
category, user's (commission) accuracy values were when monarchsflythrough the landscape thevuseboth
92.4% and 90.3%, and producer's (omission) accuracy the forest and non-forestpatches (Mastersetal., 1988).
values were 91.8% and 83.2%, based on IKONOS Shape complexity (SHAPE, GYRATE, FRAC 0.006)
panchromatic imagery and aerial photography, andassociatededgedensity(PD)werehigherforcolony
respectively. All of these measures suggest that the landscapes, but the patches were less isolated (ENN).
forest cover map is a highly accurate representation of However, there is no significant difference in the
forest cover in 2003 and a reliable source for percentage of forest cover (PLAND 0.1717) between
quantitativelyassessingforest fragmentation. the colonyandnon-colonylandscapes.
Influence of forest fragmentation on colony The top priority for monarchs seems to be the
locations. Examples of the more discriminatory colonizationofimmediatesitesthatcontainthegreatest
landscape metrics, their distributions, and significance totalforestcovereventhough manyofthese forestareas
levels in relation to variations in colonyand non-colony are fragmentedand appearto be thinned.
sites at the 100 m radial scale are shown in Table 2.
Significance levels for all other metrics are listed in the Comparison of fragmentation at site and
text. At this scale, colony sites contained more forest landscape scales. The monarchs seem to have
(PLAND) in fewer patches (PD) than the non-colony different forest composition and configuration
sites. In addition the forest patches ofthe colony sites requirements for areas inwhichtheycolonize, than the
were more complex (FRAC 0.0332, TCA), elongated more extensive areas in which theyinteract and move.
—
1
Volume 61, Number2 97
Table2. Selectedfragmentation metricsusedtocomparecolonyandnon-colonysites atthe 100m radialscale.
Landscape Metrics T-test(2-tail) Boxplot Ecologicalinterpretationof
colonylandscapes
Class Level (Forest Patches)
\ I
Edge Density(ED) 0.0385 greateredgedensity
o
__j_
c NC
—
1
Patch Density(PD) 0.0413 fewerpatches
-!-
o
c NC
—
1 1
P(Feorrceesntt)a(gePLofANLaDn)dscape 0.0455 1 -i- greaterforestcover
C NC
Landscape Level (Forest and Non-Forestpatches, indicatinglandscape heterogeneity)
Landscape Shape Index(LSI) 0.0242 less aggregated
c NC
o
;
TotalCoreArea(TCA) 0.0376 lesscorearea
1
c1 NCI
o
Aggregation Index(AI) 0.0374 ; lesspatchaggregation
1 1
1 NiC
C = Colonysite, NC = Non-colonysite
T-testsignificancedifference < 0.05
—— —
98 Journalofthe Lefidopterists' Society
Table3. Selectedfragmentationmetricsusedtocomparecolonyandnon-colonysitesatthe 1 km radial scale.
Ecologicalinterpretationof
Landscape Metrics T-test (2-tail) Boxplot colonylandscapes
Class Level (Forest Patches)
j 1
Landscape: Smallerpatches
PatchAreaDistribution (AM) 0.0311
o
i
c NC
i—
RadiusofGyration Patch: Lesselongatedand
0.0083
Distribution (AM) — compact
-i- 1-
c NC
__
1 ' 1
Fractal Dimension Index 1 1 Landscape: Morecomplexand
0.0007
Distribution(MN) convoluted
C NC
Landscape Level (Forestand Non-Forestpatches, indicatinglandscape heterogeneity)
ContiguityIndex Distribution Landscape: Lesscontiguous
0.0006
(AM) ^^
z3
c NC
—
i ,_
Li') 1
o _ Patch: Moreirregularand
(ShAaMp)e IndexDistribution 0.0007 O _ complex
cn
o _
cm
c NC
CJ _ o
CM _
EuclideanNearestNeighbor 0.0068 - - — Patch: Lessisolation
Distance Distribution (MN) o I
cn - =.i 1 —' 1
1 ' -
NC
Distributionmetricsmeasuretheaggregatepropertiesofthepatches: FRAGSTATScomputesthefollowing: (1) mean
(MN), (2)area-weightedmean(AM), (3) median (MD), (4) range (RA), (5) standarddeviation(SD), and (6)coefficientof
variation (CV).
T-testsignificancedifference < 0.05.
Volume 61, Number2 99
The metrics used for this analysis are summarized in (GYRATE), complex (FRAC), irregular (SHAPE), and
Table 4. less compact (CIRCLE) forest patches than the 1 km
In general, the immediate colony sites contained a colony landscapes. However, they also had larger core
greater percentage of forest cover (PLAND), had areas (CORE) and were less isolated (ENN) than the
greater edge densities (ED), had more elongated extendedcolonylandscapes.Theimmediatecolonysites
Table4. Comparisonoflandscapemetricsbetweenimmediateandextendedcolonylandscapescales.
T-test EcologicalInterpretationofLandscapes
Landscape Metrics (2-tail) (immediatecolony>or< extendedcolony)
Class-LeyelMetrics (ForestPatches)
PLAND PercentageLandCover(Forest) 0.002 >percentageforestcover
ED EdgeDensity' 0.004 >edgedensity
GYRATE RadiusofGyrationDistribution(MN) 0.011 > elongationandcompaction
SHAPE ShapeIndexDistribution(MN) 0.000 >irregularityandcomplexity
FRAC FractalDimensionIndexDistribution(MN) 0.011 >complexityandconvolution
FRAC FractalDimensionIndexDistribution(SD) 0.000 > heterogeneityinpatchfractaldimensions
CIRCLE RelatedCircumscribingCircleDistribution(MN) 0.017 <circularityandcompaction
CORE CoreAreaIndexDistribution(MN) 0.036 >patchcorearea
ENN EuclideanNearestNeighborDistanceDistribution(MN)0.019 <patchisolation
COHESION PatchCohesionIndex 0.020 <connectivityandgreaterdivision
Landscape-Level(ForestandNon-Forestpatches,indicatinglandscapeheterogeneity')
PD PatchDensity 0.006 >patchdensity
AREA PatchAreaDistribution(MN) 0.042 <patchsize
GYRATE RadiusofGyrationDistribution(MN) 0.004 > elongationandcompaction
GYRATE RadiusofGyrationDistribution(SD) 0.002 > uniformityingyration
SHAPE ShapeIndexDistribution(MN) 0.014 > irregularity'andcomplexity
' • 'IIK CoreAreaIndexDistribution(MN) 0.009 >patchcorearea
CORE CoreAreaIndexDistribution(SD) 0.000 >heterogeneityinpatchcorearea
COHESION PatchCohesionIndex 0.000 < connectivityandgreaterdivision
Distributionmetricsmeasuretheaggregatepropertiesofthepatches: FRAGSTATScomputesdiefollowing: (1)mean(MN).(2)area-
weightedmean(AM),(3)median(MD),(4)range(RA),(5)standarddeviation(SD),and(6)coefficientofvariation(CV).
T-testsignificancedifference <0.05