Assessing the Indicator Properties of Species Assemblages ... · Ecological Applications, 2(2),...

Assessing the Indicator Properties of Species Assemblages for Natural Areas MonitoringAuthor(s): Claire KremenSource: Ecological Applications, Vol. 2, No. 2 (May, 1992), pp. 203-217Published by: Ecological Society of AmericaStable URL: http://www.jstor.org/stable/1941776 .

Accessed: 07/02/2014 15:50

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 toEcological Applications.

http://www.jstor.org

This content downloaded from 128.32.85.74 on Fri, 7 Feb 2014 15:50:07 PMAll use subject to JSTOR Terms and Conditions

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

http://www.jstor.org/stable/1941776?origin=JSTOR-pdf

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


Ecological Applications, 2(2), 1992, pp. 203-217 ? 1992 by the Ecological Society of America

ASSESSING THE INDICATOR PROPERTIES OF SPECIES ASSEMBLAGES FOR NATURAL AREAS MONITORING'

CLAIRE KREMEN Center for Conservation Biology, Stanford University, Stanford, California 94305 USA

Abstract. The diversity of organisms and complexity of ecosystems prevent thorough inventory and monitoring of protected areas, yet sound databases are needed to manage ecosystems for long-term persistence. One strategy is therefore to focus monitoring on indicator organisms, but guidelines are lacking for selecting appropriate species or groups. This paper presents a simple protocol based on ordination techniques for establishing the indicator properties of a group of organisms and for selecting an indicator species subset for more intensive monitoring. Use of ordination allows inclusion of many more taxa than have been traditionally used for natural areas monitoring, and need not rely on detailed knowledge of species biology. As an example, I studied the indicator properties of a butterfly taxocene in a rain forest in Madagascar. Butterflies have been suggested as particularly good environmental indicators due to their sensitivity to micro-climate and light level changes, and their interactions as larvae and adults with different sets of host plants. The indicator properties of butterfly assemblages were evaluated in this study with respect to a known pattern of environmental heterogeneity along topographic/moisture and disturbance gradients. Butterfly assemblages were found to be excellent indicators of heterogeneity due to the topographic/moisture gradient, limited indicators of heterogeneity due to anthropogenic disturbance, and poor indicators of plant diversity. The protocol defined in this study is widely applicable to other groups of organisms, spatial scales, and environmental gradients. By examining the environmental correlates of the distribution of species assemblages, this protocol can assess the indicator properties of target species groups.

Key words: diversity; dominance; ecological monitoring; indicator species assemblage; Madagas- car; natural areas conservation; ordination; rarity; Satyrinae; tropical butterflies.

INTRODUCrION

Most ecosystems today are subject to one or more forms of anthropogenic disturbance, especially pollu- tion and acidification, habitat modification and fragmentation, and invasions by introduced species (Soule and Wilcox 1980, Burgess and Sharpe 1981, Peters and Darling 1985, Soule 1986, Schreiber and Newman 1988, Carleton 1989, Fajer 1989, Fajer et al. 1989, Klein 1989). Given the pervasive spatial and rapid temporal scale of current anthropogenic environmental changes, methods are needed for choosing appropriate species or species assemblages for establishing conservation priorities and monitoring biotic responses to local and global environmental change (Kimball and Levin 1985, Soule 1990).

Outside of a substantial literature on single or multi- species indicators of specific environmental contami- nants (Cairns 1985, 1986), few practical guidelines currently exist for selecting indicators for monitoring natural areas. A recent review critically evaluated selection of vertebrate indicators, concluding that no single traditional criterion (e.g., high sensitivity to habitat modification, large size, habitat specialization, low population or species turnover rates, or large area re- quirements) can be safely used for monitoring popu-

lation trends or habitat quality (Landres et al. 1988). Given the difficulties inherent in using one or a few species as indirect assays of complex ecosystem structure and function (Ward 1978, Kimball and Levin 1985, Cairns 1986, Soule 1987, Landres et al. 1988, Noss 1990), further work is needed to establish and test cri- teria for selecting indicators.

Much previous work involving indicators has relied on utilizing one or a few species (e.g., Management Indicator Species of the Forest Service, Landres et al. 1988). The failing of this approach is its narrow focus, which can result in protection of one organism at the expense of others (Kushlan 1979, Landres et al. 1988). By contrast, use of a greater variety of indicator species could provide more fine-grained information (Noss 1990): the degree of detail gained would in turn depend on variation in microhabitat use, niche breadth, ecological function, and response to environmental change among members of the indicator assemblage. By using ordination techniques, one can easily examine the distributions of many species simultaneously and their relationship to environmental parameters (Gauch 1982a, Ter Braak 1987, Peet et al. 1988). These pow- erful multivariate tools allow assessment of the indicator properties of a much wider array of organisms, and could be used to broaden monitoring concepts to include detection of environmental patterns based on the response of a group of species.

I Manuscript received 7 January 1991; revised and accepted 15 August 1991.



204 CLAIRE KREMEN Ecological Applications Vol. 2, No. 2

This paper presents an analytical method based on ordination techniques for assessing the indicator properties of a given species assemblage (guild, taxon, taxocene, or community). The method first tests whether the distributional data from the chosen species assemblage indicate environmental patterns at the geograph- ic scale of interest. It then determines which environmental parameters the species assemblage serves to indicate. Finally, it provides a framework for selecting the most important indicator species from within the original species assemblage.

Butterflies as indicators The taxon Rhopalocera (butterflies and skippers) was

the group chosen for this study of a rain forest habitat in Madagascar. Little ecological information exists on these species or their habitat, and some of the taxa are poorly defined at the species level. This study therefore required a method for choosing indicator species that did not depend on detailed knowledge of their biology. This property of the method developed below makes it particularly important for use in testing and selecting indicators in regions where the ecological information- base is poor.

In addition, the choice of the Rhopalocera allowed a test of the claim, advanced by several authors, that butterflies have a particular value as ecological indicators (Gilbert 1980, 1984, Pyle 1980, Brown 1982, Murphy et al. 1990). On a practical basis, butterflies (in comparison to other insect taxa) have a manageable level of diversity, are better known taxonomically, and can in many areas be reliably identified in the field (Pollard 1977, Thomas 1983, Thomas and Mallorie 1985, Murphy and Wilcox 1986). On biological grounds, it has been suggested that butterfly diversity could be used as an index of plant diversity (Pyle 1980), since coevolution between butterflies and their larval host-plants has in some cases led to high butterfly- plant specificity (Ehrlich and Raven 1964, Gilbert and Smiley 1978). However, positive correlations between butterfly and plant diversity may in fact be the excep- tion (e.g., Thomas and Mallorie 1985) rather than the rule (e.g., this study, Sharp et al. 1974, Vane-Wright 1978; M. Singer, personal communication, P. Brussard, personal communication). Of perhaps more interest is that butterflies interact with plants both as larval her- bivores and as adult pollinators, potentially influencing plant population dynamics in both interactions (Gil- bert 1980, Rausher and Feeny 1980, Jennersten 1988). Thus, with sufficient system-specific knowledge, one might be able to predict changes in plant populations from knowledge of butterfly population changes. Fi- nally, populations of butterflies are strongly influenced by local weather, microclimates and light levels acting in a variety of ways at all stages of the life cycle (e.g., oviposition site selection and larval development, Rausher 1979; larval development and host plant phenology, Weiss et al. 1987, 1988; availability of flight

time for adult courtship and oviposition, Watt et al. 1968, Kingsolver 1983a, b; severe climate and population extinction, Ehrlich et al. 1972). Since butterfly populations respond to habitat modifications affecting local climates and light levels, it has been suggested that changes in butterfly populations could serve to herald local or global climate change (Murphy et al. 1 990).

While studies of butterflies have contributed to conservation biology in both a population and community context (e.g., Brown 1982, Ehrlich and Murphy 1987, Murphy et al. 1990, Singleton and Courtney 1991), no studies to date have evaluated the indicator properties of butterfly species or assemblages. This study exam- ines the ability of butterfly community data to reveal patterns of habitat heterogeneity due to topography and anthropogenic disturbance. Variation related to these two parameters strongly influences micro-climate and plant community composition (Swanson et al. 1988), and thus provides one appropriate assay of some of the proposed indicator properties of butterflies.

METHODS

Field site The field site was located in the southeastern mon-

tane rain forest of Madagascar, 7 km southwest of Ra- nomafana on route 45 in the province of Fianarantsoa. The area is topographically diverse, with steeply wood- ed hills, dissected by numerous streams draining into the Namorona River. The site had a tree species richness on the order of 95 species/0.2 ha (D. Overdorff, unpublished data) including Ficus spp. (Moraceae), Eu- genia spp. (Myrtaceae), Weinmannia spp. (Cunoni- aceae), Symphonia spp. (Guttiferae), Ravensara and Ocotea spp. (Lauraceae), Pittosporum spp. (Pittospora- ceae), and many others. Common understory plants include many Psychotrea and other Rubiaceae, Pan- danus (Pandanaceae), palms, grasses and bamboos, tree- ferns, and ferns. The forest is rich in lianas and epiphytes, and receives a total rainfall of ;2300-2600 mm per year. The soils are acid, and of low natural fertility (P. Sanchez, personal communication).

Much of the forest at the study site experienced a patchwork cycle of slash and burn farming within the past 50-100 yr. Slashed field sizes were small (50 x 50 m) and regeneration of a diverse rain forest community in the absence of further disturbance has been rapid (C. Kremen, unpublished observations). For example, the site supports 12 species of lemur, one of the highest levels of primate diversity anywhere in the world (P. Wright, personal communication). The understory vegetation is also relatively sparse, except where recent disturbance has led to the invasion of the introduced guyava, Psidium cattleianum.

In this study, two forest types were studied within a distance of 4.2 km: forest at Vatoharanana that had been selectively logged 25 yr previously (referred to from now on as "old forest"), and forest at Talatakely



May 1992 ASSESSING INDICATOR PROPERTIES 205

that had been selectively logged within the previous 2 yr ("new forest"). The old forest was farther from the main road, and probably never experienced the same intensity of logging as the new forest. In both forest types, 100-m transects were situation along trails on ridges (new forest, N = 3; old forest, N = 2) and stream banks (new forest, N = 3; old forest, N = 2). Three additional transects were located within the new forest in highly disturbed areas slashed to provide logging trails. In addition to disturbance regime, the old forest was found at a slightly higher elevation than the new forest (50 to 100 m) and may receive higher rainfall (D. Overdorif, personal communication). Both forest types had closed canopy, but the old forest contained a much higher proportion of trees >20 m in height (11.5%) than the new forest (2.5%) (F. White and D. Overdorif, unpublished data).

Butterfly sampling The study was begun at the end of the cool, low-

rainfall "winter." Increasing mean daily temperatures and monthly rainfall at this time appeared to stimulate eclosion in a wide variety of butterfly species. During each of five 1 0-d sampling periods between the end of August and the beginning of November, each of the 13 transects was sampled for butterflies twice, 1 h during a morning time period (1000- 1200) and 1 h during an afternoon time period (1200-1400). Morning and afternoon collecting dates were randomized between sites in any one sampling period, and transects were only sampled under sunny conditions. An equal sample effort was thereby devoted to each transect, for a total of 130 person-hours of sampling time.

Sampling was conducted by continuously walking back and forth along the transect for 1 h, collecting all butterflies that were observed within a 5-m band on either side of the transect (modification of Pollard 1977, Thomas 1983). The sampling method assessed the relative abundance of understory butterfly species. While it ignored the canopy (i.e., Charaxes species) and un- derestimated the fast-flying mid-level to canopy species (most papilionids, some nymphalids), it was consistent between transects. Observations and understory and canopy trap-data collected during the same sampling period permitted an assessment of which species were not quantitatively represented in these samples.

Butterfly identification

Butterflies in most families were identified by the author using D'Abrera (1980), Paulian (1956), and Paulian and Viette (1968). Dr. Robert K. Robbins and Dr. Donald Harvey (National Museum of Natural His- tory, Washington, D.C.) assisted with identifications of lycaenid and riodinid species, respectively. Dr. Jacques Pierre (Musee Nationale d'Histoire Naturelle, Paris) assisted with identifications in the family Acreai- dae. Skippers were identified by the author using Viette (1956) and Evans (1937).

Taxonomy of the Malagasy Satyrinae (Nymphali- dae) is poorly established; hence morpho-species were designated according to characteristics of the male gen- italia, a standard character set for defining taxa in this subfamily (Miller 1968; R. Robbins, personal communication). Once morpho-species were resolved according to genitalic characters, the remaining specimens were identified by associated wing pattern characteristics. Names given to morpho-species are preceded by question marks, and represent an attempt to synonymize these species with established names.

Habitat characterization At each transect, three randomly located plots of 25

m2 each were established for plant sampling. All plants within these plots > 1 cm diameter at breast height (dbh) were identified by a local guide using common Malagasy names. Voucher collections made in this area since 1986 have allowed the identification of many of these species (D. Overdorif and P. Wright, personal communication). Measurements of dbh and estimates of height to top of crown were taken on all plants > 1 m in height. These measurements and the plant identifications were used to determine average transect values for plant species diversity, plant basal area, and height of the canopy. Altitudes at each transect were measured with an altimeter.

Flowing phenology was examined at each transect four times during the study. Flower abundance and richness were assessed along the length of the transect in ten 10 m wide by 5 m deep plots, spaced alternately every 10 m on either side of the transect. All flowers (whether lianas, epiphytes, herbs, shrubs, or trees) were noted by three observers equipped with 7 x 35 power binoculars. Species were identified by Malagasy names, and detailed notes on floral characteristics were taken to ensure consistent identifications between plots. The number of flowering individuals was counted for each flower type in each plot to derive an estimate of floral abundance (number of individuals in flower at a given time) and floral richness (number of types of plants in flower at a given time). In general, few flowers were observed in any one of the ten 50-nM2 plots; therefore the data for the 10 plots were pooled at each site. Av- erage floral abundance and richness at a site refer to these measurements averaged over time.

Analysis

Ordination.-Species abundance matrices for butterfly or plant data were analyzed by ordination, using Detrended Correspondence Analysis (DCA; detrended by polynomials, CANOCO program, Ter Braak 1988). DCA displays the patterns of covariation in species distributions, by reducing the dimensionality of N species among M samples to just a few ordination axes (Knox and Peet 1989). The eigenvalue associated with each axis can be thought of as the proportion of vari-



206 CLAIRE KREMEN Ecological Applications Vol. 2. No. 2

ation in sample or species dispersion explained by that axis (Gauch 1982a). Detrended correspondence analysis was chosen over other available techniques because: (1) this algorithm removes several artifacts such as the arch effect and compression of gradients at ends, (2) it ordinates sites and species simultaneously within the same ordination space by reciprocal averaging, and (3) it requires no a priori assumptions about directions of environmental gradients or weighting of species or sites (Gauch 1982a, Peet et al. 1988, Knox and Peet 1989). DCA, an indirect gradient analysis technique, ordinates samples based only on the distribution of species abundance values. In the absence of other information, it is therefore a good preliminary technique for assessing the ability of floral or faunal distributional data to indicate a particular environmental pattern.

For analysis of the butterfly data set by DCA, the abundance of each species was partitioned into male and female abundances, since males and females of a species may fly in different habitats (e.g., oviposition vs. feeding habitat). Relative abundance was estimated as the sum of individuals of each sex of each species captured during the five sampling periods at each site. Relative abundances for the plant data set were estimated as the total number of individuals of a morpho- species > 1 cm dbh observed at a site (the sums of abundance measures from the three separate quadrants totaling 75 m2).

Axis interpretation. -Canonical Correspondence Analysis (CCA) was used to examine the relationship between species distributions and environmental parameters. In CCA, a direct gradient analysis technique, axis interpretation is performed within the ordination algorithm using a set of supplied environmental variables (Ter Braak 1987). A major difference between CCA and DCA is that ordination of species and samples are constrained to lie along axes determined by the environmental variables. The significance of a particular environmental variable can be assessed through Monte Carlo testing (bootstrapping) of the axis associated with that variable, using the axis eigenvalue as the test statistic (Ter Braak and Prentice 1988). Thus, in CCA, one can test the significance of an environmental variable in establishing the ordination pattern.

Environmental variables included in the CCA analysis of butterfly data were altitude, topography, and level of disturbance (see Table 1 for scoring systems for disturbance and topography). Vegetative and plant community characteristics were canopy height, stem basal area per square metre, plant species richness, DCA site ordination scores based on plant community data, and average richness and abundance of plants in flower. A characteristic of butterfly species assemblages was dominance (Berger-Parker index, d = Nma,,/N, where Nmax is the number of individuals of the most abundant species, Magurran 1988). For axis interpretation of the plant DCA analysis, environmental variables considered were topography and level of disturbance.

TABLE 1. Environmental variables used in Canonical Cor- respondence Analysis (CCA) of butterfly data.

Nom- Environ- inal mental van- variable able Description

Topography* 0 ridges 1 slopes 2 streams

Disturbance* 0 selectively logged 25 yr ago 1 selectively logged 2 yr ago 2 slashed for logging trails

Altitude metres (measured) Plant species Margalefs index

richness Canopy height metres (estimated) Floral number of individuals in flower abundance (averaged over time)

Floral number of species in flower richness (averaged over time)

Butterfly Berger-Parker index dominance * Variables also used in CCA of plant community data.

RESULTS

Sample effort Cumulative species diversity began leveling off un-

der the sampling scheme used in this study after the fifth sample period (a total of 10 person-hours of sampling time per transect). Fig. 1, which shows a sample effort curve for the five sample periods of the study, includes additional data from a sixth and seventh period sampled at 9 of the 13 sites. When different habitat types were considered separately, cumulative species diversity also began to level off at the fifth sample periods (Fig. 1). This suggests that sampling effective- ness was similar in each of the habitat types despite topographic and vegetative differences that might have rendered certain habitat types easier to sample than others.

The possibility that sample curves would rise again with continued sampling cannot be excluded, since a new phenological guild of species could begin flying later in the season. However, inventory taken during a January-March field season (C. Kremen, unpublished data from 1990) added only a few more species to the list compiled from July to November (1988 data), despite distinct increases in temperature and rainfall during this period.

The data from the five sample periods used in the ordination analysis contained a total of 57 species of Rhopalocera (Appendix) among 659 specimens. Using the maximum likelihood method of Cohen (1959, 1961, in Krebs 1989), the distribution of species' abundances was not significantly different from a lognormal, with a mode of 13 species in the 4th octave of abundance (data analyzed included all seven sample periods with 59 species distributed among 8 octaves of abundance, N= 882, x2 = 5.852, df = 5, P > .5).




Environmental patterns indicated by butterfly species abundances

Detrended Correspondence Analysis (DCA) of the butterfly species abundance by samples matrix resulted in strong clustering of the samples along the first axis into two clusters at either ends of the axis, representing ridge and stream groups, and a third, middle, cluster of highly disturbed logging trails (D sites) (Fig. 2A). Within ridge and stream clusters, samples also clustered according to whether they were located in the forest selectively logged 25 yr ago (old forest) vs. forest selectively logged within the previous 2 yr (new forest). Eigenvalues for the first two principal axes were high (0.676, 0.436). The ordination based on butterfly data effectively clustered sites according to the patterns of known environmental heterogeneity due to the topographic/moisture and disturbance gradients.

In the above ordination, sexes of each species were considered separately. When DCA was run on data in which male and female abundance values were com- bined for each species, the first eigenvalue declined from 0.676 to 0.618, although the overall ordering of sites along this axis was qualitatively similar (data not shown). Thus, the separation of species abundance data by sex resulted in a slightly better ordination of the sites, suggesting that the distribution of males vs. females among sites conveyed additional information. All subsequent ordination analyses therefore separated species abundance values by sex.

The size effect Previous studies have questioned whether ordina-

tion and clustering techniques reveal actual covariation of species distributions or result from a sampling artifact in which species cluster according to their frequency of occurrence among samples (the "size effect" of Jackson et al. 1989). In DCA, ordination of samples and species occurs simultaneously by reciprocal averaging; thus the "size effect," if present, could influence the clustering of sites and affect the environmental interpretation. Prior to axis interpretation, it was important to determine whether such an artifact had contributed to establishing the ordination pattern.

For each of the DCA axes, I ran Spearman's rank correlation analyses (Statgraphics 4.0; Statistical Graphics Corporation 1989) between species' frequen- cies of occurrence among sites and species' DCA scores. No significant correlations were observed, although the second axis showed a weak trend toward a negative correlation (r = -0.1802, P = .0877). Similarly, no significant correlations were observed between species' overall masses and species' DCA scores on either axis. These analyses therefore rule out the possibility of the DCA ordination being trivially based on a "size effect."

Axis interpretation: topography To interpret the axes produced by DCA, I ran ca-

nonical correspondence analysis, a direct gradient anal-

ysis ordination technique. Out of eight supplied environmental variables (including one descriptor of butterfly species assemblages, see Table 1), only the nominal variables defining the topographic/moisture gradient significantly explained the variation along the first CCA axis (P = .01, exact probability from 99 Monte Carlo runs). This axis had an eigenvalue of 0.616, and the relative ordering of sites closely resem- bled that produced by DCA (Fig. 2B, note orientation of the axes appears reversed in CCA relative to DCA, but axis orientation is arbitrary [Knox and Peet 1989]). Given the similarity between the ordering of sites along the first axis in both ordinations, one can assume that topography also significantly explains variation among sample and species scores in the first axis of the DCA ordination.

Axis interpretation: disturbance In Fig. 2B, the significant first axis is plotted for ease

of viewing against a nonsignificant axis. The vectors indicate the strength and direction of the topographic/ moisture and disturbance gradients. Lack of significance of the axis defined principally by disturbance in the CCA analysis may be due to the coding of disturbance, which was based only on known anthropogenic effects (time since selective logging or clearing), and did not consider the contribution of natural disturbance (e.g., size and shape of stream beds and treefall gaps).

Ridge areas are likely to experience different natural disturbance regimes from stream areas (Swanson et al. 1988), and the intensity of anthropogenic disturbance from selective logging probably also differed between these two habitats. In addition, the large separation of

70 All habitats .

60 Disturbed habitats A

a) Stream habitats.

5 50 Ridge habitats a

(.D 40- C.)

0 -

520- E

1 0

0

Person-Hours per Transect FIG. 1. Sample effort curves for all habitats, disturbed

habitats (logging trails), stream habitats, and ridge habitats. Curves began to level off in all habitat types after 20 person- hours of sampling time per transect (five sample periods con- sisting of one morning and one afternoon hour).




A. DA B.0OA 400- .40 400 - ~~~ ~ ~~~~~~~DCA *40~CCA

ridge stream disturbance (P .34) 200 200 _

CM1 disturbed MAdi

-0 -20 -0 0 10 20 30 40 30-200 -200 00 20 30 C3

DCA axis 1 CCA axis 1

FIG. 2. Ordination of sampling sites by butterfly data. O = sites located in the old forest (selectively logged 25 yr ago); OR = old ridge, OS = old stream. * = sites in the new forest (selectively logged 2 yr ago); R = ridge, S = stream. * = sites along slashed logging trails; D = disturbed. (A) Detrended Correspondence Analysis ordination plot (DCA), showing ridge, stream, and disturbed area clusters. Eigenvalues are 0.676 and 0.436 for the first and second axis. (B) Canonical Correspondence Analysis (CCA) ordination plot, constrained according to topography and disturbance variables (see Table 1); arrows indicate the strength and direction of gradients. The ordering of sites on the first axis was significant (A = 0.616, P = .01); from the diagram this axis can be interpreted as topography. As no subsequent axes were found to be significant, the disturbance gradient in this ordination is not significant.

ridge vs. stream sites in the DCA ordination in Fig. 2A indicates that the butterfly assemblages occupying ridge vs. stream sites are distinct. Given these consid- erations, a separate CCA analysis was conducted for each of two site clusters defined empirically by DCA axis 1 (Fig. 2A): sites with values <- 100 (ridge cluster) and sites with values > + 100 (stream cluster).

The ridge cluster contained all of the new and old ridge sites, plus one of the logging trail sites (Dl). The first axis of CCA of the data from these sites alone was significantly related to the disturbance variables (X = 0.469, P = .01), but not to the six other environmental variables. In contrast, the stream cluster, containing two of the new stream sites and all of the old stream sites, showed no significant relationship to disturbance in CCA analysis (X = 0.541, P = .32), or to any of the other environmental variables.

Identifying butterfly indicator species assemblages

In correspondence analysis, species ordination scores represent the optimal localization of the species within the ordination space, and the abundance or probability of occurrence of a species decreases with distance from the location of its score. I used the DCA species ordination diagram as an aid to find assemblages of species characteristic of points along the topographic/ moisture and disturbance gradients.

Fig. 3 shows the DCA species scores for the 12 most abundant species (the scores are separated by sex as in the analysis, but only the male score is shown if male

and female scores are side by side). The ordination space is identical to that shown for the sites ordination (Fig. 2A), and the location of habitat clusters (new stream, old stream, etc.) is indicated in the diagram. A complex of species, including Henotesia ?iboina and Strabena andriana (Nymphalidae: Satyrinae) and Sari- bia ?perroti (Riodinidae) had ordination optima within the ridge cluster. Another species complex, including Henotesia ?angulifascia, H. ?undulosa (male), H. ?subsimilis (male), H. ?anceps and H. ?pauper, had ordination optima within the stream cluster.

The relationship between the DCA ordination scores and species distributional data can be seen by com- paring Table 2 with Fig. 3. Henotesia species ?anceps, ?subsimilis, and ?pauper were each consistently found at the new stream sites, but were almost never found at the old stream sites. Their ordination optima fall within the domain of the new stream sites. Male H.

TABLE 2. Distribution of Henotesia species among stream sites located in new and old forest.*

Abundance (individuals) Speciest Si S2 S3 OSI OS2

Henotesia ?anceps 18 4 0 0 0 Henotesia ?subsimils 16 5 1 1 0 Henotesia ?pauper 20 1 2 0 1 Henotesia ?undulosa 19 3 0 6 6 Henotesia ?angulifascia 5 5 3 29 18

* S1-S3: new stream sites; OS 1-OS2: old stream sites. t See Appendix footnote t for information on Malagasy

butterfly taxa.




?undulosa were most frequently observed at new stream sites, but also regularly occurred at old stream sites, and its ordination optimum is intermediate in position between the new and old stream domains. The female also frequented disturbed sites, and its optimum reflects this. While H. ?angulifascia was observed frequently at new stream sites, its position within the old stream domain reflects its extreme dominance at these sites (comprising 44-66% of the total butterfly abundance).

Interpreting the species ordination diagram Typically, many of the species found at the center

of the DCA ordination are ubiquitous species, bimod- ally distributed species, or species whose distribution otherwise departs from a unimodal response curve (Ter Braak and Prentice 1988). In Fig. 3, Mylothris phileris Pieridae) is a good example of this phenomenon. De- spite its placement in the domain of the disturbed sites, this species occurred abundantly in all of the habitat types sampled.

In contrast, species found at the edges of the ordination diagram are frequently "rare" species, and occur there because they are associated with "extreme" environmental conditions (Ter Braak and Prentice 1988). In Fig. 4, the species scores of the "rare" species (i.e., species individually contributing < 1% ofthe total sample abundance) are plotted. Fig. 4A shows scores for those species whose "rarity" was known by independent observations to be due to sampling error (i.e., species that were frequently observed in one or more of the habitat types at the study site but were not frequently captured during the study sample hours; C. Kremen, personal observations). As expected, ordination scores for these species were spread over the entire ordination diagram due to random sampling error.

Ordination scores for the remaining "rare" species (those that appeared to be low in abundance during both 1988 and 1990 field seasons) are plotted in Fig. 4B. A subset of these species had scores at the edges of the ordination diagram. In particular, several Hen- otesia species were noted that were found only at either old or new ridge sites (see H. ?parva, H. ?obscura, and H. ?turbans). The distributions of these species between sites were consistent although the number of observations was small, and independent records of these species support their habitual occurrence at these or similar sites. H. ?parva and H. ?obscura occurred at both the old ridge sites, and nowhere else, while H. ?turbans occurred at two out of three of the new ridge sites, and nowhere else. These species might qualify as either genuinely rare species characteristic ofthese habitats, or as species that were sampled at the edges of their range. In the latter case, the two species found only at the old ridge sites (H. ?parva and H. ?obscura) may have optima under less disturbed conditions (e.g., primary forest). Further work could distinguish between these two hypotheses to determine whether these

species would make good indicators of undisturbed conditions.

Dominance and diversity characteristics relative to DCA ordination

Characteristic levels of dominance were associated with different site types. In and of itself, ranking sites by dominance (Berger-Parker index) organized the sites: low dominance characterized newly disturbed sites plus new stream sites; and high dominance characterized new ridge, old ridge, and old stream sites (Fig. 5A). Species richness was negatively correlated with dominance (Fig. 5B, r = -0.56, P = .047). No density compensation was observed; sites that were low in species richness did not have a compensatory increase in abundance of one or more species (Fig. 5C).

The dominance index distinguished significantly (ANOVA, Berger-Parker, F = 15.504, P = .0292) between butterfly species assemblages found at old vs. new stream sites, although species richness did not distinguish between the two types of sites (F = 1.681, P = .2856). In contrast, dominance did not distinguish between old and new ridge sites (F= 2.516, P = .2109), although species richness did (F = 30.943, P = .01 15).

400-

*Hen angM

_) :Hen iboM Os SarperM R *Hen angF *Hen undM O ovsacM 5OR HensubF

O 0 ?r andM D HenundF

- oMyiphiM S Hen subM *AcrstrM Hen pauM *

Hen ancM -200- a Jan eurM

-300 -200 -100 0 100 200 300 400 DCA axis 1

FIG. 3. Detrended Correspondence Analysis (DCA) ordination showing the 12 most abundant species. The ordination space is identical to the DCA sites ordination in Fig. 2A. The letters indicate the approximate average locations of new ridge (R), old ridge (OR), disturbed logging trails (D), new stream (S), and old stream (OS). Species scores represent the optima of Gaussian response surfaces. Here they are separated by sex, as in the analysis; however, only the male score is shown for species whose male (M) and female (F) scores are located side by side. Species names are placed as close to the marker as possible, usually to the right. Acr str = Acraea strattipocles, Hen anc = Henotesia ?anceps, Hen ang = Hen- otesia ?angulifascia, Hen ibo = Henotesia ?iboina, Hen sub = Henotesia ?subsimilis, Hen pau = Henotesia ?pauper, Hov sac = Hovala saclavus, Jun eur = Junonia eurodoce, Myl phi = Mylothris phileris, Sar per = Saribia ?perroti, and Str = Stra- bena andriana.




A. r* B. l 400

B. 400_

- - o HenturM U

200- 200-

R O R Os Hen obsM

o o P OR 0 HenparM DO * S I U eR D

* OS a

-200 * -200

-300 -200 -100 0 100 200 300 400 -300 -200 -100 0 100 200 300 400 DCA axis 1 DCA axis 1

FIG. 4. Detrended Correspondence Analysis (DCA) ordination showing the rare species (< 1% of total sample abundance). As in Fig. 3, the letters indicate the approximate average locations of new ridge (R), old ridge (OR), disturbed logging trails (D), new stream (S), and old stream (OS). Species scores, shown by markers, represent the optima of Gaussian response surfaces. Here they are separated by sex, as in the analysis; however, only the male score is shown except when male (M) and female (F) scores did not occur side by side. * = 1 species, LI = 2 species, 0 = 4 species, A = 5 species. (A) Species known from additional observations to be rare due to sampling error (i.e., species commonly seen but seldom captured); scores are distributed across the ordination space. Species are: Acraea fornax (M and F markers shown), Appias sabina, Catopsilia thauruma, Celaenorrhinus humbloti (M and F), Eagris sabadius (M and F), Eicochrysops sanguigutta (M and F), Graphium cyrnus, Henotesia vola (M and F), Heteropsis drepana (M and F), Houlbertia sp., Junonia goudoti, Neptis saclava, Papilio oribazus, Pseudacrea glaucina, Pseudacrea lucretia (M and F). (B) Species that appeared to be truly low in abundance during both 1988 and 1990 field seasons. The rare species near the edges of the ordination diagram may indicate the presence of longer gradients. The species located at the edge of the topography gradient are labeled by name. Henotesia ?obscura (Hen obs) and Henotesia ?parva (Hen par) were only found at old ridge sites, and may be more characteristic of a less disturbed point on the disturbance gradient (e.g., primary forest). Henotesia ?turbans (Hen tur) was found only at new ridge sites. The other markers identify rare species not at the edge of the topography gradient. Species are: Acraea hova, Aterica rabena, Henotesia ankova, Iolaus argentarius, Mylothris smithii, Strabena ?aurivilliusi, Strabena smithii.

Environmental patterns indicated by plant species abundance data

In comparison with the butterfly DCA ordination, the plant DCA ordination was less successful in re- vealing the known environmental gradients (Fig. 6). To test whether plant community data were significantly related to the two known gradients, a CCA was performed using the topographic and disturbance variables (Table 1). The first CCA axis was not significantly related to either of these variables (X = 0.399, P = .47), and there was therefore no need to look at subsequent axes.

Relationship between butterfly species assemblages and plant communities

To examine the relationship between the butterfly and the plant ordinations, DCA site scores from the plant ordination were also included as one of the environmental variables for CCA analysis of the butterfly data; however, no significant relationship was found. As noted earlier, none of the other environmental variables related to vegetative diversity or structure were found to explain significantly the variation along CCA axes of the butterfly data.

To see whether local butterfly diversity was a good indicator of plant species richness or diversity, I used linear regression. Butterfly species richness was not a good predictor of either plant species richness (Mar- galef's index, F = 0.608, r2 = -0.05, P = .45) or plant species diversity (Shannon's index, F = 0.257, r2 = 0.02, P = .62). However, butterfly richness showed a weak relationship with average floral abundance (F= 4.349, r2 = 0.28, P = .061) and was significantly related to average floral richness (F = 18.65, r2 = 0.63, P = .0012) (Fig. 7). Floral richness and abundance were both highest in sunny regions by the larger streams (S2 and S3) and the logging trails (D1-D3), which also had the greatest butterfly species diversity and abundance (Fig. 5C).

DISCUSSION

Ordination techniques are exploratory tools that have frequently been used in community ecology to reveal environmental patterns and to generate ecological hypotheses based on the covariation of species distributions (Gauch 1982a, b, Digby and Kempton 1987). As techniques that examine the relationship between species distributions and environmental gradients, they are inherently useful for testing the indicator properties




of a group of organisms, especially in poorly known systems. However, limitations of these techniques, including ambiguity of interpretation (Gauch 1 982a) and lack of statistical tests, have perhaps prevented a greater utilization (Ter Braak and Prentice 1988). Recent advances that allow bootstrapping of DCA (Knox and Peet 1989 and R. G. Knox, personal communication) and CCA (Ter Braak and Prentice 1988) ordination axes have opened the door for significance testing, and can now be used creatively for testing as well as gen- erating hypotheses. In particular, this facility greatly enhances the utility of ordination techniques for testing

the indicator properties of an assemblage of species and selecting appropriate sets of indicators from within the larger group.

Why and what to monitor?

Before elaborating on a protocol for testing and selecting indicators for natural areas monitoring (see below), it is important to establish the goals of a monitoring program and the desirable qualities of indicators. The general function of a monitoring program is to provide data that can be used for the scientific management of reserves for restoration or maintenance of the composition, structure, and function of natural ecosystems (Franklin et al. 1981, Noss 1990). An ideal monitoring program should therefore be made up of two components: (1) a descriptive component that es- tablishes the baseline in a large, buffered, "pristine" natural area, including study of "normal" fluctuations over time, and (2) an experimental component that evaluates the effects of specific management activities over time using standard ecological experimental design. The details of a monitoring program will be highly specific to a given reserve, depending, for example, on its "naturalness" (how different it is from baseline), size, shape, biotic and physiographic heterogeneity, and on the impacts of anthropogenic and natural disturbance regimes, global changes, and management practices.

Properties of species or species assemblages can serve as indicators at genetic, population-species or community-ecosystem levels for monitoring composition, structure, and function in natural areas; however, many other indicators, both biotic and abiotic, are available for monitoring natural ecosystems at all biological levels (genetic to landscape) (Noss 1990). It is therefore important to determine the unique value of monitoring properties of species or species assemblages. First, monitoring species or species assemblages allows the most direct assessment of fundamental management goals such as the maintenance of viable populations and native biodiversity (Noss 1990). Second, monitoring at this level can provide direct or indirect as- sessments of ecological function (Gilbert 1980), although such information is likely to be narrowly focused (e.g., studies of specific ecological interactions, as in a predator-prey cycle) as compared with ecosystem studies at the landscape level (e.g., studies of functional processes, as in nutrient cycling).

When monitoring of populations or communities is conducted within the context of known environmental change (e.g., due to human or natural disturbances, global changes, or management practices), it can then provide a basis for improved management decision- making. There is little point in using indicators to mea- sure an environmental change that can be measured directly using a pH scale, thermometer, or LANDSAT image, for example; instead, the purpose and value of using indicator organisms is to observe the biotic re-

A.100

80

C 60

40 0

20 o

S2 S3 D2 Dl Si R2 Ri D3 OR1 OR2 OS2 R3 OSi

B. 35 Sites

u) - ADJ 0 30 0 C eC 25 2D

2 5 l- OS 3

'602 D3 20 -2

1 5 CR2 -S2

Z 10 SR1 0. R2 10051

X RI3

C O I -.

0 0 20 40 60 80 100

Dominance

CI. 35. Chrceitc fbtefydvriyand srctr

co30- 0k 25 - ls C A0

020 -CR S2 R

forest (S) d R si

rig ie;OS=odsra sies (B)S Speierihsss

0. 5 C) R3

0 0 0.05 0.1 0.15 0.2

Relative Abundance FiG. 5. Characteristics of butterfly diversity and structure

of species assemblages at each site. (A) Dominance (Berger- Parker index) is lowest in the sunniest sites: streams in new forest (S) and disturbed sites (D). R = ridge sites; OR = old ridge sites; OS = old stream sites. (B) Species richness is negatively correlated with dominance (r = -0.56, P = .041); sunny new streams and disturbed sites have the highest species richness. (C) Species richness is positively correlated with the total relative abundance of butterflies at each site (r = 0.786, P = .001). At the spatial scale of this study, no density compensation for butterfly abundance was observed.



212 CLAIRE KREMEN Ecological Applications Vol. 2. No. 2

sponse to environmental stress, and especially to provide early warnings of natural responses to environmental impacts (Noss 1990). If studies are well designed, such information should also be useful in developing or refining restoration or management plans, and in monitoring the success of these efforts.

An indicator should "be capable of providing a con- tinuous assessment over a wide range of stress" (Noss 1990: 357). Inherently, species assemblages should provide finer gauges of biotic responses than single species, just as precision is increased by placing finer gradations upon a scale. Certain properties of assemblages can be defined that might enhance the likelihood of wide-scale sensitivity to a specific environmental stress. For example, assemblages that include species covering a wide range of spatial heterogeneity (low to high vagility, narrow to broad distribution) and microhabitat specificity (specialist to generalist), and that collectively occupy a broad range of micro-habitats, might be expected to display a greater range of sensitivities to habitat fragmentation and/or modification over time (Terborgh 1974).

For completeness, a monitoring program might use separate indicator species assemblages representing different taxonomic and/or functional groups to monitor different environmental impacts, as will be dealt with more fully in a separate paper. This paper is concerned instead with testing the indicator properties of a uni- form taxonomic/functional group, and selecting a subset of appropriate indicators from within the group for addressing a specific management problem. As the protocol summarized below is widely applicable to different species groups, as well as different spatial scales

and environmental gradients, its objective is to en- courage the testing, selection, and monitoring of multi- species indicator groups from among a broad range of taxonomic groups. This protocol can be used in the absence of detailed autoecological or community data, and it is probably most appropriate to situations where such data are lacking.

Establishing the indicator properties of a species assemblage

The first step in establishing the indicator properties of a species assemblage is to ask whether the chosen group of species is appropriate for indicating a specific ecological pattern at the desired spatial and/or temporal scale. This can be accomplished using indirect gradient analysis techniques such as DCA, which are based simply on patterns of species covariation (Knox and Peet 1989). The pattern of species covariation ei-

200 *S1 PLANTS

150

100 AD2 AD3

C 50 _" oOSl.S2

mS3 0R2 OR1

0 OS Q -~~~~o E Rl D -50 m.R3

AD1 -100

-150 -R2

-200 i . l -200 -100 0 100 200 300

DCA Axis 1

FIG. 6. Detrended Correspondence Analysis (DCA) ordination of sampling sites by plant community data. As in Fig. 2, [ = sites located in the old forest; OR = old ridge, OS = old stream. * = sites in the new forest; R = ridge, S = stream. A = sites along logging trails; D = disturbed. Eigen- values are: axis 1, X = 0.483; axis 2, X = 0.444. Canonical correspondence constrained along topographic and disturbance variables did not produce a significant first axis.

A. 35 ND1 a) 30

C .2) 25

20 - * D2 8D3 /S3 aD) 20 .0 0 ~~~S2 aD 0. 15 OR- C/ 0S R2

1%I0 -Si c10 - R2.

R7OS1

5 R3

0 0 20 40 60 80 100 120 140

Average Floral Abundance (no. plants in flower)

B. 35 iDl

*D1 g 30

.C 25 3 *D2 C.) co 20 S

a) 15 VR1V 0. Si VS

>, 10 _ Rvs aE nR3 g 5

00 2 4 6 8 10

Average Floral Richness (no. types of plants in flower)

FIG. 7. Butterfly richness in relation to floral abundance and richness. Floral abundance and richness values are av- erages of four phenological samples taken during the study period. D = disturbed areas (logging trails), R = new ridge sites, OR = old ridge sites, S = new stream sites, OS = old stream sites. (A) Butterfly richness was weakly dependent on average floral abundance (r2 = 0.28, P = .061). Sunny areas (D1-D3 and S1-S2) had the greatest floral abundance. (B) Butterfly richness was significantly related to average floral richness (r2 = 0.63, P = .0012). Sunny areas had the greatest floral richness.




ther will or will not reflect the ecological pattern or environmental impact of interest. More than one potential indicator property can be assessed simultaneously, dependent on the sampling design, as for topography and disturbance in this study. In addition, the results of a DCA ordination analysis may suggest other indicator properties of the assemblage associated with environmental gradients or factors not explicitly considered within the study design.

Alternative hypotheses of indicator properties can be tested in the second step, replacing the need for ad hoc arguments to justify indicator properties. A direct gradient analysis technique such as CCA can be used to interpret the ordination axes by looking at the significance of correlations between species and environmental data. The CCA analysis can be used to match a species assemblage to environmental factor(s) for which it is a good indicator.

Once the indicator properties of an assemblage are established, the ordination analyses can be employed at two biological levels for monitoring over time. At the community level, the entire assemblage can be monitored, and CCA ordination analysis used to track the community response to the environmental factor of interest (for an example of time-series analysis using ordination, see Wiegleb et al. 1989). For example, bird assemblages known to be good indicators of habitat heterogeneity associated with elevation could be monitored in permanent plots along an elevational gradient over a period of years. Each yeafrs data from a sample plot would then be ordinated as a point in sample space. Imagine that with time each plot shifted along the first CCA axis in the direction of an elevation loss. Therefore, as the true elevation at a plot is unchanged, this indicates that the new species composition at the plot is more similar to lower elevation sites from earlier years. Such a result might be predicted, for example, due to the effects of global warming. As temperatures increase, low-elevation faunas adapted to warmer climates can be expected to replace high-elevation faunas (Murphy and Weiss 1991). The effect of year on changes in species composition along the gradient can also be tested for significance.

At the population/species level, the ordination diagrams can be used to select a subset of the species as indicators for more intensive monitoring (e.g., studies of spatial dispersion, population fluctuations, adapta- tion, gene flow, Noss 1990). In DCA and CCA ordination diagrams, species placements within a plot represent the optima of unimodal species distributions with respect to environmental gradients (Ter Braak and Prentice 1988). In choosing potential indicator species, one looks for species whose optima fall at reference points for the environmental gradient of interest. The use of indirect (DCA) vs. direct (CCA) gradient ordinations should depend on knowledge of the system. CCA ordination plots should be chosen if the impor- tance of a particular gradient is well established, since

species ordination values would then be constrained along the gradient of interest. However, if knowledge of environmental gradients is minimal, DCA plots might be preferred, because this unconstrained ordination may show the response of species to environmental gradients not explicitly considered within the study design.

The selection of indicator species from an ordination diagram of species scores represents a hypothesis about the properties of these species in relation both to the entire assemblage and to environmental gradients. The reliability of these choices can be assessed through either spatial or temporal replication of the study, or through autoecological studies.

Rare species contribute little to establishing the ordination axes (Ter Braak and Prentice 1988). However, it may be useful to identify such species, because their "rarity" may result from being sampled far from their own environmental optima. Presence, in this case, may indicate the existence of a longer gradient or a rare habitat type (Terborgh 1974, Rabinowitz et al. 1986). In an ordination diagram, such species tend to be found at the edges of the diagram (e.g., Fig. 4B). In contrast, ordination scores of species whose rarity results simply from random sampling error will be randomly spread across the diagram (e.g., Fig. 4A). In this study, several uncommon species associated only with old ridge sites (Henotesia ?parva and H. ?obscura) were located at the edges of the ordination diagram (Fig. 4B). These species may prove to be indicators for more undisturbed conditions (e.g., primary forest) than were included in this study.

Indicator properties of a butterfly species assemblage

I used the protocol described above to assess the indicator properties of a butterfly assemblage from a rain forest habitat in Madagascar. The study looked at the response of butterfly communities to variation along topographic/moisture and disturbance gradients in space. Potential indicator properties established in this "snapshot" spatial study would, by inference, be likely to be valid for temporal variation in these same environmental factors.

The DCA ordination of the butterfly data reflected local patterns of habitat heterogeneity associated with both the topographic/moisture and disturbance gradients, which suggests that this assemblage could be considered as an appropriate indicator of these types and this spatial scale of habitat heterogeneity. How- ever, Monte Carlo tests of CCA axes were significant only for the axis principally defined by topography/ moisture. Since the influence of topography was so strong, it may have masked less obvious gradients related to disturbance at each end of the topographic scale. For example, in the DCA ordination shown in Fig. 2A, "gradients" of disturbance ran in opposite directions in the ridge vs. the stream clusters. By an-




alyzing each cluster separately using CCA, it was possible to assess the significance of disturbance in isolation from topography. Disturbance was then found to be a significant variable in CCA ordinations of the ridge cluster, although not for the stream cluster.

To select indicator species I used the DCA ordination, since the knowledge of one gradient of interest, disturbance due to anthropogenic effects, was limited to the historical record of length of time since selective logging. In general, small assemblages of species rather than individual species could be said to characterize a particular point on a gradient (see the Appendix for species names). Similarly, taxocene composition, including level of dominance, identity of dominant species, and species richness, were more characteristic of particular habitat types than were individual species.

Many of the species identified as potential indicators (Fig. 3, 4B, and Appendix) are members of the genus Henotesia (Satyrinae), a species-rich genus containing 41 endemic and one nonendemic species in Madagas- car (D'Abrera 1980). The members of this genus and its tribe (Mycalesini) exhibit some primitive characteristics in Madagascar linking them with Indo-Ma- layan rather than African relatives, and suggesting an early arrival in Madagascar (late Cretaceous or early Tertiary), followed by a long period of evolution in isolation (Miller 1968). This genus appears to have experienced an extensive evolutionary radiation in Madagascar and exhibits high beta-diversity. Many species appear to be restricted to particular forest micro-habitats (canopy vs. understory, ridge vs. riparian), elevational zones, and successional stages, while others are more widely distributed, both in range and habitat type (this study, R. Van Buskirk and C. Kremen, unpublished data, C. Kremen, unpublished manuscript). This range of habitat specificity and the diversity of micro-habitats occupied by members ofthis genus may allow them to serve as an indicator assemblage with a wide range of sensitivities to habitat fragmentation and/ or modification. Species-rich genera resulting from evolutionary radiations and having high beta-diversity may frequently make good indicator assemblages, as will be elaborated more fully elsewhere.

Butterflies as ecological indicators

Other authors have suggested that butterfly diversity could provide an index of plant diversity, since larval host-plant relationships are frequently so specific (Murphy and Wilcox 1986). Interestingly, in this study, butterfly community data were not found to be a good indicator of any environmental gradients related to measured parameters of vegetation structure, plant diversity, or community composition. Nor was butterfly species richness a good predictor of plant richness or diversity at the spatial scale of this study.

While butterfly diversity was not correlated with plant diversity in general, it was strongly correlated with the average diversity of plants in flower. Regions of high

floral and butterfly diversity also supported a higher total abundance of butterflies (compare Figs. 5C and 7B). Not surprisingly, these sites corresponded to the sunniest, most disturbed areas (logging trails and good- sized streams). Judging by both abundance and species composition of butterflies at these sites, these abundant and diverse nectar resources may have attracted forest species to the forest edge (Lovejoy et al. 1986; Brown, personal communication), so that the pool of species observed to occur there included both resident disturbed area species (Pieridae: Eurema floricola; Ly- caenidae: Cacyreus darius, Leptotes rabenafer; Nym- phalidae: Junonia eurodoce) and visiting forest species (Papilionidae: Papilio delalandei, P. oribazus; Nym- phalidae: Pseudacrea lucretia). While forest edges had higher diversity than either disturbed or forest habitat alone, massive, widespread clearing of forests to create edges would not enhance landscape-level diversity, since it reduces habitat area for forest specialists (e.g., Hen- otesia ?anceps, H. ?parva; see also Wilcove et al. 1986).

In this study, significance tests of CCA ordination axes allowed evaluation of the hypothesis that butterflies are good ecological indicators (Gilbert 1980, 1984, Pyle 1980, Brown 1982, Murphy et al. 1990). Are butterflies in fact good ecological indicators? In some sense, the presence, absence or abundance level of any organism must always indicate something about the biotic or abiotic environment, and the question is therefore trivial. The question should be: what does the presence/absence/abundance of species X or the composition of species assemblage Yindicate? In this study, butterfly taxocene composition proved to be an excellent indicator of heterogeneity due to anthropogenic disturbance by logging (although coding of disturbance levels may have been a factor obscuring the relationships between the disturbance gradient and butterfly data), and a poor indicator of plant species richness and diversity. Given their strong response to topographic/moisture effects, one likely utility of butterflies as ecological indicators will be for monitoring the biotic response to climate change, as has also been suggested elsewhere (Weiss and Murphy 1988, Murphy et al. 1 990).

Conclusion The modification of natural environments is accel-

erating at an alarming rate, and the responses of organisms to these changes provide information not only on the viability of global life support systems, but also on the efficacy of our protected area networks in main- taining biological diversity. A substantial research effort is needed to select appropriate indicators and to design monitoring programs. Ordination techniques provide a simple method for establishing the indicator properties of target assemblages of organisms. Contin- ued use of these techniques in testing properties of assemblages and selecting multi-species indicator groups for intensive monitoring could increase the tax-




ic diversity and sensitivity of natural areas monitoring programs.

ACKNOWLEDGMENTS

This manuscript benefited greatly from the comments of P. Brussard, R. Colwell, A. Moldenke, D. Murphy, and an anonymous reviewer. The work would not have been possible without the assistance of the Malagasy Ministries of Higher Education and Water and Forests, who granted permission to conduct this work at the Ranomafana National Park, P. Wright who provided access to the Duke University field station, and P. Daniels who arranged logistics. R. Van Epps and L. Rasabo helped with field work. D. Harvey and E. Holland assisted with preparing and drawing butterfly geni- talia. R. Knox, A. Moldenke, and M. Palmer were invaluable resources on ordination techniques. Research grants from the Douroucouli Foundation and the Duke University Research Council, an equipment grant and Short-term Visitor's grant from the Smithsonian Institution are gratefully acknowledged. The author was a staff scientist at the Xerces Society and the Duke University Center for Tropical Conservation while this work was being conducted, and is grateful for the support and encouragement of these organizations.

LITERATURE CITED

Brown, K. S. 1982. Paleoecology and regional patterns of evolution in neotropical forest butterflies. Pages 255-308 in G. T. Prance, editor. Biological diversification in the tropics. Columbia University Press, New York, New York, USA.

Burgess, R. L., and B. M. Sharpe. 1981. Forest-island dynamics in man-dominated landscapes. Springer-Verlag, New York, New York, USA.

Cairns, J. 1985. Multi-species toxicity testing. Pergamon, New York, New York, USA.

. 1986. The myth of the most sensitive species. BioScience 36:670-672.

Carleton, J. T. 1989. Man's role in changing the face of the ocean: biological invasions and implications for conservation of near-shore environments. Conservation Biology 3:265-274.

Cohen, A. C. J. 1959. Simplified estimators for the normal distribution when samples are singly censored or truncated. Technometrics 1:217-237.

. 1961. Tables for maximum likelihood estimates: singly truncated and singly censored sample. Technomet- rics 3:535-541.

D'Abrera, B. 1980. Butterflies of the Afro-Tropical region. Lansdowne Editions, East Melbourne, Australia.

Digby, P. G. N., and R. A. Kempton. 1987. Multivariate analysis of ecological communities. Chapman and Hall, London, England.

Ehrlich, P. R., D. E. Breedlove, P. F. Brussard, and M. A. Sharp. 1972. Weather and the "regulation" of subalpine populations. Ecology 53:243-247.

Ehrlich, P. R., and D. D. Murphy. 1987. Conservation lessons from long-term studies of checkerspot butterflies. Con- servation Biology 1: 122-13 1.

Ehrlich, P. R., and P. H. Raven. 1964. Butterflies and plants: a study in coevolution. Evolution 18:586-608.

Evans, W. H. 1937. A catalogue of the African Hesperiidae indicating the classification and the nomenclature adopted in the British Museum. British Museum, London, England.

Fajer, E. D. 1989. How enriched carbon dioxide environments may alter biotic systems even in the absence of cli- matic changes. Conservation Biology 3:318-320.

Fajer, E. D., M. D. Bowers, and F. A. Bazzaz. 1989. The effects of enriched carbon dioxide atmospheres on plant- insect herbivore interactions. Science 243:1198-1200.

Franklin, J. F., K. Cromack, W. Denison et al. 1981. Eco- logical characteristics of old-growth Douglas-fir forests. United States Forest Service General Technical Report PNW-118.

Gauch, H. G. 1982a. Multivariate analysis in community ecology. Cambridge University Press, Cambridge, England.

. 1982b. Noise reduction by eigenvector ordinations. Ecology 63:1643-1649.

Gilbert, L. E. 1980. Food web organization and the conservation of neotropical diversity. Pages 1 1-34 in M. E. Soule and B. A. Wilcox, editors. Conservation biology: an evolutionary-ecological perspective. Sinauer Associates, Sunderland, Massachusetts, USA.

1984. The biology of butterfly communities. Pages 41-54 in R. I. Vane-Wright and P. R. Ackery, editors. The biology of butterflies. Princeton University Press, Prince- ton, New Jersey, USA.

Gilbert, L. E., and J. T. Smiley. 1978. Determination of local diversity in phytophagous insects: host specialists in tropical environments. Symposium of the Royal Entomo- logical Society of London 9:89-104.

Jackson, D. A., K. M. Somers, and H. H. Harvey. 1989. Similarity coefficients: measures of co-occurrence and as- sociation or simply measures of occurrence? American Nat- uralist 133:436-453.

Jennersten, 0. 1988. Pollination in Dianthus deltoides (Car- yophyllaceae): effects of habitat fragmentation on visitation and seed set. Conservation Biology 2:359-367.

Kimball, K. D., and S. A. Levin. 1985. Limitations of lab- oratory bioassays: the need for ecosystem-level testing. BioScience 35:165-17 1.

Kingsolver, J. G. 1983a. Thermoregulation and flight in Colias butterflies: elevational patterns and mechanistic limitations. Ecology 64:534-545.

1983b. Ecological significance of flight activity in Colias butterflies: implications for reproductive strategy and population structure. Ecology 64:546-551.

Klein, B. C. 1989. Effects of forest fragmentation on dung and carrion beetle communities in central Amazonia. Ecol- ogy 70:1715-1725.

Knox, R. G., and R. K. Peet. 1989. Bootstrapped ordination: a method for estimating sampling effects in indirect gradient analysis. Vegetatio 80:153-165.

Krebs, C. J. 1989. Ecological methodology. Harper & Row, New York, New York, USA.

Kushlan, J. A. 1979. Design and management of continental wildlife reserves: lessons from the Everglades. Biological Conservation 15:281-290.

Landres, P. B., J. Verner, and J. W. Thomas. 1988. Eco- logical uses of vertebrate indicator species: a critique. Con- servation Biology 2:316-329.

Lovejoy, T. E., R. 0. Bierregaard, A. B. Rylands, J. R. Mal- colm, C. E. Quintela, L. H. Harper, K. S. Brown, A. H. Powell, G. V. N. Powell, H. 0. R. Schubart, and M. B. Hays. 1986. Edge and other effects of isolation on Amazon forest fragments. Pages 257-285 in M. E. Soule, editor. Conservation biology: the science of scarcity and diversity. Sinauer Associates, Sunderland, Massachusetts, USA.

Magurran, A. E. 1988. Ecological diversity and its mea- surement. Princeton University Press, Princeton, New Jer- sey, USA.

Miller, L. D. 1968. The higher classification, phylogeny and zoogeography of the Satyridae (Lepidoptera). Volume 24. American Entomological Society at the Academy of Nat- ural Sciences, Philadelphia, Pennsylvania, USA.

Murphy, D. D., K. E. Freas, and S. B. Weiss. 1990. An environment-metapopulation approach to population viability analysis for a threatened invertebrate. Conservation Biology 4:41-51.

Murphy, D. D., and S. B. Weiss. 1991. Predicting the effects




of climate change on biological diversity in westem North America: species losses and mechanisms. In R. L. Peters, editor. The consequences of greenhouse warming on biological diversity. Yale University Press, New Haven, Con- necticut, USA, in press.

Murphy, D. D., and B. A. Wilcox. 1986. Butterflv diversity in natural habitat fragments: a test of the validity of vertebrate-based management. Pages 287-292 in J. Verner, M. L. Morrison, and C. J. Ralph, editors. Wildlife 2000, mod- eling habitat relationships of terrestrial vertebrates. Uni- versity of Wisconsin Press, Madison, Wisconsin, USA.

Noss, R. F. 1990. Indicators for monitoring biodiversity: a hierarchical approach. Conservation Biology 4:355-364.

Paulian, R. 1956. Faune de Madagascar, II. Insectes Lepi- dopteres, Danaidae, Nymphalidae, Acraeidae. Institut de Recherche Scientifique, Antananarivo, Madagascar.

Paulian, R., and P. Viette. 1968. Faune de Madagascar, XXVII. Insectes Lepidopteres Papilionidae. Institut de Re- cherche Scientifique, Antananarivo, Madagascar.

Peet, R. K., R. G. Knox, J. S. Case, and R. B. Allen. 1988. Putting things in order: the advantages of detrended correspondence analysis. American Naturalist 131:924-934.

Peters, R. L., and J. D. S. Darling. 1985. The greenhouse effect and nature reserves. BioScience 35:707-717.

Pollard, E. 1977. A method for assessing changes in the abundance of butterflies. Biological Conservation 12:115- 124.

Pyle, R. M. 1980. Butterfly eco-geography and biological conservation in Washington. Atala 8:1-26.

Rabinowitz, D., S. Cairns, and T. Dillon. 1986. Seven forms of rarity and their frequency in the flora of the British Isles. Pages 182-204 in M. E. Soule, editor. Conservation biology: the science of scarcity and diversity. Sinauer Associates, Sunderland, Massachusetts, USA.

Rausher, M. D. 1979. Larval habitat suitability and oviposition preference in three related butterflies. Ecology 60: 503-511.

Rausher, M. D., and P. Feeny. 1980. Herbivory, plant density, and plant reproductive success: the effect of Battus philenor on Aristolochia reticulata. Ecology 61:905-917.

Schreiber, R. K., and J. R. Newman. 1988. Acid precipi- tation effects on forest habitats: implications for wildlife. Conservation Biology 2:249-259.

Sharp, M. A., D. R. Parks, and P. R. Ehrlich. 1974. Plant resources and butterfly habitat selection. Ecology 55:870- 875.

Singleton, T. A., and S. P. Courtney. 1991. Critical resources for an endangered butterfly: using habitat selection to solve management problems. Conservation Biology, in press.

Soule, D. F. 1987. Marine organisms as indicators: reality or wishful thinking. Pages 1-13 in D. F. Soule and G. S. Kleppel, editors. Marine organisms as indicators. Springer- Verlag, New York, New York, USA.

Soule, M. E. 1986. Conservation biology: the science of scarcity and diversity. Sinauer Associates, Sunderland, Massachusetts, USA.

1990. The onslaught of alien species, and other challenges in the coming decades. Conservation Biology 3: 233-240.

Soule, M. E., and B. A. Wilcox. 1980. Conservation biology: an evolutionary-ecological perspective. Sinauer Associates, Sunderland, Massachusetts, USA.

Statistical Graphics Corporation. 1989. Statgraphics. STSC, USA.

Swanson, F. J., T. K. Kratz, N. Caine, and R. G. Woodman- see. 1988. Landform effects on ecosystem patterns and processes. BioScience 18:92-98.

Terborgh, J. 1974. Preservation of natural diversity: the problem of extinction prone species. BioScience 24:715- 722.

Ter Braak, C. J. F. 1987. The analysis of vegetation-environment relationships by canonical correspondence analysis. Vegetatio 69:69-77.

. 1988. CANOCO-an extension of DECORANA to analyze species-environment relationships. Vegetatio 75: 159-160.

Ter Braak, C. J. F., and I. C. Prentice. 1988. A theory of gradient analysis. Advances in Ecological Research 18:271- 317.

Thomas, C. D., and H. C. Mallorie. 1985. Rarity, species richness and conservation: butterflies of the Atlas moun- tains in Morocco. Biological Conservation 33:95-117.

Thomas, J. A. 1983. A quick method forestimatingbutterfly numbers during surveys. Biological Conservation 27:195- 211.

Vane-Wright, R. I. 1978. Ecological and behavioural origins of diversity in butterflies. Symposium of the Royal Ento- mological Society of London 9:56-70.

Viette, P. 1956. Faune de Madagascar, III. Insectes Lepi- dopteres Hesperiidae. Institut der Recherche Scientifique, Antananarivo, Madagascar.

Ward, D. V. 1978. Biological environmental impact studies: theory and methods. Academic Press, New York, New York, USA.

Watt, W. B., F. S. Chew, L. R. G. Snyder, A. G. Watt, and D. E. Rothschild. 1968. Population structure of pierid butterflies. I. Numbers and movements of some montane Colias species. Oecologia (Berlin) 27:1-22.

Weiss, S. B., D. D. Murphy, and R. R. White. 1988. Sun, slope and butterflies: topographic determinants of habitat quality for Euphydryas editha. Ecology 69:1486-1496.

Weiss, S. B., R. R. White, D. D. Murphy, and P. R. Ehrlich. 1987. Growth and dispersal of larvae of the checkerspot butterfly Euphydryas editha. Oikos 50:161-166.

Wiegleb, G., H. Herr, and D. Todeskino. 1989. Ten years of vegetation dynamics in two rivulets in Lower Saxony (FRG). Vegetatio 82:163-178.

Wilcove, D. S., C. H. Mclellan, and A. P. Dobson. 1986. Habitat fragmentation in the temperature zone. Pages 237- 256 in M. E. Soule, editor. Conservation biology: the science of scarcity and diversity. Sinauer Associates, Sunder- land, Massachusetts, USA.




APPENDIX Butterfly and skipper species list for the study.

Potential Potential Species indicator* Species indicator*

Papilionidae Acraea hova Graphium cyrnus Acraea strattipocles Papilio delalandei Nymphalinae Papilio oribazus Aterica rabena

Pieridae Junonia eurodoce Appias sabina confusa Junonia goudoti Catopsilia thauruma Neptis kikideli Eurema floricola floricola Neptis saclava saclava Leptosia nupta viettei Pseudacrea glaucina Mylothris phileris Pseudacrea lucretia apaturoides Mylothris smithii Salamis anteva

Nymphalidae Salamis dupreii

Satyrinaet Riodinidaet Henotesia ?anceps S Saribia ?perroti R Henotesia ?angulifascia OS Saribia ?tepahi Henotesia ?ankova Lycaenidae Henotesia ?iboina R Cacyreus darius Henotesia ?obscura OR Iolaus argentarlus Henotesia ?parva OR Iolaus mermeros Henotesia ?pauper S Eichochrsops sanguigutta Henotesia ?subsimilis Eptotes r afer Henotesia ?turbans R Leptotes rabenafer Henotesia ?undulosa Hesperiidae Henotesia ?vola Celaenorrhinus humbloti Heteropsis drepana Eagris sabadius Houlbertia sp. Fulda coroller Strabena andriana OR Galerga hyposticta Strabena ?aurivilliusi Hovala dispar Strabena rakoto Hovala pardalina Strabena smithii Hovala saclavus OR Strabena triopthalma Miraja varians Strabena sp. Tagiades insularis

Acraeinae Acraea cuva villettei Acraea fornax

* S = stream; OS = old stream; R = ridge; OR = old ridge. t The taxonomy of some Malagasy butterfly taxa is poorly established. Species names preceded by a question mark refer

to morphospecies and represent an attempt to synonymize these forms with established names.



Assessing the Indicator Properties of Species Assemblages ... · Ecological Applications, 2(2),...

Documents

Transcript of Assessing the Indicator Properties of Species Assemblages ... · Ecological Applications, 2(2),...