Biodiversity

and Conservation

4,35-55

(1995)

Preservation of biodiversity in small rainforest patches: rapid evaluations using butterfly trapping GRETCHEN Energy

PAUL Center

C. DAILY*

and Resources

Group,

T-4, Room

100, University

of California,

CA 94720 USA

R. EHRLICH for Conservation

Biology,

Stanford

University,

Stanford,

CA 94305 USA

Received 3 January 1994; revised and accepted 11 July 1994 Determining the capacity of small forest remnants to support biodiversity is of critical importance, especially in the tropics where high rates of land conversion coincide with extraordinarily high species richness and endemism. Using fruit-baited traps, we conducted rapid evaluations in 1993 and 1994 of the forest butterlly diversity of seven small patches (3-30 ha) and a single remaining large patch (227 ha) of Costa Rican mid-elevation moist forest. Our results suggest that even recently isolated 20-30 ha fragments of primary forest retain surprisingly depauperate butterfly faunas relative to that supported by the 227 ha patch only OS-l.0 km away. If forest butterflies are an index of the diversity of small-bodied organisms in general, preservation of the latter may require unexpectedly large patches. In 1994 we also surveyed a 16 ha botanical garden, situated between and contiguous to both the 227 ha patch and an exceptionally species-rich 25 ha patch. In the garden, we discovered adults of many butterfly species associated with forest interior, suggesting that even heavily managed systems of largely exotic plants (such as agricultural systems) could be designed to serve as corridors for butterflies and perhaps some other groups of organisms. We discuss some implications for a planned restoration of biotic connections between lowland and montane forests in southern Costa Rica. biodiversity; butterflies; conservation; corridors; restoration; speciesrichness;tropical moist forest fragments Keywords:

Introduction The rainforests of the world are rapidly being reduced to a few large tracts with varying degrees of protection, and numerous relatively small, usually degraded, patches embedded in extensive areas converted to agricultural use (e.g. Skole and Tucker, 1993). The inevitable continuation of this trend raises two important conservation-related questions: how effective are these small patches in supporting biodiversity, and how can they be made more effective (e.g. Bierregaard et al., 1992)? Some answersare clear from first principles a 20 ha patch will not maintain a population of jaguars, Sumatran rhinos, or monkey-eating eagles (e.g. Terborgh, 1976). But will such a patch maintain diverse faunas of, say, butterflies, beetles, ants, frogs, rodents, or passerine birds? The answers to these questions have far-reaching implications for the nature of the *To whom correspondence should be addressed. 0960-3115 0 1995 Chapman & Hall

36

Daily and Ehrlich

social, economic, and political incentives and commitment required to protect and restore Earth’s biodiversity. The small patches of rainforest that persist in agricultural landscapes typically cover terrain that is not desired, or even marginally suitable, for other uses; less commonly, they may be preserved for religious reasons (e.g. Gadgil and Guha, 1992). The prospects for protecting existing small patches from outright clearing (as opposed to protection from less severe or indirect impacts) remain relatively good, and even augmenting their numbers may be feasible in many parts of the world without overwhelming difficulty. But, if large tracts are needed to protect the ‘little things that run the world’ (Wilson, 1987), then the preservation of many elements of biodiversity will be much more problematic. Unfortunately, the time and logistic support required for comprehensive ecological investigations on which to base conservation decisions are rarely available today. especially in the tropics. These circumstances make crucial the development and validation of techniques to rapidly compare the capacity of different sites to support biodiversity. We describe here two rapid evaluations of insect diversity, conducted in different years in the same series of small (3-30 ha) remnant forest patches and in a more extensive, protected forest (227 ha). We used butterflies as a surrogate for insects in general, and butterllies that are attracted to decaying fruit as surrogates for all butterflies. This paper thus provides a preliminary assessment of the effectiveness of remnant tropical forest patches in supporting the diversity of small-bodied organisms, and starts laying a groundwork for devising scientifically sound rapid evaluation techniques. In addition, our initial results may provide some guidance for an actual restoration project now being planned. Materials

and methods

The work was done in the vicinity of the Las Cruces Biological Field Station of the Organization of Tropical Studies (OTS/OET), Coto Brus, Costa Rica from 16 March to 20 April 1993 and from 12 February to 5 March 1994. Our basic experimental design involved sampling the butterfly faunas in a series of forest patches of various sizes, while controlling for other inter-patch differences as much as possible. Patches

The Las Cruces Biological Field Station encompasses our 227 ha ‘control’ tract of primary Pacific mid-elevation forest (211 ha) and mature second growth (16 ha), which constitutes the largest forest within a ca. 10 km radius. It is surrounded by a landscape of pastures (some badly eroded), coffee plantations, and towns, dotted with small patches of both primary and secondary forest of variable size (0.1-30 ha) and integrity. The Las Cruces forest is contiguous with the managed Wilson Botanical Garden (16 ha). We identified potential small study patches on recent (1992) aerial photographs and then visited each to determine accessibility and similarity to other possible sites (in terms of elevation, level of disturbance, surrounding habitat, etc.). We did not include patches less than 3 ha in extent; no patches in the 40-200 ha size range remain in the vicinity of Las Cruces. We selected seven small patches, six of which were situated along a ridge that traverses the farming landscape between Las Cruces and the large forest tract of the Guaymi-Coto Brus Indian Reservation (Fig. 1). These six patches were surrounded by pasture under

Butter-y

biodiversity

in rainforest

37

patches

heavy use at the time of the study, and were accessible by dirt road (requiring 4-wheel drive) followed by hiking. The seventh small patch, Cascada, abutted an abandoned pasture, and was separated from the Las Cruces forest by a tree-lined, paved road and the managed Wilson Garden. These forest patches have been isolated for 18-33 years. With the help of Dr Luis Diego Gomez, we characterized various aspects of the patches, including successional stage and disturbance (Table 1). For reference, in each table, the patches are listed in an order that constitutes an a priori hypothesis of decreasing habitat quality for forest butterflies, based on a subjective evaluation of the combined effects of size, degree of isolation, and within-patch characteristics. The sizes of and distances between patches were estimated from the aerial photographs after making rough topographic corrections. We measured slopes with a Silva Precision Compass and Clinometer (Ranger Type 15CL); canopy cover above traps with a Robert Lemmon Model-C spherical densiometer; and, elevation with a Swift altimeter/barometer model no. 478. Trap design

We made 40 Van Someren-Rydon traps (Fig. 2) cylinders of grey nylon netting (65 cm high and 25 cm in diameter), sewn onto a frame of two wire loops, closed at the top, open at the bottom, with a Velcro-fastened slit in the side for removing trapped insects. Each trap had a sheet of plywood (40 X 40 cm’) suspended 6 cm below the netting, with a disposable plastic dish 15cm in diameter in the centre to hold bait. The bait consisted of mashed, rotting bananas, a liberal dose of molasses, and a dash of rum. Trapping

routine in the eight patches

Five traps were hung in each patch in 1993 and four (in four of the original trap locations) were hung in each patch in 1994. Approximately 2 h were required to emplace the traps in a

Labrador

stream

-

dirt toad paved road

m

buildings

0.5

I

Figure 1. Map of the forest patches.

-

-

km

2s

30

20

5

5

3

3

Cascade

Ridge Road Ravine

Vaca Vaca

Lower Gamboa Swamp

t Jpper Gamboa Swamp

Labrador Stream

Labrador Slope

I390

1370

2.4

2.2

1.1

1.0

1400

1400

0.5

1.4

05

10

1400

1420

1200

1140

elevation (ml

Approximate distance from nearest > 200ha patch (km)

20-30

20-30

15

c 10

Cl5

s 10

15-20

1.5

Disturbance at trap locations” (%I Primary forest with some edge disturban~ Highly disturbed primary forest with advanced second growth Primary forest; some edge areas disturbed (< 20%) due to cattle Primary forest; some edge areas disturbed (c 30% ) due to cattle and logging Very advanced second growth with remnant emergent trees Highly disturbed primary forest Highly disturbed second growth smaller edge: area than Labrador Slope Highly disturbed second growth

Successional stage and other characteristics

Mean and range of canopy closure at trap locations

34

(32-%,

1s (12-18)

23 ( 1S-28)

(Ii%,

(IZO,

Mean and range of slope at trap locations

‘Percent disturbance is a measure of the percentage of tree species missing because of anthropogenic disturbance; thus, natural disturbance (e.g.. due to tree fall or stream course) is not reflected in this measure. Most anthropogenic disturbance is caused by selective logging and cattle grazing.

227

Estimated size (ha)

of the study patches

Las Cruces (unmanaged area only)

Patch

Table 1. Descriptions

a 2 3

B 5

ki

Butterfly

biodiversity

in rainforest

patches

39

Figure 2. Schematic of a butterfly trap. single site. They were generally suspended l-3 m above the ground, depending on slope and surrounding vegetation. Experience with traps hung ca. 5 m above the ground yielded no apparent difference in fauna, but individuals frequently escaped upon lowering the trap (personal experience; H. Sparrow, pers. comm.). To control for edge effects partially, and because precipitous slopes made trapping in the centres of some patches virtually impossible, we placed the traps in each site along a rough transect near and parallel to a forest-pasture edge. Exact trap placement was determined by the presence of at least a small light-gap (because canopy species appear to descend in small clearings) (see Table l), the availability of a suitable hanging limb, and accessibility (especially with respect to slope). The mean trap-edge distance was 23 (f 17) m. Traps were ordinarily baited on one day and cleared two (and occasionally one or three) days later. Logistical difficulties and weather (an unusually early start to the rainy season in 1993) accounted for the variation; about one and a half hours was required to identify and release the captured butterflies in a site. To detect the transfer of individuals between patches and to determine the rate of recapture within a single patch, we marked uniquely all individuals of species other than Cissia satyrina, using a Sanford fine point permanent marker. Cissia satyrina were so numerous that the additional time required for marking them would have substantially reduced sampling intensity. In 1993, each patch was sampled between four and nine times, varying with patch size

40

Daily and Ehrlich

and logistical difficulties. More frequent sampling of the larger patches was done to equalize partially the sampling effort per unit area, but our key statistical comparison for that year is based conservatively on the first five trap days. In 1994, each patch was sampled six or seven times as part of the regular protocol. Trapping routine in the Cuscada-Wilson

Garden-Las

Cruces Complex

To investigate the possibility that the managed garden serves as a corridor (if not suitable habitat) between Cascada and Las Cruces, all traps were transferred from the ridge patches to the Wilson Garden-Las Cruces area after completing the patch sampling experiment in 1994. The traps along the Las Cruces ‘traditional’ transect and the Cascada transect were left in place. This also permitted an estimation of(i) the variance in butterfly diversity sampled within the Las Cruces forest; and, (ii) the relative total diversity in Las Cruces based on a sampling intensity per unit area more similar to (but still much lower than) that in the smaller patches. Eight of the relocated traps were distributed through the garden grounds. four each near the edges opposite the Cascada and Las Cruces patches. Twelve additional traps were placed in two sets of six in Las Cruces over 50m from forest edge, at either end of the traditional transect. This series of 20 traps in total, along with the Cascada and the traditional Las Cruces traps, were surveyed on 5 days each. Unless otherwise specified. the following analyses treat separately the two sets of Cascada and Las Cruces data collected in 1994. Analysis

The statistical methods used are described in Sokal and Rohlf (1981). Most comparisons of the tabulated data are by R x C tests of independence, for which William’s correction of G-values is used. Rank correlations were made using Kendall’s coefficient of rank correlation, tau.

Results Surveying of the eight patches

Substantial interpatch differences in butterfly fauna (summarized in Table 2; refer also to Appendix 1 and 2) became apparent early in the study. Species accumulation curves for 1993 are presented both for all species (Fig. 3) and for all except satyrines (Fig. 4). The vast majority of satyrines are grass-feeders as larvae, and roughly 70% of the satyrine species we captured are characteristic of open areas or forest edges (DeVries, 1987). Morph0 peleides (Morphinae) and the two Caligo species (Brassolinae) are considered forest butterflies (DeVries, 1987) but all are seen at edges. Thus, the fruit-eating charaxines and nymphalines, especially, and to a lesser extent the morphines and brassolines, are probably better indicators of forest-interior conditions than the satyrines. The large Las Cruces forest and adjacent Cascada patch yielded much richer forest-interior butterfly faunas than any of the other patches. In 1993,13 and 17 individuals in the subfamilies Charaxinae and Nymphalinae (representing at least six species in each case) were trapped in Las Cruces and Cascada, respectively. In contrast, at most two charaxine individuals and no nymphaline individuals were trapped in any of the other

of nymphaline

of Cissia satyrina

Number

Number

“Approximate

trapped

1993 1994

20 55

7 25

1993 1994

4 2 0

2

8

3 0

a8

34

214

1993 1994

1993 1994 1993 1994

1993 1994 1993 1994

9 7 34 45

the escape of a few individuals

number

reflect

trapped

trapped

captures

individuals

species

in total site

numbers

Percent C. satyrina of captures within

of nymphaline

Number

individuals

of charaxine

trapped”

trapped

1993 1994 1993 1994

Las Cruces

statistics for the patches

species trapped”

of species

Number

number

Total

of individuals

of charaxine

number

Total

of visits

trapping

Number

number

Total

Table 2. Summary

from

8 6

2 4

6 2 2

93 89

101 74

certain

0 1

2 0 0 1

28 7 2 0

6 6 108 83

the traps before

9 13

214

214 216 24 4

9 7 63 46

Cascada

Ridge Road Ravine

species

92 77

155 49

0 0

1 Z=l 0 0

7 25 1 1

I 7 167 63

Vaca Vaca

identification

98 96

96 52

0 0

0 1 0 0

3 3 0 1

4 7 98 54

Upper Gamboa Swamp

0 0

0 0

had been made.

29 23 83 71

0 0

0 0

58 84 97 90

0 1

5 10 0 1

3 5 0 0 0 0

5 7 34 32

Labrador Slope

5 6 60 93

Lower Gamboa Swamp

20 57 67 93

0 1

0 1

1 0

5 4 1 0

5 6 29 61

Labrador Stream

;f: 2 % f? s& c

5

R k -. 3 a

s. 2.’ (D 2

Daily and Ehrlich

42 15 -

loTotal Number Of

Species 5-

0

’

I

1

I

1

2

3

I

I

4 5 Sample Day

1

I

I

r

6

7

8

9

Figure 3. Species accumulation curves among patches. 1993, all species. * Las Cruces; x Cascada; 0 Ridge Road Ravine; o Vaca Vaca; * Upper Gamboa Swamp: 0 Lower Gamboa Swamp; l Labrador Stream; W Labrador Slope.

patches. The smallest patches (5 ha in area and smaller) supported especially depauperate faunas dominated by satyrines, most notably C. satyrina. In 1994, at an earlier stage of a much dryer season than in 1993, the apparent species richness of Las Cruces was diminished relative to the previous year (Figs 5 and 6). Although more charaxine species and fewer C. satyrina were present than in any of the small patches other than Cascada, the difference between Las Cruces and these patches

Total Number of Species

Sample Day Figure 4. Species accumulation curves among patches, 1993, excluding satyrines. k Las Cruces; x Cascada; 0 Ridge Road Ravine: o Vaca Vaca; * Upper Gamboa Swamp; 0 Lower Gamboa Swamp: l Labrador Stream; n Labrador Slope.

Butterfly

biodiversity

in rainforest

patches

43

Total

Number of Species

”

i

rl

4

i

;

fi

;

Sample Day

Figure 5. Speciesaccumulation curves among patches, 1994, all species.* Las Cruces;x Cascada;0 Ridge Road Ravine; 0 Vaca Vaca; * Upper Gamboa Swamp; 0 Lower Gamboa Swamp; l Labrador Stream; n Labrador Slope. was lessstriking and no nymphalines were captured in the initial seven samplesfrom 1994 (Table 2). We obtained contrasting results from the subsequent five days of sampling Las Cruces in 1994 (described in the following subsection). Lumping all species, in 1993 there was a weak positive rank correlation between butterfly diversity and patch size after 5 days of trapping at each site (tau = 0.628; 12= 8; tayo.oq = 0.643; p = 0.05), assuming that Upper Gamboa Swamp would have yielded no IO-

Total

Number of 5Species

0

I1

,2

I‘

I

3 4 Sample Day

I

5

1

6

I

7

Figure 6. Speciesaccumulationcurves amongpatches,1994,excluding satyrines. * Las Cruces; x Cascada; 0 Ridge Road Ravine; 0 Vaca Vaca; * Upper Gamboa Swamp; 0 Lower Gamboa Swamp; 0 Labrador Stream; W Labrador Slope.

44

Daily and Ehrlich

additional species after a fifth day of trapping. Lumping all species after six days of trapping at each site in 1994, there was no significant correlation between diversity and patch size (tau = 0.519: n = 8; tat+,,,,, = 0.571; p > 0.10). In both years, however, the abundance of charaxines and nymphalines was substantially higher in Las Cruces and Cascada relative to the other patches (Table 2: G-tests; 1993: d.f. = 2, G = 53.2,~ + 0.001; 1994: d.f. = 2, G = 21.688.~ < 0.001). The monotony of the trapping in the six patches other than Las Cruces and Cascada was striking (Table 2 and Appendix 1): the fraction of the catch made up by C. satyrina was close to 90% or more in four of the patches in both years. As many as 16 individual C‘. satyrina were found simultaneously in a single trap and more than 150 captures were made in Vaca Vaca in 1993. In contrast, in 1993 the Las Cruces and Cascada traps yielded only seven and eight C. satyrina captures (20 and 9%) respectively. This difference is highly significant in both years (G- tests; 1993: d.f. = 7; G = 362.566;~ 6 0.001, where Las Cruces and Cascada make the entire contribution to the difference: 1994: d.f. = 7; G = 542.55: p < 0.001, where Las Cruces and Cascada are significantly different from all other patches with the exception of a non-significant difference between Las Cruces and satyrine-rich Labrador Slope (d.f. = 1: G = 6.156; p > 0.0s). The two Cal&o species were relatively common in all of the patches in both years. Other brassolines occurred sporadically. One Opsiphanes cassina was taken at Upper Gamboa Swamp in 1993 and one at Ridge Road Ravine in 1994. Opsiphanes quiteria was taken only once in each year, in Cascada and in Las Cruces. Single captures of the rare Catobfepia orgetorix were made at Ridge Road Ravine in 1993, and then in Cascada and Labrador Slope in 1994. In 1993. one individual of M. peleides was found in each of the four largest patches except Cascada, which yielded 14 individuals; one individual was also found in Labrador Slope, the most degraded patch. In 1994. this butterfly was observed only in Las Cruces and Cascada. Recapture rates (calculated as the number of recaptures divided by the total number of captures for a species in a patch) were generally low. lndividuals of most species were never recaptured. The Morph0 and Caligo species had the highest recapture rates, on average, ranging between 25 and 50%. We did not detect the transfer of any marked individuals between patches. In over 300 person-hours of walking through pastures to and from the patches (by the authors), only three Morph0 peleides individuals and a Memphis individual, probably M. beatrix, were observed crossing open pasture. The only other species found in the traps and observed in the open were C. renatu and C. satyrina. Surveying of the Cascada- Wilson Garden-Las

Cruces complex

Five days of additional sampling in 1994 in the Cascada-Wilson Garden-Las Cruces complex yielded four basic results (Table 3 and Appendix 2). First, the diversity in Cascada was comparable to that sampled there in the two earlier rounds (in 1993-94). Second, the diversity sampled along the traditional Las Cruces transect was somewhat higher than that sampled there just days earlier in 1994, though not as high as in any of the three Cascada samples (Tables 2 and 3). There were two charaxine and two nymphalinc individuals (each representing a different species). and also relatively few C. satvrina. Third, the species richness sampled along the two new Las Cruces transects (traps Sl-Sh and 51-56). situated over 50m from an edge. was on a par with the richest sampling

Butterfly

biodiversity

in rainforest

patches

45

sequences in Las Cruces and in Cascada (over the same number of sampling days). There is significant variation in the relative abundance of subfamilies among the three sampling transects in Las Cruces (G-test; d.f. = 20, G = 161.36, p G 0.001). Finally, our data permit an evaluation of the hypothesis that the managed Wilson Botanical Garden serves as a corridor between Las Cruces and Cascada. In support of the hypothesis, the eight traps (Gl-G8) in the Garden revealed the presence of a fauna nearly as rich as that found in Cascada and Las Cruces as a whole. A total of six charaxine species and two nymphaline species (represented by seven and three individuals, respectively) was obtained (Table 3). In addition, Las Cruces and Cascada share a significantly higher fraction of species (20 of 37) than do Las Cruces and Vaca Vaca (eight of 33) (Table 4; G-test, d.f. = 1, G = 6.471, p < 0.01). Vaca Vaca is comparable to Cascada in size and in degree of time since isolation, but differs in habitat type separating it from Las Cruces (degraded pasture as opposed to botanical garden). All of the species (eight of nine total) that Vaca Vaca shares with Las Cruces were sampled in Cascada as well; seven of those eight species were also found in Ridge Road Ravine. Thus, the subset of species found in Vaca Vaca is not unique; only one of the seven species shared among the four sites is a charaxine or nymphaline (Memphis beatrix, the most abundant and widely distributed charaxine). Lastly, the fraction of species shared between the Garden and Las Cruces (28%) and between the Garden and Cascada (26%) is roughly equivalent to that shared between faunas trapped along different transects within the same patch (23%, on average, for the three Las Cruces transects and the two Garden transects) (G-test, d.f. = 2, G = 0.340, p > 0.50).

On the other hand, this degree of overlap (roughly 25Y0 ) is no higher than that observed between Las Cruces and Vaca Vaca and could be interpreted as evidence against the corridor hypothesis. This result could also simply be attributed to relative undersampling of the garden fauna. The garden is rich in the subfamilies containing species found by DeVries (1991) in Las Cruces, but not detected in our comparatively heavy surveying done in Las Cruces. Further information bearing on the corridor hypothesis is discussed below. Discussion BUTTERFLY

DIVERSITY

IN THE PATCHES

As is clear from the species accumulation curves (Figs 3-6) we did not appear to reach saturation at any site. Indeed, we know from DeVries (1991) and our 1994 results that, for example, about a dozen species of Charaxinae are present in the Las Cruces forest that we did not trap in 1993. Nine recorded by DeVries (1991) were not found either year, while several trapped by us are not reported by him. Nonetheless, a marked difference between the sites was manifest almost immediately. Our patch protocol in both years was conservative with respect to this difference among patches, for two reasons. First, our sampling intensity per unit area was much higher in the small patches than in the large one (0.02 versus 1.67 traps per ha at the extreme). Second, equidistant trap placement from the edge meant that virtually all of the habitats in small patches was sampled, while the interior of larger patches was not sampled at all. Additional sampling in the larger sites partially adjusted the effort per unit area and did increase observed differences in species diversity. Our 1994 sampling of the deeper interior of the Las Cruces forest enhanced, as expected, the pronounced differences in butterfly fauna

Dai1.y and Ehrlich

between the largest and smallest patches. Nonetheless, conservatively excluding the results of the data yielded from the additional trap days or the more thorough geographical sampling of Las Cruces does not change the basic result. The six depauperate patches were characterized by a striking lack of charaxines and nymphalines, combined with a great abundance of Cissia satyrina. DeVries (1987) states that C. satyrina is ‘common and widespread in all forest habitats,’ and that ‘individuals fly in the forest understory.’ The role of this species as a possible indicator of forest disturbance clearly merits further study. The two Cafigo species (Brassolinae) were also quite common in these six patches. Interestingly, brassolines are characteristically crepuscular, and conceivably may find open pastures less of a barrier than strictly diurnal species. Our data indicate that Cascada supports a butterfly fauna of richness and diversity similar to that of the large Las Cruces patch. Vaca Vaca, although nearly equivalent in size, as well as in recency and degree of isolation, supported a comparatively depauperate fauna. This difference, along with the results of trapping in the Wilson Botanical Garden. suggests that the Garden serves as a corridor for many species. Considering that most of the Garden is highly managed (including mowing, weeding, and selective use of pesticides), and is made up largely of exotic plants, it is unlikely to constitute suitable habitat for the larvae of many butterfly species. However, an abundance of fruiting shrubs and trees (absent from the surrounding pastures) apparently makes it quite attractive to adults. Aside from size and associated level of habitat heterogeneity, elevation and climate would seem the most likely candidates for explaining the great differences between both Cascada and Las Cruces and the other patches. The latter group of patches were approximately 200 m higher and receive more precipitation because the ridge above them (Fila Paraguas) intercepts weather moving in from the west. Trees hold more epiphytes and fog often sweeps over the ridge toward the middle or end of the day, usually dissipating a few hours later. Under such conditions, there can be at least two fewer hours of sunlight per day than at the Las Cruces and Cascada sites. However, a potential consequent reduction in butterfly activity (and hence trapping rate) relative to the lower-elevation sites does not seem sufficient to account for the differences in species richness among patches; much of the trapping in both years, but especially in 1994, was done under dry and sunny conditions. It is also highly unlikely that this difference in elevation and climate restricts the range of many non-satyrines found in our species-rich sites. In fact, eight of the 1.5 trapped charaxines, five of the six trapped nymphalines, the one morphine, and six of the seven trapped brassolines are known to occur at 1400m or higher, several as high as 1800m (DeVries, 1987; Ehrlich et al., 1994; Sparrow et al., in press). Extensions of previously reported ranges are being rapidly discovered. For example, six of the seven charaxines not presently known to occur at 2 1400m elevation are reported as having unknown or restricted distributions to 1000 m or lower in DeVries (1987); we found all at 2 1200 m. Morever, although the elevational difference between both Cascada and Las Cruces and the ridge patches might cause some of the differences in their butterfly faunas, it clearly does not explain why the faunas of the smaller and more isolated of the patches on the ridge are depauperate in comparison with the larger, less-isolated ridge patches. Unfortunately, the limited number of remaining forest patches in the vicinity of Las Cruces did not allow an evaluation of the relative effects of area and isolation on fauna1 diversity.

~u#er~y

biodiversiiy

in r~inforest

patches

47

Table 3. Summary trapping statistics for the Cascada-Wilson Garden-Las Cruces complex, 1994 Gl-G4

Garden Cascada near IIb Cascada Total number of visits Total number of traps Total number of individuals trapped Total number of species trapped” Number

of charaxine

5

26

Sl-S6

in

near Las

Las

Cruces

Cruces

Traditional 51-56 Las Cruces in Traps LXS IIb Cruces

4

5 4

5 6

5 4

5 6

18

28

33

18

21

312

11

12

315

9

212

a3

4

2

23

2

23

7

5

2

16

2

3

1

1

4

2

3

1

2

7

2

6

4

9

4

6

7

23

22

32

12

33

33

species

trapped” Number of charaxine individuals trapped Number of nymphaline species trapped Number of nyphaline individuals trapped Number

5 4

G5-G8

Garden

of Cissa sa~rina

captures Percent C. satyrina in total number of captures within site

aApproximate numbers reflect the escape of a few individuals from the traps before certain species identification had been made. bII refers to the second round of sampling these traps in 1994; the first round is reported in Table 1.

Internal consistency in the rapid evaluations

Resampling the forest butterfly fauna in the system of patches in 1994 allowed us to evaluate the robustness of the results obtained in 1993. The effort expended in 1993 - 12 person-weeks of intensive field work in total - was probably greater than that which is available in aid of many tropical conservation decisions; were the results replicable? Our principle conclusions based solely on the 1993 data were that the 3-30 ha patches other than Cascada lacked a significant component of the biodiversity represented by forest butterflies. The 1994 results supported the same general conclusion, helping to establish the potential utility and reliability of this rapid evaluation technique. Nevertheless, it is important to consider possible causes of the reduction in apparent species richness of Las Cruces at the traditional traps in 1994 relative to 1993. The sampling in 1993 was done comparatively later in the year and in an unusually wet ‘dry season’. The most likely explanation for our result is the noticeable dryness of the Las Cruces traditional transect in 1994 relative to 1993 and relative to moister areas in that forest to which adult butterflies could retreat. Indeed, in 1994 we found that butterfly diversity was higher along two other transects in the moister interior. Moreover, each of the other patches (for which no noteworthy difference in species

48

Daily and Ehrlich

richness was observed between 1993 and 1994) is situated along one or more steep, wet ravines. The traps in these sites were situated within 20 m of a stream bed: half were hung adjacent to or right in them, since near-vertical slopes left no other alternatives. In contrast, the Las Cruces traditional transect was over 1OOm from a stream bed. Our conservative sampling scheme involved situating all traps roughly equidistant from forest/pasture edge; in future rapid evaluations using this technique, it would seem advisable to control for air humidity and soil moisture to the best possible extent as well. Further sampling at other times of year could either enhance or diminish the differences in fauna observed between sites. While the results from these rapid evaluations must be considered preliminary. they are sufficiently striking as to make possible a general interpretation at this time. They suggest that in southern Costa Rica even recently isolated fragments of primary forest in the 3-30 ha size range with average separation distances of 0.5-l .Okm do not retain appreciable butterfly species richness or diversity. The smallest patches we studied may actually function as sinks (Pulliam. 1988). incapable of sustaining populations of the Archaeoprepona, Memphis, Morpho. and Chligo species observed in them. We cannot even be sure that some of those species will not eventually disappear from Cascada and Las Cruces, which may still be undergoing a slow fauna1 collapse. It is interesting to compare our butterfly results with those obtained by Newmark (1991) in an investigation of understory bird distributions at a series of sites in Tanzania consisting of one large (521 ha) control forest block and nine smaller (0.1-30 ha) patches. There, the large control site had 26 species, and the next two largest patches (30 and 9.5 ha, 300m apart in a tight archipelago of small patches) had 18 and 14 species, respectively. The seven remaining (2.7. 1.6, and five < 1 ha) all had nine or fewer species. This pattern features a decline in diversity with patch size very similar to the one we found. To the extent that butterfly species diversity is an index of the diversity of small-bodied organisms, our findings suggest that large, or at least a series of several small patches connected by appropriate corridors. will be required to preserve a great deal of that diversity. The effectiveness of the latter will depend heavily upon the influence of habitat edges. Both forest interior insects and birds are sensitive to microclimatic changes associated with edges, such as temperature and humidity (e.g. Janzen and Schoener, 1968: Janzen, 1973: Sisk 1993a; Kaspari. 1994). A conservative standard of 250m edge penetration (Lovejoy et al., 1986; Sisk, 1993b; Skole and Tucker, 1993) would mean that all of the smaller patches in both our study and that of Newmark (1991) consist entirely of edge habitat. Buttefly

trapping as an evaluation of total forest biodiversity

Butterflies seem potentially good indicators of forest biodiversity (e.g. Scoble, 1992) for several reasons. First, the general habitat requirement of most non-satyrine butterflies that depend on decaying fruit for sustenance (and are thus most attracted to rotten fruit baits) is forest interior, including the canopy of interior trees. That is also the locus of most forest biodiversity. Second. such forest interior species make up a substantial fraction of the butterfly fauna. Over 40% of the comparatively well-known papilionids, pierids, and nymphalids feed exclusively on rotting fruit as adults (DeVries. 1987). Third, rain-forest interior butterflies utilize a wide variety of plant families (DeVries. 1987). Forest butterflies should, therefore. ‘sample’ the flora. The quality of that sample. however, remains to be widely tested (see e.g., Holloway, 1980: Kremen, 1992). Fourth. their distributions and abundances are clearly related to forest structure and microclimate

Butterfly

biodiversity

49

in rainforest patches

Table 4. Overlap in species composition between samples

Site” Cascada Vaca Vaca Garden Garden

Number of species trapped in siteb 25 9 18

18

% species shared 54% (20/37) 24% (8/33) 26% (9/34) 28% (11139)

Number of species trapped in siteb

Site”

32 32 25 32

Las Cruces Las Cruces Cascada Las Cruces

“Fauna collected during all sampling periods is included. bExcluding species known to be unique to the sample, but which escaped before certain identification had been made.

(Kremen, 1992). Finally, while abundances may vary radically, many forest interior species fly as adults for much of the dry season (DeVries, 1987). This reduces the possibility of missing a substantial portion of the fauna because of mis-timed sampling and makes reliable relative comparisons among sites as long as the sites are sampled in the same stage of the flight season. As compared to sampling many other groups of organisms, the relatively extensive knowledge of and ease of both trapping and identifying butterflies makes efforts to hone their use as tools in conservation biology extremely important (e.g., Hodgson, 1993; Sparrow et al., 1994). We should note, however, that even the relatively limited sampling we carried out was very time-consuming and labour-intensive. Considerations for a proposed restoration project

Under the leadership of Luis Diego Gomez, Director of the Las Cruces Field Station, a project has been initiated with the goal of eventually connecting the lowland rain forests of the Osa Peninsula (and Corcovado National Park) with the highland forests of the Cordillera Central in La Amistad National Park. Key to reestablishing a connection is the Las Cruces forest, situated between the extensive forest of the Guaymi-Coto Brus Reservation and the southern edge of the extensive forest of the Cordillera Talamanca, 32 km northeast at the border of the Zona Protectora La Amistad. There are manifold uncertainties associated with the conservation value of corridors in general (e.g. Simberloff and Cox, 1987; Ness, 1987; Saunders and Hobbs, 1991; Simberloff et al.. 1992) and of their effectiveness relative to series of ‘stepping-stone’ patches (e.g., Beckon, 1993). For example, would the establishment of a broad corridor of continuous forest be required to restore the former floristic and faunistic gradient between shoreline and the Cordillera? Or would simply increasing the size and/or number of patches in the corridor area suffice, counting on the normal vagility of organisms to maintain metapopulation structures secure from extinction? Sufficient information is not available to predict with assurance whether patch or potential corridor areas would be sustainable for any of the organisms involved. Restoring a series of additional disconnected patches in the 30 hectare size range would seem less helpful than we had anticipated before our study. Considering the likely function of the Wilson Garden as a corridor to Cascada, the development of managed, agroforestry

50

Daily and Ehrlich

connections (with abundant arboreal fruit sources) where restoration of natural forest is not possible appears to be a sensible initial objective (Daily and Ehrlich, 1994). Clearly, much more research would be desirable to determine the minimum physical dimensions and biological characteristics of an effective (and socioeconomically feasible) corridor. For example, while butterflies may be reliable indicators of forest interior species richness, they may in fact be especially poor at indicating the suitability of alternative corridor designs to other taxa. This need for more basic information characterizes problems of, and opportunities for, biodiversity preservation the world over and requires decision-making in the face of great uncertainty. Although funds allocated to this project might theoretically conserve more biodiversity if expended in less-protected parts of the tropics, the leadership and opportunity now exist at this site. We are convinced that small amounts of information are better than none at all. and hope that this study contributes to steering this and similar projects in the right direction.

Acknowledgements We owe an enormous debt to George and Yvonne Burtness who in 1994 worked tirelessly. providing the most reliable and enthusiastic research assistance imaginable. We are also especially grateful to Dr Luis Diego Gomez, Gail Hewson Gomez, and Ana Maria Herra of the OTS Las Cruces Research Station for their continued cheerful hospitality and assistance. Luis made the botanical evaluation of each patch. Thomas Sisk and Helen Sparrow helped prepare the traps and offered advice based on their related work at the nearby Las Alturas Station, established by the Center for Conservation Biology and now operated by Centro para la Biologia de la Conservation, Costa Rica. Susan Alexander. Kelsey Wirth, and Wren Wirth provided invaluable assistance with various aspects of the project. Helpful comments on early drafts of the manuscript were provided by Luis Diego Gomez (OTS, Costa Rica), Alan Launer, Dennis Murphy, Steve Rottenborn, Thomas Sisk, and Helen Sparrow (Center for Conservation Biology, Stanford University), and anonymous reviewers; especially careful and useful criticism was provided by Michael Beck and Daniel Simberloff (Florida State University). This work was supported by the Winslow and Heinz Foundations and by the generosity of Peter and Helen Bing.

References Beckon, W.N. (1993) The effect on insularity on the diversity of land birds in the Fiji islands: implications for refuge design. Oecologia 94,318-29. Bierregaard, R.O. Jr., Lovejoy, T.E., Kapos, V., dos Santos, A.A. and Hutchings, R.W. (1992) The biological dynamics of tropical rainforest fragments. BioSci. 42, X.59-66. DeVries P.J. (1987) The Butterflies of Costa Rica and Their Natural History. Princeton: Princeton University Press. DeVries. P.J. (1991) The Butterflies of Jardin Botanic0 Wilson and Adjacent Forest. Mimeo. Daily, G. and Ehrlich. P. (1994) Butterflies used as rapid assessment tool. Amigos Newsletter. Organization for Tropical Studies, 40, 12-14. Ehrlich, P., Sparrow, H., Sisk. T. and Daily, G. (1994) Notes on butterfly distributions in southern Costa Rica (Lepidoptera: Papilionoidea). Trop. Lepidop. 5. 21-3.

Butterfly

biodiversity

in rainforest

patches

51

Gadgil, M. and Guha, R. (1992) This Fissured Land: An Ecological History of India. Oxford: Oxford University Press. Hodgson,J.G. (1993)Commonnessand rarity in British butterflies, J. Appl. Ecol. 30,407-27. Holloway, J.D. (1980)Insect surveys- An approachto environmentalmonitoring. Afti XZZ Congr. Naz. Ital. Entornol.,

Roma: 239-61.

Janzen, D.H. (1973)Sweet samplesof tropical foliage insects:effects of seasons,vegetation types, elevation, time of day, and insularity. Ecol. 54, 687-708. Janzen, D.H. and Schoener,T.W. (1968) Differences in insect abundanceand diversity between wetter and drier sitesduring a tropical dry season.Ecol. 49,96-110. Kaspari, M. (1994)Body size and microclimateusein neotropical granivorousants. Oecologia (in press). Kremen, C. (1992) Assessingthe indicator properties of speciesassemblages for natural areas monitoring. Ecol. App. 2,203-17. Lovejoy, T.E., Bierregaard, R.O. Jr., Rylands, A.B., Malcolm, J.R., Quintela, C.E., Harper, L.H., Brown, K.S. Jr., Powell, A.H., Powell, G.V.N., Schubart,H.O.R. and Hays, M.B. (1986)Edge and other effects of isolationon Amazon forest fragments,In Conservation biology: the science of scarcity and diversity (M.E. Soule,ed.) pp. 257-85.Sunderland,Ma: SinauerAssociates. Newmark, W.D. (1991)Tropical forest fragmentationand the local extinction of understory birdsin the Eastern UsambaraMountains, Tanzania. Conserv. Biol. $67-78. Noss,R.F. (1987) Corridors in real landscapes:a reply to Simberloff and Cox. Conserv. Biol. 1, 159-64. Pulliam, R. (1988) Sources,sinks,and population regulation.Am. Nat. 132,652-61. Saunders,D.A. and Hobbs, R.J., eds (1991)The Role of Corridors. ChoppingNorton, New South Wales,Australia: Surrey Beatty. Scoble, M.J. (1992) The Lepidoptera: Form, Function and Diversity, Oxford: Oxford University Press. Simberloff,D. andCox, J. (1987)Consequences andcostsof conservationcorridors. Conserv. Biol. 1, 63-71. Simberloff, D., Farr, J.A., Cox, J. and Mehlman, D.W. (1992) Movement corridors: conservation bargainsor poor investments?Conserv. Biol. 6,493-504. Sisk,T. (1993a)Distribution andabundanceof understorybirdsalongdisturbancegradientsin Costa Rica: edgeeffects and community organization. Manuscript. Sisk T. (1993b)Edge effects following tropical deforestation exert complex influenceson habitat quality and biodiversity. Manuscript. Skole, D. and Tucker, C. (1993)Tropical deforestation and habitat fragmentation in the Amazon: satellitedata from 1978to 1988.Science 260, 1905-10. Smythe, N. (1978) The natural history of the Central American agouti (Dayprocta punctata). Smithsonian Contr. Zool., number 257. Sokal, R.R. and Rohlf, F.J. (1981)Biometry. New York: Freeman. Sparrow, H., Sisk, T., Ehrlich P. and Murphy, D. (1994).Techniquesand guidelinesfor monitoring Neotropical butterflies. Conserv. Biol. 8,800-9. Terborgh, J. (1976) Island biogeography and conservation: strategy and limitations. Science 193, 1029-30. Wilson, E.O. (1987) The little things that run the world: the importance and conservation of invertebrates. Conserv. Biol. 1, 334-41.

52

Daily and Ehrlich

Appendices Appedi~

1. Butterfly

diversity in the patch study Index of relative abundance Ridge Upper Lower Las Road Vaca Gamboa Gamboa Labrador Labrador Stream Cruces Cascada Ravine Vaca Swamp Swamp Slope -------93 94 93 94 93 94 93 94 93 94 93 94 93 94 93 94

Species Charaxinae Archaeoprepona demophon Archaeoprepona demophoon Archaeoprepona meander Archaeoprepona phaedra Consul electra Memphis arginussa Memphis beatrix Memphis glycerium Memphis Laura Memphis morvus Memphis niedhoeferi Memphis oenomais Memphis proserpina Memphis xenocles Memphis sp. Prepona omphale

Unidentified

I

3

11 4

11

3

11 13

4

3

4

4

11

4 4

2 2 4 2 2

4 4

4

4

2

Nymphalinae Catonephele mexicana Catonephele numilia Colobura dirce Erisia melaina Hamadryas fornax Tigridia acesta

2 4

II

2

‘l

4

4 2

4

Morphinae Morph0

peleides

2

11 31

14

3

13 7 2

14 27 13

29 7 4

3 3

5

3

Brassolinae Caligo atreus Caligo eurilochus Catoblepia orgetorix Eryphanis aesacus Opsiphanes cassina Opsiphanes quiteria Opsiphanes tamarindi Opsiphanes sp.

17 17 ?I 3 7

5

4

3 8

4 12 4 3

4

5

2 2 3

Satyrinae Chloreuptichia arnea Cissia gigas Cissia hermes Cissia hesione Cissia nr. labe Cissia metaleuca Cissia pol.vphemus

14 7 2 4 11 4

4

4 4

Butterfy

biodiversity

Species Cissia renata Cissia satyrina Cissia sp. Cithaerias menander Cyllopsis hedamanni Cyllopsis rogersi Pedaliodes manis Taygetis andromeda Taygetis (nr.) virgilia

in rainforest

53

patches

Index of relative abundance Upper Lower Ridge Las Road Vaca Gamboa Gamboa Labrador Labrador Cruces Cascada Ravine Vaca Swamp Swamp Slope Stream -------93 94 93 94 93 94 93 94 93 94 93 94 93 94 93 94 10

16

89 18

21 337 308 443 175 480 186 232 350 145 4 4 7

82 80 238 4 4 4

7 2

3

4 4

8

8

6

The index of relative abundance is calculated as (total number individuals captured) x 100 / {(number of visits to site) x (number of traps in site)).

Daily and Ehrlich Appendix

2. Butterfly

diversity in the Cascada-Wilson

Garden-Las

Cruces study, 1994

Index of relative abundance

Cascada II

Species

Gl-G4 Garden near Cascada

G.5-G8 Garden near Las Cruces

Sl-S6 in Las Cruces

Traditional Las Cruces traps II

Jl-Jh in Las Cruces

Charaxinae Archaeoprepona demophorz Archaeoprepona demophoon Archaeoprepona meander Archaeoprepona phaedra Consul electra Memphis arginussa Memphis beatrix Memphis glycerium Memphis laura Memphis morvus Memphis niedhoeferi Memphis oenomais Memphis proserpina Memphis xenocles Memphis sp. Prepona omphale

5 s

5

s 5 7 3

20

3 3

3

13

17 10

5 10

10 3

s

Unidentified Nymphalinae Catonephele mexicana Catonephele numilia Colobura dirce Erisia melaina Hamadryas fornax Tigridia acesta

s s s

IO

s

13 3 3

5

3

5

3

3

Morphinae Morph0

5

peleides

Brassolinae Caligo atreus Caligo eurilochus Catoblepia orgetorix Eryphanis aesacus Opsiphanes cassina Opsiphanes quiteria Opsiphanes tamarindi Opsiphanes sp.

s

20 IS

10

.:

3

10 3

5

Satyrinae Chloreuptichia arnea Cissia gigas Cissia hermes Cissiu hesione Cissia nr. labe Cissia metaleuca Cissiu polyphemus Cissiu renata Cissia satyrina Cissa sp.

25 10 5

20

10

5

3 3

1s

3

10 3

25 5 30

20

4s

13

30

3 23

Butterfly

biodiversity

in rainforest

patches

55 Index of relative abundance

Species Cithaerias menander Cyllopsis hedamanni Cyllopsis rogersi Pedaliodes manis Taygetis andromeda Taygetis (nr.) virgilia

Cascada II

Gl-G4 Garden near Cascada

G5-G8 Garden near Las Cruces

Sl-S6 in Las Cruces

Traditional Las Cruces traps II

51-56 in Las Cruces

5 5

3 5

3

The index of relative abundance is calculated as (total number of individuals captured) x 100 I {(number of visits to site) x (number of traps in site)}.