Use of fruit bait traps for monitoring of butterflies ( Lepidoptera : Nymphalidae )

Abstrae!: There exisls great interest in using fruit-feeding adult nymphalid butterflies to monitor changes in tropical forest ecosystems. We intensively sampled the butterfly fauna of mid-elevation tropical moisl forest in southem Costa Rica with fruit bait traps to address a series of practical issues eonceming the development of a robust, efficient sam­ pling programo Variation in the number of captures and escapes of butterflies at the traps was better explained by the time of day than by the age of bait. Species' escape rates varied widely, suggesting that short term, less intensive sur­ veys aimed at determining presence or absence of species may be biased. Individuals did not appear 10 become "trap­ happy" or lO recognize the traps as food sources. Considering the tradeoff between numbers of traps and frequency of trap servicing, Ihe most efficient sampling regime appears to be baiting and sampling the traps once every other day.

Tropical forests are being altered and destroyed at rapid rate, yet the impacts on bio diversity and ecosystem services are only beginning to be understood (Lawton and May 1995, Pimm et al. 1995, Daily 1997).Practical, reliable methods of evaluating changes in biodi versity and ecosystem function are needed to ínform public policy (Noss 1990).There is great inlerest in developing monitoring pro grams of tropical ecosystems using frugivorous adu!t butterfly communities sampled with bait traps (Holloway 1980, Lovejoy et al. 1986,  Brown 1991, Kremen 1992, Scoble 1992,  Sparrow et al. 1994, Daily and Ehrlich 1995).A number of important questions about the utility of this approach remain open, however, includ ing the extent to which the samp!ing routine can introduce bias into a monitoring programo We have trapped butterflies extensively in a region of southern Costa Rica in primary and secondary forest and agricultural areas for over 2000 trap days.In previous studies, we used these surveys to address issues such as the applicability of butterflies as indicator species of other taxa and the impact of forest fragmen tation on butterfly diversity (e.g., Daily andEhrlich 1995, Daily andEhrlich 1996).We have also begun an intensive mark-recapture study to assess the use of forest fragments as corridors by butterflies.
Certainly, bait trapping provides a biased look at the butterfly fauna itself.First, there are many butterfly species that are never cap tured in the traps.Second, among those cap tured in the traps, sorne are probably more strongly attracted than others; thus, the relative abundances of species caught almost certainly do not reflect the relative abundances of fruit feeding species in the region.Yet despite these drawbacks, bait trapping offers a useful way to monitor changes in species abundances over time, to compare species composition and abun dances between sites, and to track the move ment of individuals.
Using the relatively well-known butterfly fauna of mid-elevation tropical moist forest in southern Costa Rica (e.g., DeVries 1987), we designed a sampling regime to determine: (1) what time of day butterflies typically enter the traps; (2) the propensity of different species to escape; (3) the influence of age of bait on trap ping and escape rate; (4) the tendency of indi viduals to become "trap-happy;" and (5) the amount of informatíon lost by less intensive surveys.
We hope that this preliminary study sheds light on the most obvious and easily addressed problems with the fruit-trapping method.WhiJe our main purpose was to provide background for refining trapping techniques, we also sam pled in two habitat types to see if there were striking differences in the faunas present.This study is not designed to examine these potential differences in detail, however.Many researchers have addressed the influence of habitat type on butterfly assemblages elsewhere (e.g., Brown andHutchings 1997, DeVries et al. 1997).

Area:
The study was conducted in the vicin ity of the Las Cruces Biological Field Station of the Organization of Tropical Studies (OTS/OET), Coto Brus, Costa Rica in March and April 1996.The Las Cruces forest consists of primary Pacific mid-elevation forest (211 ha) and mature second growth (16 ha) adjacent to the managed Wilson Botanical Garden.
Trap design: We used 18 Van Someren Rydon traps (De Vries 1987), cylinders of grey nylon nettting (65 cm high and 25 cm in diam eter) sewn onto a frame of two wire hoops, closed and covered by a plastic dish at the top, open at the bottom, and with a Velcro-fastened slit in the side for removing trapped insects.Each trap had a sheet of plywood (40 x 40 cm 2 ) suspended a few centimeters below the netting, with an ínverted disposable plastic dish 15 cm in diameter in the center on which the bait was placed.The bait was covered with a second inverted dish perforated with hoJes roughly 1 cm in diameter and fixed into positíon with string to prevent robbery by birds and mammals yet still permit feeding by butterflies.The bait consisted of mashed, rotting bananas, a liberal dose of molasses, and a dash of rumo The bait was prepared on the afternoon prior to the morning of use and typically became well fer mented in that time.
Trap location ami monitoring scheme: Nine traps were placed in mature second growth of the Las Cruces forest (hereafter forest siles), and nine traps were placed on a farm of coffee and overgrown pasture adjacent to the Wilson Botanical Garden (hereafter pasture sites).Each trap was checked three t imes per day: early morning (05:30 -08:00), midday (10:00-13:30), and late afternoon (15:00 -17:00).Dawn was at 05:30 and dusk at 17:30.The traps were baited every other day during the morning check, alter nating each day between the forest and pasture sites.The traps were open for 14 days and baited seven times.AH butterflies werecleared from the trap at each baiting, but not before.
. Each butterfly.captured was identified to species, and all but the superabundant species in the genus Cissia (Satyrinae) were uniquely marked with a feh peno We would not have been able to check the traps as often with the additional time required to mark all the Cissia individuals.
Individuals that were capturedand had been previously marked were noted as recaptures.Numbers of captures inelude both newly cap tured and recaptured individuals.Marked butter flies were recorded as escapes if they disap peared from the trap between trap checks; sorne individual s (five) escaped from the hand during marking, but these were not included in the analysis of escape frequencies.
After feeding on the bait, the butterflies usu ally hung quietly from the top or sides of the trap.Handling often caused them to fly around inside the trap and sometimes escape.Therefore, immediately after marking any newly captured butterflies, the trap was held closed until all the butterflies were resting again.Once a butterfly was marked, we did not handle it during subsequent trap checks.
Statistical analyses: We used the Kmskal Wallis statistic to test for differences between both capture and escape rates across the time of day and Wi1coxon signed-rank tests to examine the effect of bait age.For these analyses, the data were pooled across all the days sampled and the replicates were the eighteen traps.To test whether more individuals were recaptured than expected by chance, we compared the dis tribution of the number of times each individual was captured to the Poisson using a chi-square test distribution (combining the categories which had expected values below one).To cal culate evenness of species abundance in the dif ferent habitats, we used the index J' = -í P i ln(P i )/lnS, where P i is the proportional abun dance of species i, and S is the total number of species (Pielou 1966).'

RESULTS
A total of 337 individual s representing 43 species of butterflies were caught (Table 1).All species were fruit-feeding nymphalids of the subfamilies Charaxinae, Nymphalinae, Morphinae, Brassolinae, and Satyrinae.Nine species of Cissia were captured (Table 2), and this genus comprised 54% of the indi v id l lals captured.We recorded 185 non-Cissia captures; of these, 146 were newly captured and marked individuals, 34 were recaptured, and five escaped before we could mark them.
Time oC day: The number of individuals (excluding Cissia species) captured by the traps before the morning, midday, and afternoon observations differs significantly (Kruskal Wallis; H=I1.092, p=O.OO4, d.f.=2) (Table 3).Most of the species captured are described as diurnal in the I l terature (e.g., DeVries 1987).Qur results confirmed this; many more individ uals arrived during the day (between morning midday and midday-afternoon checks), than at either dusk or dawn (between the afternoon and morning checks).The few exceptions corre spond with the known periods of activity of the species involved.For instance, Opsiphanes  were also observed in the morníng and midday traps only, although their time of peak activity is not reported in DeVries (1987).Of the remaining 29 non-Cissía species, the vast majority of new captures were made at midday and afternoon (Table 1).
Of the 180 marked captures, 73 (41 %) escaped between recording them and clearing of the trap.The number of individual s that escaped did not vary significantly with respect to time of day (Kruskal-Wallis; H=5.002, p=0.082, dJ.=2), although escapes followed the same trend as capture activity; that is, most escaped during the day rather than between dusk and dawn (Table 3).  .Reeapmre s : 'inie p6tentlaFf�ríridiwiduals tb: "learn!! that thei .trap�'ftfe reIiaole fQOt1 $owees .cotild blas ,;�, dlversity invéntory,y¡: li.er,:��we: address to two aspects of this problem.We refer to "trap-happiness" as a tendency of individual s to return repeatedly to the same trap.We define a separate term, "trap-recognition," as the ten dency of individuals to recognize any trap as a food source.There is no sign of either trap-hap piness or trap-recognition.The frequency of recaptures falls off sharply with increasíng number of captures: 146 individuals were cap tured once, 27 twice, 5 tbree times, and 1 four times.This distribution is not significantly dif ferent from a Poisson distribution ( )(2 =2.049, p=0.355, dJ.=2).In other words, Índividuals are not being recaptured more than expected by chanceo The frequent movement between traps is further evidence against trap-happiness; 15 out of 34 (44%) of the recaptured individuals moved between different traps.The only indi vidual that was captured four times was Memphis glycerium, and the captures were at tbree different traps.Detecting significant dif ferences among species in their tendency to be recaptured proved difficult b ecause of the small number of individuals recorded.It is interesting to note, however, that while they have a greater tendency to escape, Morpho peleides and M. theseus also appear more likely to show trap recognition than the other species; of the eight M. peleides captures recorded, three were recaptures, and two of the five M. theseus cap tures were recaptures.Trap-recognition in these species has been previously noted (Young 1973).
Sampling intensity: We can explore the efficiency of different samplíng intensities by comparing our actual routine (checking tbree times a day) to less intensive routines by remov ing the results of sorne of the trap checks.For instance, if we had checked the traps only every morning, we would have recorded 76% of the captures and 91 % of the species richness that we detected in the more intensive routine.If we had checked the traps only when baiting every two days, more than half (55%) of the individu als and most (85%) of the species richness of the actual routíne would have been represented.We can also directly compare the efficiency of daily versus every-two-day sarnpling; checks every other day would have recorded 72% of the individuals and 94% of the species from daily observations.
Diffe rences between nabitats: More butter flies were caught in the pasture site than in the forest (229 and 142 captures, respectively).In particular, more Cissia individuals were cap tured in the pasture, although the percentage of Cissia captured was lower in the pasture than the forest (41 % versus 65%).Both species rich ness and Pielou's evenness index were higher in the pasture than forest (37 and 23 species, J' = 0.851 and 0.711, respectively).Excluding the Cissia species, a genus thought to be associated wíth human disturbance (Singer and Ehrlich 1991), species nurnbers remained higher in the pasture (28 versus 16 species), although Pielou's index was then similar for the two habítats: 0.866 in the pasture and 0.876 in the forest.

DISCUSSION
A number of practical suggestions about these trapping methods can be offered on the basis of this study.For instance, variatíon in captures and escapes seerns to be explained well by the time of day, rather than bait age, except immediately after baiting.Thus, baiting every day would only slightly íncrease the nurnber of captures (by an estimated 20% frorn this study) and does not seem likely to alter the qualitative results of a diversity inventory.From previous experience, we suspect that baiting every tbree days would not be sufficient, as rain and evapo ration of the alcohol appear lo greatly dimínish the attractiveness of the bait by the third day.Considering the time required to prepare the bait and rebait the traps, baiting every two days seems to offer a reasonable compromise between effort and informatíon.Furthermore, only a small amount of informatíon would have been 10st by checking the traps only when bait ing compared to checking daily or to checking three times a day.As with baiting every two days, reducing the sampling to every other day probably would not change the qualitative resuJtsof an inventory.For most studies then, the effort used to bqit or check the traps mote ofien would be better allocated lo sampling ¡n more habitats or trapping over a longer time periodo Perhaps not unexpectedly, escape rates appear to vary greatly among species.Thus, infrequent trap checks and short term trapping may lead to an undersampling of species with _ high rates of escape.Cer;tain1y, further work is necessary to refine knowledge of species dif ferences in bait attractiveness, trap-happíness, and trap-recognition.Other variables that we did not address here, but that also sllrely influ ence the results of a trapping program, are sea sonality and trap height; however, within a given study it should be practical 10 standard ize these factors.
An exarnination of the list of species caught in the two habitats reveals striking differences in habitatpreferences among sorne sp�cies (Table s 1 and 2).itor changes in tropical forest ecosystems.We intensively sampled the butterfly fauna of mid elev�tion tropicalmoist forest in southern Costa R ica with fruit bait traps to address a series of practica1 issues conceming the development of a. robust, efficieIlt sampling programo Variation in the numbet of captures and escapes of butterflies at the traps was better explained by the time of day than b)'the ag� of bait.Species' escaperates varied wide1y, suggesting that short terrn, less intensive surveys aimed at deterrniníhgpresence or absenc.e of speci�s may be biased.Individuals did not ap¡rear tobecome '�trap-ñappY" o� to rec ognizé the traps as Jood souices.Consid�ring the tradeoff between numbers of traps and fre quency of traP servicing, the most efficienlsam� pling regime appears.tobe baiting• and sampling the traps once every other day .

RESUMEN
Existe gran interés en monitoreár las poblaciones de mariposas tropicales mediante trampas cebadas.Estudiamos de manera intensiva una localidadde bosquetropical húme do de mediana altitud en el sur de Costa Rica.La variación en el número de capturas y escapes de las trampas sé asoció más con la hora del día que con la edad del cebo.Los muestreos a corto plazo son poco confiables pues las especies dif ieren en su tasa de escapes.Los individuos no parecieron hacerse adictos a las trampas o reconocerlas como fuentes de alimentación.La atención a las trampas de día de por medio resultó eficiente, ) is considered crepuscu lar, and all new arr ivals were observed in the morning.Caligo atreus (Brassolinae) is also crepuscular and was captured only in the morn ing and at midday.Like Caligo atreus, Taygetis virgilia and Taygetis andromeda (Satyrinae) malee exploratory flights then.These deceptive Iy simple questions are crucial for addtessing issues of tropical biodiversity consérvation.

TABLE 1
Number and percent of captures by species trapped, number of captures that were recaptures, number of captures by time of day, and percent of captures that escaped.
For instance, within the Cissia butter flies, C. renata was• fourid .only in pasture, whereas most of the C. satyrind and C. hesione individuals were captured in forest These results correspond well witbI:iabitát preferences notedin DeVries (1987).Similarly,MOlphothe seus was trapped only in pasture .and M. pelei des, Qnly in foresf..In contrast io'ihe Cissia species, .however, bptlÍ' of the Úorhpo spe�ies are thought to be.assQciated with forest habitats (Young 1973).Altllough the sample si,ze.here is smaH, this pattem highlights an important süite of questions that remain to be answered regard ing the use of open, human-dominated hahitats by forest species (Daily in press).How many of thespecies found in open areas coüld survive without forest nearby?Are forest species enter ing open country during part of the da y; perhaps when temperature and humidity conditions more likely resemble the forest understory?From this study, it appears that veryfew butterfly species, forest specialists or otherwise, are moving in open areas at dawn andlor dusk Perhaps mid moming is simply a timeof high-activity for for est butterfly species, and they are more likely to