Conservation organizations often must rely on data collected quickly and cheaply to make informed decisions in unstudied regions. Butterflies represent an opportunity in this respect, in that many species can typically be sampled and identified in a short time and provide an indication of habitat or conservation value as well. During nine days of sampling in June 2008, we found and identified 84 butterfly species and 1,856 butterfly individuals at Playa El Icacal, Department of La Unión, El Salvador, using transect counts. Through species richness estimators and a ratio extrapolation based on a list of species expected to be found onsite, we sampled 40–60% of the butterfly community present. Species richness at the site is estimated to be between 100 and 200 species. Sites with small patches of dry forest (La Bocana and La Laguna) had higher species richness than sites without dry forest (Hacienda Casco, El Manglar, and El Esteron). While two weeks is not enough time to fully document the butterfly community in 20 km2 of neotropical coastal countryside, we were able to provide a valuable estimate of species richness and provide some information as to which areas in the region hold the most conservation value. We suggest using a list of widespread species with modified habitat associations for use as an inventory index for ratio extrapolation and discuss specific guidelines for future butterfly rapid assessments in Mesoamerica.
Introduction
Tropical biologists face particularly great conservation challenges under the mounting threats of anthropogenic disturbance to biodiversity. To fully inventory diversity in tropical environments requires nearly impossible levels of time and effort [1]. Insects are a perfect example of these challenges. Worldwide species estimates range anywhere from 5–30 million [2] and species richness reaches its height in the tropics. Furthermore, even though insects make up the majority of animal species, they remain mostly undiscovered and are frequently omitted from conservation assessments [34–5].
In a rapid conservation assessment one method frequently employed by biologists to avoid the logistically impossible task of sampling an entire community, is that of indicator species which serve as a proxy for overall biodiversity [6]. Several insect taxa have been examined for utility as indicators for the state of environmental, ecological, and biodiversity of various ecosystems at multiple spatial scales (see [7] for review). Butterflies are often used as bioindicators of ecosystem health and as surrogates for overall biodiversity [8]. Sensitivity to changes in microclimate and habitat make them particularly good indicators for monitoring of natural areas undergoing change [9, 10]. As with any indicator taxa, the relationship between butterfly diversity and the diversity of other species is imperfect (for examples see [1112–13]). However, butterflies have great potential as indicators for use in conservation efforts as their taxonomy, distributions, and natural history are better described than for any other insect taxa [14].
The appeal of using indicator taxa is one of logistics. By focusing on one particular set of species in a location rather than all of the species, considerable time and money (resources that can be extremely limited) can be saved. Even when focusing on indicator taxa, however, time and money are still typically insufficient to sample a given community [15]. With respect to butterflies then, how much sampling effort is required for an adequate assessment of a butterfly community? Can the conservation value of sites and habitats be distinguished from one another given short sampling periods?
In this study we sampled butterflies in five different sites consisting of different habitats (mixtures of dry forest, pasture, and mangrove) at Playa El Icacal in the Department of La Unión, El Salvador. We compared the butterfly communities and species richness at the different sites and assessed the extent to which we were able to inventory the species present at Playa El Icacal completely. We used an integrated approach to butterfly sampling including multiple habitats, assessment of sampling effort, and multiple data collection techniques. Such an approach is essential for tropical conservation programs aimed at Lepidoptera [16]. We then demonstrated the potential utility of studies undertaken with limited sampling time and concluded that while not ideal, studies of low sampling effort can provide comparable diversity estimates and even give some indication as to which sites in a region have superior or inferior habitat qualities with respect to butterflies and potentially to insects or biodiversity more generally.
Methods
Study site and organisms
Playa El Icacal is a coastal countryside community in the department of La Unión, El Salvador. We sampled five sites within Playa El Icacal (Figs. 1 and 2). La Bocana consists of a dry forest habitat and also a large open pasture. La Laguna also has dry forest and a pasture dotted with shrubs and isolated trees. Hacienda Casco is a roadside habitat with a matrix of pasture, shrubs, and small trees. El Manglar is a roadside transect that is surrounded by a mangrove forest and cattle pasture on either side. Finally, El Esteron is a large pasture with isolated trees and shrubs that is directly adjacent to the western estuary of El Icacal.
We sampled all butterflies of Papilionoidea and Hesperioidea. Identification of species within the Nymphalidae, Pieridae, Papilionidae, and Riodininae followed DeVries [17, 18] and the identification of species within Lycaenidae and Hesperidae followed Glassberg [19]. Here we use the nomenclature of Lamas [20].
Sampling protocol
Frugivorous butterflies were trapped using cylindrical Van Someren-Ryndon traps baited with rotting bananas and beer [21]. Six traps were placed in each of the two sites with forest cover (La Bocana and La Laguna) approximately 2–3 meters off the ground. Suitable places for fruit-baited traps in the other three sites (Casco, El Esteron, and El Manglar) were not available or too logistically difficult for set up (e.g., mangroves in El Manglar were difficult to access and maneuver within, which complicated efforts to establish fruit-baited traps). Traps were checked daily for seven days (from June 17 to June 23) for a total of approximately 2,000 trap hours (12 traps used for seven days).
One transect for each site was constructed based on available trails and focused on prominent nectar sources, mud puddles, and other resource-rich locations within each site [22,23]. While there was no specified distance for each of the transects, we contained each of the sites within an area of about 250 m2. We instead focused on time, and each transect was sampled at least every other day and for at least one hour.
Collection occurred during hours of weather favorable for butterfly flight (sunny and negligible wind) and was undertaken solely by the authors, both sufficiently familiar with Central American butterflies to identify them with only few errors (though surely we made them!). We noted all butterflies seen during each transect. Butterflies requiring identification were either photographed or captured with a hand net and examined upon returning from the field. Nearly all specimens taken from the field were identifiable with the exception of approximately 10 butterflies (mostly Hesperidae) that were too damaged for accurate identification. Butterflies that were too fast or too far to reliably identify were not counted. Specimens were deposited at the Museum for Natural History of El Salvador and digital photographs were taken to a photographic database kept and maintained by SalvaNATURA.
Species richness and diversity analysis
Effort and sampling efficiency were assessed by calculating species richness estimators using EstimateS software [24] based on individual-based species accumulation curves [25]. Species richness estimates were compared between sites using a rarefaction curve in EcoSim [26].
We generated a list of species we would expect to find at the site based on Thomas [27] which classifies 74 species within the Nymphalidae, Pieridae, and Papilionidae of Costa Rica as being widespread (ranging throughout Central America south to at least Brazil) and found in modified habitats. For this reason we expected to find most, if not all, of these species at Playa El Icacal. We could then extrapolate to estimate species richness using a taxonomic hierarchical ratio [28] that assumes the focal taxa (in this case Nymphalidae, Pieridae, and Papilionidae of broad distribution and disturbed habitats) respond to collection time in the same way as that of the target taxa (in this case the entire butterfly fauna). Therefore, we estimated species richness in this way by using the relationship that # species found/# of species total = # of species found within the index/# of species total in the index. The Nymphalidae, Pieridae, and Papilionidae represent the best known and most visible of the butterflies and so sampling is often biased in favor of these species [29]. Additionally, good field guides and natural history information are generally lacking for many tropical lycaenid and hesperid species. Due to this bias, the ratio extrapolation may be an underestimate of the species present.
The above methods describe estimation of species richness of the study area and within sites. These methods ignore the evenness of communities. We therefore additionally calculated Fisher's alpha [30] for each of the sites which combines both richness and evenness attributes of the communities [31]. We calculated Fisher's alpha in R [32] using the VEGAN package [33]. Given the small sampling period we were unable to effectively examine quantitatively diversity patterns within and between Papilionoidea and Hesperioidea and instead focused on the butterfly community as a whole.
Community and species composition analysis
We performed a community-level analysis using nonmetric multidimensional scaling (NMDS) [34]. The ordination was based on a Bray-Curtis dissimilarity matrix of ecological distance [35]. We performed the NMDS in SYSTAT 12.0 (SYSTAT Software, Inc. 2007).
Table 1
Species richness estimators of Playa El Icacal. The ACE is the “Abundance-based Coverage Estimator” and the ICE is the “Incidence-based Coverage Estimator”. The Chao 1 and Chao 2 are estimators based on Chao [45], and Jack 1 and Jack 2 are respectively the first and second-order Jackknife richness estimators. For a review of these and other species richness estimators see Magurran [31].
Results
Species sampled
We sampled 84 butterfly species and 1,856 butterfly individuals at Playa El Icacal in nine days over a total of 58 collecting/netting hours along the transects (Appendix 1). We sampled five species and 14 individuals in the baited fruit traps (1 Zaretis ellops, 4 Hamadryas februa, 1 Hamadryas guatemalena, 6 Temenis laothoe, and 1 Opsiphanes tamarindi). Three of these (Z. ellops, T. laothoe, and O. tamarindi) were not recorded by any other method.
State of the inventory and species richness
We calculated six estimators of species richness based on the data (Table 1). The largest estimate of species richness at Playa El Icacal over all five sites was 138 using a Chao 2 estimate. This estimate would suggest that the 84 species we detected represented 61% of the species richness present during our visit. Chao 1, the least conservative estimate of sampling effort (and lowest estimator of species richness) suggested that 87% of the present butterfly fauna were detected.
Of the 84 species we found, 31 were defined by Thomas [27] as widespread and of modified habitats (Appendix 2). We therefore sampled 42% (31/74) of the species that we expected to find. Using the ratio extrapolation (84/x=.42), this would suggest that there are approximately 200 species of butterfly (Papilionoidea and Hesperioidea) at Playa El Icacal.
Habitat and site comparison
Based on rarefaction curves, the two sites with tracts of dry forest (La Bocana and La Laguna) had higher species richness than other sites (Fig. 3). The 95% confidence intervals for species richness at 200 individuals of both La Laguna (34–41 species) and La Bocana (33–42 species) have statistically significant higher levels than El Esteron (26–30 species; Fig. 3) but not significantly higher than El Manglar (31–33 species) and Hacienda Casco (29–34 species).
Diversity and community analysis
Fisher's alpha for the sites was highest in La Bocana (13.01) and La Laguna (11.74) mirroring the pattern found in species richness. El Manglar (10.97) and Hacienda Casco (10.76) had the next highest diversity with El Esteron (9.32) containing the lowest estimate of diversity. The NMDS showed clustering of La Bocana and La Laguna as well as clustering of Hacienda Casco and El Esteron (Fig. 4). El Manglar fell out as a unique assemblage, dissimilar to the four other sites (Fig. 4).
Discussion
Sampling effort and inventory completeness
In nine days of sampling we sampled approximately 40–60% of the estimated butterfly fauna at Playa El Icacal. Though this is a small amount of time for sampling such a diverse community we were able to find significant differences between sampled sites in addition to being able to provide an estimate of the species richness in the area. Had we sampled even a day or two less than the nine days we sampled, site differences might have been difficult to detect; La Bocana and El Esteron, for example, only differed significantly after the last 20–40 individuals sampled in rarefaction (Fig. 3). Based on the community-level analysis we were also able to identify El Manglar as a unique site even though species richness was lower than La Laguna and La Bocana (Fig. 4).
Certainly more time spent sampling would have improved both species richness estimates of the area as well as site differences in the butterfly community. Another nine days would have shrunk the confidence intervals, and sampling throughout the season (approximately May-November) would have done so even more. Another strategic approach would have been to sample again for another 7–9 days later in the season to better account for seasonality effects. Nevertheless, we were restricted here to nine days but were still able to provide a reasonable and valuable picture of the butterfly diversity in the area.
Fig. 3.
Individual based rarefaction curves [25] for each of the different sites (top) and with 95% confidence intervals for El Esteron and La Bocana (bottom). Note that the La Bocana curve is truncated in order to evaluate the confidence intervals at the smaller individual sample size of El Esteron.
Evaluation of habitats and suitable sampling methods
Taken together these results corroborate a growing consensus that human dominated landscapes can often support diverse assemblages of butterflies [23, 36, 37]. This is not to say that natural habitat is unimportant nor do we imply that the butterfly community sampled here is equivalent to a community within a large forest preserve. The data show that the sites with some amount of forest (La Bocana and La Laguna) had the highest diversity and provided habitat for the most species of butterflies. Other insect species have not been sampled at Playa El Icacal but in the absence of such data, we could use this assessment as a biodiversity indicator and predict insect species patterns to follow similar patterns as those found in the butterflies. Though correlations between butterflies and other insects are decidedly imperfect, in the absence of more complete insect survey data we suspect that La Bocana and La Laguna are likely to be important sites for insect conservation in the Playa El Icacal region. Certainly, further surveys of insect communities at Playa El Icacal to test these correlations and relationships in an adaptive management context would be ideal [38].
Families and subfamilies of butterflies tend to have different habitat requirements and it can be informative to examine within taxa patterns. For example, Zaretis ellops of the Charaxinae subfamily is associated with undisturbed habitat [27], and the capture of this individual only at La Bocana highlights the importance of the dry forest habitat at Playa El Icacal. Also, we caught one individual of Opsiphanes tamarindi (Brassolinae) in La Bocana, another forest-associated species and the only representative of the Brassolinae. We found no individuals of the Ithomiinae subfamily. Ithomiinae have been targeted as useful indicator taxa for butterflies as they tend to have very forest-specific habitat requirements [39]. It is unclear whether their absence at Playa El Icacal is a result of limited sampling given the high seasonality of the Ithomiinae [40] or if it is due to the lack of good forest habitat. El Manglar was distinguished in the NMDS ordination as being unique to the other sites (Fig. 4). Compositionally the site was unique in a couple of ways including large numbers of Danaus individuals as well as a more diverse lycaenid fauna (Appendix 1).
Baited fruit traps are frequently cited as a tool for rapid assessment of neotropical butterfly communities [41]. Using these traps we found three species (Z. ellops, O. tamarindi, and T. laothoe) that we would have otherwise not seen. However, only a very small fraction (6%) of the species was represented by the fruit-baited trap captures with only seven trapping days, though additional traps might have also improved capture rates. For rapid assessments of only a few months and in habitats other than rainforest, it is advisable to supplement fruit-baited trapping with hand-net capture.
Diversity estimates: uncertainty and assumptions
A more complete survey of an area such as Playa El Icacal would require more sampling time, as the seasonality of butterflies is one factor likely to complicate sampling efforts. For example, dry forest habitats, such as those found at Playa El Icacal, are often characterized by a succession of butterfly assemblages throughout the year [17]. Not only do dry season and wet season assemblages differ from one another, but also the composition of butterflies at the beginning of the wet season typically differs from the composition of species mid-way through the wet season.
To illustrate this point, consider a similar butterfly inventory undertaken in comparable habitats (though baited fruit traps were not used) in July and in December of 2006 at nearby Salamar, Department of Usulután, El Salvador, about 15 kilometers west of El Icacal [42]. After nine days in Salamar, 34 species were found. Most of these species were from the families Nymphalidae, Pieridae, and Papilionidae. There were 20 species shared between the Salamar and the present surveys. We did not find several species that we expected to find, though they were found by Mendéz and Funes [42]. These included Siproeta stelenes, Heliconius hecale, and Eueides isabella. Also interestingly, we did not find Anartia fatima until the fourth day of sampling despite this species being one of the most commonly seen butterflies in Central America [17]. These species are often very common in most Mesoamerican habitats and their absence from our sampling is more likely indicative of seasonal differences than an actual absence of these species at Playa El Icacal.
The species richness estimators are based on the assumption that the community being sampled is present at all times of sampling. But, given the seasonality of the butterflies, many species may not have been flying during the sampling period. The ratio extrapolation based on the exécted species list is reflective of all expected butterflies in a given habitat rather than solely of the present community at the time of sampling. It is therefore possibly more resistant to seasonality problems than the species richness estimators (ICE, Chao 1, Chao2, Jack 1, etc.). This could explain the 20% (or greater) difference in the completeness estimates for the inventory and suggests that in short time periods that don't reflect a community's seasonality, the ratio extrapolation approach may be more robust to such problems.
The ratio extrapolation approach has its own assumptions and uncertainties however. First, it assumes a strong correlation between finding the focal species and the target species (see Methods and Materials section). In this instance it seems likely that the focal species (broadly distributed species of disturbed habitat within Nymphalidae, Pieridae, and Papilionidae) are easier to find than the non-focal species given the detection bias towards the focal species. Therefore, it is quite possible that 200 species is an underestimate. On the other hand, it is also possible that a lack of correlation between the non-focal and focal species exists not because of detection bias but instead because of “presence” bias. It is possible that non-focal species are distributed differently than focal species and these groups are simply uncorrelated with one another. This could then overestimate the number of species, as a large percentage of the non-focal species predicted to be present at the site would simply not be there. Given these uncertainties and assumptions, we feel that 200 species is a good first approximation of the number of species present at Playa El Icacal but also predict the actual number of species is likely to be more or less based on the accuracy of the focal and non-focal taxa correlation.
Implications for conservation
There are an estimated 540 species of butterflies (Papilionoidea and Hesperioidea) at El Imposible, El Salvador's largest national park, which consists of tracts (about 9,000 acres) of tropical dry and moist forest covering a large altitudinal range from near sea level to over 1,400 m a.s.l. [43]. We certainly would not expect to find comparable numbers of species at a site as small as Playa El Icacal, without moist tropical forest and a large altitudinal range. However, the 100–200 butterfly species that we did estimate are impressive given the lower variability in habitat and greater fragmentation of native habitat (Fig. 5).
Fig. 5.
A sample of the butterfly species found at Playa El Icacal. Arawacus sito (top left), Dryas iulia (top right), Heraclides cresphontes (bottom left), and Timochares trifasciata (bottom right). Photos by T. Bonebrake and R. Sorto.
There are a variety of techniques and approaches to monitoring of neotropical butterfly communities [see 16]. In this study, we provide further guidelines with special emphasis on time-limited rapid assessments:
Multiple data collection techniques (including baited traps, hand-net captures, visual censuses, and digital photography) enhance the ability to find species.
Species accumulation curves and richness estimators can provide not only useful information about the number of species in a region, but also an estimate of the completeness of an assessment. However, seasonal changes can result in an underestimate of species richness.
Using common species lists, or expected species list, as an index for ratio extrapolation can be another approach to estimating assessment completeness (and species richness). However, the accuracy of this extrapolation is dependent upon the correlation between focal (butterflies on the list) and target species (all butterflies) which introduces error into estimates.
We have provided a list based on designations of Costa Rican butterflies by Thomas [27], which could be used as a list of butterflies expected to be found in most Mesoamerican locations with even small amounts of suitable, non-paved habitat.
Comparing diversity between habitats is very sensitive to sample size, but using rarefaction curves allows for accurate comparisons between habitats even when sample sizes are low or disparate.
Community-level analysis through ordination can be a useful means of describing species composition and distinguishing unique community assemblages.
Time allowing, exhaustive and extensive sampling, both temporally and spatially, is the only way to understand and document a butterfly community fully, particularly in such complicated ecosystems as tropical forests [44]. Under resource-limited circumstances this is rarely possible. However, important and accurate information can be attained in a short time when careful attention is paid to the uncertainty and flaws inherent in such time-limited sampling efforts. Furthermore, this is the only survey of any insect community undertaken at Playa El Icacal to date. While the utility of butterflies as indicators of biodiversity continues to be tested and refined, we feel that the results presented here will helpfully guide conservation efforts in this area. We also hope that perhaps this study will inspire a more thorough investigation of the insect and ultimately the ecological community of this Salvadoran coastal countryside and others like it.
Acknowledgements
We are grateful for the support of this research by SalvaNATURA. We thank the National Science Foundation (OISE-0832204) for additional funding support. Comments and suggestions from C. Boggs, P. Ehrlich, O. Komar, E. Pringle, and two anonymous reviewers improved the manuscript by an order of magnitude. V. Henríquez helped create the Playa Icacal map. K. Lara masterfully organized and executed field accommodations and facilitated the collection permit process. We thank the Ministerio de Medio Ambiente y Recursos Naturales (MARN) in El Salvador for granting the permits.
References
Appendices
Appendix 1.
Number of individuals sampled for each species by each site during nine days at Playa El Icacal along with species richness total from each of the sites.
* Species sampled by fruit-baited traps.
Appendix 2.
A list of 74 butterflies we expected to find based on designations by Thomas [27] of widespread butterfly species often found in modified habitats. Species in bold represent species we sampled at Playa El Icacal.
