Lian Pin Koh, Department of Ecology and Evolutionary Biology, Princeton University, 106 A Guyot Hall, Princeton, New Jersey 08544, USA (e-mail: email@example.com).
1South-east Asia has the highest relative rate of habitat loss and degradation in the humid tropics. The responses of less ‘charismatic’ groups, including butterflies, to habitat disturbance remain relatively poorly understood. Many South-east Asian butterflies are endemic to the region and face global extinction if current levels of deforestation were to continue.
2Here, I highlight South-east Asia as a region urgently in need of butterfly conservation research and review empirical studies of the responses of South-east Asian butterflies to land use change. Additionally, I discuss some methodological pitfalls for such studies. Furthermore, I argue for the importance of identifying the ecological correlates of sensitivity of butterfly species to forest modification and the potential biological mechanisms underlying their responses to land use change.
3There has been no consensus among previous studies on the effects of land use change on butterfly communities in South-east Asia. Of the 20 studies I reviewed, seven reported higher species richness/diversity in undisturbed (or the least disturbed) forest than in disturbed habitats, nine reported the opposite trend, three reported no difference and one reported a strong influence of seasonality on the impacts of logging.
4Some of these studies may contain inherent methodological biases resulting from the failure to control for sampling effects, the lack of consideration for the spatial scale of analysis and incomplete sampling of the vertical strata in tropical rainforests.
5Synthesis and applications. Empirical studies of the effects of land use change on tropical forest insects are sorely lacking from South-east Asia. Butterflies are an ideal taxonomic group for such investigations. Future studies should be designed carefully to avoid the methodological pitfalls highlighted here. Determining the ecological correlates of sensitivity of butterflies to forest modification is important for the pre-emptive identification of species of conservation concern and for generating testable hypotheses on the differential responses of species to forest modification. Experimental studies are needed to determine the mechanisms underlying the responses of species to land use change in order to develop effective strategies for the conservation of butterflies in human disturbed landscapes.
Tropical forests, which contain much of the Earth's remaining biological diversity, are also experiencing unprecedented rates of deforestation (Laurance 1999; Brooks et al. 2002). South-east Asia has the highest relative rate of net forest loss (0·71%) and degradation (0·42%) in the humid tropics (Achard et al. 2002), and could lose up to three-quarters of its original forests and almost half its species by 2100 (Brook et al. 2003). Although the effects of deforestation on the mega-fauna of South-east Asia have been well reported (e.g. Brooks et al. 1997; Sodhi et al. 2005), the responses of less ‘charismatic’ groups, such as insects, to habitat disturbance remain relatively poorly understood (Sayer & Whitmore 1991; Dunn 2005). This is alarming, considering the global dominance of insects among animal communities in terms of species richness, abundance and biomass (Wilson 1987), as well as the overarching importance of insects in providing ecosystem services for human societies (e.g. pollination of crops; Nabhan & Buchmann 1997).
Butterflies (Lepidoptera: Papilionoidea) are among the best-studied insect groups in South-east Asia in terms of taxonomy and biogeography (D’Abrera 1982, 1985, 1986; Corbet & Pendlebury 1992; Igarashi & Fukuda 1997; Igarashi & Fukuda 2000). They are highly sensitive to habitat disturbance and have been used commonly as an indicator taxon for ecological research (e.g. Kremen 1994; Koh & Sodhi 2004). Many South-east Asian butterflies are endemic to the region and face the grim prospect of global extinction if current levels of deforestation were to continue. From a conservation perspective, forest butterfly species (i.e. those recorded previously only from primary forests) deserve the highest conservation and research attention.
Here, I highlight South-east Asia as a region urgently in need of butterfly conservation research and review empirical studies of the community level responses of South-east Asian butterflies to land use change. Additionally, I discuss some methodological pitfalls for such studies. Furthermore, I argue for the importance of identifying the ecological correlates of sensitivity of butterfly species to forest modification and the potential biological mechanisms underlying their responses to land use change.
Mittermeier et al. (2004) identified 34 ‘biodiversity hotspots’ in the world as areas containing high concentrations of endemic species and undergoing immense habitat loss. South-east Asia overlaps with four of these biodiversity hotspots (i.e. Indo-Burma, Sundaland, Wallacea and the Philippines) (Fig. 1). Each of these hotspots has a unique and complex geological history that probably influenced the contemporary geographical range and local distribution of its species (for a detailed discussion, see Sodhi et al. 2004).
Four vascular plant, one fish, one bird and five mammal species have been listed as ‘extinct’ or ‘extinct in the wild’ in South-east Asia on the International Union for the Conservation of Nature and Natural Resources Red List (IUCN 2006). Although these numbers are not currently alarming, the level of endangerment of extant species reveals the seriousness of threats, such as land use change, faced by the regional biotas. For example, 47 insect species (8% of total threatened insect species in the world that have been assessed) are listed as threatened on the IUCN Red List [including categories of ‘critically endangered’ (CR), ‘endangered’ (EN) and ‘vulnerable’ (VU) IUCN 2006]. Because of the high proportion of endemic species in South-east Asia, the loss of many of these threatened species would probably result in global extinctions. For example, 43% of the butterfly species on Sulawesi are endemic (Fermon et al. 2005). Data on the conservation status of insects are scarce because the IUCN Red List is highly biased towards larger and better-studied taxa (e.g. vertebrates; Rodrigues et al. 2006). Of the 47 insect species listed above, 43 are butterflies (CR: 1, EN: 14, VU: 28) from only two families (Papilionidae and Nymphalidae).
Using the databases biosis Previews and Web of Science and the online search engine Google Scholar (http://scholar.google.com/), I searched for peer-reviewed research articles published between 1945 and 2006 which reported the effects of land use change (e.g. logging, agriculture) on butterflies in South-east Asia. A total of 20 relevant studies were gathered, several of which compared butterfly richness or diversity across multiple types of land use change. Overall, there is little consensus on the effects of land use change on butterfly species richness or diversity in South-east Asia, as illustrated by the following studies.
Willott et al. (2000) compared butterfly species richness and diversity between a primary lowland rainforest and a selectively logged secondary forest (6 years old) in the Danum Valley Field Centre in northern Borneo. They found that the primary forest had significantly lower species richness (121 vs. 151), Fisher's α diversity [31 ± 3 (95% confidence interval) vs. 40 ± 4] and rarified species richness (120 ± 3 vs. 143 ± 4) than the logged forest. In a separate study conducted at the same location, Hamer et al. (2005) reported that a primary forest had significantly higher Simpson's [0·93 ± 0·003 (standard error) vs. 0·91 ± 0·005] and Shannon's diversity (3·01 ± 0·07 vs. 2·82 ± 0·08) than a selectively logged forest (10–12 years old) during the dry season. The opposite trend was detected during the wet season (Simpson's diversity: 0·89 ± 0·005 vs. 0·91 ± 0·005; Shannon's diversity: 2·64 ± 0·08 vs. 2·80 ± 0·07). At Kinabalu Park, located ∼180 km north-west of Danum Valley, Beck & Schulze (2000) compared butterfly diversity across a wide range of land uses, including primary forest, selectively logged forests of different ages and a farmland. They found that Fisher's α diversity was significantly higher in a primary forest than a 5-year-old secondary forest and a farmland, but was not significantly different between the primary forest and two older secondary forests (10 and 15 years old).
Spitzer et al. (1993) investigated the effects of agricultural expansion on butterfly communities in the Tam Dao mountains of northern Vietnam (lowlands to 1200 m elevation). They showed that closed canopy forests had lower species richness (early wet season: 54 vs. 59; late wet season: 23 vs. 33) and Shannon's diversity (early wet season: 8 vs. 24; late wet season: 5 vs. 17) than cultivated areas. In a separate study at the same site, they reported that closed canopy forests had lower butterfly species richness (24 vs. 42) and Shannon's diversity (6·95 vs. 22·27) than artificial forest gaps created by logging (Spitzer et al. 1997). A different trend was reported from Thailand and Sulawesi. In the Huay Kha Khaeng Wildlife Sanctuary in western Thailand, Ghazoul (2002) found that the mean number of species per transect (14·3 ± 0·8 vs. 13·0 ± 0·4) and Simpson's diversity (0·958 ± 0·005 vs. 0·937 ± 0·005) were significantly higher in undisturbed than in selectively logged forests. In Lore Lindu National Park in central Sulawesi, Schulze et al. (2004) and Veddeler et al. (2005) showed that undisturbed mature forests had significantly higher species richness than selectively logged forests and plantations.
Other studies conducted in South-east Asia have also presented conflicting findings in terms of the impacts of land use change on butterfly species richness or diversity (Appendix 1). Of the 20 studies I reviewed, seven reported higher species richness or diversity in undisturbed (or the least disturbed) forest than in disturbed habitats (e.g. Veddeler et al. 2005), nine reported the opposite trend (e.g. Fermon et al. 2005), three reported no difference (e.g. Hamer et al. 2003) and one reported a strong influence of seasonality on the impacts of logging (Hamer et al. 2005). Studies from other tropical regions (e.g. Costa Rica) also variedly reported negative (e.g. Perfecto et al. 2003), positive (e.g. Lewis et al. 1998) and no effect (e.g. Lewis 2001) of various types of land use change on butterfly richness or diversity. While it is possible that these inconsistencies reflect the complex responses of butterflies to land use change, it is likely that they are at least partly the result of methodological biases inherent in some of these studies.
When sampling a biological community, the number of species observed (i.e. species richness) invariably increases with both sampling effort and the number of individuals sampled (i.e. sample size) (Magurran 1988). Unless sampling saturation has been reached, communities are not directly comparable in terms of species richness or diversity without controlling for sampling effects (Magurran 1988; Denslow 1995; Gotelli & Colwell 2001). To circumvent this problem, communities have been compared using numerical species richness (i.e. number of species per specified number of individuals; Kempton 1979), species density (i.e. number of species per specified area, Hurlbert 1971), ‘rarefied’ species richness (i.e. expected species richness for a standardized sample size; Sanders 1968; Simberloff 1972), expected species accumulation curves (i.e. sample-based resampling curves; Gotelli & Colwell 2001; Colwell et al. 2004, 2005), nonparametric estimators of ‘true’ species richness (examples and references found in Colwell & Coddington 1994; Chazdon et al. 1998), and non-sampling-based extrapolations (e.g. Michaelis–Menton means, Raaijmakers 1987; Colwell & Coddington 1994; Colwell et al. 2004). These methods have been discussed in detail elsewhere (e.g. Gotelli & Colwell 2001). Only eight of the 20 studies I reviewed here have used at least one of the methods above to quantify and compare butterfly species richness or diversity between land uses. It is impossible to assess the validity of the conclusions of studies that did not account for sampling effects.
Hamer & Hill (2000) and Hill & Hamer (2004) re-analysed data from past studies to examine the relationship between the spatial scale at which samples were analysed, and the effects of habitat disturbance on butterflies and moths. They concluded that the probability of observing a positive effect of disturbance onto species richness and diversity increased with decreasing spatial scale. Eleven of 12 studies conducted at spatial scales of < 1 ha reported increases in species diversity following disturbance, whereas 10 of 15 larger scale studies (> 3·1 ha) reported higher diversity in undisturbed than disturbed forests (Hamer & Hill 2000; Hill & Hamer 2004). These patterns may be explained by the differential effects of land use change on different hierarchical levels of diversity (i.e. point, alpha and gamma diversities, sensu Whittaker 1972). For example, if species are more patchily distributed in an intact forest than in a logged forest, the former would be expected to have lower point diversities (i.e. diversity of a sample) than the latter. On the other hand, because the intact forest has a higher diversity of microhabitats (e.g. forest canopy, understorey and gap) than the logged forest, the former would be expected to have a higher alpha diversity (i.e. diversity of a habitat) than the latter. Additionally, as a logged landscape probably contains a higher diversity of habitats (e.g. logging roads, timber collection points, logged forests, unlogged forest patches) than the intact forest, one would expect a higher gamma diversity (i.e. diversity of a landscape) in the former than the latter. The spatial scale at which comparative analyses of species richness or diversity are performed should be stated explicitly in the objectives of a study and any conclusions drawn from such studies must be framed within the appropriate spatial context.
Disturbance is known to disrupt the vertical stratification of butterfly communities in tropical forests (DeVries 1988; Schulze et al. 2001; Fermon et al. 2005). In studies where butterflies were sampled only from the understorey, species richness or diversity may have been inflated in disturbed habitats due to the presence of canopy species in the understorey of disturbed sites. This hypothesis is supported by the fact that in studies where both canopy and understorey butterfly assemblages were sampled, land use change was consistently reported to have a negative impact on species richness (e.g. DeVries et al. 1997; Perfecto et al. 2003; Dumbrell & Hill 2005). For valid comparisons to be made, studies should sample from all vertical strata in a tropical forest. Where this is logistically impossible, the results of the study should be discussed with this potential sampling bias in mind.
Ecological correlates of sensitivity
The community-level responses of organisms to land use change are ultimately the consequence of how each species is adapted to its natural environment and how it responds to changes in biotic and abiotic factors following forest modification. Recently, the comparative approach has been used to investigate how traits possessed by species may predispose them to extinction (McKinney 1997; Purvis et al. 2000). Such correlative analyses serve two important purposes in the context of land use change. First, they may allow us to identify pre-emptively species likely to be at risk from forest disturbance, using ecological traits that are easily measurable or readily available. Secondly, they may generate testable hypotheses as to why different species respond as they do to forest disturbance. Koh et al. (2004b) compared ecological traits between extinct and extant species of butterflies in Singapore, an island state that has lost > 98% of its original forest cover. Their analyses show that butterfly species that were more host-specific were more likely to go extinct, and that for a given host specificity, forest specialists were more likely to be driven to extinction than habitat generalists.
Traits that are potentially important for butterflies include:
2Forest specialization: forest butterflies are more likely to be impacted by forest modification than those that are known to occur in disturbed habitats. This hypothesis is supported by two studies conducted in Singapore, which showed that forest specialists were more likely to avoid urban habitats than habitat generalists (Koh & Sodhi 2004), and forest specialization as an ecological trait was one of the strongest predictors of a species becoming extinct in Singapore (Koh et al. 2004b).
3Microhabitat specialization: the butterfly fauna in tropical rainforests are vertically stratified (DeVries 1988; DeVries et al. 1997; Schulze et al. 2001; Fermon et al. 2005). In terms of species composition, butterflies in disturbed habitats are more similar to those in the forest canopy or gap than those in the understorey of intact forests (e.g. DeVries 1987; Wood & Gillman 1998; Hill et al. 2001). DeVries (1987) noted that butterflies treat forest disturbance as if ‘the canopy had come to the ground’. These studies suggest that forest understorey butterflies may be more impacted by forest modification than forest canopy or gap species.
4Larval host specificity: land use change often results in a decrease in plant richness (Foody & Cutler 2003). The more host-specific a butterfly species is, the less probable it will find a suitable host plant in disturbed habitats. Also, to escape their parasitoids, highly host-specific butterflies tend to specialize on rare plants (Weseloh 1993), which are less likely to occur in disturbed habitats.
5Adult feeding guild: the survival and fecundity of butterflies are likely to depend on the quantity and quality of adult food resource (i.e. floral nectar, rotting fruits) (Boggs 2003), which will probably vary among different land uses.
6Flight ability: butterflies capable of evading predators through fast and erratic flight are more likely to occur in habitats with high solar radiation and ambient temperature than slow-flying species (Chai & Srygley 1990; Srygley & Chai 1990). The reasons for this are discussed in the next section.
7Body size: body size is a commonly investigated trait in both ecological and palaeontological studies of extinction proneness (McKinney 1997; references therein). However, because body size is expected to scale with other ecological attributes (e.g. population growth rate, Gaston & Lawton 1988), it may be difficult to interpret biologically. Nevertheless, because body size is a convenient trait to measure, it is useful for identifying species at risk from land use change.
8Body surface area to volume ratio: surface area to volume ratio may affect the rate of radiant heat exchange between the body and the environment, which may influence the sensitivity of species to different habitats that vary in microclimate (Casey 1993).
9Coloration: body and wing coloration has important anti-predatory and heat exchange functions for butterfly larvae and adults. Chemically defended species (and their mimics) use conspicuous coloration to advertise their unpalatability to potential predators (Chai 1990). On the other hand, cryptic species have darker coloration that matches the background of their natural environment to escape detection by visual predators. Coloration could also influence the rate of heat exchange between the butterfly and its surroundings (Casey 1993). Larvae with dark coloration may enhance radiant heat absorption, whereas those with light coloration may enhance heat reflectance. Furthermore, adult butterflies may use a combination of reflectance basking behaviour (e.g. holding wings at particular angles to the body) and wing melanization patterns to control the amount of heat reflected by the wings to the body (Kingsolver 1985a,b; Kingsolver & Wiernasz 1990). The effectiveness of colouration as anti-predatory strategy or heat exchange mechanism is likely to be affected by changes in the microclimate following forest modification.
A reaction norm is the expression of the potential phenotypes of a given genotype along an environmental gradient (Stearns 1989; Woltereck 1909). Thermal reaction norms that allow for higher growth rates at lower ambient temperatures have evolved repeatedly in ectothermic organisms inhabiting relatively colder environments (Angilletta et al. 2003). One of the possible evolutionary trajectories to become specialized for growth at low temperatures is through changes in physiology, with the consequent reduction in growth rates at other temperatures (i.e. specialist–generalist trade-off, Angilletta et al. 2003). In the context of land use change, the thermal specialization of forest butterflies for the cooler forest microclimate may explain their sensitivity to forest modification. For example, if the maximum ambient temperature in disturbed habitats exceeds the critical thermal limits (Angilletta et al. 2002) of forest butterflies, it may directly cause mortality. This would be true particularly for butterfly larvae, which have limited thermoregulatory responses (Casey 1993). Additionally, the high temperatures in disturbed habitats may result in reduced growth rates (e.g. from reduced consumption rates, Kingsolver & Woods 1997), which may prolong larval exposure to natural enemies and thereby increase the probability of predation or parasitism (Benrey & Denno 1997). Reduced growth rates may also result in the eclosion of smaller female adults, which has been linked to lower potential fecundity (e.g. Tammaru et al. 1996). Any of these effects could have implications for the population dynamics and viability of forest species in disturbed habitats.
specialization for nutrient resource
The conversion of mature forests to other land uses is expected to result in compositional changes of the plant community (Tilman 1987). Since butterflies are known to be highly host specific and would not be expected to occur where their larval host plants are absent (Ehrlich & Raven 1964), the loss of host plant species following land use change would likely result in the loss of the affiliated butterfly species. This effect of butterfly–host plant coextinction is likely to be exacerbated by the tendency for highly host specific butterflies to specialize on rare plants in order to escape from parasitism (Weseloh 1993). Koh et al. (2004a) showed that the extinction probability of a butterfly species was significantly correlated with both the proportion of its host plants that are extinct and the threat of extinction faced by its extant host plants. In another study, Koh & Sodhi (2004) reported that the species richness of butterflies across 21 urban parks (1–53 ha) in Singapore was better predicted by the number of larval host plant species occurring in the park than any other explanatory variable, including park area and isolation from forests. These findings suggest that host specialization has a profound influence on the responses of butterfly species to land use change.
While not as resource specific as their larval stage, the distribution of adult butterflies may also be affected by the availability of their food resource (Boggs 2003). Different guilds of adult butterflies are known to feed on different nutrient sources, including floral nectar, pollen, rotting fruit, urine, perspiration, dung, and carrion (Gilbert & Singer 1975). The response of different feeding guilds to land use change may be influenced by the quantity and quality of their food resource, which are likely to vary across different habitats (Hamer et al. 2006).
specialization of anti-predatory strategy
The anti-predatory strategies of butterfly larvae and adults are likely to have evolved in response to the foraging strategies of their predators (including parasitoids) and environmental conditions of their natural habitat. Many chemically defended butterfly larvae and adults are aposematic (Chai 1990), and store plant compounds that are vertebrate toxins such as cardenolides, alkaloids, cyanogens and azoxyglycosides (Montllor & Bernays 1993; references therein). Land use change may affect the effectiveness of unpalatability as an anti-predatory strategy if the host plants from which these butterflies sequester their toxins are unavailable in disturbed habitats.
Another anti-predatory strategy likely to be affected by land use change is crypsis. Butterfly larvae or adults that depend on crypsis to escape detection by visual predators (e.g. birds) live typically among the shrubs and leaf litter of the forest understorey (Endler 1986; Brakefield & French 1999). Forest modifications that result in increases in light availability would probably reduce the effectiveness of their camouflage, which may result in higher mortality from predation.
Empirical studies of the effects of land use change on tropical forest insects are sorely lacking, particularly from South-east Asia. Butterflies are an ideal taxonomic group for such investigations. The lack of consensus among studies makes it impossible to conclude to what extent land use change affects butterfly species richness or diversity. Future community level studies should be designed carefully to avoid the methodological pitfalls highlighted here. Determining the ecological correlates of sensitivity of butterflies to forest modification is important for the pre-emptive identification of butterfly species of conservation concern and for generating testable hypotheses on the differential responses of species. Experimental studies are needed to determine the mechanisms underlying the responses of species to land use change in order to develop effective strategies for the conservation of butterflies in human disturbed landscapes.
I thank David Wilcove, Trond Larsen, Rachael Winfree, Tien Ming Lee, Navjot Sodhi, Juanita Choo-Koh, Keith Hamer and two anonymous referees for their insightful comments and suggestions. This work was supported by grants from the Department of Ecology and Evolutionary Biology and the Princeton Environmental Institute, Princeton University.
Table Appendix 1.. A review of past studies of the effects of anthropogenic land use change on butterflies in South-east Asia. Based on the area of the unit of comparison (e.g. transect or habitat), studies were classified into spatial scales of either small (< 1 ha) or large (> 3·1 ha) following Hamer & Hill (2000), and Hill & Hamer (2004)
Buru island, Indonesia
Unlogged lowland monsoon forest vs. selectively logged forest (5 years old)
Large (∼8–9·2 ha)
Margalef's, Fisher's α and Berger–Parker's diversities significantly higher in unlogged than logged forest
Primary forest vs. old secondary forest (logged 8–9 years ago) vs. young secondary forest (logged 4–5 years ago)
Visual transects and hand netting
Large (∼3·3–5·4 ha)
Rarefied species richness was significantly higher in young secondary than primary forest; 2005 analysis reported richness of generalist species was significantly higher in open road sites than in primary forest, while richness of herb specialists was significantly higher in old and young secondary forests than primary forest
Closed forest vs. transition zone (mosaic of vegetation patches at varying successional stages) vs. ruderal habitat (cultivated)
Small (∼0·5 ha)
Absolute species richness highest in transition zone, and lowest in closed forest; Shannon diversity lowest in closed forest, and highest in ruderal habitat during early wet season, and in transition zone during late wet season; no test of statistical significance performed
Disturbance gradient comprising closed forest, moderately disturbed forest, heavily disturbed forest, secondary forest, and agriculture and clearing lands
Small (∼0·5 ha)
Absolute species richness highest in heavily disturbed forest, and lowest in closed forest (no statistical test performed); Shannon's diversity significantly higher in agriculture and clearing lands than closed forest