Rapid assessments of tropical dung beetle and butterfly assemblages: contrasting trends along a forest disturbance gradient


Owen T. Lewis, Department of Zoology, University of Oxford, South Parks Road, Oxford OX1 3PS, U.K. E-mail: owen.lewis@zoo.ox.ac.uk, Tel: 01865 271162, Fax: 01865 310447


Abstract.  1. We carried out rapid assessments of the richness and diversity of fruit-feeding butterflies (sampled with baited traps) and dung beetles (sampled with buffalo dung-baited pitfall traps) at 20 sites across an anthropogenic forest disturbance gradient in Ba Be National Park, Vietnam.

2. We investigated measures of diversity, richness, and functional composition for individual taxa in relation to the degree of disturbance, and verified whether dung beetles and butterflies showed congruent trends.

3. For butterflies, overall species richness increased with forest disturbance, but the richness of rare species decreased. Species diversity was uncorrelated with disturbance.

4. In dung beetles, species richness was unrelated to forest disturbance, but species diversity increased with forest disturbance. The richness of dung beetles in the telecoprid (roller) guild declined with forest disturbance.

5. There was no significant correlation between dung beetles and butterflies across sites for either species richness or species diversity.

6. Apparent effects of disturbance were thus sensitive to the particular metric used (species richness or diversity), the taxonomic group studied (butterflies or dung beetles), and the functional group investigated (different guilds of dung beetle).


It is widely recognised that conservation biologists should take advantage of insect diversity as a rich data source for planning and management in tropical forests (Kremen et al., 1993). Conservation decisions require information on species composition, species richness, and diversity, but the formidable diversity of the tropical forest insect fauna makes it difficult to compile thorough species inventories (Godfray et al., 1999; Putz et al., 2001; Lewis & Basset, 2007). On the assumption that focal taxa can act as predictors for diversity patterns in unstudied groups, an alternative approach is to focus on individual ‘biodiversity indicator’ taxa (Pearson, 1994). A variety of criteria for selection of appropriate indicator groups has been proposed (McGeoch, 2007).

However, there is considerable concern that single-taxon studies may not provide useful information for conservation planning .The few studies available suggest that correlations in richness and diversity among taxa are weak (Lawton et al., 1998; Barlow et al., 2007; Basset et al., 2008a,b). Particular indicator groups have different merits and it is unlikely that any single group will adequately represent the overall diversity (Kremen et al., 1993; Pearson, 1994; Brown, 1997; Osborn et al., 1999). For this reason, it is widely recommended that multiple taxa be included in such studies, especially if these have contrasting ecological roles and habitat specialisations (Kremen et al., 1993). However, relatively few studies have used a multi-taxon approach in tropical forests.

Here we focus on two insect taxa, butterflies and dung beetles, which have been used widely as indicators in tropical forests. Butterflies fulfil many of the criteria proposed to define useful indicator groups: they have short generation times, are day-flying, diverse, and easily identifiable. Furthermore, they are closely associated with other resource and ecosystem characteristics (Brown, 1991) and thus might be expected to act as ecological indicators and also reflect diversity in other groups (biodiversity indicators; see Basset et al., 1998). Butterfly responses to forest disturbance have been investigated at a range of spatial scales (e.g. Bowman et al., 1990; DeVries et al., 1997; Hamer et al., 1997; Lewis, 2001; Ghazoul, 2002; Cleary, 2003; Lien & Yuan, 2003; Spitzer et al., 1993).

Dung beetles (Coleoptera: Scarabaeidae), have also been viewed as a useful indicator group (Halffter & Favila, 1993; Spector & Forsyth, 1998; Nichols et al., 2007). They have a major functional role in tropical forest ecosystems through nutrient cycling, secondary seed dispersal, and control of vertebrate infection through removal of sources of infection (Nichols et al., 2008). They reflect the structural (Davis & Sutton, 1998) and botanical (Andresen, 2003) composition of their habitats, as well as the number of mammal species present and the quantity of their dung (Estrada et al., 1999). Both butterflies and dung beetles have been intensively studied compared with most tropical insect taxa, and they have relatively well understood taxonomy and ecology. They are ubiquitous in tropical regions and can be sampled using simple, low-cost methods. Collection can be standardised geographically and identification can be carried out thoroughly compared with other invertebrate groups.

We investigated species richness and diversity of dung beetles and fruit-feeding butterflies across a human-induced disturbance gradient in Ba Be National Park, Vietnam. We used a rapid inventory approach that corresponds to the temporal and spatial scale of many conservation studies. Our objective was to assess how measures of diversity and richness from this sort of rapid assessment are affected by disturbance in each group, and whether the two groups show congruent responses. As summary statistics such as species richness and diversity may mask changes to species composition with important conservation or functional consequences (Lewis & Basset, 2007), for each group we investigated the community response to disturbance in more detail. For dung beetles, we assessed whether disturbance affects species composition in terms of functional guilds. For butterflies, we investigated whether species with restricted distributions (endemics) tend to be associated with less disturbed habitats.


Study site

Ba Be national park is located in the Bac Can province in north-eastern Vietnam, 254 km north of Hanoi (22°24′N, 105°37′E), within the northern Indochina subtropical forests. Topography of Ba Be is characterised by steep limestone hills, interspersed with flatter non-limestone areas. The area has a variety of land use types, including patches of primary forest, areas of vegetation that have re-grown following partial clearance, and patches subject to permanent and shifting agriculture. The forest can be classified into two main types: limestone forest and lowland evergreen forest. The climate is monsoon tropical, with an annual precipitation of approximately 1300 mm, a mean annual temperature of 22 °C and a cool, dry season from October to March (Nguyen et al., 2000). Ba Be is one among the 12 highest priority sites for biodiversity conservation in Vietnam (Government of SRV/GEF, 1994). The forests of Ba Be have been historically subject to shifting agriculture, and although there has been legislation against this practice in the National Park since 1994, much of the forest remains highly disturbed. This study was undertaken near the park headquarters in lowland evergreen forest. Butterflies of Ba Be have been particularly well studied, with several rare and restricted range species recorded (Monastyrskii, 2004). We are unaware of previous studies on dung beetles of this area.

Butterfly and dung beetle sampling

Butterfly and dung beetle assemblages were assessed at 15 sampling points along a 1.5 km straight-line transect through a gradient of vegetation representing a range of forest disturbance levels. All sites studied were forested, with mature trees present, but the sites varied in terms of openness, tree density, and other characteristics (see Disturbance measurements, below). Trapping took place on 28 days between 4 August and 5 September 2005, corresponding to the later part of the rainy season. Traps were evenly spaced at 100 m intervals. Although both dung beetles and fruit-feeding butterflies will move distances greater than 100 m, traps spaced by 100 m will most probably capture insects present in their immediate surroundings (Lewis, 2001; Larsen & Forsyth, 2005). One butterfly trap and one dung beetle trap was placed at each of the 15 sampling points. Traps were checked and emptied between 0800 and 1100 h each day, and re-baited every second day.

Butterflies were trapped with Van Someren-Rydon traps (De Vries, 1987), suspended 1–2 m above the ground. Traps were 30 cm in diameter and 110 cm in height, and were baited with fermenting mashed bananas. This bait is attractive to some members of the globally distributed Nymphalidae family (DeVries et al., 1997). Butterflies which could be identified in the field using a field guide (Monastyrskii & Devyatkin, 2002) were released. Butterflies which could not be identified in the field were killed by freezing and preserved for later identification. Dung beetles were trapped at each site using dung-baited pitfall traps comprising 1 l plastic buckets sunk into the ground so that the top of the bucket was level with the ground surface. The traps were half filled with water, and a few drops of scentless detergent were added. Buffalo (Bubalus bubalis) dung (50 g) was placed in a muslin bag and suspended from an angled stick over the centre of the bucket. Specimens were removed from the traps daily, transferred to Whirl-Pak® (Nasco Fort Atkinson, Wisconsin, USA) bags and preserved with 75% alcohol. The preserved specimens were taken to the laboratory for identification. Dung beetles were sorted to species or morphospecies, counted and measured. Species were identified by D.J. Mann using Balthasar (1963a,b), Kabakov and Napolov (1999) and references therein, and reference collections in the Oxford University Museum of Natural History. Butterflies were identified to species using Monastyrskii and Devyatkin (2002) and using the reference collections of A.L. Monastyrskii. Differences in sample size can lead to biased estimates of species richness (Lande et al., 2003). To assess the completeness of sampling, species accumulation curves were plotted for all 15 trap locations, separately for each taxon. Rarefaction curves (with 95% confidence intervals) were plotted from the pooled data for each taxon, using past software (http://folk.uio.no/ohammer/past/). Actual species richness values for each sampling point were plotted on the same axes as the rarefaction curve to provide a visual representation of the extent that observed samples deviate from those expected in a random sample of the same size from the community as a whole.

Disturbance measurements

Canopy openness was measured at each trap position, using the ‘canopy-scope’ (Brown et al., 2000), a transparent plastic sheet marked with dots at 3-cm intervals in a 5 × 5 grid. The canopy-scope is held 20 cm from the eye, pointed at the largest canopy gap, and the number of dots falling within the gap is counted. The canopy-scope does not give an absolute measure of canopy openness, but the scale of measurement is ordinal. Measurements were taken at nine points systematically placed within 0.25 ha around each trap location, with the trap location at the central point. Eight to ten canopy-scope measurements in 0.25 ha have been found to be sufficient to estimate canopy openness (Hale & Brown, 2005).

Stand basal area was estimated with a point sampling technique and angle gauge, using the methods described by Grosenbaugh (1952, 1967). From each fixed point (trap location) a fixed reference angle was compared with the angle subtended by the trunk of a tree at breast height (1.3 m). If the subtended angle was larger than the reference angle, then the tree was included in the count. If it was the same as the reference angle it was counted but given a half value. Rotating through 360°, 10 trees were assessed in this way at each trap location. To estimate stand basal area per hectare, the number of trees which fall into this size class were multiplied by a factor specific for the gauge angle. The distances from the centre of the village to each sampling point were measured using a GarminTM Etrex Vista Global Positioning System device (Garmin International Inc., Olathe, Kansas, USA).

Data analysis

As sample sizes varied markedly among sampling points and the lowest numbers of individuals caught at any sampling location was just seven for butterflies and 11 for dung beetles, it was not possible to use rarefaction to standardise sampling effort for statistical analysis. Instead, we used the rate at which new species are added to our inventory to predict asymptotic species richness (Colwell & Coddington, 1994). Values were computed using EstimateS (Version 7.5, R. K. Colwell, http://purl.oclc.org/estimates). Chao 2 estimators (Chao, 1984) were used as these can provide accurate predictions of species richness from small samples (Silva & Coddington, 1996). As disturbance may alter the relative abundances of species independent of changes in species richness, a diversity index (Fisher’s alpha) was also calculated (using PAST). Fisher’s alpha was chosen as it is relatively unaffected by small sample sizes (Magurran, 2004).

Measures of disturbance were highly inter-correlated. Rather than subjectively choosing a single disturbance measure for analysis, we used principal components analysis (PCA) to extract components of the variation in forest structure which best describe the relationship between sampling points based on the measured attributes. This generated new continuous explanatory variables as components of the variation. The first principal component (PC1) accounted for 87.5% of the variability within the explanatory variables. This factor increased with (in order from greatest to least importance) decreasing stand basal area, decreasing distance from the village, and increasing canopy openness. Therefore, a high PC1 score represents highly disturbed forest close to the village, with a low stand basal area and high canopy openness. Inspection of eigenvalues and scree plots indicated that inclusion of additional components provided little extra information, and PC1 was therefore used as a surrogate for disturbance in further analyses. Disturbance levels were also independently ranked from subjective visual assessment of sampling plots from 1 (most disturbed) to 15 (least disturbed). These ranks were highly correlated with PC1 (= −0.94, P < 0.001). Because these ranks increase with disturbance, to make the figures easier to interpret, they are plotted in preference to PC1 in the figures.

General Linear Models (GLMs) were used for both dung beetles and butterflies to assess the effect of our measure of disturbance, PC1, on the response variables: (i) recorded species richness; (ii) asymptotic richness (Chao 2); and (iii) species diversity (Fisher’s alpha). For all GLMs in this study, assumptions were checked through model criticism. Data for dung beetle species richness were square root-transformed to ensure homogeneity of variance. As theory suggests that species richness may peak at intermediate levels of disturbance (Connell, 1978), we tested for quadratic effects of disturbance on species richness by fitting the quadratic term (PC12) to our model.

Pearson rank correlations were calculated to test for relationships across sites between the two taxa for each of the three response variables. To investigate whether rare species show contrasting trends to overall species richness, we repeated the analysis for rare butterfly species richness (both raw values and Chao 2 asymptotic estimates), where rare species were defined as those with biogeographic ranges restricted to the eastern Himalayas, southern China, and northern Indochina (range category 1 in Table 2), and/or considered to be of particular conservation concern by Monastyrskii and Devyatkin (2002). Data were square root-transformed to ensure homogeneity of variance of residuals. To determine whether guild structure and the potential functional activity of dung beetles differed according to disturbance level, we used GLMs to assess the effect of disturbance (PC1) separately for two dung beetle guilds, paracoprids and telecoprids. Paracoprid nesters (tunnellers) bury below the dung pile, taking dung with them, while telecoprid nesters (rollers) make a dung ball and roll it away from the source pile before burying it (Hanski & Cambefort, 1991).

Table 2.   Numbers of dung beetles of each species recorded at each site. Asterisks indicate species that are rollers; all other taxa are tunnellers.
Trap site:156133121411724108519Total
Rank disturbance level:123456789101112131415
Caccobius (Caccophilus) simplex Boucomont0000000020000002
Cassolus aff. nudus Sharp0001010000000013
Catharsius molossus (Linnaeus)2200110100001008
Copris cariniceps Felsche020111608047202144
Copris carinicus Gillet0204151224478211364
Drepanocerus aff. sinicus Harold00000011323002214
Drepanocerus sp. 10200000000000002
Liatongus vertagus (Fabricius)1000000000100013
Microcopris aff. reflexus (Fabricius)0000000100000001
Microcopris propinquus (Felsche)1006614012411112110280
Ochicanthon obscurum* (Boucomont)0000000111000003
Onitis falcatus (Wulfen)61210100000006017
Onthophagus (Gibbonthophagus) aff. luridipennis Bohemann44101010000000011
Onthophagus (Gibbonthophagus) aff. remotus Harold040517521133235948116188
Onthophagus (Gibbonthophagus) aff. taurinus White020110113417360544
Onthophagus (Gibbonthophagus) proletarius (Harold)2110000000000509
Onthophagus (Indachorius) magnini Paulian0000001000000001
Onthophagus (Indachorius) sp. 150000100001000002
Onthophagus (Macronthophagus) cludtsi Ochi0000000100000001
Onthophagus (Onthophagiellus) sp. 181000000000000001
Onthophagus (Onthophagus) deflexicollis Lansberge group0110000100000003
Onthophagus (Onthophagus) orientalis Harold group0000000000000101
Onthophagus (Onthophagus) pacificus Lansberge0001100020030007
Onthophagus (Onthophagus) purpurascens Boucomont1000100000000002
Onthophagus (Onthophagus) sp. 040100000001000002
Onthophagus (Onthophagus) sp. 050020000000000002
Onthophagus (Onthophagus) sp. 06000000000230000023
Onthophagus (Onthophagus) sp. 070010000100000002
Onthophagus (Onthophagus) sp. 080020000000000002
Onthophagus (Onthophagus) sp. 091000000000000001
Onthophagus (Onthophagus) sp. 101000000000000001
Onthophagus (Onthophagus) sp. 120000100000000001
Onthophagus (Onthophagus) sp. 130000000000101002
Onthophagus (Onthophagus) sp. 140000000000010001
Onthophagus (Onthophagus) sp. 171000000000000001
Onthophagus (Onthophagus) sp. 212000000000000002
Onthophagus (Onthophagus) sp. 240000000001100002
Onthophagus (Paraphanaeomorphus) comittoides Kabakov0000000002000103
Onthophagus (Paraphanaeomorphus) sp. 221000000000001002
Onthophagus (Parascatonomus) anceyi Boucomont & Gillet01006004553020026
Onthophagus (Parascatonomus) muticifrons Endrödi0000400100001006
Onthophagus (Parascatonomus) rudis Sharp0000100100000002
Onthophagus (Serrophorus) sp. 110000000000000011
Onthophagus (Sunenaga) anguliceps Boucomont000140110114070231
Paragymnopleurus brahminus* (Waterhouse)000056356311559866
Paragymnopleurus sinuatus* (Olivier)1001000000000002
Sisyphus laoticus* Arrow0010000000000001
Synapsis yama* Gillet0000000001200306
Total (species richness)14128101688171016138111011 
Total (abundance)2523112276461255639013822444132700


Summary information

In total, 584 butterflies of 48 species and 700 dung beetles of 47 species were recorded. Tables 1 and 2 show the number of individuals of each butterfly and dung beetle species trapped at each sampling point. For butterflies, species accumulation curves for some locations appeared to approach an asymptote. In contrast, dung beetle species richness was still increasing at the end of the sampling period, even in the best-sampled locations. Rarefaction curves for the pooled data across all 15 sampling points are shown in Fig. 1, with observed values for each sampling point plotted on the same axes. For the butterfly data, values for most sampling points fall within the 95% confidence interval, suggesting low heterogeneity in richness among locations: the number of species trapped at each point is not higher or lower than that expected if species richness in all the plots was identical. For dung beetles, there were several plots (rank disturbance levels 6, 9, 11 and 14) in which the number of species captured was lower than expected by chance if species richness in all the plots had been identical, but there was no clear trend for this to be related to the levels of disturbance.

Table 1.   Numbers of butterflies of each species recorded at each site. Asterisks denote species of particular conservation concern or habitat specific species, as defined by Monastyrskii and Devyatkin (2002). The column headed ‘R’ gives biogeographic range categories, in decreasing order of endemicity: (1) Eastern Himalayas, South China, Northern Indochina; (2) Indo-Malayan mainland; (3) Entire Indo-Malayan region; (4) Indo-Malayan and Australasian regions; (5) Old World tropics.
Trap site: 156133121411724108519 
Rank disturbance level: 123456789101112131415 
 R               Total
Melanitis leda (Linnaeus)523252925168111465131105182
Melanitis phedima (Cramer)313270613110850038
Elymnias patna (Westwood)20000100010022107
Elymnias hypermnestra (Linnaeus)31010000001010004
Elymnias malelas (Hewitson)20000000000002002
Penthema michallati* Janet10000000011102016
Lethe confusa Aurivillius34100000000000005
Neope muirheadi*(C. & R. Felder)10000000000020002
Orsotriaena medus (Fabricius)43112592223110101061
Mycalesis inopia* Fruhstorfer10000201001000015
Mycalesis anaxias* Hewitson20000000000130004
Mycalesis gotama Moore20020000000000002
Mycalesis perseus (Fabricius)30001000000000001
Mycalesis mineus (Linnaeus)3312765223020600048
Mycalesis perseoides (Moore)208754231000300134
Mycalesis intermedia (Moore)216332203000400024
Mycalesis mucianus Fruhstorfer101002104000410013
Mycalesis sangaica Butler20110100000000003
Mycalesis malsara Moore238121110100410023
Mycalesis annamitica Fruhstorfer214012003000400015
Ypthima baldus (Fabricius)30220000000000004
Ypthima huebneri (Kirby)20000100000000001
Zipaetis unipupillata* Lee10000000010000102
Ariadne ariadne (Linnaeus)35000000000000005
Junonia almana (Linneaus)31000000000000001
Junonia lemonias (Linnaeus)21000000100000002
Hypolimnas bolina (Linneaus)41000000000000001
Kallima inachus (Doyere)20000100000000001
Neptis hylas (Linnaeus)38100000000000009
Pantoporia hordonia (Stoll)31110000000000003
Tanaecia niepelti* (Strand)10001000000000001
Euthalia phemius (Doubleday)31000000000000001
Lexias pardalis (Moore)30000000000000011
Stibochiona nicea (G.R. Gray)221000102011210011
Rohana tonkiniana (Fruhstorfer)21220000000000005
Herona marathus* Doubleday20000000000001001
Polyura athamas (Drury)33211000000000007
Stichophthalama suffusa tonkiniana* Fruhstorfer100000001010214110
Stichophthalama fruhstorferi* Rober10000000100100002
Thauria lathyi* Fruhstorfer10000000100011014
Discophora sondaica Biosduval30220111101000009
Discophora deo Stichel20010000000000001
Zemeros flegyas (Cramer)301010003000610012
Prosotas nora C. Felder40010000000000001
Anthene emolus (Godart)30030200000000005
Loxura athymnus (Stoll)30100000000000001
Yasoda tripunctata (Hewitson)20000000110010003
Euploea mulciber (Cramer)30000001000000001
Total (species richness) 1920181215109168105181247 
Total (abundance) 6493916243261342111696729711584
Figure 1.

 Rarefaction curves for pooled data on (a) butterflies and (b) dung beetles, with observed abundance and richness values for individual sites plotted on the same axes. Dotted lines indicate 95% confidence intervals.

Effects of disturbance on species richness

Species richness for butterflies was significantly higher in more disturbed habitats (F1,13 = 15.34, P = 0.002), but no such relationship was observed for dung beetles (F1,13 = 2.94, P = 0.110; Fig. 2a). The direction and statistical significance of these results remained unchanged if the asymptotic estimator Chao 2 was used in place of uncorrected species richness (butterflies: F1,13 = 13.36, P = 0.003; dung beetles: F1,13 = 0.03, P = 0.858; Fig. 2b).

Figure 2.

 Effects of disturbance on butterflies (closed symbols) and dung beetles (open symbols): (a) raw species richness; (b) asymptotic richness.

In contrast, when analysis was restricted to rare butterfly species richness, there was a significant negative relationship with disturbance (F1,13 = 18.50, P = 0.001), with fewer rare species in disturbed areas. All results remained quantitatively unchanged when Chao 2 values were weighted by their 95% confidence intervals to take into account biases that might be caused by differences in sample size. The quadratic term for disturbance was not significant (P > 0.05 in all cases) for either taxa, for both response variables, providing no evidence for elevated richness at intermediate levels of disturbance. Fisher’s alpha diversity index showed no significant relationship with disturbance for butterflies (F1,13 = 0.15, P = 0.705), but for dung beetles there was a significant increase in diversity with disturbance (F1,13 = 16.84, P = 0.001). There was no significant correlation in richness or diversity values between the two taxa across sites (Chao 2: r = 0.403, P = 0.136; Fisher’s alpha: = 0.069, P = 0.808).

Effects of disturbance on dung beetle community structure

The number of roller species was significantly lower in more disturbed sites (F1,13 = 47.24, P < 0.001). No significant relationship with disturbance was found for tunneller species richness (F1,13 = 1.72, P = 0.212).


Contrasting responses of different taxa

Our study demonstrates that species richness and diversity can be sensitive to forest disturbance, even over a small scale, where adjacent sampling points were separated by just 100 m. However, trends in diversity and richness were not congruent between taxa. Species richness (but not species diversity) was found to increase with increasing disturbance for butterflies. Conversely, species diversity (but not species richness) was found to increase with disturbance for dung beetles.

It is perhaps inevitable, but not always widely acknowledged, that taxonomic groups with differing sensitivities show contrasting responses to habitat disturbance. In a widely cited study, Lawton et al. (1998) investigated species richness patterns across a tropical land-use gradient in Cameroon. The habitats investigated ranged from highly modified grasslands to relatively intact closed-canopy forest. They found that trends in species richness across the disturbance gradient varied widely for different taxonomic groups, suggesting that conservation monitoring and assessment based on indicator taxa may not be useful. However, because the analyses reported by Lawton et al. (1998) use raw species richness values, it is difficult to know the extent to which the results are influenced by variations in numbers of individuals sampled. Additionally, because the disturbance gradient studied by Lawton et al. (1998) included extremely degraded (non-forest) habitats, unlike the current study, their results may not be representative of less severe habitat modification within forests where some degree of canopy cover is maintained. A recent study by Barlow et al. (2007) also documented weak inter-correlations in diversity across an Amazonian land-use gradient for 15 higher taxa (eight of them invertebrates, and including dung beetles and butterflies). Barlow et al. (2007) found that alternative metrics such as community similarity showed greater consistency across taxa, and this study and similar analyses by Basset et al. (2008a,b) for invertebrates across a habitat disturbance gradient in Gabon argue for a broadening of the metrics used by conservation biologists in such assessments to include, for example, aspects of beta diversity and functional attributes of communities.

Varying responses within individual taxa

It has been observed frequently, particularly for butterflies, that disturbance in tropical forests can lead to higher richness of generalist and widespread species at the expense of closed-canopy specialist species (Bowman et al.,1990; Leps & Spitzer, 1990; Lewis et al., 1998; Spitzer et al., 1993). While this can lead to a net increase in overall species richness, it will typically result in a decline in conservation value, as generalist taxa tend to be widespread species of low conservation concern. In the current study, butterfly species confined to less disturbed habitats tended to be rarer species with smaller geographic ranges, indicating the importance of such habitats despite lower levels of diversity and reinforcing the need to take species identity into account when assessing human impacts and setting conservation priorities.

For dung beetles, different functional groups can also show contrasting responses (Halffter & Favila, 1993; Favila & Halffter, 1997), and overall species richness can be insensitive to disturbance (McGeoch et al., 2002). In the current study, we found fewer roller species at more disturbed localities. This is consistent with results from some neotropical assemblages (Vulinec, 2002) but not others (Favila & Halffter, 1997). Tropical dung beetle communities are characterised by strong competition to partition resources that are patchily distributed in time and space (Cambefort & Hanski, 1991). The variety of nesting behaviours and modes of resource use of tropical beetles mean that tight species packing is possible (Cambefort & Hanski, 1991), and perhaps explain contrasting responses of different guilds to disturbance. Again, studies focusing on the effects of disturbance on tropical insects should be cautious when reporting only ‘total’ species richness or diversity values; where possible, information on responses of different guilds or functional groups should also be reported and analysed.

Diversity and richness measures

A wide variety of metrics are used in species richness and diversity comparisons, despite numerous studies attempting to identify the most appropriate diversity index to use under different circumstances (e.g. Magurran, 2004). Our results were highly contingent on the particular metrics used. In particular, results for both butterflies and dung beetles using diversity indices conflict with the results using species richness values (see Ghazoul, 2002 for an equivalent example involving butterflies). These differences were observed whether or not we used species richness measures that addressed the issue of incomplete sampling. As with most studies of megadiverse insect faunas, species accumulation curves had not reached asymptotes for all sampling sites (Fig. 1); asymptotic estimators such as Chao 2, to some extent, address this potential source of bias. In our study they gave similar results to uncorrected ‘raw’ species richness values, although this will not always be the case (Gotelli & Colwell, 2001). A species diversity measure, which reflects differences in the relative abundance of species and so provides more information about community composition than can be provided by species richness alone (Gotelli & Colwell, 2001), showed contrasting trends to species richness. There was no relationship between disturbance and species diversity values of butterflies in this study, perhaps because of varying patterns of species accumulation across the sampling plots; interpretation of diversity indices can be unreliable in situations where sampling levels are uneven. Similar to species richness values, diversity index values are sensitive to sample size, and the index used should ideally be chosen with respect to its biological relevance (Washington, 1984), as there are numerous ways of emphasising different aspects of the species abundance relationship (Magurran, 2004). Thus, it may be helpful to include, where possible, both richness and diversity measures in studies quantifying disturbance effects on tropical rainforest insects.

Sample sizes and the utility of indicators

In common with many tropical forest insect inventories (even those involving very large sample sizes), our measures of species richness were incomplete: species accumulation curves did not reach an asymptote. Many dung beetle taxa show marked temporal variations in activity, even over short time periods (Andresen, 2008). Because our sampling was seasonally restricted, additional taxa would probably be recorded if sampling was spread over a more extended period, even without a concurrent increase in total sampling effort. For these reasons, caution is needed in interpreting absolute measures of diversity and richness in this and similar studies, and it is possible that samples in different seasons will lead to different conclusions about disturbance effects (e.g. Hamer et al., 2005). Nonetheless, because of the difficulties in undertaking complete, replicated insect inventories in tropical forests such ‘rapid’ biodiversity inventory methods are likely to remain a widespread conservation tool where the interest is in comparative assessments of sites or treatments rather than quantifying absolute measures of richness or diversity. There is usually a trade-off between the completeness of individual diversity assessments and the number of replicates that can be assessed: logistically, near-complete inventories will be feasible for only a small number of localities. Their appropriateness can be assessed by comparing the results of more complete inventories with values calculated using smaller subsets of the data, and such studies provide reassurance that reliable rankings of sites can be achieved using levels of sampling effort substantially lower than those required to generate near-complete inventories (e.g. Fisher, 1999). However, much of the financial burden associated with biological diversity surveys in the tropics comes from taxonomic assistance and travel, rather than time spent sampling in the field (Gardner et al., 2008). Depending on the spatial scale and specific goals of individual studies, different compromises will be necessary to generate the required data sets.

While the temporal and spatial scale of our sampling was quite limited, these aspects of the study were consistent with much of the literature on this topic and therefore should be informative about the wider utility of studies aimed at using indicator groups as estimators for diversity in other taxa (e.g. Brown, 1991;Kati et al., 2004). However, several authors have demonstrated that caution is required in carrying out such extrapolations (e.g. Lawton et al., 1998), particularly if analyses are restricted to species richness alone (Barlow et al., 2007; Basset et al., 2008a,b). The current study, focused on two of the most popular indicator groups, supports this view: metrics widely used to study diversity may appear contradictory, even within individual taxonomic groups. Disturbance and species diversity can be inversely related and other aspects of community structure and considerations of species rarity should play a part in the measurement of tropical forest biodiversity for conservation purposes.


We thank Mr Dzien and F. Pottess for permission to work in Ba Be national park and for use of the Ba Be ecological research station. Five anonymous reviewers provided helpful comments. Rarefaction analyses were conducted using past written by P.D. Ryan, D.A.T. Harper and J.S. Whalley. Asymptotic estimators were calculated using EstimateS 7.5, written by R.K Colwell. OTL is funded by a Royal Society University Research Fellowship.