Temporal shifts in dung beetle community structure within a protected area of tropical wet forest: a 35-year study and its implications for long-term conservation

Authors


*Correspondence author. E-mail: federico.escobarf@gmail.com

Summary

  • 1Throughout much of the tropics, habitat loss is increasing and intensifying on the unprotected land surrounding conservation areas. The influence of these land-use changes on biodiversity is poorly understood. This study used data on dung beetles, a taxonomic group widely acknowledged to be an effective ecological indicator of anthropogenic disturbance, to evaluate temporal changes in diversity inside a natural protected area.
  • 2Using data from quantitative sampling events over the last 35 years along with an exhaustive review of the information available in museums and the literature, we present evidence suggesting that the dung beetles community structure has shifted dramatically over time at La Selva Biological Station, Costa Rica.
  • 3To date, 50 dung beetle species have been reported from La Selva. Of these, 10 (20%) were consistently collected within the study timeframe while 21 species (42%) were uncommon. Our results indicate a tendency toward decreasing species richness and changes in species composition over time.
  • 4Analysis of the community structure revealed a decrease in diversity (H′), an increase in dominance (D) and a decrease in evenness (J) over the 35-year period; all of which can be linked to an increase in the dominance of one species (Onthophagus acuminatus). These changes were also reflected in the proportional abundance of major species guilds.
  • 5Synthesis and applications. Despite the relatively low human impact within La Selva, this study suggests that the dung beetle community has changed as result of habitat loss in the surrounding landscape and the progressive increasing isolation of the reserve over the last 35 years. Our findings, together with studies of other biota in the reserve, indicate a worrying decline in the conservation value of La Selva in recent decades. This shift in the diversity and composition of native forest biota needs to be taken into account in future studies that continue to rely on La Selva as providing an intact baseline for comparative research. More importantly, we suggest that the size of the reserve may need to be increased if its ecological integrity is to be restored. This study provides further evidence that isolated protected areas may often fail to safeguard biodiversity in the long term, and that to be viable, conservation strategies urgently need to adopt a wider landscape perspective.

Introduction

How and why species diversity changes in a given location are questions that have long fascinated ecologists. The comparison of species diversity at different spatial and temporal scales, and in different ecological and biogeographical settings, is one of the ways we attempt to understand the mechanisms that determine the structure of communities and their functional regulation (Fukami & Wardle 2005). As such, these comparisons provide a valuable tool for wildlife management, as well as in the planning and development of conservation programmes (Turner et al. 2003). In particular the efficacy of conservation efforts will be limited unless we pay attention to precisely how human activities alter species diversity within and outside protected areas (DeFries et al. 2005).

Protected areas represent a central component of many conservation strategies, yet habitat loss and degradation in neighbouring areas could reduce their conservation value. Around the world, many protected areas are currently under imminent threat from logging, clearing for agriculture, and other types of land-use change (Hansen & DeFries 2007). According to DeFries et al. (2005), since 1980, 66% of 198 reserves in the tropics have been losing the forest habitat in their immediate surroundings (within 50 km of their borders) at a rate of 5% or more per decade. Many empirical studies have established that area reduction and isolation of native habitat frequently have negative effects on biodiversity and ecological functioning (Fisher & Lindenmayer 2007). Therefore, long-term studies are key to understanding the conservation value of protected areas that are situated in landscapes undergoing intense anthropogenic modification.

Temporal community dynamics can be evaluated by assessing the continual presence of the component species and whether their abundances are constant (Magurran & Philip 2001). It has been suggested that all communities are dynamic, and that this is related to environmental variability, habitat type and spatial distribution, as well as to the dispersal capacity of the species and the intensity of interspecific interactions (Shuri 2007). The amount of variation detected therefore depends on the temporal and spatial scales at which species are observed, the degree of taxonomic resolution and on the definition of the community (Rahel 1990).

Despite the importance of invertebrates for many critical ecological functions, their dominant contribution towards the total animal biomass of most natural systems, their unparalleled contribution to the totality of biodiversity, and their potential use in conservation planning (Kremen et al. 1993), long-term ecological studies of the majority of species are extremely scarce. Scarabaeinae dung beetles have traditionally been used as a focal taxon for biodiversity and conservation studies in many regions of the world, but there is only one long-term study on dung beetles in protected areas (Howden & Howden 2001). By using dung as a food and nesting resource, they provide several important ecological services such as waste removal, soil nutrient recycling, secondary seed dispersal and vertebrate parasite suppression (Nichols et al. 2008). Numerous studies indicate that the vegetation structure and soil type, as well as the spatial and temporal availability of excrement, affect the dung beetle community at the local level (see Hanski & Cambefort 1991). Overlaid on these natural patterns, human activities such as deforestation, and the intensification of agriculture and cattle ranching, also play a key role in shaping extant dung beetle assemblages (Nichols et al. 2007).

Using data from a series of dung beetle samples collected over the last 35 years in the La Selva Biological Station in Costa Rica, along with an exhaustive review of the information available in biological collections and the literature, the goal of this study was to further our understanding of the long-term ecological integrity of this iconic tropical conservation area that has suffered from the progressive loss and degradation forest areas in the neighbouring. Specifically, we address the following questions:

  • 1How has dung beetle species diversity changed over time?
  • 2How has the relative species abundance structure changed?
  • 3What is the temporal pattern of species turnover?
  • 4What changes occur in the structure of major ecological guilds over time?

Our analyses provide a novel case study of the importance of implementing long-term ecological monitoring programmes in order to assess the viability of protected-area strategies for conservation.

Methods

study area

The study was conducted at the La Selva Biological Station (10°26′ N and 83°59′ W) located in northern Heredia, on the Atlantic slope of the Central Cordillera at the confluence of the Sarapiquí and Puerto Viejo Rivers. La Selva encompasses 1600 ha of tropical wet forest between 30 and 200 m above sea level. Mean annual precipitation reaches 4000 mm distributed more or less uniformly throughout the year and mean annual temperature is 24 °C. Close to 55% of the total area is covered by mature forest, and it is in this type of vegetation that the majority of the studies on successional dynamics at La Selva have been carried out (e.g. Lieberman & Lieberman 1987). The remainder is comprised of a variety of habitats, including selectively logged primary forest (7%), early successional pasture (18%), young secondary forest (11%), and abandoned plantations (8%) (Hartshorn & Hammel 1994).

The Finca La Selva (area: 800 ha) was originally purchased in 1953 by the eminent tropical ecologist L. R. Holdridge, who planned to develop it as a commercial forest and fruit plantation with the goal of putting his ideas about tropical wet forest sustainability into practice (McDade & Hartshorn 1994). The main plantations were for cacao Theobroma cacao, pejibaye palm Bactris gasipaes, banana Musa paradisiaca and the timber tree, laurel Cordia alliodora on the alluvial terraces of the Río Puerto Viejo at the northern end of the property. Subsequently, La Selva was bought by the Organization for Tropical Studies (OTS) in 1968, which also acquired some of the proprieties around it and continues to administer the reserve today. Except for the abandoned plantation on the alluvial terraces and an area of succession strips and young vegetation forest from pasture along the eastern border of the reserve, the core area of La Selva has no known recent history of human disturbance (McDade & Hartshorn 1994). In contrast, the surroundings of La Selva have undergone dramatic land-use changes in the last four decades, especially through the elimination of large tracts of forest to make way for cattle ranches. In the Atlantic lowlands, this has been responsible for the transformation of approximately 200 000 ha between 1963 and 1983 (Butterfield 1994). According to Sánchez-Azofeifa et al. (1999), the annual deforestation rate in the region was 3·2% in the Sarapiquí region between 1976 and 1996. As a result, the number of forest fragments increased from 537 to 1231 and the average forest remnant size decreased from 96 to 25 ha. Today, La Selva is an isolated portion of forest connected to the Braulio Carrillo National Park at its south-eastern edge, and surrounded on other sides by pastures, small farms and commercial plantations.

sampling and data

The data set is comprised of three sampling periods: 1969, 1993–1994 and 2004. All sampling sites, except those used for the extensive supplementary inventory, were located in the far north-east of the reserve over an area of approximately 2·5 km2. Dung beetles were always captured using pitfall traps buried flush with the ground and baited with excrement (human or pig, both omnivorous species) or with carrion (decomposing fish or meat); the most effective baits for dung beetle studies in the Neotropics. All of the specimens collected during the three samples were checked and identified by one of the authors (Á. Solís) to avoid taxonomic bias.

The 1969 data set represents the first quantitative sampling of the dung beetle fauna of La Selva. Sampling was done using 26 traps (9 cm diameter, 11 cm depth) in September and October at three sites: mature forest (10 traps: seven with human excrement and three with carrion); secondary forest (eight traps: five with human excrement and three with carrion); and an abandoned cacao plantation (eight traps: six with human excrement and two with carrion). Traps were buried along a linear transect 50 m apart and left for 24 to 36 h before collection (mean ± SD: 30 ± 8·5 h).

The 1993–1994 data come from an intensive monthly sampling programme carried out between April 1993 and March 1994 by Arthropods of La Selva Project (ALAS) inside the biological station. Four sites were sampled: two sites in mature forest, one in secondary forest and one in a transition zone between mature forest and secondary forest. At each site, eight traps (7 cm diameter, 9·5 cm depth) were set in a linear transect separated by 100 m, baited with pig excrement and left for 24 h. For the temporal comparison, we used the data obtained between September and November 1993 in order to have a similar sampling effort and make the data comparable with the sampling carried out in 1969.

The 2004 part of the study focused on two objectives. The first was to sample close to the sites used in 1969 and have a similar sampling effort (Table 1). Sampling was done during the month of October at three sites: one mature forest and two sites in secondary forest, with 10 traps (10 cm diameter, 11 cm depth) separated by 50 m along a linear transect. Five of the traps were baited with human excrement and the other five with carrion for ~30 h. The second objective was to carry out an extensive inventory of the dung beetles throughout the reserve. Ten sites were selected, representing at least seven different types of vegetation. At each site, 15 traps, 50 m apart, were set along a linear transect: five with human excrement, five with carrion and five with rotting fruit (a mixture of banana and papaya). To complement this inventory, a site was chosen in the mature forest for the installation of 15 canopy traps at a height of over 20 m in five trees that were more than 100 m apart. In each tree, three traps were set with different types of bait (human excrement, carrion, fruit). In addition, two flight interception traps were set in the understorey, separated by 250 m. All traps used for the extensive inventory were left in the field for 48 h before collecting the specimens captured.

Table 1.  Total and site average (± SD) species richness and abundance sampled during each time period. Estimators of species richness are also given. *Inventory completeness is the percentage of observed species with respect to the number of species predicted by the estimators
SampleSampling effort (hour trap−1)Richness (mean ± SD)Abundance (mean ± SD)Estimators ± SD (%)*
Chao1Chao2
Intensive inventory844826 (11·7 ± 4·3)1990 (180·9 ± 179·3)30·0 ± 5·3 (86·6)28·7 ± 3·5 (90·6)
Extensive inventory720028 (9·8 ± 4·4)1524 (152·4 ± 163·8)30·5 ± 3·1 (91·8)32·9 ± 5·0 (85·0)
196978025 (13·0 ± 1·0)217 (71·6 ± 19·5)27·2 ± 2·5 (91·6)33·6 ± 6·8 (74·3)
199376818 (10·5 ± 2·7)262 (65·5 ± 65·0)19·0 ± 1·8 (94·7)20·0 ± 2·7 (90·2)
200490018 (12·0 ± 3·6)687 (229·0 ± 226·5)28 ± 10·2 (64·3)30·5 ± 17·1 (59·0)
All 3 years224829 (11·7 ± 2·36)1164 (116·4 ± 130·6)29·6 ± 6·9 (98·0)30·1 ± 8·42 (95·0)

In order to obtain the most complete species list possible for La Selva and its surroundings (< 3 km), we used two additional sources of data: (i) the entomological collections of the Instituto Nacional de la Biodiversidad (INBio) and the ALAS Project, and (ii) the available literature, particularly taxonomic revisions and studies of the geographical distribution of the coprophagous beetles of Costa Rica carried out in the last 10 years (Kohlmann 1996; Kohlmann & Solís 1997, 2001; Solís & Kohlmann 2002, 2004; Kohlmann et al. 2007).

data analysis

To evaluate the efficiency of the sampling events, we used species accumulation curves with the number of individuals collected (individual-based rarefaction) as a measure of sampling effort, as recommended by Gotelli & Colwell (2001) in order to avoid bias in the comparisons that might result from differences in the intensity of sampling events or differences in overall abundance between sites. For the direct statistical comparison of the accumulation curves, we calculated the number of species observed ± the 95% confidence interval using the analytical formula proposed by Colwell, Mao & Chang (2004). We used two non-parametric estimators to estimate total species richness (Chao1 and Chao2), which are the most appropriate for small sample sizes (Colwell & Coddington 1994). Inventory completeness was measured as the percentage of observed species with respect to the number of species predicted by the estimators.

We used the Bray–Curtis index based on abundance to examine the changes in species composition over time. The value of the index is 1 when the species compositions of the data being compared are identical and the index drops to zero when there are no species in common between samples (Magurran 2004). We also counted the observed number of shared species. Given that any measure of beta diversity depends on the number of shared and exclusive species between samples, we also calculated the estimated number of shared species using the procedure proposed by Chao et al. (2005) with the routine provided in EstimateS version 7·5 (Colwell 2005). The reason for comparing the observed and estimated number of shared species is to obtain additional information regarding the exactness of the measure of observed beta diversity (Chao et al. 2005).

To quantify the contribution of the sites (spatial) and sampling year (temporal) to the total diversity, we used the additive partitioning model (Lande 1996), in which total observed species richness is distributed as follows: γ = αsite + βsite + βyear, where α is local diversity and β is species turnover. The additive distribution model expresses α and β diversities in the same units, allowing for the direct evaluation of the relative contribution of each to total diversity (γ) (Crist et al. 2003). The probability that the observed values of α and β diversity could have been obtained as a random outcome was checked using bootstrapping (10 000 iterations based on species abundance) with the PARTITION program (available at http://www.users.muohio.edu/cristto/partition.htm). This randomization procedure generates a null distribution of estimated α and β values, making it possible to statistically compare the observed values for each scale of analysis.

Changes in community structure were analysed by comparing the distribution pattern of species abundance (Magurran 2004). For each year, we calculated Shannon (H′ = –∑pi ln pi) and Simpson [D = ∑(ni(ni − 1))/(N(N − 1))] and evenness (J = H′/Log S), where pi is the proportion of individuals belonging to the ith species, ni is the number of individuals of each species, N is the total number of individuals captured and S is the total number of species (Magurran 2004). The values of the indices were compared statistically using the null model developed by Solow (1993) and implemented in Species Diversity and Richness version 3·0 program (Henderson & Seaby 2002).

In addition, abundance rankings were used to assess whether the species’ ranks were correlated through time (Rahel 1990). The null hypothesis that the abundance ranks are not correlated over time was tested using the concordance coefficient (W), a non-parametric correlation statistic for multiple samples (Sokal & Rohlf 1981). The values from this statistic range from 0 to 1, with 0 indicating no concordance among the ranks of abundance over time (i.e. the community is unstable), and 1 indicating complete concordance in the abundance ranks.

Finally, species were classified into guilds using two criteria: by size (total length) as large (> 10 mm) or small (< 10 mm), and by food relocation habit as roller (R), tunneler (T) and dweller (D). We used a goodness-of-fit test to examine whether the proportion of guild abundance, a group of species that use the resource in the same way, was the same between sampling years. To detect those guilds for which the proportion differed among years, we subdivided the contingency table and reanalysed the data with the same test (Zar 1996). Where expected frequency was < 5, we used the goodness-of-fit test with a Monte Carlo simulation (10 000 times) of a multinomial sampling distribution with the VassarStat program (Lowry 2007). Bonferroni test by the Dunn–Šidák method was applied to correct the significance level of alpha (α) from sequential non-independent comparisons (Sokal & Rohlf 1981).

Results

Based on all the data sources available, the dung beetle fauna of La Selva and surrounding areas is comprised of at least 50 species (16 genera; see Supporting Information Appendix S1). Of this total, only 10 species (20%, the core fauna) have been consistently recorded over the last 35 years (i.e. appeared in all three sampling events), while 21 species (42%) were occasional, that is, only captured in one sampling event (nine species) or appeared only in the supplementary records found in the INBio and ALAS Project's insect collections, or reported in the literature (12 species).

Regardless of whether sampling was intense in time (i.e. the 1993–1994 inventory) or in space (i.e. the 10 sites sampled during 2004), none of the sampling events captured more than 28 species (Table 1). In both cases, the curves approached asymptotes and the species richness estimators indicated that more than 85% of the species present had been captured (Fig. 1a; Table 1). A visual inspection reveals that the 1993 and 2004 species accumulation curves approached asymptotes, but the 1969 curve remained in the exponential growth phase (Fig. 1b). In 1969, a total of 25 species were captured, while in 1993 and 2004 the number of species was lower and identical (18 species; Table 1; Fig. 1b). The estimators indicate that in 1969 an average of 83% of the species present were caught and in 1993 and 2004 92·4% and 61·5%, respectively, were caught (Table 1). When the same analysis was done using the 3 years of sampling data pooled together, the cumulative number of species was 29, and according to the estimators, 96·5% of the species present were captured (Fig. 1a and Table 1).

Figure 1.

Smoothed species accumulation curves (± CI 95%, dotted line) using the number of individuals collected as a substitute for sampling effort for (a) the intensive inventory and the extensive inventory, for (b) each of the sampling years used in the temporal comparison and (c) species richness (± CI 95%, whiskers) for each year at the lowest value of abundance (1969: 217 individuals).

changes over time

Comparing the species accumulation curves at the lowest abundance value (1969: 217 individuals), rarefied species richness seems to have decreased over time from 25 species to about 13 species (Fig. 1c). However, the overlap of the confidence intervals (CI: 95%) suggests that there were no statistical differences in species richness among years. Similarly, a non-parametric one-way analysis of variance indicated that both the mean of number of species per site (Kruskal–Wallis; χ2 = 3·01, P = 0·22, d.f. = 2) and the mean number of individuals (Kruskal–Wallis; χ2 = 1·95, P = 0·37, d.f. = 2) did not differ over time.

The analysis of species composition indicates that similarity between years was never greater than 62%. According to the Bray–Curtis index, similarity values decreased over time (Fig. 2). In all comparisons, the number of shared species was lower than expected, most notably between 1969 and 2004 (Fig. 2). The observed mean number of shared species between years was 14·3 ± 2·5 (SD), while the expected was 24·5 ± 12·3.

Figure 2.

Species turnover between sampling years using the Bray–Curtis Index (circles). The observed number of shared species (white bars) and the expected number of shared species (grey bars) are also given. The numbers indicate species lost (L) and gained (G) for each comparison.

The total diversity (γ = 29) for the 3 years together expressed in an additive way indicated that local diversity accounted for 47% (α site = 13·6 species; P < 0·0001). Between-site diversity accounted for 19·6% (βsite = 5·7 species; P < 0·0001), while between sampling-year diversity accounted for 33·4% (βyear = 9·3 species; P = 0·1).

The concordance coefficient was low (W = 0·008; P = 0·78), suggesting that the abundance ranks were not correlated over time. Analysis of community structure indicates a significant decrease in diversity (H′), a notable increase in dominance (D), and consequently, a reduction in evenness (J) over time. This tendency is corroborated by comparing the shapes of the species abundance distribution curves (Fig. 3). The greatest contrast is observed between 1969 and 2004, and results from the increase in the representation of Onthophagus acuminatus (from 37% in 1969 to 60% in 2004) in the community (Fig. 3). There was a notable decrease in species of low abundance over time. In 1969, 68% of the species were singletons and doubletons (abundance ≤ 2 individuals); while in 1993 and 2004 these only represented 38% (Fig. 3).

Figure 3.

Distribution of species abundance over time. Values for each of the parameters used to describe changes in the community are given: H′, Shannon index; D, Simpson index (1/D); J, Evenness. Numbers with different letters are significantly different: a,b,c: P < 0·05; d,e,f: P < 0·004; g,h,i: P ≤ 0·004. Probability values were obtained using Solow's test (1993). *The position of Onthophagus acuminatus in each year is shown. The arrow indicates singleton and doubleton species (abundance ≤ 2).

With respect to guild membership, of all the species collected in La Selva, 38 (76%) are tunnelers (11 large and 27 small), eight species (16%) are rollers (four large and four small) and four species (8%) are dwellers (two large and two small) (Supporting Information Appendix S1). The proportion of individuals in each guild varied over time (χ2 = 3825·01, P < 0·0001, d.f. = 5; Fig. 4). Small tunnelers increased significantly in abundance (χ2 = 420·32, P < 0·0001, d.f. = 2), while in the other guilds, there was a tendency to decrease in abundance: large rollers (χ2 = 66·62, P < 0·0036, d.f. = 2) and small rollers (χ2 = 10·33, P < 0·005, d.f. = 2). There were no statistical differences for the other guilds after the Bonferroni correction: large tunnelers (χ2 = 5·89, P = 0·052, d.f. = 2), large dwellers (χ2 = 6·25, P < 0·032, d.f. = 2) and small dwellers (χ2 = 3·0, P < 0·32, d.f. = 2) (Fig. 4).

Figure 4.

Proportional changes in dung beetle guild abundance over time.

Discussion

temporal changes in the dung beetle community

The data included in this study represent two levels of diversity in space and time. The first is the cumulative list of all the species reported to date for La Selva. This number (50 species) is the gamma diversity or historical inventory. This type of list, because of its cumulative nature, favours the inclusion of tourist species, that is, species whose presence in La Selva is occasional and transitory. The second level of variation in diversity is the richness of species captured at La Selva over a relatively short period of sampling. In total, all of the quantitative sampling recorded 76% of all the known species (Supporting Information Appendix S1), whereas each sampling event only detected an average of 43% (mean ± SD: 23 ± 4·2, n = 5), and for none of the events did the estimators predict more than 67% of all the known species (Table 1). On the other hand, the species accumulation curve predicted that in 1969, the dung beetle community would be richer in species compared to 2004. The asymptotic behaviour of the accumulation curve and notable underestimation of species richness for 2004 can be explained by the increase in dominance and decrease in evenness of the dung beetle community (for a discussion of this, see Thompson & Withers 2003). These results and the discrepancy between the two levels of diversity suggest that the dung beetle community at La Selva is both spatially and temporally very dynamic.

Although the sampling protocols were not identical, sampling effort was similar over time, and therefore, we believe that the results suggest that the observed changes in the dung beetle community at La Selva during the last 35 years are real. It is impossible to separate the importance of local (e.g. forest regeneration) and regional effects on dung beetle community structure in our study, particularly those effects resulting from increasingly intense human activity around La Selva. However, in addition to isolation effects, we cannot exclude the possibility that changes in the structure of the beetle community – mainly the loss of diversity and compositional shifts – are related to other effects, including changes in the local climate or interactions between climatic and landscape vegetation changes (Whitfield et al. 2007). Isolation and associated edge effects could result in changes in the environmental conditions inside La Selva and limit the ability of beetles from nearby patches of forest to colonize the reserve. The consequence of this isolation process may be more accentuated in the far north-east of the reserve. Currently, the reserve is only connected to other forested areas on its southern boundary, where it joins the Braulio Carrillo National Park. However, even where the reserve is connected to the National Park, altitude increases abruptly, such that the lowland forest species may be effectively isolated.

The rapid expansion of cattle farming and the intensification of agriculture together with the concomitant loss and deterioration in suitable habitat have been implicated in the decline of dung beetle diversity in many tropical areas (Nichols et al. 2007). Although human impact inside the La Selva conservation area is low (McDade & Hartshorn 1994), the process of isolation and associated edge effects at the boundaries of the reserve could be responsible for modifying the habitat quality of the reserve for individual dung beetle species in terms of food availability and microclimate. These changes may be favourable to some trophic or taxonomic groups, as what occurs with Onthophagus acuminatus. This species is widely distributed throughout Central America and is able to exploit a wide variety of resources and habitats (Kohlmann & Solís 2001). However, other species may be negatively affected, especially those species that live exclusively in the forest (e.g. Gardner et al. 2008). Similarly, Sigel, Sherry & Young (2006) report a decrease in the forest bird populations with a concomitant increase in the abundance of open-habitat birds at La Selva, results they attribute to increases in isolation and to the amount of disturbed habitat in the surrounding landscape during the last 40 years. In contrast, Whitfield et al. (2007) suggest that the dramatic decline in the diversity and total density of amphibians and lizards since 1970 inside La Selva cannot be explained by habitat loss. They argue that a shift in the local climatic conditions is the most likely explanation.

Both the diversity partitioning the community and similarity analysis indicate that the mean species turnover through time is approximately 35% (i.e. about nine species in each sampling period), with turnover much more affected by species loss (mean ± SD: 8·33 ± 2·52) than by species gain (3·33 ± 2·08) (Fig. 2) over time. This helps to explain the differences found between the observed and expected number of shared species, particularly for the period between 1969 and 2004 when eight species were lost and only one was gained (Fig. 2), and the gradual impoverishment of the dung beetle community over time.

The suggestion that dung beetle communities in La Selva are instable is based on the marked change in the abundance distributions over time and the persistence of only a small fraction of the total species set throughout independent sampling periods. According to Shuri (2007), environments that are more variable over time favour species with a range of environmental tolerances and promote turnover. In La Selva, the dung beetle species that persisted in the community were found in a wide selection of habitats, including those resulting from human activity, while apparently rare and endemic species were preferentially collected in well-preserved forest (Supporting Information Appendix S1). This is similar to the findings for the Welder Wildlife Refuge in Sinton, southern Texas, at the limit of tropical conditions, where only 40% of the dung beetle species persisted during a 30-year period and there was a notable decrease in the relative abundance of native species (Howden & Howden 2001). In a study of the temporal dynamics of a dung beetle community in a montane landscape in Veracruz, Halffter et al. (2007) reported that turnover was higher and persistence was lower over time in well-preserved areas of forest, secondary forest and coffee plantations when compared to pastures. This suggests that the temporal dynamics of the community depends in part on the conservation of spatial heterogeneity at the landscape context.

Between 1969 and 2004, there was an increase in small coprophagous tunneler abundance in the extreme north-east of La Selva, while the reverse occurred for large tunnelers and particularly large rollers (Fig. 4). Increases in disturbance resulting from human activities can lead to a loss of large species (e.g. Gardner et al. 2008) and the hyperabundance of some small generalist species as reported for several tropical areas (Nichols et al. 2007). The selective local extinction of beetle species in response to habitat loss and other changes resulting from human activities can quickly lead to the interruption of important ecological functions, such as nutrient recycling and secondary seed dispersal (Nichols et al. 2008).

conclusions and further research

Our results show that the dung beetle community at La Selva has changed significantly in the past 35 years. The progressive isolation derived from habitat loss in the surroundings of La Selva and climate changes inside the reserve (see Whitfield et al. 2007) present two possible, non-exclusive, explanations for the declining species diversity. We recommend that these shifts in community structure and diversity be taken into account when La Selva is used as an ‘intact’ forest control for comparing patterns of dung beetle communities across the wider landscape scale. Our results together with similar patterns reported for birds (Sigel et al. 2006) and leaf-litter amphibians and reptiles (Whitfield et al. 2007) indicate a worrying reduction in the conservation value of La Selva over time. Therefore, we believe that the current size of La Selva (1600 hectares) is unlikely to be sufficient to protect the full complement of native forest biota. The removal of particular functional groups may also have dramatic cascading effects on ecological integrity of the forest. Thus, in the case of dung beetles, if an important fraction or entire guilds of species are lost (especially large species, as our data suggest), detrimental effects on the ecological services mediated by dung beetles could follow (Nichols et al. 2008).

Our results also highlight the need to: (i) continue monitoring the dung beetle community at La Selva using standardized sampling protocols; (ii) conduct research on the mechanisms causing the decline in dung beetle diversity; specifically the link between changes in landscape structure, isolation and edge effects, and the population density of mammals (i.e. dung supply); (iii) better understand the dung beetle meta-population dynamics in the landscape context (i.e. also taking into account the habitat that surrounds the protected area); and (iv) study the potential cascading effect of declines in key species and guilds on dung-beetle-mediated ecological services. The future prospects for La Selva to operate as a safe haven for biodiversity depend on including the reserve in regional conservation efforts. La Selva is already connected to the Braulio Carrillo National Park; however, we recommend that La Selva be further enlarged, with special attention to the northern boundary, which is currently the most isolated portion of the Reserve.

Protected areas are a critical part of conservation strategy throughout the world and an important element of the Ecosystem Approach to develop sustainable rural economies (Sayer & Maginnis 2005). However, the effectiveness of reserves to operate as viable reservoirs of biodiversity in the future could be severely threatened unless land managers can successfully integrate the role of protected areas into the management of the wider landscape (Halffter 2005). Although long-term data such as those available at La Selva do not exist for many sites in the tropics, it is likely that similar declines in species diversity are occurring within other isolated areas elsewhere.

Acknowledgements

The 1969 collections were funded by a research grant from the OTS. Thanks to J. Longino for providing the 1993–1994 dung beetle data base. We thank Pedro Reyes for his collaboration in the 1969 sampling excursion, and Carol Hameli and Darren Mann for logistical assistance with the 2004 fieldwork. We thank E. Andresen and two anonymous reviewers who helped to improve the manuscript. Bianca Delfosse translated the manuscript from the original in Spanish and Toby Gardner made valuable comments on the final version. F. E. appreciates support from University of Pretoria (Postdoctoral Fellowship Programme) to allow writing the first version of this article. This study was funded by ORCYT-UNESCO project 883·612·2.

Ancillary