Introduction
Amongst the most significant factors shaping an insect community are the spatial distribution and temporal stability of the habitat it occupies (Southwood, Reference Southwood1977). The occurrence of a habitat in ephemeral patches affects the dynamics of a community by preventing the complete mixing of individuals between patches (Sevenster, Reference Sevenster1996), which often results in uneven, or aggregated, distributions of individuals. Suitable resource patches may be sparsely distributed and the dispersal movements of individual inhabitants become important in shaping the community structure; the probability of colonisation may be determined by the dispersal and search ability of a species (Karieva, Reference Karieva, Diamond and Case1986). Where patches are frequent and close together, dispersal ability may be less important in shaping community structure.
In addition to dispersal, the probability of initial colonisation may be determined by factors such as the weather and habitat characteristics (Valiela, Reference Valiela1974); and, as a result, there may be pronounced stochastic variation in the mixture of species that find a given patch (Levin, Reference Levin1974). Following colonisation, the subsequent species composition may depend more on survival, competitive ability and predation within the community, all of which may become increasingly important as time progresses (Beaver, Reference Beaver1977). Levels of predation and competition may be high and act as further mechanisms contributing coexistence and maintenance of high numbers of species (Hanski, Reference Hanski1981; Ives, Reference Ives1991; Sevenster, Reference Sevenster1996; Hartley & Shorrocks, Reference Hartley and Shorrocks2002). As a result, the biological communities associated with patchy, ephemeral habitats are characterised by a large number of species, often with high dispersal ability, allowing colonisation of new habitat patches as they appear (Beaver, Reference Beaver1977; Chesson, Reference Chesson, Diamond and Case1986). Colonising species typically begin to inhabit a suitable habitat patch almost immediately. A strong succession of colonisation then often occurs, with specialised fauna arriving at new resource patches and giving way to more generalist species as the resource becomes incorporated into the surrounding environment (Hanski, Reference Hanski, Gray, Crawley and Edwards1987; Menendez & Gutierrez, Reference Menendez and Gutierrez1999). Larger systems of patches may show relatively higher levels of community stability, emphasising the importance of scale in ecological studies (Hanski, Reference Hanski, Gray, Crawley and Edwards1987). Successional change in habitat patches alters their suitability to new colonisers so that, as new patches arise, old ones become hostile and development within a patch may become time-limited (Southwood, Reference Southwood1977).
Herbivore dung is a good example of such an ephemeral habitat, used by a large and diverse array of organisms. It is often highly abundant and can be relatively predictable in its occurrence compared with other patchily distributed resources, such as carrion, particularly where a large population of dung-producing vertebrates is present. Many insects, particularly among the Coleoptera and Diptera, exploit this valuable resource by feeding on dung throughout their larval stages (Strong, Reference Strong1992). Understanding the assembly and structure of dung-colonizing invertebrate communities is important because the invertebrates that contribute to dung decomposition are vulnerable to a range of agricultural practices, especially the treatment of livestock with insecticides and anthelmintics (Floate, Reference Floate1998; Floate et al., Reference Floate, Wardhaugh, Boxall and Sherratt2005; Bang et al., Reference Bang, Lee, Na and Wall2007). Following administration, these compounds may be metabolised only slowly in the body and pass largely unmodified into the faeces (Sommer, et al., Reference Sommer, Steffansen, Overgaard Nielsen, Gronvold, Vagn Jensen, Brochner Jespersen, Springborg and Nansen1992; Strong & Wall, Reference Strong and Wall1994; Floate, Reference Floate1998). This exposes the larvae and adults of many dung-colonising insects to toxic residues (Sherratt et al., Reference Sherratt, Macdougall, Wratten and Forbes1998). Reductions in insect abundance are likely to be qualitative and ecological effects are likely to be subtle; hence, a precise understanding of the forces that structure the insect community in dung is required if the effects of insecticide residues are to be assessed or predicted.
The aim of the present work, therefore, was to examine the distribution of the insects colonising cattle dung, to uncover patterns within and between taxa in the degree of aggregation and to use these data to consider mechanisms by which such aggregated assemblages may be produced.
Materials and methods
Artificial, 1.5 kg cow pats, created from fresh dairy-cattle dung, were used to study populations of dung insects. Ten fresh pats were placed out each week in an area of permanent cattle pasture, with 24 batches placed out between the 21st May and 29th October 2001. Batches of pats were placed in a linear array, with at least 1 m between each pat. Pats were left exposed for seven days, following which they were retrieved and returned to the laboratory where colonizing insects were collected and identified. More complete details of the methods used and a species list with seasonal abundance patterns are given in Lee & Wall (Reference Lee and Wall2006).
Aggregation analysis
Aggregation analysis was undertaken for all species or taxa identified and quantified as the index of dispersion (ID), calculated as the variance multiplied by the number of pats minus one, divided by the mean number of individuals per pat (Southwood, Reference Southwood1977). To explore the nature of any aggregation in more detail, further analysis was then undertaken for the larvae of six taxa of dung colonizing Diptera: Sepsis spp., Chloromyia formosa (Scopoli), Mesembrina, meridiana Linnaeus, Hydrotaea spp., Scatophaga stercoraria (Linnaeus) and Musca autumnalis DeGeer; and four taxa of Coleoptera: two genera of larvae (Aphodius spp. and Sphaeridium spp.) and two species of adult beetle, Aphodius fimetarius (Linnaeus) and Cercyon lateralis (Marsham). These taxa were selected on the basis of their significance in the dung colonizing community and the fact that they represented a range of availability and abundance characteristics (Lee & Wall, Reference Lee and Wall2006). For these ten taxa, normal, Poisson, logarithmic and negative exponential models were fitted to the frequency distribution of the number of each taxon per pat by maximum likelihood estimation (Statgraphics, Centurion XV, Herndon, Virginia). Chi-square analysis was used to compare the observed and expected best-fit frequency distributions. Where a negative-binomial distribution was fitted and considered to give a good fit to the data, the coefficient of aggregation, k, was also determined. This is an inverse measure of the degree of ‘clumping’ of the population within patches. Low levels of k indicate a high degree of aggregation.
Modelling the distribution of taxa
Using all the invertebrate taxa identified from the pats, simulations were performed (Excel, Poptools, Microsoft), based on the assumption that colonization is a random binomial event in which each species colonises a pat independently of all other species. For this, for each taxon the probability of colonization, P, is given by the proportion of pats that were seen to be colonised by that taxon in any one week. Monte Carlo simulations were then carried out, in which a randomised binomial outcome with probability P is generated in over 1000 iterations for each of the taxa recorded. The simulation gives an estimate of the frequency distribution of taxa that would be expected in each batch of ten pats, in each week examined, if colonisation occurs at random and independently of other taxa. The simulation was repeated for weeks one, ten and 20, to give a representation of the range of relative abundances seen over the summer field season. The predictions were then pooled and compared to the observed numbers of taxa identified in these three weeks; the observed and expected numbers were compared by χ2 analysis.
Results
In total, 145,454 insects were collected from the 240 artificial cow pats, with a median of 333 (188–795) individuals per pat in 52 different taxa. The number of taxa present in each batch of ten pats varied as the season progressed (fig. 1). Batches placed out during the early summer consistently contained 35–40 taxa, rising to a peak of 43 in pats placed out in week 15 (3rd September) and then falling sharply towards the end of the study (fig. 1). The last five batches, placed out during October, contained the fewest taxa (no more than 27 in all cases). However, individual pats contained on average only half the taxa present in the entire batch, although the overall pattern closely followed that measured in total (fig. 1). The lowest number of taxa in a single pat was 6, (week 24) and the greatest number was 31 (week 12). Negligible numbers of earthworms were found in any of the pats.
The total number of taxa recorded from each batch of ten artificial cow pats (○) and the number of taxa recorded from each individual artificial cow pat (•) placed out in cattle pastures between May and November 2001. Weeks are numbered throughout the season beginning on May 1st (week one).
Aggregation analysis
Twenty-two of the 23 dipteran taxa showed significantly aggregated distributions between pats, the one exception being M. meridiana, the distribution of which was not significantly different from a Poisson model (table 1). Although present throughout the study, M. meridiana only occurred in low abundance within individual pats (fig. 2). For the other dipteran taxa subjected to detailed analysis, the frequency distribution of individuals between pats was not significantly different from a negative binomial model (fig. 2).
The frequency distributions of numbers per pat, of six species or genera of dung-breeding Diptera collected from artificial cow pats placed out in cattle pastures between May and November 2001. a, Sepsis spp.; b, Chloromyia formosa; c, Mesembrina, meridian; d, Hydrotaea spp.; e, Scatophaga stercoraria; f, Musca autumnalis.
The total number of each of ten selected taxa collected from artificial cow pats placed out in cattle pasture in southwest England between May and November 2001 and the index of dispersion (ID). Also, the best fit model which describes the frequency distribution of the numbers of this taxon per pat, the negative binomial index of aggregation, k, and the output values of chi-squared analysis resulting from the statistical comparison of the observed distribution against the distribution fitted.
All of the four larval coleopteran taxa and twenty of the 22 adult coleopteran taxa were significantly aggregated. The distributions of the remaining two taxa, the scarabaeid dung beetles O. coenobita and A. fimetarius, were not significantly different from random, as indicated by the index of dispersion. Of the four taxa selected for more detailed analysis, the distributions of the two genera of coleopteran larvae were not significantly different from a negative binomial model (table 1, fig 3). In contrast, the distributions of the two species of adult beetle, A. fimetarius and C. lateralis, were significantly more over-dispersed than expected from the negative binomial (fig. 3); but the latter distribution was nevertheless a closer description of these data than a Poisson model, both frequency distributions giving low values for the fitted coefficient of aggregation, k (table 1).
The frequency distributions of numbers per pat, of four species or genera of dung-breeding Coleoptera collected from artificial cow pats placed out in cattle pastures between May and November 2001. A, adults of Cercyon lateralis; b, adults of Aphodius fimetarius; c, larvae of Aphodius spp.; d, larvae of Sphaeridium spp.
Modelling the distribution of taxa
Given the observation that the distribution of most taxa between pats can be adequately modelled by a negative binomial distribution, a Monte Carlo simulation was undertaken based on the assumption that colonization is a random binomial event in which each species colonises a pat independently of all other species, the probability of which, P, can be estimated from the proportion of pats seen to be colonised by each taxon in any one week. This analysis shows that if insects colonised pats at random, independently of the presence of other taxa, the observed frequency distribution of taxa should follow an approximately normal distribution (fig. 4). The frequency distribution of observed taxa in weeks one, ten and 20 was similar in shape (fig. 4) and was not significantly different from that expected by chance (Kolmogorov-Smirnov, two-tailed test, K–S=1.26, P=0.09).
The relative frequency of taxa observed in groups of ten artificial cow pats (hatched bars) placed out in cattle pastures in weeks one, ten and 20 between May and November 2001 and the relative frequency expected in the same weeks from a random binomial colonisation model (open bars).
Discussion
In a resource-limited, patchy environment, unless insects develop strategies for avoidance, high levels of competition and predation might be expected to lead to exclusion and low levels of diversity. However, stability and coexistence may be facilitated by aggregation, particularly where colonisers are small relative to resource patches and cannot monopolise a single patch (Atkinson & Shorrocks, Reference Atkinson and Shorrocks1984; Shorrocks, et al., Reference Shorrocks, Rosewell, Edwards and Atkinson1984; Hanski & Cambefort, Reference Hanski and Cambefort1991; Ives, Reference Ives1991; Heard & Remer, Reference Heard and Remer1997; Woodcock et al., Reference Woodcock, Watt and Leather2002). Contagious distributions have been identified in the insects colonising a range of patchy environments, such as fruit (Atkinson & Shorrocks, Reference Atkinson and Shorrocks1984; Rosewell et al., Reference Rosewell, Shorrocks and Edwards1990), mushrooms (Heard, Reference Heard1998) and carrion (Beaver, Reference Beaver1977; Ives, Reference Ives1991).
In the present study, the dung-colonising community identified was characterised by a wide diversity of species that varied as the season progressed, as has been observed in previous studies (Hammer, Reference Hammer1941; Mohr, Reference Mohr1943; Laurence, Reference Laurence1954; Valiela, Reference Valiela1974). The total number of taxa present per batch was greater than 35 for much of the study period, rising to a peak of 43 during early September following which it declined sharply through the autumn to the end of the study. Other studies have found that species richness among dung beetles in southern Ireland was highest during the early summer. Finn et al. (Reference Finn, Gittings and Giller1999) and Hanski (Reference Hanski1980) observed the maximum species diversity of Aphodius species during May and June in Oxfordshire. However, notably here, individual pats contained only approximately half the number of taxa present in the entire batch. The majority of dipteran and coleopteran taxa showed significant aggregation which appeared broadly binomial in distribution; and, based on this observation, a Monte Carlo simulation, which generated an expected distribution frequency if colonization occurred at random, was not different from that observed. This result supports the proposal that colonization and assembly of the dung-inhabiting community observed is a stochastic event.
In a study of Diptera breeding in fallen fruit, Atkinson & Shorrocks (Reference Atkinson and Shorrocks1984) found that the distributions of only two out of 34 species examined were not significantly aggregated. One of these two species was a muscid belonging to the Phaoniinae. The authors suggested that the lack of contagion in this species may be due to cannibalism, or a mechanism to prevent it, since the larvae of Phaoniinae are generally predacious. Notably, of the Diptera considered here, M. meridiana is considered predatory and was the one taxon that was not significantly aggregated.
Since most insect larvae cannot disperse from the dung, their distribution between pats would be expected to be strongly determined by the levels of mortality within the pat and the oviposition behaviour of adults (Beaver, Reference Beaver1977). Females of most insects deposit their eggs in clutches, and clutch size is a fundamental aspect of the life-history of both the adult female and offspring (Godfray & Parker, Reference Godfray and Parker1992; Withers et al., Reference Withers, Madie and Harris1997). Different habitat patch types influence the range of strategies that are successful according to the average patch quality and the distance between individual patches (Beaver, Reference Beaver1977; Hanski, Reference Hanski, Gray, Crawley and Edwards1987; Withers, et al., Reference Withers, Madie and Harris1997; Thiel & Hoffmeister, Reference Thiel and Hoffmeister2004). Accordingly, species breeding in rarer, less predictable but high quality habitat patches should typically deposit their eggs in large clutches, while smaller clutch sizes are expected where patches are more frequent and predictable or of lower quality (Godfray et al., Reference Godfray, Partridge and Harvey1991; Heard, Reference Heard1998). Adult mortality is also an important factor, as it will influence the average number of patches that are encountered in a female's lifetime (Skinner, Reference Skinner1985; Withers et al., Reference Withers, Madie and Harris1997).
Several carrion-breeding blowflies of the genera Lucilia and Calliphora produce clutches of between 200 or more eggs (Ives, Reference Ives1991; Wall, Reference Wall1993), resulting in high levels of aggregation and considerable risk of exceeding the carrying-capacity of a carcass, often leading to severe competition (Holter, Reference Holter1979; Ives, Reference Ives1991). Most dung-breeding Diptera oviposit in much smaller clutches than carrion-breeding species, rarely exceeding 70 eggs and often considerably fewer (Wall et al., Reference Wall, Anderson and Lee2008). Dung is a more predictable and abundant resource than carrion, and so dung-breeding species can expect to encounter a greater number of available oviposition sites in their lifetime. The abundance of predatory flies, such as S. stercoraria, around fresh cow pats may also influence the clutch size of other dung-breeding Diptera, since an increased risk of predation during oviposition favours reduced clutch sizes (Skinner, Reference Skinner1985). However, such comparisons must be treated with caution because of the variety of other ecological variables that differ between the two groups, many of which may contribute to differences in clutch size (Godfray et al., Reference Godfray, Partridge and Harvey1991).
Other causes of aggregation may also exist. For example, Hammer (Reference Hammer1941) found that H. irritans adults aggregate on certain cows, so that their visits to individual pats may also be non-random and Blanckenhorn et al. (Reference Blanckenhorn, Morf and Reuter1999) showed that S. stercoraria adults were regularly distributed between pats, while Sepsis cynipsea were not. Hence, there may be a wide variety of additional factors, associated with mating behaviour, predation, or the density-dependent distribution of clutches, which determine the degree of aggregation observed and the resultant community structure. Little is known about whether patch sampling is employed by some colonizing species to assess local insect abundance prior to oviposition, or whether patch density (as opposed to pat abundance sensu stricto) will influence aggregation.
Overall, the results of this study demonstrate the aggregation of individuals within resource patches and the stochastic colonization and assemblage of the dung-inhabiting community. In terms of practical implications, the highly aggregated distributions highlight the possibility that the populations of even highly abundant insects will be more susceptible to the deleterious effects of insecticidal residues in dung than if they were evenly distributed, if by chance they colonize a pat containing chemical residues from a recently treated animal.
Acknowledgements
We would like to thank Mr John Keedwell for permission to use his farm, the University of Bristol for financial support to CL and Dr Eric Morgan for statistical advice.
