Collection and Identification of Focal Species
To establish experimental treatments, live dung beetles were collected by hand and using dung-baited pitfall trapping in sites in southern and western UK during April–June 2010. Dung beetles were identified using Jessop (1986) and classified into three functional groups: (A) soil-ovipositing endocoprids, (B) dung-ovipositing endocoprids and (C) paracoprids, following Doube (1990) and based on life-history traits from Landin (1961) and Finn & Gittings (2003). Adult beetles feed on the liquid fraction of the dung, while larvae feed on solid particles. Adult dung-ovipositing endocoprid beetles tunnel within the dung pat and eggs are laid within the pat. Soil-ovipositing endocoprid adults also feed within the pat but their eggs are laid in the underlying or surrounding soil and larvae move to the pat. Paracoprid species dig brood chambers in the soil below the pat and provision them with dung. Eggs are laid and larvae develop within these brood chambers.
Beetles were housed in mixed-sex containers in a well-lit, ventilated shed until the start of the experiment. They were fed with fresh cattle dung from grass-fed animals that had not been treated with anthelmintics for >6 months. Species included in the experiments were randomly chosen from those collected in sufficient numbers.
Dung for the mesocosm experiments was obtained from a group of 12 mature, organic Welsh Black cattle at Penweathers Farm, St. Davids, Wales (51°53′7·5″, 5°15′48·9″). The cattle had been wintered outdoors and had not been treated with anthelmintics for at least 2 years. Dung quality can vary depending on the pastures grazed by cattle (Beynon et al. 2012). To avoid confounding pasture-to-pasture differences with anthelmintic treatment, dung from the same group of cattle grazing the same pasture, but collected pre- and post-treatment, was used in the mesocosm experiments. Pre-treatment fresh dung (<6 h old) was collected on the day before treatment and on the day of treatment (10 May 2010). Cattle were then treated with an ivermectin pour-on (Animec™) at the recommended dose of 10 mL kg−1 live mass. Post-treatment dung was collected one and 2 days after treatment, corresponding to peak faecal excretion of pour-on ivermectin (Sommer & Steffansen 1993; Herd, Sams & Ashcraft 1996). One- and 2-day old dung was combined 1 : 1 by mass. Dung was frozen to −20 °C until required, when it was thawed at ambient temperatures. Pre-treatment and post-treatment dung samples were separately homogenized using a mortar mixer.
Experimental mesocosms were constructed using 14-L, 35-cm diameter cylindrical buckets. Three 2-cm-diameter drainage holes (covered with 0·7-mm plastic mesh) were drilled in the base and a 2-cm-diameter hole for fitting an emergence trap was drilled at a height of 16 cm and sealed with tape. Each mesocosm was filled to a depth of c. 13 cm with hand-compressed sandy loam topsoil and topped with a disc of Lolium perenne Linnaeus turf (c. 3 cm thick), which was free from macro-invertebrates (S. Beynon, unpublished data). Mesocosms were allocated to random positions on an 8 × 7 m rectangular grid, spaced at 1 m (Fig. S1, Supporting Information) in an unimproved meadow at Lower Moor, St Davids, Wales (51°52′30·5″N, 5°16′38·5″W).
The effects of varying species richness within functional groups on rates of dung decomposition were explored in synthetically assembled dung beetle assemblages. The experiment included one, two or three species per functional group per replicate while holding total dung beetle biomass constant (Table 1). Mesocosm beetle biomass (156·6 mg dry mass) was set by the monoculture biomass (n = 6) of the heaviest species, Aphodius fossor Linnaeus (26·10 mg dry mass). The biomass of beetles used in the experiment per unit dung mass was equivalent to that observed under natural field conditions in the UK (S. Beynon, pers. obs.). Where fewer than five individuals per species were allocated to a mesocosm, beetles were sexed to ensure both females and males were represented. While large paracoprids (e.g. Geotrupidae) can contribute disproportionately to dung removal (Slade et al. 2007; Rosenlew & Roslin 2008), we excluded them from the experiment. Biomass discrepancies made the inclusion of geotrupid species impractical, for example, a single Geotrupes spiniger is more than 88 times the mass of one Aphodius granarius (Gittings & Giller 1997; Rosenlew & Roslin 2008). Thus, restricting the study to smaller species allowed us to investigate more subtle biodiversity–ecosystem functioning relationships, while retaining realistic assemblages of beetles in our treatments.
Table 1. Experimental beetle species
|Species||Functional group||Individual dry mass (mg)||Species richness experimental treatments|
|Aphodius ater De Geer||Dung-ovipositing endocoprid||5·43a||29||14||10|
|Aphodius granarius Linnaeus||Dung-ovipositing endocoprid||2·98a||53||0||18|
|Aphodius pedellus De Geer||Dung-ovipositing endocoprid||9·40b||17||8||6|
|Aphodius depressus Kugelann||Soil-ovipositing endocoprid||9·10, n = 21c||17||9||6|
|Aphodius fossor Linnaeus||Soil-ovipositing endocoprid||26·10b||6||3||2|
|Aphodius luridus Fabricius||Soil-ovipositing endocoprid||15·69, n = 11c||10||0||3|
|Aphodius erraticus Linnaeus||Paracoprid||9·01a||17||0||6|
|Onthophagus joannae Goljan||Paracoprid||9·47, n = 10c||17||8||6|
|Onthophagus similis Scriba||Paracoprid||13·87, n = 11c||11||6||4|
As there were too many permutations to study all possible species combinations, species identities for the one- and two-species mesocosms were drawn at random from the pool of available species so that a randomly selected one-, two- and three-species combination was included for each functional group (Table 2). Each mixture was replicated three times. This design was repeated using either ivermectin-treated or untreated dung, giving a total of 54 mesocosms containing 1140 beetles. While this design leaves many species combinations unstudied, it allows us to quantify the role of species richness, independent of functional group richness. Five beetle-free mesocosms per anthelmintic treatment (ivermectin or control) were also included to check for direct effects of dung treatment on decomposition.
Table 2. Mesocosm species identities
|Functional group||Three-species mesocosm||Two-species mesocosm||Monoculture mesocosm|
|Dung-ovipositing endocoprid|| Aphodius ater, Aphodius granarius, Aphodius pedellus || Aphodius ater & Aphodius pedellus || Aphodius granarius |
|Soil-ovipositing endocoprid|| Aphodius depressus, Aphodius fossor, Aphodius luridus || Aphodius depressus & Aphodius fossor || Aphodius depressus |
|Paracoprid|| Aphodius erraticus, Onthophagus joannae, Onthophagus similis || Onthophagus joannae & Onthophagus similis || Onthophagus similis |
To assess species-specific survival due to ivermectin, a second experiment was established using replicated single-species mesocosms of all species included in the first experiment. Total mesocosm beetle dry biomass was again 156·6 mg. Mesocosms for each species were replicated with either ivermectin or control dung (n = 3 for both treatments). Monoculture mesocosms included in the species richness experiment were also used in the analyses of this experiment. Thus, 36 additional mesocosms containing 576 additional beetles were included, giving a total of 54 mesocosms containing 1062 beetles for analyses.
At the start of both experiments (7 June 2010), a single 600-g, 18-cm-diameter dung pat, made from the collected dung, was placed in each mesocosm, supported by a 25-cm-diameter circle of 2-cm wire mesh. The appropriate number of beetles was then added. Anthelmintic treatments (ivermectin or control) and beetle treatments were randomly allocated to mesocosms. Mesocosms were covered with fine plastic mesh, secured with elastic cord. Grass within the mesocosm was cut every 2 weeks (June–November 2010) to c. 4 cm.
For the first 4 weeks of the experiments, pats were lifted from the mesocosms on the wire mesh and weighed weekly to calculate ‘short-term’ decomposition rates corresponding to the main period of beetle feeding and nesting activity (Landin 1959, 1961). Pats were then re-weighed after 36 weeks, in March 2011 (corresponding to the start of the next grazing season). This final ‘long-term’ measure is relevant from an applied perspective, as any dung remaining at the start of the following grazing season would reduce the area of grass available for grazing and may retard palatable spring grass growth (Bang et al. 2005). Wet masses were used to measure dung decomposition as they allow repeated, non-destructive sampling and provide a reliable measure of decomposition rates (Wall & Strong 1987; Slade et al. 2007).
In both experiments, beetle survival was assessed using emergence traps at 2·5 weeks, when adult beetles naturally start to leave the dung (Landin 1959, 1961; Gittings & Giller 1997). Tape covering the exit holes was removed and clear plastic collecting bottles were connected using a clear plastic tube of 4 cm length and 2 cm diameter. At this time, the fine mesh covering the mesocosm was replaced with light-reducing black material to encourage beetles to move into the collecting tube and bottle. The fine mesh was replaced in November, once all beetles had emerged, and black material again replaced the fine mesh in February 2011, for collection of offspring emerging the following year. Emergence traps were removed, and the experiment terminated in November 2011. A measure of tibial wear (Tyndale-Biscoe 1978) was used to separate the original beetles used to establish the experiment from their offspring. Fitting emergence traps also ensured that adult beetles were not artificially forced to inhabit pats that they would leave under natural conditions.
Two measures of decomposition were analysed separately: decomposition rate over the first 4 weeks of the experiment, and dung wet mass removed at 36 weeks. Log10-transformed dung pat masses over the first 4 weeks were regressed against time for each replicate, and the fitted slope values were used as a measure of decomposition rate. These log-linear regressions consistently provided the best-fitting description of the relationship between pat mass and time, with adjusted R2 values ranging from 0·884 to 0·996. Species richness effects on dung decomposition were assessed by comparing monocultures, two-species polycultures and three-species polycultures of each functional group. Each measure of decomposition was modelled as a function of anthelmintic treatment, functional group and species richness, with all response variables treated as categorical.
The effect of ivermectin on total adult beetle survival per mesocosm was assessed with logistic regression (generalized linear models with binomial errors). Adult survival was taken as the proportion of initial adult beetles emerging from mesocosms within 6 weeks of the start of the experiment. Ivermectin impacts on offspring emergence (the number of emerging offspring) were analysed with generalized linear models (with Poisson errors) following Warton & Hui (2011). We checked for overdispersion in all cases, with quasipoisson or quasibinomial errors used where overdispersion was evident. Species-specific adult survival and offspring emergence in monoculture (where there were essentially 15 independent experiments) were modelled separately with generalized linear models as a function of anthelmintic treatment, which was treated as a categorical variable.
The two measures of performance analysed (adult survival and offspring emergence, which is a function of adult survival and fecundity plus offspring survival) as well as the two measures of dung decomposition (short-term decomposition rate and long-term decomposition) are not completely independent. However, we consider that analysing the suite of responses allows us to document more fully the potential effects of ivermectin on dung beetle assemblages and associated ecosystem services.
Following model criticism, linear models were used in all analyses measuring dung decomposition. All interaction terms were fitted, and model simplification was carried out using Akaike's Information Criterion (AIC) to test the goodness-of-fit of the model. Where changes were non-significant, interactions or variables were dropped until the minimum adequate model was obtained (Crawley 2007). Post hoc treatment contrast coefficients and the Tukey's HSD (Honestly Significant Differences) test were used to explore significant results. Data were analysed using the statistical package r 2.10.0 (R Development Core Team 2006).