The effects of livestock grazing on foliar arthropods associated with bird diet in upland grasslands of Scotland

Authors


Correspondence author: P. Dennis, Institute of Rural Sciences, Aberystwyth University, Llanbadarn Campus, Aberystwyth SY23 3AL, Wales, UK. E-mail: pdd@aber.ac.uk.

Summary

  • 1Upland biotopes have conservation importance for their typical plant and animal species. Recently, the condition of upland habitats has deteriorated with associated declines in many upland birds. Grazing by increased densities of sheep has been implicated in these changes. Studies in lowland agricultural land have shown a link between declines in bird populations and the availability of arthropod prey.
  • 2We studied the effect of three grazing regimes and an ungrazed control, each replicated six times in a total of 24 3·3-ha plots, on the numbers and overall biomass of foliar arthropods in upland grassland in the Southern Highlands.
  • 3The three grazed treatments of the experiment were stocked with the commercial stocking density of sheep, one-third of the commercial stocking density by sheep only and one-third of the commercial stocking density by sheep and cattle.
  • 4Arthropod groups recognized as components of the diet of moorland birds were sampled by motorized suction sampler and sweep net. Arthropod numbers were unaffected by grazing treatments 6 months after grazing commenced. Significant grazing treatment effects on spiders, bugs and beetles were observed in years 2 and 3, with higher numbers in the less-grazed treatments, but no such effect on brachyceran flies, caterpillars and craneflies.
  • 5A residual maximum likelihood (REML) analysis related the numbers of spiders, bugs, beetles, craneflies and caterpillars to either the stocking density of sheep or an interaction of sheep with year. The analysis related bugs and brachyceran flies to an interaction between cattle stocking density and year.
  • 6Estimated total biomass of foliar arthropods increased significantly with decreasing grazing intensity in years 2 and 3 and biomass in the ungrazed treatment was approximately twice that in the commercially grazed treatment. The REML analysis related biomass to the stocking density of sheep and both the stocking density of sheep and of cattle in an interaction with year.
  • 7Synthesis and applications. We demonstrated that the stocking density and inclusion of cattle grazing affects the numbers and biomass of particular foliar arthropods in the uplands of Scotland. Grazing management is important not only for the conservation of arthropods per se but also as food for insectivorous birds of conservation concern.

Introduction

Upland biotopes are of conservation importance for their typical, associated plant arthropods and birds of acid grassland, heath and mire (Usher & Thompson 1988). Heather moorland is a priority habitat for conservation (Thompson, Hester & Usher 1995a), but the conservation status of semi-natural grasslands has not received the same level of attention. This may be because semi-natural grasslands cover large areas of upland Britain (e.g. 1·2 m ha of acid grasslands and 20–23 800 ha calcareous grassland; UK Biodiversity Action Plan (website)) and degraded forms of this habitat often replace former areas of heather moorland (Anderson & Yalden 1981).

Land use in these upland areas has undergone significant changes in the last few decades. Between 1975 and 1999, the number of sheep in upland Britain increased by 25%, while cattle numbers declined by 22% in Scotland during the same period (Fuller & Gough 1999). Hence, the overall increase in stocking densities of livestock in the uplands has coincided with a general shift away from mixed herbivores towards domination by sheep (Sydes & Miller 1988). The increased density of sheep has been implicated in the deterioration of habitat condition, recorded as dramatic changes in vegetation and declines in the abundance and/or breeding effort of bird species (Woods & Cadbury 1987; Ratcliffe 1991; Thompson et al. 1995b; Fuller & Gough 1999; Evans et al. 2006). Hence, there have been calls to reduce sheep numbers to halt the decline of moorland biodiversity (Thompson et al. 1995b).

There is currently little information available on the mechanisms that link grazing management to changes in bird populations in the uplands. However, arthropod food availability is currently under consideration as a key factor (Pearce-Higgins & Grant 2002; Pearce-Higgins & Yalden 2004; Evans, Redpath & Evans 2005a; Buchanan et al. 2006; Evans et al. 2006; Pearce-Higgins & Grant 2006). There has been no experimental work to establish causality in the relationship, although an association between grazing and various taxa of arthropods has been observed (Gardner et al. 1997; Dennis, Young & Gordon 1998; Dennis, Young & Bentley 2001). In lowland agricultural land, studies have also demonstrated a correlative link between declines in bird populations and a reduced abundance of the arthropod prey (Vickery et al. 2001; Benton et al. 2002; Robinson & Sutherland 2002). Different species of grazer at equivalent stocking densities enclosed on the same vegetation have considerably different effects on vegetation structure and botanical composition (Clarke, Welch & Gordon 1995; Dennis et al. 1997; Hester & Baillie 1998) which also affects the abundance of arthropods (Dennis et al. 1997; Dennis et al. 2001). The effects of high stocking densities of particular species of grazers may also culminate in the same consequence for the survival and breeding success of insectivorous birds as observed in the lowlands.

Intensive grazing management reduces the abundance of most invertebrate groups in grasslands (Hutchinson & King 1980), but the direction and magnitude of response of different arthropod groups to grazing varies according to taxa and trophic level (Gibson et al. 1992a; Dennis et al. 1997, 1998; Hartley & Jones 2003). For example, the dung fauna responds to higher stocking densities (Hanski & Cambefort 1991) and shifts from sheep to cattle grazing (Lumaret, Kadiri & Bertrand 1992). Also, the caterpillars of certain thermophylic butterflies depend on higher stocking densities of domestic grazers to maintain a short grass sward and the necessary thermal regime on the host plants for successful development (Thomas 1990). The abundance of many of the arthropod groups associated with grasslands has been related to increased complexity of plant architecture (Gibson, Hambler & Brown 1992b; Dennis et al. 2001) and increases in vegetation height (Morris & Plant 1983; Purvis & Curry 1981; Dennis et al. 1997) and height heterogeneity (Dennis et al. 1998; Vickery et al. 2001) that results from differences in grazing management. Kruess & Tscharntke (2002) found that grazing pressure strongly affected the number of species and abundance of Coleoptera, Hemiptera and parasitic Hymenoptera, even if plant species richness was unaffected. Abandonment or relaxation of grazing quickly results in changes in the abundance and species composition of invertebrates in grasslands (Morris 1971, 1990; Gibson et al. 1992a). For instance, acid grasslands support numerous species of leafhoppers and increased vegetation height increased overall abundance (Waloff & Solomon 1973), while spiders increase in abundance due to both the greater plant architecture for web building and increased prey availability (Coulson 1988).

Here we designed a large-scale experiment to investigate the effects of both changes in stocking density and differences in species of livestock on foliar arthropods. The experiment has already demonstrated that low intensity grazing by a mixture of sheep and cattle can positively affect egg size (Evans et al. 2005b) and breeding abundance (Evans et al. 2006) of a common insectivorous passerine, Anthus pratensis (L.) meadow pipit. The availability of arthropods as food is thought to be an important factor (Vickery et al. 2001), although the mechanisms in upland, seminatural grasslands are unclear. Recent experimental work by Devereux et al. (2004) suggested, in agricultural grasslands, that short swards are a more profitable foraging habitat for some soil and surface invertebrate-feeding birds. Sward density, however, had no effect on the foraging behaviour, intake rates or intake efficiency of common starlings Sturnus vulgaris (Devereux et al. 2006). In the uplands, prey availability is likely to be primarily a function of both food abundance and accessibility and both are affected by grazing. We tested whether the abundance of the main arthropod taxa associated with plant foliage, especially those that feature in the diet of upland birds, are affected by (a) sheep density or (b) species of grazer, namely sheep alone or a mixture of sheep and cattle. We also tested whether the biomass of foliar arthropods changed significantly in response to these grazing regimes.

Materials and methods

study site

This study was carried out at the Glen Finglas estate (56°16′ N, 4°24′ W), located within the Loch Lomond and the Trossachs National Park, Southern Highlands of Scotland. Samples were collected from three different paired replicate blocks composed of eight experimental plots, the three groups of plots separated by about 5 km. Altitude varied between 220 m for the lowest replicate block and 500 m for the highest. The area was unintensified, indigenous, acid grassland and mire. Dominant vegetation types were National Vegetation Classification (NVC) groups (Rodwell 1991, 1992) M23 (Juncus effusus-acutiflorus–Galium palustre rush-pasture), M25 (Molinia caerulea–Potentilla erecta mire), U4 (Festuca ovina–Agrostis capillaris–Galium saxatile grassland) and U5 (Nardus stricta–G. saxatile grassland). Small areas were covered by bracken fern (Pteridium aquilinum, NVC group U20), but there were no trees save a few single Betula pubescens Ehrh. downy birch, Salix aurita L. eared willow and Sorbus aucuparia L. rowan on the lowest-altitude plots. The climate is cool and humid; the nearest official meteorological station at Ardtalnaig, 33 km to the north-east, records the 1971–2000 mean annual daily temperatures as 4·9 °C (minimum) and 11·9 °C (maximum), with a mean annual precipitation of 1, 344 mm [Meteorological Office (website)].

grazing experiment

The experiment was arranged across three locations, each containing two replicate blocks, each block composed of four plots. Each of the 24 plots was 3·3 ha in extent, and was fenced with stock-proof Rylock post-and-wire fencing in winter 2002–03. Plots allocated for cattle in addition to sheep grazing were topped with barbed rather than plain wire. The eight plots within each paired replicate block were arranged adjacent to one another. Grazing treatments were assigned randomly to the four plots in each of the six replicate blocks as follows:

  • 1Commercial density of sheep grazing. Nine ewes per plot (= 2·7 ewes ha−1).
  • 2Low density of sheep grazing. Three ewes per plot (= 0·9 ewe ha−1) which continued the existing management of the grassland.
  • 3Low density of mixed grazing. Two ewes per plot; in addition two cows each with a single, suckling calf were grazed for 4 weeks in autumn, so that the total stocking density as grazing livestock units and hence offtake was the same as that of treatment 2.
  • 4No livestock grazing (control).

The plots were grazed by breeding ewes throughout the year except during lambing, when these were replaced by young ewes until shearing time. Sheep were removed entirely during severe winter weather or when the ewes were removed for dipping. The grazing treatments commenced in January 2003. Cattle were grazed annually in late summer from 2003, hence the measured responses of vegetation, arthropods and meadow pipits were not affected by the contribution of cattle grazing until the pipit breeding season of 2004 and later years.

arthropod sampling

A total of 25 sampling points were placed at the intersections of a grid composed of squares of 40 m-length sides superimposed upon a map of the experimental site, the precise coordinates were calculated using a geographical information system and located in each plot using a geographical positioning system (GPS) handset. Samples were collected using a ‘D-Vac’ suction sampler (Dietrick 1961) equipped with a circular nozzle of 34·3 cm diameter. Each sample consisted of five pooled subsamples, each with 45 s suction time. The total ground area sampled was thus 0·462 m2. In the treatment years 2003–05, sweep net samples were also collected. A transect of 20 m length and 0·5 m width was orientated along the contour of the hill and centred on each of the 40 sampling points in each plot used for suction sampling but offset to the magnetic north by 2 m to avoid interference between sampling methods. Sweep netting was not undertaken in wet and windy conditions. The transect was swept vigorously side to side while walking steadily along the length of each transect using a net with a pentagonal frame, side length 20 cm. Samples were transferred to plastic bags and frozen, and transferred later to 75% ethanol for storage.

The samples were collected in rounds of five randomly chosen sample points per plot on each visit, to avoid bias due to different blocks being sampled at different times. Samples were collected 13 June–11 July 2002 (baseline, pretreatment set of data), 25 April–25 June 2003, 20 May–2 July 2004 and 24 May–7 June 2005 (five d-vac samples were collected instead of 25 samples per plot in 2005 compared with previous years). We did not attempt to collect samples if it was raining or the ground was very wet, nor in strong winds. Identification and counting was focused upon the arthropod groups that were both sampled effectively by these methods and prominent in the diet of bird species that breed in the uplands (Buchanan et al. 2006).

Estimating arthropod abundance and diversity based on a single sampling method is problematical (Standen 2000). In particular, suction samples do not sample large arthropods efficiently. Mobile groups such as flies tend to fly away from the noisy sampling apparatus and large spiders and beetles tend to cling too tightly to the vegetation to be dislodged by the suction. Sunderland & Topping (1995) found that a D-Vac suction sampler collected 33–97% of all spiders in cereal fields, but their samples did not include large-sized species. Sweep netting was used to augment suction samples because this method sampled larger arthropod species more efficiently, especially spiders, craneflies and moth caterpillars.

The larvae of craneflies were recorded as a major prey of upland birds (Buchanan et al. 2006), and soil sampling of these highly aggregated larvae proved time-consuming and difficult. Instead, and with information from soil samples taken in 2002–04 that Tipula pagana was the dominant cranefly species at Glen Finglas, a further 600 sweep net samples were collected using an identical procedure to that described above in November 2005, when adults were emerging from the soil. Females are flightless and hence this indicated the distribution of larvae in the spring 2005 bird breeding season and suggested where oviposition would take place and provide further cranefly aggregations in 2006.

biomass estimates

An estimate of mean biomass (wet weight) per specimen in the different taxonomic groups was obtained by weighing the specimens representing each group, taken from 10 randomly chosen suction samples collected in 2004. The specimens were wrapped in paper towels to dry off excess alcohol but not dried further before weighing.

statistical analyses

The abundance of key arthropod groups (Araneae (spiders), Hemiptera (true bugs), Coleoptera (beetles) and Diptera: Brachycera (flies), Diptera: Tipuloidea (craneflies) and Lepidoptera (moth caterpillars and adults) from each experimental plot were averaged for each year, 2002–05. No sweep net samples were collected in the pretreatment period of sampling, so analyses of craneflies and moth caterpillars are for 2003–05. Prior to the analysis of the arthropod biomass data over the 4 years, caterpillars were excluded from the 2003–05 data sets as these were not recorded in 2002. Univariate analyses were carried out using residual maximum likelihood (REML; Elston 1998; GenStat 9·1©; Payne 2006). The REML analysis consisted of a general test of treatment effects with a fixed model of treatment + year + treatment × year.

A more detailed investigation of specific contrasts was calculated with REML to allow for the time lags in treatment implementation: i.e. the arthropods could not respond to cattle grazing until 2004. Covariates were introduced as single degree of freedom (d.f.) contrasts for year (linear, quadratic and cubic), linear and quadratic terms for sheep density and a linear term for cattle grazing days in the previous autumn. Categorical fixed effects were also included for treatment and treatment × year to identify the generally small amounts of remaining variation due to these terms unaccounted for by the contrasts. For consistency, the random model was standardized across all analyses and consisted of the terms block + plot + block × year + plot × year. The plot × year term identified individual data elements, and was declared in a fashion that incorporated a lag 1 autoregressive term within plots, to allow for the repeated measures aspect of the data, and year-specific residual variances to allow for heteroscedasticity between years. Significance tests involved comparing scaled Wald statistics (divided by numerator d.f.) against F statistics with 15 residual degrees of freedom, a conservative value taken from the analysis of variance table constructed from the experimental design (Elston 1998).

The abundance data of T. pagana obtained from additional sweep net assessment of craneflies in November 2005 were analysed using a one-way analysis of variance with grazing treatment as the factor applied to log (n + 1)-transformed data (GenStat 9·1©; Lawes Agricultural Trust 2006).

Results

The statistical analysis is based on a total of 247 480 specimens collected over the 4 years of the study. The total number of specimens of each target taxonomic group collected each year is presented in Table 1. The analysis proceeded to compare the abundance of spiders, true bugs, beetles and brachyceran flies collected by suction sampler, 2002–05 and moth caterpillars and cranefly adults collected by sweep net sampling, 2003–05.

Table 1.  Total number of specimens of arthropod groups that feature in moorland bird diets collected each year 2002–05 from suction (Araneae, Hemiptera, Coleoptera and Diptera: Brachycera) and sweep net samples (Diptera: Tipuloidea and Lepidoptera)
Method/yearAraneaeHemipteraColeopteraDiptera: BrachyceraDiptera: TipuloideaLepidoptera
  • *

    2005 figures for suction samples have been multiplied by five because five instead of 25 samples per plot were collected in 2005 compared with 2002–04. NC, not counted.

Suction
 200217 07211 92747585240   702Few, NC
 200317 28228 52653103749 4 013 113
 200424 92070 86249896162 6 552 305
 2005*13 46039 0852960 76511 090  30
Sweep net
 2003   896  3041 2423211 1 811 753
 2004 2 25810 902 6165721 4 2731312
 2005NCNCNCNC   2501465
Sweep net – November 2005NCNCNCNC 2 279NC

A similar mean abundance was recorded in the pretreatment sampling of 2002 within the areas demarcated for the imposition of the four experimental treatments for spiders, true bugs, beetles and brachyceran flies (Fig. 1, Table 2a). No significant effect on abundance of these taxa was observed within 6 months (2003), but a significant effect of grazing treatment was detected using the REML linear mixed-model treatment model both at 18 months and 30 months for all but the brachycerans (Fig. 1, Table 2a). In 2004, the mean abundance of spiders was significantly higher in the ungrazed and low stocking density treatment composed of sheep and cattle compared with the commercial stocking density treatment (Fig. 1a, Table 2a). The abundance of spiders in the low stocking density treatment composed of only sheep was intermediate between these two groupings but not significantly different from either. In 2005, overall foliar arthropod numbers were lower, due possibly to poor weather and spider abundance, and were significantly higher only in the ungrazed compared with all grazed treatments (Fig. 1a, Table 2a). The experimental treatments exerted a very similar effect on the abundance of true bugs and beetles (Fig. 1b,c, Table 2a). In 2004 there was a significant difference between the lower abundance of true bugs and of beetles recorded in the commercial stocking density and the low stocking density composed of sheep-only treatments and the higher abundance of the ungrazed and low stocking density composed of sheep and cattle treatments. In 2005, true bugs and beetles had the same significantly higher abundance in the ungrazed compared with all grazed treatments, as described earlier for the spiders. There was no significant effect of experimental treatments on the abundance of brachyceran flies, although significant changes in abundance were observed between years (Fig. 1d, Table 2a).

Figure 1.

Mean ± 1 SD number of (a) spiders, (b) true bugs, (c) beetles and (d) robust flies sampled in each treatment for 4 years by motorized suction sampler (pretreatment year 2002). Experimental treatments: (1), sheep at 2·7 ewes ha–1; (2), sheep at 0·9 ewe ha–1, (3), sheep and cattle equivalent to 0·9 ha–1, and (4), ungrazed.

Table 2.  Wald statistics (divided by numerator degrees of freedom) generated from the residual maximum likelihood (REML), linear mixed model output for major arthropod taxonomic groups in (a) suction samples, 2002–05; (b) summer sweep net samples, 2003–05 and (c) the biomass of arthropods (n.d.f., d.d.f.: numerator and denominator degrees of freedom, respectively, Wald/n.d.f. and probability values, NS = P = 0·05)
 n.d.f., d.d.f.Wald/n.d.f.Significance, P
(a) Arthropod group
 Araneae
  Treatment3, 15 14·45< 0·001
  Year3, 15 11·45< 0·001
  Treatment × year9, 15  8·54< 0·001
 Hemiptera
  Treatment3, 15  2·92< 0·05
  Year3, 15 60·88< 0·001
  Treatment × year9, 15  3·39< 0·001
 Coleoptera
  Treatment3, 15562< 0·001
  Year3, 15 23·83< 0·001
  Treatment × year9, 15  2·01< 0·05
 Diptera: Brachycera
  Treatment3, 15  1·65NS
  Year3, 15 54·09< 0·001
  Treatment × year9, 15  1·62NS
(b) Arthropod group
 Diptera: Tipuloidea
  Treatment3, 15  1·83NS
  Year2, 15  5·01< 0·01
  Treatment × year6, 15  1·74NS
 Lepidoptera
  Treatment3, 15  1·64NS
  Year2, 15  1·85NS
  Treatment × year6, 15  1·38NS
(c) Arthropod biomass (g)
  Treatment3, 15  5·39= 0·001
  Year3, 15 57·08< 0·001
  Treatment × year9, 15  4·98< 0·001

Further analysis with the REML linear mixed-model specific contrasts model was used to determine effects of individual factors of grazing by sheep or cattle and possible interactions with time since the establishment of the experiment. Spiders, true bugs and beetles were significantly related to sheep stocking density and the interaction between sheep stocking density and year (Table 3). Neither true bugs, spiders nor beetles related significantly to cattle stocking density but true bugs were related to the interaction of cattle with year (Table 3). Despite no overall treatment effect (REML linear mixed-model treatment model account, above), brachyceran flies were associated with the linear derivative of cattle stocking density × year but not to sheep stocking density alone or interacting with year (Table 3).

Table 3.  Additional selected arthropod responses to linear and quadratic derivatives of sheep or cattle stocking density, year and their interactions within the Glen Finglas grazing experiment, 2002–05, calculated using the REML linear mixed model, specific contrasts model
Arthropod groupFactorModelWald/n.d.f. statisticSignificance, P
  1. Sheep: sheep stocking density; cattle: cattle stocking density; year: duration of experiment. Numerator and denominator degrees of freedom were 1, 15, respectively, for all arthropod groups.

SpidersSheepLinear32·47< 0·001
SheepQuadratic10·51< 0·001
Sheep × yearLinear40·59< 0·001
CattleLinear 1·64NS
True bugsSheepLinear 7·23< 0·01
Sheep × yearLinear 9·12< 0·01
CattleLinear 0·71NS
Cattle × yearLinear 8·75< 0·01
BeetlesSheepLinear 8·17< 0·01
SheepQuadratic 6·16< 0·05
Sheep × yearLinear 4·93< 0·05
CattleLinear 0·19NS
Brachyceran fliesSheepLinear 0·53NS
Sheep × yearLinear 0NS
Cattle × yearLinear 4·78< 0·05
CranefliesSheepLinear 2·07NS
Sheep × yearLinear × quadratic 5·18< 0·05
Cattle × yearLinear 1·33NS
Moth caterpillarsSheepLinear 3·61NS
Sheep × yearLinear 5·70< 0·05
CattleLinear 0NS
Arthropod massSheepLinear12·78< 0·001
Sheep × yearLinear18·38< 0·001
Cattle × yearLinear 8·87< 0·01

The REML linear mixed-model treatment model applied to the abundance of the cranefly adults and moth caterpillars collected by sweep net showed no significant effect of grazing treatment at 6 months to 30 months after the start of the experiment (2003–05), although significant changes in abundance were observed for both groups between years (Fig. 2a,b, Table 2b). Cranefly adult numbers and moth caterpillars were related significantly to sheep stocking density interacting with year but not sheep stocking density alone or an interaction of cattle stocking density × year in the REML linear mixed-model specific contrasts model (Table 3). Caterpillar numbers were inversely related to stocking density (Fig. 2a). Interestingly, the abundance of T. pagana in autumn sweep net survey suggested a slight depression in numbers with stocking density but again, this did not translate to a statistically significant effect of grazing treatment (Fig. 3, F3,20 = 0·67).

Figure 2.

Mean ± 1 SD number of (a) moth caterpillars and (b) craneflies in sweep net samples in each of four grazing treatments 2003–05. Caterpillars were not counted in 2002 but numbers were very low. Experimental treatments: (1), sheep at 2·7 ewes ha–1; (2), sheep at 0·9 ewe ha–1, (3), sheep and cattle equivalent to 0·9 ha–1, and (4), ungrazed.

Figure 3.

Mean ± 1 SD estimated biomass of arthropods (wet weight) per sample, divided by treatment (treatment-to-be in 2002) and year. Experimental treatments: (1), sheep at 2·7 ewes ha–1; (2), sheep at 0·9 ewe ha–1, (3), sheep and cattle equivalent to 0·9 ha–1, and (4), ungrazed.

Arthropod biomass diverged from the pretreatment values after 18 and 30 months (Fig. 3, Table 2c). Arthropod mass was significantly related to sheep stocking density, an interaction between sheep stocking density and year and an interaction between cattle stocking density and year (Table 3). Arthropod biomass was significantly higher in the ungrazed and low stocking density treatment composed of sheep and cattle compared with the commercial stocking density treatment and low stocking density treatment composed of only sheep (Fig. 3, Table 2c). The arthropod mass available per unit area in the ungrazed and low stocking density treatment composed of sheep and cattle was c. double that recorded in the commercial sheep grazed treatment (Fig. 3). The general reduction in biomass from 2004 to 2005 represented annual variability of 38·3%, possibly the influence of annual weather conditions; this annual variability was of a similar magnitude to within-year variability between grazing treatments of greatest contrast, namely 45·7% in 2004 and 37·6% in 2005 (Fig. 3, Table 2c).

Discussion

There is increasing evidence of the contribution of particular arthropod groups to the diet of upland birds (Pearce-Higgins & Yalden 2004; Buchanan et al. 2006) and growing concern that the steady decline in populations of upland breeding birds may relate to food supply and in turn to recent grazing practice (Pearce-Higgins & Grant 2002, 2006). Despite decades of increasing stocking densities of grazing livestock with a trend of an increasing proportion of sheep compared with cattle (Fuller & Gough 1999), recent falls in market prices and a shift from livestock production subsidies to environmental stewardship payments has driven a reduction in stocking densities but with less incentive to maintain cattle as part of the grazing system (Evans, Gaskell & Winter 2003).

The Glen Finglas grazing experiment has demonstrated for the first time in upland acid grassland the consequences for the abundance of key arthropod groups of both a reduced stocking density and the effects of providing this with sheep compared with a combination of sheep and cattle. From 18 months after the initial experimental manipulation of grazing regime, sheep stocking density or an interaction of sheep with year were significantly related to the overall abundance of spiders, true bugs, beetles, adult craneflies and moth caterpillars. A significant response specifically to the interaction of cattle stocking density with year was identified for the density of true bugs, brachyceran flies and overall arthropod biomass. Further years of the experiment will confirm the true extent of the effect of grazing species rather than gross stocking density on a broader group of arthropods. Spiders, beetles, flies, true bugs and moths represent five of the seven groups that have been recognized as important in the diet of upland breeding birds of conservation concern (Buchanan et al. 2006).

Our results indicate considerable annual variability in arthropod abundance for any given treatment that was of a similar order to the differences observed between contrasting treatments by 2004–05. Similarly, biomass of foliar arthropods in the lower stocking density treatment grazed by sheep and cattle and the ungrazed treatment was nearly twice that recorded in the commercially grazed treatment in 2004. The percentage reduction in arthropod biomass between years was close to the observed increased biomass in the ungrazed and low stocking density treatments grazed with sheep and cattle compared with the commercial stocking density of sheep. Two explanations may account for the observed annual variability. Relatively cooler and wetter weather during the breeding season may affect the productivity, especially of foliar arthropods, although the experiment will need to run for six or more years to formally analyse weather as a factor. If so, the potential benefits of reducing stocking densities, and the grazing of cattle in addition to sheep, may be diminished in years experiencing poor weather conditions. Alternatively, vegetation structure and arthropod numbers and biomass may respond to the period of grazing by cattle. There may be an optimal period beyond which annual cattle grazing diminishes habitat quality for arthropods. Again, further years of study are required to test whether ‘pulse’ grazing might be a better strategy for managing wet grassland and mire. If the potential benefits that arise from the reduced stocking densities and retention of cattle grazing observed thus far in this experiment are dampened by poor weather, this may become an increasing concern if such weather conditions are more likely to prevail in western and northern parts of Britain with the progress of climate change.

Gibson et al. (1992b) reported that the succession of spiders on ex-arable grasslands where cultivation had ceased was still ongoing seven years after cessation of grazing. Spiders and particularly Linyphiidae, are particularly sensitive to the effects of grazing management because they are so dependent on vegetation structure (e.g. Dennis et al. 1998). Morris (1990) reported that some arthropod species increased strongly a short time after cessation or reduction of grazing, only to collapse later on when the system approached a new equilibrium. The longer term response of arthropod abundance to the reduction in grazing ‘intensity’ may be limited by a decline in the quality of the host plants consumed by herbivorous insects, e.g. sucked by plant bugs. The increased proportion of plant litter and smaller inputs of faeces and urine may alter the productivity and nutrient value plants (McNaughton et al. 1997) and hence the leaves or phloem consumed by such insect herbivores (Waloff 1980). Cranefly abundance will decline in response to increased vegetation height because females cannot access the lower stem – soil interface to oviposit (McCracken & Tallowin 2004). A decrease in the contribution of leatherjackets to available arthropod biomass should be considered a fundamental qualitative difference in prey availability, irrespective of an observed overall increase in arthropod biomass. Craneflies appear to be essential for several rare, especially wading bird species in the uplands (Buchanan et al. 2006).

Sheep, especially at low stocking density tend to restrict their grazing to a few vegetation patches composed of preferred plant species hence less favoured patches are effectively ungrazed in typically heterogeneous upland vegetation (Hester & Baillie 1998). Several previous experiments on other types of upland vegetation corroborate the general changes in arthropod abundance associated with lower stocking densities of livestock (Dennis et al. 1998) although rarely dealing with the broad range of taxa associated with the avian food chain (Buchanan et al. 2006). Similar responses of arthropods have been demonstrated in calcareous grasslands (Morris 1971, 1990; Morris & Plant 1983), ex-arable land (Gibson et al. 1992a) and Nardus grasslands (Dennis et al. 1997). The apparent benefits to arthropod abundance albeit for selected groups of arthropods, may also tail off with time after changes to grazing due to the differences in the quality of vegetation. The ease of accessibility of foraging birds to available arthropod prey is also strongly associated with the vegetation structure and the favourable effect on parameters of meadow pipit breeding success in the mixed livestock grazed treatment (e.g. in 2005, 5 ± 0·67 se. breeding pairs plot−1 cf. 3·33 ± 0·42 se. in the intensively grazed treatment, Evans et al. 2006) may indicate that optimal habitat represents a trade-off between food abundance and foraging suitability.

These results are significant in terms of understanding how increases in livestock grazing might lead to decreases in some upland bird populations. Here, we demonstrated that the potential abundance of food for birds was affected by livestock grazing, with arthropod biomass lowest in intensively grazed areas during all three years of the experiment. Livestock grazing alters habitat structure and thus the suitability of the sward for nesting and feeding birds (Vickery et al. 2001). Habitats with a complex vegetation structure, as a result of herbivore grazing, can support a higher diversity of bird species (Martin & Possingham 2005). The foraging behaviour of many grassland birds is influenced heavily by sward height, which modifies prey availability, and hence grazed areas are often preferred by invertebrate feeders (Vickery et al. 2001). Too much or too little grazing, on the other hand, might lead to reduced food availability and thus be detrimental to birds (Evans et al. 2005b). Vegetation structure is also likely to influence foraging site selection, because this can affect predation risk and foraging efficiency (Whittingham & Evans 2004; Butler et al. 2005), although more specific research on avian foraging behaviour in seminatural upland habitats is needed.

This experiment has generated quantitative evidence of the contribution of cattle to the maintenance of structural diversity and arthropod abundance in grazed ecosystems. The reason is that cattle graze more generally and are less selective compared with sheep (Gordon & Illius 1988; Hodgson et al. 1991) and hence do not contract the area that they graze to a few vegetation patches to the same extent as sheep (Grant et al. 1985, 1996). Further work will focus upon the detailed response of specific arthropod taxa, on one hand, and an analysis of vegetation and arthropod characteristics associated with the apparent increased breeding success of meadow pipits on the other hand.

Acknowledgements

This study formed part of the Grazing and Upland Birds project funded by the Scottish Executive Environment and Rural Affairs Department (SEERAD). Timothy Conner, Sally Burgess, Ian Joyce and Elaine McEwan assisted with the collection and sorting of samples. We thank The Woodland Trust, Scotland for permission to use the Glen Finglas Estate. Thanks also to David Elston, BioSS for input on the experimental design, writing the Genstat procedure for the REML variance components analysis and for providing general statistical advice, and to Göran Högstedt, University of Bergen, for comments on an earlier version of the manuscript.

Ancillary