Testing biotic indicator taxa: the sensitivity of insectivorous mammals and their prey to the intensification of lowland agriculture


Correspondence author: M. J. O. Pocock, School of Biological Sciences, University of Bristol, Woodland Road, Bristol BS8 1UG, UK. Fax +44 117 3317985; e-mail: michael.pocock@bristol.ac.uk


  • 1Changes to agricultural policy aim to extensify agriculture and increase biodiversity. However, it is not known how sensitive many taxa are to intensification. Sensitive taxa could be used as biotic indicators, to assess changes over time and the effectiveness of policy changes.
  • 2We sampled shrews, bats and their prey (beetles, flies and moths) on matched pairs of sites and assessed the response in their abundance to aspects of intensification: increased agrochemical inputs (using abundance on matched organic and conventional cereal fields as a proxy), the switch from hay to silage (using abundance on matched hay and silage fields) and boundary loss (using abundance in the field and near the boundary as a proxy, in cereal and grass separately). We quantified the abundance-derived sensitivity of the taxa in order to assess their use as biotic indicators.
  • 3There was substantial variation in the sensitivity of taxa to the three aspects of intensification. Most estimates (51%) of sensitivity to boundary loss were significant, but only 8% for increased agrochemical inputs and 16% for the switch from hay to silage. Sensitivity to one aspect of intensification was not significantly related to sensitivity to another.
  • 4Insectivorous mammals were relatively insensitive to increased agrochemical inputs and the switch from hay to silage, but strongly sensitive to boundary loss.
  • 5Taxa with significant sensitivity to increased agrochemical inputs included some Carabidae and Diptera. We found fewer significant differences in abundance between organic and conventional farms than previous workers, probably because we controlled for boundary characteristics, suggesting that the quality of field boundaries is important in influencing biodiversity differences between organic and conventional farms.
  • 6The switch from hay to silage had a positive effect on some Coleoptera and Diptera but a substantial negative effect on Hepialidae (Lepidoptera).
  • 7Synthesis and applications. The recorded sensitivity of taxa to changes in agricultural practices is highly variable. Therefore, the selection of biotic indicator taxa of agricultural intensification is not straightforward. If surveys of biotic indicator taxa are used to assess the effectiveness of changes in agricultural practice, empirical evidence should be used to select suitable taxa.


Since the mid-20th century in Europe and North America, agricultural intensification has had profound, detrimental impacts on biodiversity (Aebischer 1991; Pimentel et al. 1992; Krebs et al. 1999; Robinson & Sutherland 2002). Agricultural intensification is the process of increasing productivity through increased inputs, and even species dependent on farmland may decline as agriculture intensifies (Aebischer 1991; Krebs et al. 1999; Robinson & Sutherland 2002). However, the mechanisms by which intensification of farming affects biodiversity are diverse (Krebs et al. 1999; Vickery et al. 2001; Clough, Kruess & Tscharntke 2007). Trends associated with agricultural intensification in lowland farms in the European Union (EU) include: more intensive management of arable and grass fields (e.g. increased cutting and applications of inorganic fertilizer and pesticides), loss and degradation of non-farmed land (e.g. woodland, ponds and field boundaries) and reduced landscape heterogeneity at various spatial scales (Chamberlain et al. 2000; Robinson & Sutherland 2002; Hole et al. 2005). These trends interact in their effects on biodiversity (Chamberlain et al. 2000; Hole et al. 2005). The three key aspects of intensification that we considered were: increased agrochemical inputs in cereal fields, the switch in grass production from hay to silage, and the loss of field boundaries (in cereal and grass sites).

agrochemical inputs in cereal fields

Agrochemicals are an increasingly important component of modern agriculture (Chamberlain et al. 2000). However, the uptake of organic farming, in which the use of chemical pesticides and inorganic fertilizer is prohibited (Lampkin 1999), is increasing (Anonymous 2005). Organic farming provides many biodiversity benefits through the increased abundance and species richness of plants and animals (Bengtsson, Ahnström & Weibull 2005; Fuller et al. 2005; Hole et al. 2005). Several researchers have compared organic and conventional farms and showed that organic farms differ not only in agrochemical use but also have smaller fields, more non-crop habitats, more frequent crop rotations and larger, wider and more sympathetically managed hedges than conventional farms (Bengtsson, Ahnström & Weibull 2005; Fuller et al. 2005). Such habitat differences may explain differences in biodiversity (Krebs et al. 1999; Robinson & Sutherland 2002). However, if these differences are recognized and taken into account (through the experimental design and/or statistically), comparisons between organic and conventional farms may be useful research tools for understanding the effects of increased agrochemical inputs.

switch from hay to silage

In lowland grass fields in the UK, the ratio of hay : silage land use has decreased from 4 : 1 in 1970 to 1 : 4 in 1990 (Chamberlain et al. 2000; Vickery et al. 2001). In hay production, one late cut of grass is followed by aftermath grazing and sometimes organic fertilizer. In hay fields there are few inputs and there is no resowing; they are consequently relatively species-rich. Silage, however, is a high-yielding intensive crop, with high inputs of organic and inorganic fertilizer on improved pasture or recently sown leys, from which one or more cuts are taken each year (Chamberlain et al. 2000). Moderate applications of organic fertilizer may increase the abundance of soil invertebrates, but inorganic fertilizer and the early first cut of silage have a negative effect on invertebrate abundance and consequently may affect insectivorous animals (Vickery et al. 2001).

loss of field boundaries

Field boundaries are important for farmland biodiversity. Hedgerows are primary habitats for many species, while others use them as refuges, for example mammals for shelter and some Carabidae for overwintering (Pimentel et al. 1992). Hedgerows in the UK were removed at a consistent rate in 1960–93 to increase farming efficiency (Robinson & Sutherland 2002), although the rate of hedgerow loss has decreased more recently (Macdonald & Johnson 2000).

biotic indicators of agricultural change

In the last few decades, agri-environment schemes (AES) have been introduced with the intent of ameliorating the negative effects of agricultural intensification on biodiversity. In the EU, AES currently cost 1·6 billion euros year−1 (Donald & Evans 2006). Such broad, shallow approaches may not benefit uncommon species (Kleijn et al. 2006) but do bring biodiversity benefits (Donald & Evans 2006; Potts et al. 2006). However, it is important that the effectiveness of these policies is monitored.

The use of biotic indicator taxa has been proposed to evaluate the effects of large-scale changes, such as the widespread implementation of AES, on biodiversity (Büchs 2003). Indicator taxa are sensitive to the stresses that affect biodiversity and so can be used to assess habitat quality (Landres, Verner & Thomas 1988). Monitoring a suite of indicator taxa could be a cost-efficient method to provide a rapid assessment of the effectiveness of large-scale AES. However, indicators should have proven sensitivity to changes associated with agricultural intensification.

In this study, we tested the sensitivity of taxa to three aspects of intensification with a field-scale matched-pairs approach. Farmland birds and their food sources have been studied intensively (Chamberlain et al. 2000; Vickery et al. 2001; Benton et al. 2002) so we selected complementary functional and taxonomic groups: insectivorous mammals (shrews and bats) and their documented prey, Carabidae and Staphylinidae beetles active on the ground (prey of shrews; Churchfield 1990) and night-flying Coleoptera, Diptera and moths (prey of bats; Vaughan 1997). We used measures of abundance to quantify the sensitivity of the taxa to increased agrochemical inputs in cereal crops, to the switch from hay to silage, and to boundary loss. Our aim was to identify taxa that, with our sampling approach, are sensitive to these particular aspects of intensification, which could therefore serve as biotic indicators of agricultural habitat change.


field sites

We used a matched pair design to test and quantify the sensitivity to three aspects of agricultural intensification of shrews [Soricidae: Sorex spp. and Neomys fodiens (Pennant)], bats (Chiroptera) and their invertebrate prey (beetles, Diptera and moths). In a ‘natural experiment’, we made use of existing differences in land use to collect data on sensitivity (derived from measures of abundance) to habitat change. Field sites were within 60 km of Bristol, UK, within the geographical range of all species of shrew and bat in mainland Britain. To quantify sensitivity to increased agrochemical inputs, we paired 21 organically farmed cereal fields with nearby conventionally farmed fields in 2003. To quantify sensitivity to the switch from hay to silage, we paired 22 hay fields with nearby silage fields in 2004. We sampled in the field and 1 m from the inner edge of the boundary (e.g. hedgerow) at each site, and used the comparison between these sampling points as a proxy for boundary loss. All pairs were carefully matched to control for habitat variables (such as boundary characteristics) and sampling in the members of each pair was carried out simultaneously so that environmental parameters could be controlled for (see Appendix S1 in the supplementary material for details of matching). We chose a broad-based, extensive sampling regime to minimize potential site-based biases and to facilitate the potential use of our study taxa as biotic indicators in a large-scale or national study.

shrews and their prey

To measure the abundance of shrews, we used baited hair tubes (Pocock & Jennings 2006; Table 1). Shrew hairs were identified to species, and the hair tube index (number of positive tubes for each species, which is related to the abundance of shrews) was calculated (Pocock & Jennings 2006).

Table 1.  Summary of methods used to sample the taxa, including details of equipment, duration of sampling and dates. Each site was sampled once for each taxon, and members of pairs were sampled simultaneously. See also Appendix S1 in the supplementary material
TaxonMethodEquipmentDuration of sampling at each sitePosition of sample points at each siteDates (day/month/year)Time of day
SoricidaeBaited hair tubesHair tubes baited with fly pupae (Pocock & Jennings 2006)7 days40 tubes 5 m apart, on two 95 m transects, one 1 m from the field boundary and one parallel to this, 20 m into the field25/6–21/7/03, 27/8–5/9/03, 23/6–11/8/04Tubes set and collected during daylight hours
ChiropteraAcoustic samplingBat detectors (U30; Ultra Sound Advice, London, UK) connected via PCMCIA III cards (DAQCard AI-16E-4, National Instruments, Austin, TX) to laptop computers (Satellite Pro 4080XCDT and 2100, Toshiba, London, UK). Recording software (Avisoft Recorder, Avisoft Bioacoustics, Berlin, Germany) triggered when a bat flew within range. Settings in the two set-ups (used in the two members of each pair) calibrated20 min at each of four point transects, sampled in systematic order with a random start point (matched between members of pairs)Two, 1 m from the boundary and two 50 m into the field20/5–5/6/03, 11/8–4/9/03, 17/5–29/6/04, 16/8–27/8/04Starting 45 min after sunset
Carabidae and StaphylinidaePitfall trappingCovered pitfall traps, 10 cm diameter, filled with 50% water, 50% ethylene glycol7 daysOne (in 2003) or two (in 2004) traps from each boundary and field transect used for analysisAs SoricidaeTraps set and collected during daylight hours
Night-flying beetles, Diptera and mothsLight trappingTwo heath traps with 6-W blacklight bulbs. Lamps were shielded on the side facing the boundary, to avoid sampling the habitat behind the field boundary2 h 15 minOne 1 m from the boundary, one 50 m into the field, both ≥ 50 m away from bat sampling pointsAs ChiropteraStarting 15 min after sunset
Night-flying beetles, Diptera and mothsSweep nettingSweep nets100 sweeps in the air at each sample point (2 m behind the bat detector)As ChiropteraAs ChiropteraAs Chiroptera

The prey of shrews was sampled by using pitfall traps (Table 1). Carabidae were identified to species (Forsythe 2000), Staphylinidae to family.

bats and their prey

To measure the foraging activity of bats, we used bat detectors (Table 1) to record bat passes (echolocation call sequences; Fenton 1970). Sampling was continuous and broadband (encompassing all frequencies), so passes made by all bat species could be recorded. At each site, we started to record 45 min after sunset, so that bats were feeding, not travelling from roosts to feeding sites.

All sequences were inspected by using sound-analysis software (BatSound Pro v3.0; Pettersson Elektronik AB, Uppsala, Sweden). Sequences or individual calls were submitted to an artificial neural network (ANN) developed by Parsons & Jones (2000). The ANN was trained on calls made by 698 individual bats of 14 species found in our study area, and achieved an overall correct classification rate (based on the training data set) of 87% (Parsons & Jones 2000).

While bats were being recorded, sweep netting was carried out to sample for night-flying Coleoptera and Diptera and light traps were set to capture Coleoptera, moths and Diptera (Table 1). Diptera were identified to family (Unwin 1981) and moths to species (Waring & Townsend 2003). Samples from sweep netting and light traps were combined.

abundance and sensitivity

In common with most other agro-ecological workers, we did not estimate true abundance but relative activity–density for all groups. We quantified the abundance-derived sensitivity of taxa to each aspect of intensification by calculating the statistic d (the difference of log counts; Perry et al. 2003). In this study we only considered abundance-derived sensitivity and not other types outside of our definition, for example toxicological sensitivity. We adopted a similar approach to that used for the farm-scale evaluations (FSE) of genetically modified (GM) crops, as those authors used a paired design with well-developed methodology and also estimated the power of d (Perry et al. 2003; Clark, Rothery & Perry 2006).

To quantify sensitivity to increased agrochemical inputs, we combined data from the field and near the boundary to calculate the response variable for each taxon: dock(lclo)/k, where o= organic, c= conventional, k= number of pairs and lc = log[(abundance near boundary of conventional member of pair k) + (abundance in field of conventional member of pair k) + 1]. We excluded pairs in which the abundance of each taxon was one or zero. The difference between zero and d was tested with a randomization test with 4999 iterations.

The effect of intensification (doc) may have substantially differed between sample points in fields and near boundaries. When sample sizes were sufficiently large (n > 6), we were able to test for differences. The difference from zero of the response variable ioc = Σk[(lcflcb) – (loflob)]/k, where b= boundary and f= field, was tested with a randomization test with 4999 iterations. A conservative approach was adopted, so when P < 0·10 for this test, the results for margins and fields are presented separately. The statistics dhs and ihs, where h= hay and s= silage, were calculated similarly.

To quantify the sensitivity to boundary loss of the taxa considered in the analyses above, we treated data from cereal pairs and grass pairs separately and compared abundance near the boundary and in the field. We tested the difference from zero of the response variable, dbf(cereal)k(lflb)/k, where b= boundary, f= field, lb = log[(abundance near boundary in organic member of pair k) + (abundance near boundary in conventional member of pair k) + 1] and k= number of pairs of cereal. The statistic dbf(grass) was calculated similarly.

Sites in which, because of farm operations, it was only possible to sample shrews and their prey near the boundary were excluded from the estimation of dbf and i, but for doc and dhs we estimated the missing values (based on the average difference between field and margin and the abundance of each taxon in the margin) so that d was not unduly influenced by counts in margins only.

We present estimates of R= 10d, where R is the multiplicative ratio of abundance in organic/hay/boundary to abundance in conventional/silage/field (Perry et al. 2003). For example, R= 2 (or 0·25) indicates that the abundance in organic/hay/boundary was double (or one-quarter) that in conventional/silage/field. Confidence intervals (CI) of R were estimated from the standard error of d across pairs/sites, as appropriate. Average abundance (geometric mean) was back-calculated from the log-counts. We considered results significant at P < 0·05 and marginally significant at 0·05 < P < 0·10. It was not necessary to correct for multiple hypothesis testing (following Brooks et al. 2003).


Eight-hundred and fourteen hair tubes were positive for the three species of shrew, and 3146 ground-active beetles (Carabidae and Staphylinidae) were caught and identified. One-thousand and eighty-four bat passes were identified by the ANN. Of the remaining 312 passes in which the signal to noise ratio was too low for calls to be identified by the ANN, all but three were identified to species or species group following the method used by Vaughan, Jones & Harris (1996). Eight-hundred and seventy-nine night-flying Coleoptera, 14 686 Diptera and 2828 moths were caught and identified (for details of taxa see Table S1 in the supplementary material).

Eighty-two per cent of effects equal to or greater than two-fold (0·5 ≥ R ≥ 2·0) were significant at P < 0·05, so statistical power was good, as predicted (Perry et al. 2003). There were no significant relationships between any combinations of P and d on the one hand, and abundance and sample size on the other, indicating that data quantity did not bias our estimates of effect size and its significance (Clark, Rothery & Perry 2006).

sensitivity values

Sensitivity values (R) for insectivorous mammals and invertebrates (where n > 6) are given in Table 2. When i was significant at P < 0·10 we present both the combined results for completeness (Table 2) and the separate results (Table 3).

Table 2.  Estimates of the sensitivity of shrews, bats and their prey to increased agrochemical inputs in cereal crops, the switch from hay to silage, and boundary loss. Combined results (boundary + field) are shown for the first two aspects with tests of the difference between boundary and field (Pi), as described in the text. R = 10d is the multiplicative ratio of abundance in organic/hay/boundary to that in conventional/silage/field. Significance (P) was assessed by randomization tests. n= number of pairs. Where n < 7 sample sizes were deemed insufficient to calculate R and P. Where Pi  0·10 separate results for the components of R are given in Table 3. Results where P 0·10 are shown in bold. Geometric mean abundance, for sites where the taxon was recorded, is given, where measures of abundance are defined in the text
Higher-level taxonTaxonAgrochemical inputs in cereal cropsSwitch from hay to silageBoundary loss in cereal cropsBoundary loss in grass fieldsGeometric mean abundance
R (95% CI)PdndPiniR (95% CI)PdndPiniR (95% CI)PdndR (95% CI)PdndCereal sitesGrass sites
SoricidaeSorex minutus1·3 (0·8–2·0) 0·296200·037111·2 (0·8–2·0)0·36419 13·1 (2·1–4·7)<0·001164·6 (2·8–7·6)<0·001 7 5·0 2·4
Sorex araneus1·0 (0·6–1·7) 0·967190·148 91·5 (0·9–2·4)0·08718 12·5 (1·5–4·2) 0·003153·8 (2·2–6·4)<0·001 7 4·6 2·2
Neomys fodiens0·8 (0·2–3·3) 0·766 9 2 3 02·6 (1·1–5·7)<0·001 7 1 1·9 2·5
ChiropteraPipistrellus pygmaeus0·8 (0·3–1·7) 0·463 9 20·9 (0·5–1·6)0·60712 23·6 (2·9–4·6)<0·001 92·6 (1·6–4·3) 0·00312 3·0 2·4
Pipistrellus pipistrellus0·8 (0·4–1·6) 0·558140·816 80·9 (0·5–1·6)0·61716 63·4 (1·9–6·2) 0·001143·9 (1·9–7·9) 0·00316 4·9 5·4
Nyctalus/Eptesicus spp.1·0 (0·5–2·0) 0·981 9 30·7 (0·4–1·1)0·114170·685111·5 (0·8–2·8) 0·208 90·9 (0·5–1·4) 0·46917 2·3 4·1
Myotis spp.1·3 (0·6–2·9) 0·367 9 00·7 (0·4–1·1)0·12213 01·8 (0·8–3·8) 0·094 93·1 (2·0–4·9)<0·00113 1·0 1·4
Coleoptera (Carabidae)Bembidion spp.2·5 (0·9–6·9) 0·085  8  00·6 (0·4–0·9)0·043 8 20·8 (0·2–3·3) 0·627 7 4 1·4 3·5
Harpalus spp.2·0 (0·6–6·4) 0·20110 4 5 10·6 (0·3–1·6) 0·281 9 5 2·3 2·0
Pterostichus madidus0·3 (0·0–3·0) 0·369 7 5 1 00·1 (0·0–0·5) 0·014 7 111·9 1·8
Pterostichus melanarius2·1 (0·6–7·5) 0·229150·985 70·1 (0·0–0·1)< 0·00112 10·1 (0·0–0·4) 0·002130·1 (0·0–0·2) 0·028 614·1 4·9
ColeopteraStaphylinidae1·3 (0·8–1·9) 0·215170·208 70·7 (0·4–1·0)0·059200·112 71·9 (1·0–3·6) 0·063131·3 (0·7–2·5) 0·378 7 4·512·8
Coleoptera (night-flying) 1·3 (0·8–2·0) 0·221130·086 71·5 (0·8–2·8)0·201190·538 71·4 (0·9–2·3) 0·150131·7 (1·0–3·0) 0·05819 6·0 3·4
DipteraAnisopodidae 6 20·5 (0·2–0·9)0·030 8 30·7 (0·2–2·1) 0·384 62·0 (0·8–5·0) 0·130 8 2·4 3·1
Cecidomyiidae0·9 (0·5–1·5) 0·692180·382140·8 (0·5–1·3)0·294220·349153·3 (1·8–6·1)<0·001182·3 (1·6–3·2)<0·0012210·5 6·4
Ceratopogonidae1·9 (0·9–3·9)<0·001 7 11·0 (0·4–2·4)0·943 9 41·7 (1·1–2·6)<0·001 71·0 (0·5–1·8) 0·937 9 1·2 2·0
Chironomidae1·2 (0·7–2·1) 0·395190·682160·7 (0·4–1·1)0·100220·851211·3 (0·9–2·0) 0·142190·8 (0·6–1·1) 0·1322216·449·4
Culicidae1·0 (0·5–2·0) 0·995 8 2 5 11·5 (0·7–3·0) 0·209 8 5 1·9 1·4
Mycetophilidae 3 01·2 (0·6–2·3)0·687 8 1 33·0 (1·9–4·9)<0·001 8 1·4 2·0
Psychodidae1·5 (0·8–2·6) 0·182190·311120·8 (0·4–1·6)0·544190·398152·3 (1·3–4·0) 0·008193·3 (2·3–4·9)<0·00119 6·1 8·7
Sciaridae1·0 (0·6–1·8) 0·888 8 40·8 (0·3–2·3)0·575 8 11·1 (0·5–2·4) 0·801 81·3 (0·5–3·4) 0·474 8 2·4 1·6
Tipulidae1·5 (1·0–2·3) 0·07513 60·7 (0·4–1·1)0·13812 31·0 (0·6–1·8) 0·954131·8 (1·0–3·1) 0·05512 2·7 1·9
All Aschiza 4 10·7 (0·4–1·4)0·32813 4 40·4 (0·2–0·9) 0·02913 1·3 3·6
Anthomyiidae0·6 (0·4–0·9) 0·067 8 10·9 (0·3–2·9)0·956 9 20·9 (0·4–1·9) 0·749 80·5 (0·2–1·1) 0·097 9 1·4 1·9
Scathophagidae 6 10·6 (0·3–1·4)0·229170·769 71·0 (0·4–2·6) 0·985 60·6 (0·3–1·4) 0·22917 1·1 4·1
All Acalyptratae0·7 (0·5–1·2) 0·18511 51·0 (0·7–1·6)0·949190·252120·9 (0·6–1·4) 0·650110·7 (0·4–1·0) 0·05019 2·2 3·1
LepidopteraGeometridae0·6 (0·4–1·0) 0·04416 00·8 (0·5–1·2)0·24115 06·4 (3·9–10·6)<0·001164·5 (3·2–6·3)<0·00115 3·4 2·0
Hepialidae1·0 (0·3–3·2) 0·96310 22·3 (1·3–4·0)0·006150·749 93·8 (1·7–8·6) 0·010103·3 (1·7–6·1) 0·00215 3·0 7·1
Noctuidae1·2 (0·9–1·7) 0·183200·078131·1 (0·7–1·6)0·630200·220152·3 (1·6–3·3)<0·001201·4 (0·9–2·1) 0·11120 8·8 9·1
Pyralidae0·9 (0·5–1·5) 0·61510 51·3 (0·8–2·2)0·293120·695 83·0 (1·9–4·8) 0·002101·4 (1·8–2·5) 0·24112 5·1 6·0
Table 3.  Estimates of effect size and significance (P) for increased agrochemical inputs, for separate components of R where the difference (i) was significant. n= sample size
Taxon Near boundaryIn field
R (95% CI)PnR (95% CI)Pn
SoricidaeSorex minutus1·7 (1·0–2·7)0·034200·5 (0·3–1·1)0·08811
LepidopteraNoctuidae1·0 (0·7–1·3)0·894192·0 (1·1–3·6)0·02713
Coleoptera (night-flying) 1·7 (1·0–2·9)0·040100·5 (0·2–1·7)0·320 8

insectivorous mammals

The insectivorous mammals were relatively insensitive to increased agrochemical inputs and the switch from hay to silage (Tables 2 and 3). The effect of boundary loss on shrews and most bats was very strong (R≈ 3) in both cereal and grass sites, but the larger species of bat (Nyctalus/Eptesicus spp., identified as mostly Nyctalus noctula) showed no sensitivity to boundary loss.


One Coleoptera taxon showed significant sensitivity to increased agrochemical inputs [Bembidion spp., mostly Bembidion lampros (Herbst); see Table S1 in the supplementary material]; it was more abundant where agrochemicals were not used. Most Carabidae, including the most abundant species [Pterostichus melanarius (Illiger)], were approximately twice as abundant in organic as in conventionally farmed sites, although results were not significant because of large variances. Pterostichus madidus (Fabricus), however, showed the opposite trend (Table 2).

Coleoptera taxa showed a strong sensitivity to the switch from hay to silage, and all were more abundant in silage than in hay [including Bembidion spp., mostly Bembidion guttula (Fabricus); see Table S1 in the supplementary material]. Carabidae were more abundant in fields than near boundaries, two taxa significantly so, but Staphylinidae showed the opposite trend, being significantly more abundant near boundaries than in fields in cereal sites (Table 2).

Night-flying beetles were significantly more abundant near the boundary of organic than conventional sites. No other results were significant for this taxon.

diptera and moths

Most Diptera and moths were not significantly sensitive to increased agrochemical inputs (Table 2). Of those that showed significant or marginally significant sensitivities, two families were more abundant in conventionally farmed sites (Anthomyiidae and Geometridae) and two were more abundant in organic sites (Tipulidae and Ceratopogonidae).

Few Diptera and moth taxa were significantly sensitive to the switch from hay to silage (Table 2). Both Anisopodidae and Chironomidae were more abundant in silage than in hay, while Hepialidae were more abundant in hay than in silage and the effect was both large and highly significant.

Many groups showed significant sensitivity to boundary loss (Table 2). This was strongly the case for the four moth families in cereal sites and for two families in grass sites. Overall, five Dipteran taxa (all Nematocera, all of which but Tipulidae are swarm-forming) were significantly more abundant near boundaries than in fields in cereal and/or grass sites. Three Diptera taxa (none of which was Nematocera) in grass sites were significantly more abundant in fields than near boundaries, but none in cereal sites showed this result.

aspects of agricultural intensification

Overall, 42% of results were marginally significant (P < 0·10) and 31% of results were significant (P < 0·05). Half (51%) of the responses to boundary loss were significant (P < 0·05), while 8% of responses to increased agrochemical inputs and 16% to the switch from hay to silage were significant (Table 2). For our taxa, boundary loss was most influential on abundance and the increased use of agrochemicals was least influential.

Examining the four sets of results for our taxa (Table 2), there was a significant association between significant sensitivity (P < 0·05) to boundary loss in cereal and in grass (Fisher's exact test, P= 0·012) but relationships between other results were not significant. In our study, the sensitivity of a taxon to one aspect of intensification was a poor predictor of its sensitivity to other aspects.


There was substantial variation between taxa in their responses (expressed as abundance-derived sensitivity) to the three aspects of intensification considered in this study. Variation also occurred within higher taxonomic levels, so related taxa (even members of the same genus) that were tested sometimes responded differently to an aspect of intensification (Table 2). Overall, this study extends the scope of other large-scale studies of agricultural biodiversity in lowland Britain (Brooks et al. 2003; Haughton et al. 2003; Roy et al. 2003; Fuller et al. 2005).

insectivorous mammals

We studied shrews and bats partly because they show great differences in life history, mobility and habits. Bats are highly mobile and can take advantage of variation in abundance of their prey, therefore their local activity–abundance is expected to respond quickly to local changes in prey abundance. Shrews are less mobile, so differences between members of pairs are expected to be mainly population-level responses to prey abundance, as many other important habitat features (e.g. boundary characteristics) were controlled for in this study (see Appendix S1 in the supplementary material). However, neither bats nor shrews showed significant sensitivities to increased agrochemical inputs or the switch from hay to silage (except for Sorex minutus L., which was more abundant in organic sites near boundaries but not in the fields). Abundance-derived measures of sensitivity tended to be smaller in magnitude than those of most invertebrates, in contrast to the expectation that taxa at higher trophic levels show increased sensitivity (Landres, Verner & Thomas 1988).

Agricultural intensification, as a whole, may have negative impacts on populations of shrews, although there is limited evidence for declines (Harris et al. 1995; Battersby 2005). Detailed studies indicate that shrews show reduced activity in fields with high pesticide application, probably because of lack of cover (Greig-Smith 1991; Schauber, Edge & Wolff 1997). Conversely, they can benefit by feeding on invertebrates immobilized after insecticide application (Stehn, Stone & Richmond 1976), although toxicological effects, which we did not consider, could follow. However, our results demonstrate that boundary loss has a huge effect on shrew abundance, and we therefore suggest that AES promoting the replanting and careful management of hedgerows and field boundaries will have substantial positive effects on shrews and probably other small mammals.

Past declines in bat abundance in Britain (Harris et al. 1995) may be the result of agricultural intensification. Indeed, some bat species have been proposed as good biotic indicators, for example Pipistrellus pipistrellus (Schreber) is regarded as an indicator of ‘the impact of general agricultural practices on small insect populations’ (Anonymous 2004). We found no evidence that bats are sensitive to increased agrochemical inputs (although we cannot comment on possible toxicological effects) or the switch from hay to silage. However, most taxa were highly sensitive to boundary loss, possibly partly because they feed on swarming Diptera, which congregate near windbreaks such as boundaries. Previous studies have shown a positive influence of organic farming on bats (Wickramasinghe et al. 2003; Fuller et al. 2005) but our lack of significant results suggests that boundary and other non-crop habitat differences (which we controlled for) may be more important for bats than differences in field management practices such as agrochemical use.


Night-flying beetles were sensitive to increased agrochemical inputs, although this was only evident near boundaries. Carabidae and Staphylinidae were more abundant where ground cover was open, so more abundant in fields than near boundaries and in silage than in hay. Effects may be slightly over-estimated, as cereal fields and silage are also more permeable to beetles than cereal margins (near boundaries) and hay, resulting in higher activity densities (Holland & Luff 2000). For some predatory invertebrates, however, reduced floristic diversity and the early timing of grass cuts in silage is known to have negative effects (Vickery et al. 2001). Boundary loss may appear to benefit beetles, but many species invade fields from refuges in the margins, so the abundance of even the open-field specialists would probably be reduced if fields became very large.

Pesticide applications in the growing season, when Carabidae are active in the crop, directly reduce their abundance, but this is quickly mitigated by re-invasion from field margins (Greig-Smith 1991; Duffield & Aebischer 1994; Holland & Luff 2000). Herbicide application causes longer-term effects, reducing the abundance of Carabidae (Greig-Smith 1991; Sotherton 1991; Luff & Woiwod 1995; Navntoft, Esbjerg & Riedel 2006), although some species may respond positively to the reduction in ground cover (Navntoft, Esbjerg & Riedel 2006). The result of these effects on the abundance of Carabidae is complex, and site- and season-specific. This complexity may explain why we found little sensitivity to increased agrochemical inputs.

Staphylinidae show a mixed response to organic farming and pesticide applications (Aebischer 1991; Greig-Smith 1991; Brooks et al. 2003; Hole et al. 2005). In our study, responses were also mixed. Staphylinidae were more abundant near boundaries than in fields (significantly so in cereal), suggesting that vegetation cover is important, although they were less abundant in hay, which had relatively good ground cover, than in silage.

Overall, Carabidae and Staphylinidae respond positively to open habitats, which tend to be more abundant in intensive agriculture than in extensively farmed areas. Therefore, AES may cause declines in populations of some taxa, at least in the short term. However, increases in habitat heterogeneity because of AES are likely to benefit many beetle species, including those not represented in our samples.

diptera and moths

A smaller proportion of night-flying Coleoptera, Diptera and moths (bat prey) than ground-active beetles (shrew prey) were sensitive to increased agrochemical inputs and the switch from hay to silage, possibly because of their increased mobility. In contrast, the strongest effects in the FSE of GM crops were shown by more mobile groups (Haughton et al. 2003), perhaps because the smaller experimental scale (split fields rather than pairs of fields) allowed mobile taxa to select their preferred habitat.

Most farmland Diptera and moths are probably generalists, as are farmland butterflies (Asher et al. 2001), but all three groups have declined on farmland over the past 30 years (Benton et al. 2002; Conrad et al. 2004). Moths are more abundant on organic farms than on conventional farms (Wickramasinghe et al. 2004; Hole et al. 2005) but in our study they were not strongly sensitive to increased agrochemical inputs. This may be because we controlled for differences in boundary characteristics, suggesting that the quality of boundaries is very important to moths. However, Noctuidae were significantly more abundant in the fields of organic crops than in those of matched conventional crops, on which agrochemicals were used. There are several possible reasons for this: (i) adults move throughout the landscape but are killed by insecticides in conventional fields; (ii) adults prefer to forage in organic than conventional fields, perhaps because of the abundance of weed nectar sources; (iii) adults emerge in greater abundance in organic fields, either because summering larvae survive on herbaceous weeds or over-wintering stages survive better in organic fields. Hepialidae were sensitive to the switch from hay to silage, and showed the strongest sensitivity to this aspect of intensification of any group examined. They were also abundant and are easily identified, making them potentially useful indicator taxa. Noctuidae and Hepialidae deserve further investigation, particularly given the recent concern expressed over moth declines (Fox et al. 2006) and that Lepidoptera are recommended as indicator taxa (Luff & Woiwod 1995; Thomas 2005).

Diptera are abundant but underrepresented in studies of farmland biodiversity, yet they are important in the diets of bats and birds (Vaughan 1997; Benton et al. 2002) and have recently declined (Benton et al. 2002). Diptera showed much weaker responses to the switch from hay to silage than to increased agrochemical inputs, which was surprising given the habitat differences between hay and silage. Several Nematoceran families were significantly more abundant near boundaries than in cereal and grass fields. These groups tend to swarm at landmarks, and turbulence may cause accumulations near boundaries (Lewis & Dibley 1970). We identified Diptera to the family level (as we considered species identification impractical for large-scale surveys) but this may have hidden subtle taxonomic variations between sites. In the face of apparently widespread invertebrate declines (Benton et al. 2002; Fox et al. 2006), our evidence suggests that some Diptera and moths will benefit from AES, particularly from the management or re-instatement of field boundaries.

why were few taxa sensitive to increased agrochemical inputs and the switch from hay to silage?

Using matched pairs of sites allowed us to address specific hypotheses at the field scale. In the FSE of GM crops, in which a paired design was also used, the aim was to detect differences in abundance that were as small as reasonably possible (about 1·5-fold; Perry et al. 2003; Clark, Rothery & Perry 2006). Good power to detect small effects is usually desirable but costly. We were interested in detecting the largest effects and aimed to identify the taxa that were most sensitive to agricultural intensification. Although the three aspects of agricultural intensification we considered have been implicated in the decline of farmland biodiversity (Krebs et al. 1999), few estimates of sensitivity to changes in field management were significant. Indeed, many values of R were close to one and had relatively small CI (Table 2). Many other researchers using similar sampling intensities to ours (summarized in Bengtsson, Ahnström & Weibull 2005; Fuller et al. 2005; Hole et al. 2005) found more substantial (i.e. more significant and larger) differences between organic and conventional farming than we did. However, we paired for field boundary characteristics (see Appendix S1 in the supplementary material) while most other researchers using similar sampling intensity and methods to ours did not, suggesting that the state of field boundaries is very important and that our careful pairing for them may explain the lack of sensitivity we found.

Some taxa may be sensitive at critical times in their life span (e.g. immediately after emergence for invertebrates) but our extensive sampling approach was not designed to detect such effects. We aimed to include a wide range of mammalian and invertebrate taxa, and used methods applicable to large-scale surveys of these taxa. Although our approach was not designed to identify all sensitive taxa, it was successful in identifying the taxa that were most sensitive, across a range of sites, to the aspects of agricultural intensification that we considered.

insectivorous mammals and their prey as biotic indicators

We found no evidence that vertebrates were more sensitive to agricultural habitat change than their prey, i.e. that sensitivity was amplified up the food chain. Indeed invertebrates tended to be more sensitive than vertebrates, at least to changes in cereal and grass management. The most sensitive taxa identified from abundance-derived estimates of sensitivity with our extensive, multisite sampling regime are potential indicators, namely some Diptera and some Carabidae in cereal, Hepialidae and some Carabidae in grassland, and several taxa of boundary loss. No taxon that we considered was significantly sensitive to all three aspects of intensification, so none can be used singly as an indicator of agricultural intensification. A multitaxon group, including taxa sensitive to each aspect of intensification, could be used as an indicator but has the disadvantage of requiring multiple sampling methodologies. However, Hepialidae are strongly sensitive to grass management and, because they are abundant, easily identified and easily sampled, could be used as an effective indicator in grasslands. Further research on the mechanisms underlying variation in sensitivity would be beneficial, but for now it is essential that the sensitivity of all potential indicators is tested empirically, for example in the way we have demonstrated, before time and money are invested in their use as biotic indicators at larger scales.


We thank the farmers for allowing us access to their land, Emily Bennitt for assistance with field work, Stuart Parsons for running the ANN, Ray Barnett for identifying micromoths, and Stephen Harris and Gareth Jones. We particularly thank Jenny Owen for expert help in the field and laboratory. This project was funded by the Department of Environment, Food and Rural Affairs and English Nature (grant number BD2001).