Riparian field margins: can they enhance the functional structure of ground beetle (Coleoptera: Carabidae) assemblages in intensively managed grassland landscapes?

Authors


Correspondence author. E-mail: Lorna.Cole@sac.ac.uk

Summary

  1. In Europe and North America, there is growing concern that biodiversity declines associated with agricultural intensification will adversely impact the functioning and sustainability of agricultural ecosystems. Enhancing habitat heterogeneity in agricultural landscapes promotes biodiversity and, whilst erecting fences adjacent to watercourses is widely advocated as a means of mitigating diffuse pollution, associated biodiversity benefits have been largely overlooked.
  2. A range of riparian margins and their adjacent grassland fields were investigated to determine the effects of riparian management on the diversity and functional structure of carabid assemblages. Carabid assemblages of fields and open margins (i.e. unfenced watercourses) were more diverse and species rich than those of fenced margins.
  3. The functional structure of carabid assemblages in fenced margins differed from grassland fields and open margins. This disparity was greater in wide margins (i.e. fences erected over 5·4 m from watercourses) than narrow margins (i.e. fences erected within 2·6 m of watercourses). Wide margins had the highest relative proportions of carabids which had pushing body forms, were flightless, very small in size and Collembola specialists. During early summer, wide margins also had the highest proportion of carabids that overwinter as adults.
  4. The taxonomic and functional structure of carabid assemblages was more sensitive for detecting impacts of agricultural management than measurements of diversity. It is likely that this also applies to other taxa, thus emphasising the need to consider a wide range of assemblage attributes when investigating agricultural impacts on biodiversity.
  5. Synthesis and applications. Fenced riparian margins, particularly those over 5·4 m wide, harbour carabids with poor dispersal ability which are vulnerable to habitat fragmentation. While lack of management benefits sedentary species, a wider range of taxa (e.g. pollinators, foraging birds and flowering plants) are enhanced by management to obtain a more open vegetation structure (e.g. restricted grazing or mowing). It is important that management practices are implemented at a sufficiently fine spatial scale to allow recolonisation of species with restricted dispersal from adjacent undisturbed habitats. Wide riparian margins have the potential to enhance taxonomic and functional diversity at the landscape scale. Management actions must, however, be carefully balanced to ensure that they promote a wide range of taxa without unduly interfering with the margin's ability to mitigate diffuse pollution.

Introduction

Over the past 60 years the heterogeneity of agricultural landscapes has declined in Europe and North America as a result of an increase in field size, the disappearance of hedgerows and farm woodlands, and the specialisation of farming systems both at the farm and landscape level. Alongside the intensification of agricultural practices, this loss of heterogeneity has been linked to a decrease in biodiversity across a range of taxa (Benton et al. 2002; Benton, Vickery & Wilson 2003). As biodiversity is strongly linked to ecosystem functioning, there is growing concern that this decline could adversely impact the functioning of agricultural ecosystems and thus the sustainability of associated ecosystem services (Flynn et al. 2009; Firbank et al., in press). The restoration and sympathetic management of field margins in intensive agricultural landscapes improves habitat heterogeneity and enhances ecosystem services such as the conservation of biodiversity, reduction of agro-chemical drift and provision of pollinators and biological pest control agents (Olson & Wäckers 2007; Woodcock et al. 2009; McCracken et al. 2012). Situating such margins adjacent to watercourses not only targets areas which are naturally rich in biodiversity but also helps to mitigate diffuse pollution, thus providing multiple environmental benefits from land taken out of production (Cole et al. 2012).

The move towards protecting ecosystems as a whole and the associated ecosystem services they provide, has resulted in a plethora of recent studies looking at functional traits in birds (Julliard, Jiguet & Couvet 2004; Petchey et al. 2007), mammals (Flynn et al. 2009) and arthropods (Cole et al. 2002; Schweiger et al. 2007). Determining interactions between land management and functional traits not only helps to predict implications for ecosystem functioning, but may also help to establish the principal factors driving population changes. Functional analyses of arthropod communities, however, are still in their infancy due to the lack of comprehensive ecological data, the diversity of functional roles that arthropods play and the complexity of their interactions with other organisms (Schweiger et al. 2007; Flynn et al. 2009). While studies that investigate the influence of land management on functional diversity or specific ecological groups provide important information on overall trends, simplification of the data results in a loss of information and makes it difficult to determine the fundamental processes influencing observed changes.

Carabidae beetles not only show considerable interspecific diversity in ecology, but they also play key functional roles in agricultural ecosystems (e.g. controlling agricultural pests and providing food for farmland birds). They are therefore an ideal group with which to investigate functional traits (Ribera et al. 1999; Cole et al. 2002). Cole et al. (2008) found that the taxonomic structure of carabid assemblages was sensitive to riparian management, with margins open to livestock grazing containing similar assemblages to grassland fields whereas margins closed to livestock contained distinct assemblages. This disparity was greater in wide margins (i.e. >5·4 m wide) than narrow margins (i.e. ≤2·6 m). The current study investigated the influence of riparian management on carabid assemblages in intensive grassland systems using a functional rather than a taxonomic approach. A better understanding of the underlying processes driving carabid assemblages is needed to help formulate management options that capitalise on the potential of riparian margins to be multifunctional. A range of riparian margins, and their adjacent grassland fields, were investigated to test the hypothesis that fenced riparian margins, and in particular wide margins, enhance the functional structure of carabid assemblages through the introduction of more diverse vegetation structure and greater temporal stability.

Materials and methods

Study sites

Thirteen farms situated within the Cessnock catchment, Ayrshire, Scotland (N55°32′50″, W4°22′00″) were selected for study over the 3-year period, 2006–2008. The catchment was dominated by productive ryegrass, Lolium perenne L., swards encompassing livestock grazing and/or cutting for silage or hay. Grazed fields were typically grazed by cattle at high stocking levels (i.e. >1·14 livestock units ha−1) from April to September. Silage fields were cut up to three times annually with the first cut occurring in May–June. Hay fields were cut once in July. Twenty-six grassland fields were studied and within each field 1–3 sites (a total of 43 sites) were chosen to represent the range of riparian margins occurring within the landscape. Sites were classified into one of three categories (Fig. 1; Table 1): Open Sites (i.e. no fences between fields and watercourses), Narrow Sites (i.e. fences erected within 2·6 m of watercourses) and Wide Sites (i.e. fences erected more than 5·4 m from watercourses). At each site, two sampling transects were established, one (Margin transect) typically 0·5–1·5 m from the watercourse and the other (Field transect) c. 4 m into the field from either the margin fence, or in the case of Open Sites, the Open Margin transect. Additional transects, at the midpoint between the fence and watercourse, were established at Wide Sites (Wide Middle transect). Open Sites and Narrow Sites therefore each contained two transects (Open Margin, Open Field and Narrow Margin, Narrow Field, respectively) while Wide Sites contained three (Wide Margin, Wide Middle and Wide Field). Sampling adjacent to the watercourse was not feasible at one location resulting in the loss of one Wide Margin transect.

Figure 1.

Diagram showing transects occurring in the three site categories; their typical distance from the watercourse/fence and how they relate to riparian management factors. Ranges of margin widths are provided for fenced margins. The seven levels for Detailed management are given in lower case italics while the three levels for Coarse management are differentiated by lines representing pitfall transects.

Table 1. Number of transects in each transect category and, in brackets, the number of farms sampled per year
Transect categoryNumber of transects (number of farms)
200620072008Total
Open Margin9 (6)15 (10)14 (9)16 (10)
Open Field9 (6)15 (10)14 (9)16 (10)
Narrow Margin6 (3)14 (8)14 (8)14 (8)
Narrow Field6 (3)14 (8)14 (8)14 (8)
Wide Margin5 (1)12 (5)12 (5)12 (5)
Wide Middle6 (2)13 (6)13 (6)13 (6)
Wide Field6 (2)13 (6)13 (6)13 (6)
Total47 (7)96 (13)94 (12)98 (13)

Carabid sampling

At each transect, carabids were monitored by a row of nine pitfall traps (inter-trap distance 2 m; total transect length 16 m). Pitfall traps (75 mm diameter and 100 mm deep) contained 50 ml of 100% mono-propylene glycol (a killing agent and preservative) and were covered with a 15 mm mesh grid (Cole et al. 2002). Traps were installed in mid-June and were left in situ for a total of 8 weeks, although a small number were lost due to livestock trampling. Trap contents were collected in mid-July and mid-August to give 2, 4 week sampling periods annually: June–July and July–August. During collection, the samples derived from each row of nine traps were pooled.

Vegetation and management

The direct measurement method was used to accurately measure both short and long vegetation (Stewart, Bourn & Thomas 2001). Measurements were taken (to the nearest 5 mm) twice per sampling period using a graduated meter stick at ten points randomly selected along each transect and mean vegetation height calculated for each sampling period.

Data were collected on the following attributes relating to Field type and Management intensity through on-site observations and interviews with landowners: sward type, current and past land use, fertiliser and agrochemical input, grazing intensity and frequency of ploughing, harrowing and mowing. These data were then summarised to provide a Management intensity index (i.e. ordinal scale with a potential range of 0–24) for each transect, with higher values indicating greater management intensity (Cole et al. 2002).

Carabid trait data

Following Cole et al. (2002) eight functional traits were selected to provide reliable and measurable aspects of ecology, behaviour and morphology that influence how carabids interact with their environment (Flynn et al. 2009). Each trait consisted of between two and four attributes (Table 2). Fifty-three carabid species were sampled in pitfall traps and trait data were obtained for all species through extensive literature searches (Cole et al. 2002).

Table 2. Summary of the eight ecological traits investigated, showing a brief description of each trait and descriptions of attributes within a trait. The percentage of 27 genera, 53 species and 19 172 individuals showing each attribute is also provided
Ecological traitAttributes% genera with attribute% species with attribute% individuals with attribute
Size (total length)Very small (<5 mm)18·522·614·9
Small (≥5 mm–<9 mm)63·050·935·8
Medium (≥9 mm–<15 mm)18·513·226·4
Large (≥15 mm)14·813·222·8
Overwintering stageAdults only51·945·336·7
Adult & larvae or larvae only66·754·763·3
Duration of life cycle1 year92·690·682·2
2 years14·89·417·8
Diet of adultsCollembola specialists25·918·912·4
Generalist predators66·760·478·2
Mixed diet3·71·96·0
Mostly plant18·518·93·4
Diel activityDiurnal18·524·510·7
Diurnal & nocturnal14·87·631·5
Nocturnal74·167·957·8
Breeding seasonSpring/summer63·062·349·9
Autumn/winter48·137·750·1
Dispersal abilityFlightless29·617·014·0
Dimorphic wings33·334·044·9
Capable of flight70·449·141·2
Body form/locomotionRunner22·217·031·8
Pusher70·477·466·2
Digger7·45·72·0

Data analyses

To summarise associations between and within transect categories and trait attributes, a single Detrended Correspondence Analysis (DCA) was conducted on the matrix of proportions classified by sampling transect and carabid species (i.e. an analysis of relative abundance data for the two sampling periods combined: Oksanen & Minchin 1997). Centroids for each transect category (i.e. mean site scores for all sites within a transect category) and trait attribute (i.e. mean species scores for all species with a specific trait attribute) were calculated. This provided an efficient, consistent approach to represent the overall relationship between transect categories and trait attributes.

Formal analyses taking full account of the structure of data collection described above involved fitting Generalised Linear Mixed Models (GLMMs), using Residual Maximum Likelihood, a log link function and assuming Poisson distributed errors. Models were fitted to the following response variables for each Sample: beetle activity abundance (i.e. number of carabids sampled) and species richness (i.e. number of species sampled); and the counts of beetles (c.f. individuals) and species exhibiting different attributes within a trait (i.e. for each trait, data were summarised to one record per attribute per sample). In addition, Linear Mixed Models (LMMs) were fitted to the Shannon diversity index.

The hierarchical structure for random effects was, in descending order, Farm, Field, Site, Transect and Sample, thus enabling a greater precision of comparison between transects at a specific site and sampling date than between transects at different sites and dates. To deal with repeated measures and enable greater variation between years than between months within a year, interactions between these random effects and year were also included. The scale parameter in the GLMMs was always set to 1. Modelling was conducted in Genstat 12 and all factors and covariates included in models are described in Table 3.

Table 3. Summary of the fixed and random factors and covariates investigated using GLMMs/LMMs. For each factor a description and the number of levels it contains are given and, in the case of fixed factors, a description of these levels is also provided
EffectsDescriptionLevelsDescription of levels
  1. a

    Fitted as covariates and standardised (i.e. departure from the mean divided by the standard deviation) prior to fitting.

Fixed effects
Field type Land use of each field3Intensive Grazing, Hay, Silage
Coarse management Broad description of transect category3Field, Fenced Margin, Open Margin
Detailed management Detailed description of transect category7Wide Field, Narrow Field, Open Field, Open Margin, Narrow Margin, Wide Margin, Wide Middle
Year Year of sampling32006, 2007, 2008
Month Months of sampling2June–July, July–August
Vegetation height a Mean vegetation height Continuous scale of vegetation height (cm)
Management intensity a Management intensity at transect0–19Ordinal scale of management intensity
Random effects
Farm Farm where field is situated13 
Field Field with 1–3 different sampling sites26 
Site Location in a field with, for example, a narrow field and narrow margin transect43 
Transect A specific transect (e.g. Narrow Margin)98 
Sample Unique sample derived from a specific transect on a specific sampling date474 

Analyses of activity abundance and species richness were standardised for sampling effort by including log number of traps as an offset. Trait analyses were standardised for sampling effort by either including the log of the number of traps as an offset (described as analyses of counts), or by fitting a categorical variable for Sample as the first fixed effect (described as analyses of proportions). Including Sample in this way resulted in analyses which compared the proportion of individuals or species and therefore not only controlled for differences in pitfall trap number, but also helped to standardise for differences in pitfall sampling efficiency as a result of vegetation structure (Lang 2000).

Riparian management was included as a fixed effect at two levels of detail (Fig. 1; Table 3): Coarse management compared transects in fenced margins (i.e. Narrow Margin, Wide Margin and Wide Middle), Open Margins and Fields (i.e. Open Fields, Narrow Fields and Wide Fields), while Detailed management compared all seven transect categories. To detect influences of margin width over and above effects of Coarse management, Detailed management was tested for following the inclusion of Coarse management in the model. The following three combinations of fixed effects were therefore explored:

  1. Field type + Coarse management + Vegetation height + Year + Month + Management intensity
  2. Field type + Detailed management + Vegetation height + Year + Month + Management intensity
  3. Field type + Coarse management + Detailed management + Vegetation height + Year + Month + Management intensity

To determine if riparian management effects were consistent in June–July and July–August models also included interactions between Month and riparian management factors as fixed effects. Fenced margins generally have taller vegetation than Open Margins and Fields, thus potentially confounding influences of riparian management and vegetation height. To test for effects of riparian management over and above those attributable to vegetation height, the three models were also fitted with vegetation height included in the models before riparian management factors.

To investigate variation between numbers of individuals or species with the different attribute levels of a trait (e.g. very small, small, medium and large for trait Size) categorical variables were constructed with one level for each attribute (generically called Trait). The main effect of Trait was fitted as the first (i.e. for models based on counts) or second (i.e. for models based on proportions) fixed effect. Of particular interest in this paper is the influence of the above fixed effects on the balance of numbers of individuals and species with different attributes for a particular trait. In modelling terms, this is captured by the interaction between Trait and each of the other fixed effects, which were therefore also included in the models along with interactions between Trait and the hierarchy of random effects. This enabled all attribute levels of a trait to be modelled simultaneously. Significance levels were estimated using Wald tests, with denominator degrees of freedom estimated from the balance of information at different levels in the hierarchy of random effects, and hence varying between analyses of different traits.

To fully examine the data and ensure robustness of the conclusions drawn, separate analyses were conducted on the count of individuals, count of species, proportion of individuals and proportion of species with different attributes of each trait. The results and the discussion section, however, focus on the interpretation of results based on the analysis of the proportion of individuals as these were sensitive in detecting influences of riparian management but were not unduly influenced by the total number of carabids sampled. Results of additional analyses can be found in Tables S2–4 (Supporting Information).

Results

Overall trends

Over the 3 year sampling period, 19 172 carabids consisting of 53 species were recorded (Table S1, Supporting information). The dominant species were Nebria brevicollis (Fabricius) (4289 individuals), Pterostichus melanarius (Illiger) (2097 individuals), Agonum muelleri (Herbst) (1908 individuals) and Loricera pilicornis (Fabricius) (1713 individuals).

DCA of carabid species relative abundance data yielded eigenvalues of 0·499, 0·355, 0·257 and 0·222, for axes 1–4 respectively, explaining 9·4%, 6·7%, 4·8% and 4·2% of inertia. Transect categories were clearly separated along Axis 1 with Open Margin and Field having low scores, Narrow Margin having an intermediate score and Wide Margin and Wide Middle having the highest scores (Fig. 2). The ordination suggests that Wide Margin and Wide Middle transects had high proportions of species that are nocturnal, very small, flightless and Collembola specialists while Open Margins and Fields had high proportions of species that are diurnal and nocturnal, diggers, large and with a mixed diet. Trait attributes commonly found in species investing in growth over reproduction occurred in close proximity to each other (i.e. large size, 2-year life cycle and flightless) as did attributes found in species investing in reproduction over growth (i.e. small size, 1-year life cycle and capable of flight).

Figure 2.

DCA ordination based on the species relative abundance data. Ordination shows centroids for the seven Detailed management transect categories (i.e. mean site scores for all transects in a category) and trait attributes (i.e. mean species scores for species with a specific trait attribute). Overall site (i.e. mean site scores for all transects) and species (i.e. mean species scores for all species) centroids are also provided.

Influence of riparian management

All measurements of diversity were significantly influenced by Coarse management (i.e. the presence of fences). However, Detailed management was never significant following the inclusion of Coarse management in the models indicating margin width had little influence on diversity (Fig. 3; Table 4). Carabid activity abundance and species richness were higher in Fields (i.e. Narrow, Wide or Open Field) and Open Margins than fenced margins (i.e. Narrow Margin, Wide Margin or Wide Middle) and approximately twice the number of carabids were sampled in Fields and Open Margins than fenced margins. Shannon diversity was higher in Open Margins than fenced margins.

Figure 3.

Influence of Detailed management on activity abundance, species richness and Shannon diversity. Plots are based on the raw data by averaging over repeated samples for each transect. Means and standard errors of means were then determined based on replicates of each transect category.

Table 4. Results of GLMMs/LMMs conducted on diversity measurements and on trait analyses based on the proportion of individuals. Minimum (top) and maximum (bottom) approximate F-value (i.e. Wald value/ndfa), and ndf and estimated ddfb ranges for sequential tests of factors and for analyses based on traits, the interaction between Trait and factors (including interactions with Month) and the covariate Vegetation height. F-value ranges are derived from a maximum of six models
Response variable Field type Coarse management Detailed management Detailed management c Vegetationheight Coarse × Month Detailed × Month Detailedc × Month
  1. a

    Numerator degrees of freedom.

  2. b

    Denominator degrees of freedom.

  3. c

    Detailed management tested after Coarse management.

  4. × denotes interaction.

  5. ***< 0·001, **0·001 ≤ < 0·01, *0·01 ≤ < 0·05.

ndf2–62–66–184–121–32–66–184–12
Estimated ddf20–37650–39459–25560–225219–1013202–775220–809220–809
Activity abundance0·6728·09***9·82***0·860·865·05**2·82*1·70
1·0341·00***13·93***0·9225·93***5·09**2·82*1·70
Species richness0·1815·25***5·46***0·560·001·410·870·59
0·6031·60***10·85***0·5631·79***1·430·870·59
Shannon diversity0·265·09**2·47*1·140·451·170·650·39
0·5611·45***4·55***1·1412·44***1·170·650·39
Size1·733·94**2·86***2·28**3·46*2·88**1·69*1·09
1·989·38***4·78***2·48**15·44***2·92**1·69*1·09
Overwintering stage0·040·511·141·452·045·80**3·77**2·72*
0·061·651·451·464·15*5·88**3·77**2·72*
Life cycle2·050·530·670·7110·62**4·93**2·81*1·76
2·292·491·711·3317·16***4·93**2·81*1·76
Diet of adults0·753·28**2·82***2·22*3·09*2·51*2·65***2·67**
0·793·99**2·82***2·47**3·89**2·60*2·65***2·67**
Diel activity0·712·78*2·02*1·658·99***1·722·58**2·97**
0·9714·63***5·86***1·8932·31***1·792·58**2·97**
Breeding season0·360·930·790·710·095·47**4·07***3·26*
0·431·010·790·710·195·69**4·07***3·26*
Dispersal ability1·7615·89***9·51***6·04***4·07*3·26*2·02*1·28
3·58*37·71***16·82***6·37***49·30***3·51**2·02*1·28
Body form3·05*1·642·73**3·23**1·853·28*2·36**1·89
3·72*4·42**3·63***3·26**6·33**3·28*2·36**1·89

Width effects were detected for size, diet of adults and body form in analyses based on proportion or counts of individuals and for dispersal ability in all analyses. Coarse management effects were more frequently detected in analyses dealing with counts than in analyses dealing with proportions. For most trait attributes, mean counts of individuals and species occurring in the seven Detailed management categories either reflected activity abundance or showed no obvious trend, indicating that analyses based on counts were influenced by the total number of carabids sampled (Fig. S1, Supporting information). Pitfall traps operate less efficiently in tall, dense vegetation. As vegetation structure is strongly confounded with riparian management, caution must be taken when interpreting analyses based on counts. Analyses based on proportions are less influenced by the total number of carabids sampled thus helping to standardise sampling efficiency between management categories. For most trait attributes, the mean proportion of species in Detailed management categories followed similar trends to the mean proportion of individuals. Trends for the former were, however, of a lower magnitude. Results based on proportion of species tended to be less significant than those based on proportion of individuals due to summarising data to species level.

Results based on the proportion of individuals indicated highly significant influences of Coarse management for size, diet of adults, diel activity, body form and dispersal ability. With the exception of diel activity, significant effects of Detailed management were found following the inclusion of Coarse management in the models, thus indicating differences between Narrow Margin, Wide Margin and Wide Middle transects and/or Narrow, Wide or Open Field transects. The distribution of trait attributes was broadly similar in Field transects and Detailed management effects could be attributed to effects of margin width (Fig. 4). With the exception of body form, effects of margin width were weaker than effects of Coarse Management. Riparian management effects tended to persist when vegetation height was included in the models first, indicating that management effects were not solely attributable to differences in vegetation height.

Figure 4.

Influence of Detailed management on proportions of individuals displaying the different attributes for traits (a) Diel activity, (b) Size, (c) Body form, (d) Diet of adults and (e) Dispersal ability. Plots are based on the raw data by averaging over repeated samples for each transect. Means and standard errors of means were then determined based on replicates of each transect category. Lines connecting transect categories are for ease of interpretation and do not imply any true connection.

The proportion of nocturnal carabids, when compared to carabids with no diel preference, was higher in fenced margin transects than in Field or Open Margin transects (Fig. 4a). The mean proportion of very small carabids recorded in Wide Margin and Wide Middle transects was approximately double the mean in other transect categories, while the mean proportion of medium carabids in Wide Margin and Wide Middle transects was approximately half that recorded in other transects (Fig. 4b). A higher proportion of pushers and lower proportion of runners were found in Wide Margin and Middle transects when compared to all other transects (Fig. 4c). Higher proportions of Collembola specialists and lower proportions of generalist predators occurred in Wide Margin and Middle transects, and to a lesser extent Narrow Margin transects, when compared to Field and Open Margin transects (Fig. 4d).

Dispersal ability was the trait most strongly influenced by riparian management factors. The mean proportion of flightless carabids in Wide Margin and Wide Middle transects was over three times higher than in Field and Open Margin transects (Fig. 4e). The trend for Narrow Margins was similar but of lower magnitude. Conversely, the mean proportion of carabids capable of flight in Wide Margin and Wide Middle transects was approximately half that recorded in Open Margin, Narrow Margin and Field transects. Effects of riparian management were consistent for dispersal ability irrespective of whether analyses were based on proportions of individuals, proportions of species, counts of individuals or counts of species (Tables S2–4, Fig. S2, Supporting information).

Management intensity

Management intensity did not significantly influence any of the diversity measurements or traits investigated and this information is therefore omitted from Table 4. This study principally aimed to investigate the influence of riparian management and so management intensity was included in the models following Field type, riparian management factors and Vegetation height. Consequently, most variation due to management intensity had already been accounted for by these variables.

Influences of year and month

Significant influences of month were found for species richness and activity abundance. However, only activity abundance was significantly influenced by year, indicating that richness and diversity fluctuated only slightly between years. Highly significant influences of month and year were found for all traits investigated with the exception of overwintering stage. While this indicates considerable seasonal and annual fluctuations in the functional makeup of carabid assemblages, such temporal fluctuations are outside the focus of the analyses reported here.

Influence of vegetation height

Strong negative effects of vegetation height were found for species richness, Shannon diversity and activity abundance. With the exception of breeding season, vegetation height significantly influenced all traits, with the strongest effects being found for dispersal ability and diel activity (Table 4). The proportion of carabid individuals that overwinter as adults only, and have 1-year life cycles, increased with vegetation height, while the proportion overwintering as adults and larvae or larvae only, and with 2-year life cycles, decreased. Greater vegetation height increased the proportion of nocturnal carabids compared to that of diurnal carabids and carabids with no diel preference; and increased flightless carabids compared to carabids capable of flight and with dimorphic wings. For body form, the proportion of pushers increased with vegetation height compared to that of runners and diggers.

When investigating the influence of vegetation height on carabid size class, the largest trend was that increased vegetation height increased the proportion of very small carabids compared to the proportion of small and of medium carabids. The proportion of large compared to medium carabids also increased with vegetation height. Finally, when looking at the influence of vegetation height on the diet of adults, the main trend observed was that the proportion of carabids with a mostly plant diet increased with vegetation height compared to proportions in the other three dietary groups.

Influence of field type

The majority of traits investigated showed no evidence of Field type effects. While marginally significant effects were found for dispersal ability and body form, effects for dispersal ability could be solely attributed to vegetation height. For body form, differences were detected between silage/grazed fields and hay fields, but note that only three sampling sites were situated in hay fields.

Interaction between riparian management and month

While no significant riparian management effects were found for breeding season, overwintering stage and life cycle, clear interactions between riparian management and month were detected (Table 4). In June–July, Wide Margin and Wide Middle transects had a higher proportion of individuals overwintering as adults only and a lower proportion overwintering as adults and larvae or larvae only when compared to Narrow Margins, Narrow Fields or Wide Fields. The same trend was not detected in July–August. In June–July, the proportion of spring/summer breeders was lower, while the proportion of autumn/winter breeders was higher in Wide Field transects than Narrow or Open Field transects. This difference was not found in July–August.

Discussion

Impact of riparian management on functional traits

While carabid assemblages in fenced riparian margins were less diverse and species rich than in open margins and grassland fields, erecting fences adjacent to watercourses changed the functional structure of assemblages. The inclusion of these additional habitats in the intensive grassland landscape therefore enhanced the functional structure of carabid assemblages at the farm scale. Excluding livestock from field margins influences habitat structure at a range of spatial scales. At the micro-habitat scale, the lack of grazing results in taller, denser vegetation (McCracken et al. 2012) which provides a more humid and stable micro-climate. At the local scale, the vegetation is both taxonomically and architecturally more diverse, supporting a wider range of plant species and structures (i.e. grassy tussocks, dead wood, flowers and seed heads), thus benefitting a greater diversity of phytophagous insects, pollinators and their predators (Woodcock et al. 2009; Cole et al. 2012). Finally, at the landscape scale, fenced margins create habitat heterogeneity within intensive agricultural landscapes and ecological corridors that connect semi-natural habitats (Wamser et al. 2012). Enhancing habitat heterogeneity at the micro, local and landscape scales benefits a range of farmland taxa and enhances their associated ecosystem services.

Various studies have investigated the influence of margin width on biodiversity with inconclusive results. Fritch et al. (2011) found no impact of margin width on plant species richness, and Telfer et al. (2000) found that vegetation structure had a greater influence on carabid species richness than width. Perry et al. (2011) found impacts of width to be species specific for birds with some species favouring wider margins while others favoured narrow margins. In the current study area, Cole et al. (2012) found that while fencing off watercourses increased the activity abundance of many invertebrates (e.g. Heteroptera, Cicadellidae, Cercopidae and Opiliones), impacts of margin width were not apparent. In agreement with these findings, we found margin width had little influence on carabid activity abundance, species richness or diversity. However, clear influences of width were found when examining carabid functional structure. The distribution of trait attributes of carabids in narrow margins was more similar to intensively managed grassland fields than wide margins. This is in agreement with previous findings that indicate that the taxonomic structure of carabid assemblages differed most between wide riparian margins and grassland fields (Cole et al. 2008).

Wide margins, irrespective of whether they are situated adjacent to watercourses or not, are likely to be superior to narrow margins at buffering against the environmental disturbances that typically accompany intensive grassland (e.g. livestock grazing, slurry application and regular mowing) or arable (e.g. cultivation and agrochemical application) management. Wide margins are thus likely to provide more stable habitats than narrow margins. Ecological theory states that stable habitats are best exploited by large immobile species with long life-cycles while unstable habitats favour small, mobile species with short life cycles. Previous studies have found large, flightless Carabus spp. cannot cope with the sudden changes in habitat structure that accompanies intensive agricultural management (Ribera et al. 2001; Cole et al. 2002). There was, however, no evidence that wide margins (i.e. the habitat predicted to be most stable) had a higher proportion of larger carabids or carabids with 2-year life cycles. Furthermore, Carabus spp. were rare throughout the study area (i.e. represented by only 10 individuals) possibly due to the intensity of management in the surrounding landscape. Contrary to the predictions of ecological theory, wide margins contained the highest proportion of very small carabids. Small hygrophilous Bembidiini species are inherent elements of riparian assemblages and Bembidion tetracolumn Say and Bembidion mannerheimii Sahlberg were frequently recorded in wide margins. This suggests that wide riparian margins provide an important refuge for hygrophilious Bembidiini which are thought to be particularly prone to extinction (Lambeets et al. 2008).

In agreement with ecological theory, there was a higher proportion of flightless carabids in wide margins than in narrow margins, open margins or grassland fields. This is in agreement with previous research indicating positive associations between the proximity and continuity of semi-natural habitats within intensive agricultural landscapes and the diversity of carabids with low dispersal ability (Hendrickx et al. 2009; Wamser et al. 2012). This indicates that semi-natural components are particularly important in harbouring species with poor dispersal ability which are especially vulnerable to habitat fragmentation. The importance of semi-natural components within intensive agricultural landscapes in maintaining farmland biodiversity, and enhancing habitat connectivity, has been highlighted across a range of taxa including plants, arthropods and birds (Benton, Vickery & Wilson 2003).

Impact of vegetation on functional traits

Carabids have been classified as pushers, runners or diggers based on their body form and locomotory ability (Forsythe 1983). Pushers are designed to push through vegetation while runners prefer a more open habitat. A higher proportion of pushers, compared to runners, occurring in longer vegetation may therefore be expected. The proportion of pushers was also greater in wide margins than any other situation, a difference that could not simply be attributed to differences in vegetation height. The deeper litter layer in wide margins may have favoured pushers over runners, but it may also have influenced the efficiency of pitfall traps with pushers being more mobile and thus more efficiently sampled (Lang 2000).

Vegetation structure influenced diel activity in carabids with tall, dense vegetation, having higher proportions of carabids that hunt nocturnally using tactile cues, while short more open vegetation had higher proportions of carabids that hunt diurnally using visual cues. Similarly, bird species that detect prey visually (e.g. lapwing) prefer to forage in more open habitats (Vickery et al. 2001). Fencing off margins not only influences the visibility and accessibility of prey, but also its availability to birds with different feeding strategies. Fenced margins support higher densities of prey preferred by foliage gleaning birds while fields and open margins support higher densities of prey preferred by ground feeding birds (Cole et al. 2012). Impacts of riparian management on diel activity in carabids were not simply the result of differences in vegetation height, and fenced margins also provide a greater array of harbourage (e.g. grassy tussocks, leaf litter and rocks) to provide shelter for nocturnal species during the day.

In addition to circadian differences in habitat requirements, many species show distinct seasonal requirements (e.g. to provide food, shelter and breeding sites). The greater array of harbourage available in fenced margins provides aerated and stable microclimates which are utilised as overwintering sites by polyphagous predators (Pfiffner & Luka 2000). This study indicates that wide margins were important in providing harbourage for carabids that overwinter as adults and that these carabids dispersed from margins to adjacent fields during the summer. As riparian margins create habitat continuity between fields at the landscape level, they not only have the potential to enhance populations of key predators in adjacent arable fields, where they play an integral role in controlling pests, but also to facilitate their movement within the wider landscape.

Limitations and applications of functional trait analyses

In this study, differences in pitfall trapping efficiency due to vegetation structure were minimised by focusing on analyses comparing proportions of individuals rather than counts, and by looking at management effects following the removal of vegetation height effects. It could be argued that results based on the proportions of individuals may have been driven by the response of a single dominant species, or a single genus. We have consequently avoided concluding from analyses based on proportions of individuals that riparian management favours species with a particulate trait attribute. While our replicated field trial provides evidence of consistent effects of management on the balance of attribute levels for several traits, it is important to note that these effects are taking place in the context of the species present in the community. There are, however, two factors that suggest our results will hold more widely. First, for most traits, trends based on the proportion of species largely reflected those based on the proportion of individuals, thus indicating that the detected responses occurred in multiple species. Secondly, carabids show a high degree of homoplasy and, with the exception of the trait attributes mixed diet and digger (occurring in one and two genera, respectively), our trait data were relatively well balanced across genera (Table 2). Broader conclusions about management effects on a notional population of species could be achieved by conducting similar experiments in other geographical regions.

Analyses based on the taxonomic and functional structure of carabid assemblages were more sensitive for detecting impacts of agricultural management than those based on traditional diversity measurements (e.g. species richness and Shannon diversity). This is likely to be the case for other taxa (e.g. farmland birds, pollinators and flowering plants), thus adding support to the conclusions of Wamser et al. (2012) that it is important to consider a range of assemblage attributes when investigating impacts of agriculture on biodiversity to avoid drawing erroneous conclusions. Furthermore, Mokany, Ash & Roxburgh (2008) found that functional measurements of plant communities were superior in determining ecosystem processes to traditional diversity measurements, with species richness being a particularly poor determinant. Functional approaches therefore not only assist in determining the principal environmental and land management factors driving assemblages, but also in predicting subsequent impacts on ecosystem functioning.

Implications for land management

The erection of fences adjacent to watercourses to mitigate diffuse pollution from agriculture is likely to become more prevalent worldwide as countries strive to meet increasingly stringent objectives to improve water quality (e.g. Water Framework Directive in Europe and Clean Water Act in the USA). Riparian margins, particularly those over 5·4 m wide, enhanced the functional structure of carabid assemblages at the landscape scale. Wide riparian margins are also potentially superior to narrow margins in delivering a range of ecosystem services in intensive grassland and arable situations including; mitigating diffuse pollution, enhancing the ecological status of the watercourse with respect to aquatic invertebrates and provisioning of pollinators (Olson & Wäckers 2007; Greenwood et al. 2012). Through the creation of corridors of stable semi-natural habitats within intensive agricultural landscapes, wide margins provide a refuge for species with low dispersal ability thus maintaining habitat continuity and mitigating the impacts of isolation (Hendrickx et al. 2009; Wamser et al. 2012). Contiguous riparian margins also enhance the ecological status of watercourses at the catchment scale and provide corridors for the dispersal of winged adult stages of aquatic invertebrates (Greenwood et al. 2012). Discontinuity, and possibly even narrowing, of riparian margins should therefore be avoided, and margins should preferentially be situated adjacent to other semi-natural components within the landscape.

There is concern that widespread fencing adjacent to watercourses will result in the disappearance of exposed habitats required by some invertebrate riparian specialists (Alexander, Foster & Sanderson 2010). Furthermore, as many declining farmland species are intimately linked to agricultural practices the implementation of management options to open up the vegetation structure (e.g. restricted grazing, scarification or mowing) is likely to favour a wider suite of species including pollinators, foraging birds and flowering plants (McCracken et al. 2012). Such management options, however, create habitat instability, and thus, if implemented at too large a scale, could adversely impact on species with low dispersal ability. Management options should therefore be implemented at a sufficiently small spatial scale to allow recolonisation from adjacent undisturbed areas (e.g. only part of any continuous margin should undergo management within any 1 year). Furthermore, the implementation of management practices should not unduly conflict with the margin's ability to mitigate diffuse pollution (e.g. to minimise the risks of faecal contamination of downstream bathing waters, livestock grazing of margins should not coincide with the bathing season).

It is important to balance the potential ecological and environmental benefits of agri-environment options with the economic sustainability of taking land out of production. Woodcock et al. (2009) suggested that sacrificing a single field of equivalent area may be more financially viable than fencing off large areas of field margins. Landowners in the UK, however, have a tendency to prefer agri-environment options that cause minimum disruption to infield management practices (Davey et al. 2010). Many species inhabiting agricultural land require different environmental resources over time, and consequently habitat heterogeneity is key, both at the local level to support circadian differences in habitat requirements, and at the landscape level to provide seasonal requirements. The failure of current agri-environment schemes to consider the spatial arrangement of environmental resources at the landscape level, and thus to ensure that all necessary resources occur within a species' range, is a primary reason for their lack of success (Kleijn & Sutherland 2003; Firbank et al., in press). Enhancing heterogeneity and habitat continuity by directing conservation efforts towards the establishment of wide fenced riparian margins can assist land managers in integrating agronomic, biodiversity and diffuse pollution objectives within intensive agricultural landscapes. It is, however, essential to carefully balance management actions to ensure that a wide range of terrestrial and aquatic taxa are supported.

Acknowledgements

We are indebted to the farmers of Ayrshire who provided essential feedback and access to their land. SAC and Biomathematics & Statistics Scotland received financial support for this study from the Scottish Government. We are also extremely grateful to Martin Luff for his input into the carabid functional data and to two referees for their constructive comments on an earlier version of the manuscript.

Ancillary