Farm-scale spatiotemporal dynamics of predatory beetles in arable crops


  • J. M. HOLLAND,

    Corresponding author
    1. The Game Conservancy Trust, Fordingbridge, Hants SP6 1EF, UK; and
    2. Seale-Hayne Agricultural Ecology Group, School of Biological Sciences, University of Plymouth, Seale-Hayne, Newton Abbot, Devon TQ12 6NQ, UK
      Correspondence: Dr J. M. Holland, The Game Conservancy Trust, Fordingbridge, Hants SP6 1EF, UK (fax +44 1425 651026; e-mail
    Search for more papers by this author
  • C. F. G. THOMAS,

    1. Seale-Hayne Agricultural Ecology Group, School of Biological Sciences, University of Plymouth, Seale-Hayne, Newton Abbot, Devon TQ12 6NQ, UK
    Search for more papers by this author

    1. The Game Conservancy Trust, Fordingbridge, Hants SP6 1EF, UK; and
    Search for more papers by this author

    1. The Game Conservancy Trust, Fordingbridge, Hants SP6 1EF, UK; and
    Search for more papers by this author
  • H. OATEN

    1. The Game Conservancy Trust, Fordingbridge, Hants SP6 1EF, UK; and
    Search for more papers by this author

Correspondence: Dr J. M. Holland, The Game Conservancy Trust, Fordingbridge, Hants SP6 1EF, UK (fax +44 1425 651026; e-mail


  • 1The spatial dynamics of farmland invertebrates can provide essential information relevant to their management for pest control and biodiversity conservation in sustainable agriculture. Carabid beetles are one of the most important groups contributing to biological control in arable fields. Previous studies have focused on spatial dynamics within single fields and years. In this study we examined their larger scale, long-term dynamics, thereby taking into account the impact of changes in crop rotation and the influence of field size.
  • 2The spatial distributions of four beetle species were investigated at an unprecedented spatial scale in a grid of 973 pitfall trap locations across six fields encompassing 64 ha of arable land. Week-long trapping was conducted four times in the first year and twice in the two following years.
  • 3All species showed strong aggregation but the size and location of patches differed among species. The distribution of Pterostichus melanarius was stable within and between years, with a single large patch close to the field boundaries. Patches of Poecilus cupreus were also located close to field boundaries but their location changed between years. Pterostichus madidus and Philonthus cognatus distributions extended across field boundaries and were less stable, with patch locations changing between years.
  • 4Synthesis and applications. The spatial extent of a population patch for a given species was species-specific. Species overwintering in field boundaries remained in proximity to these throughout the summer, whereas patches of mid-field overwintering species were more extensive. Patches were generally stable within years but varied for some species between years. Species therefore differ in their response to crop management practices and consequently blanket management approaches for these important generalist predators of crop pests are inappropriate. For spatially stable species (e.g. Pterostichus melanarius) it may be possible to determine their specific habitat requirements and to devise predictive and protective measures to preserve populations or manipulate them at the farm-scale. More mobile species may be better at responding to pest aggregations at the farm-scale. However, operations that deplete populations, e.g. soil cultivations, should be spatially and temporally desynchronized at the farm-scale to conserve populations and enable functional biocontrol.


Invertebrates, especially insects, perform important roles in agriculture as pests, natural agents of pest control, elements of food chains, major contributors to biodiversity and general indicators of ecosystem health. The implementation of crop production methods that incorporate integrated pest management and other concessions to conserve biodiversity remains a central challenge for sustainable agriculture. The progressive development and application of practical techniques for managing insect populations depends on detailed knowledge of the basic ecology of individual species. This information, however, is relatively scant for even the most common species of interest in agricultural habitats. With the notable exception of a few long-term or large-scale studies (den Boer 1977; Aebischer 1991), most published work has been acquired in small-scale, short-term field and laboratory studies.

As theory advances it is increasingly recognized that large-scale spatial processes are of key importance in the ecology of invertebrates in agricultural habitats. Agricultural landscapes are now seen as, and have been modelled as, complex shifting mosaics of habitat patches in which a metapopulation (Levins 1969) of insects may be structured from a series of local populations, each predominantly confined to an individual field and subject to the range of agricultural operations that happen within it (Sawyer & Haynes 1985). For non-flying epigeal insects, dispersal between local populations that maintains the metapopulation is dependent on field boundary permeability, and has been shown to have a theoretical optimum depending on the frequency and intensity of pesticide use on the farm (Sherratt & Jepson 1993).

However, the paradigm of metapopulations fragmented into local populations within fields remains attached to the notion that the field is the key spatial unit, and its management the most important factor influencing population persistence. Consequently most studies of farmland invertebrates have focused on single fields. In reality, however, the field is a more or less arbitrary human-made unit superimposed on a natural topography, geology and hydrology from which important factors such as soil type, structure and drainage are derived. The natural bounds of an insect local population in farmland therefore may not necessarily be related only to the fragmentation of land by field boundaries. Although some field boundaries may be formed from natural features, they are mainly formed from other managed non-crop habitats such as hedgerows, fence lines and uncultivated strips of grassland and, occasionally, small plots of woodland or shelter-belt. The permeability of each boundary type separating adjacent fields is likely to be characteristic in its effect on the spatial dynamics of a given species and its metapopulation structure. Furthermore, barriers to dispersal between fields can affect gene flow (Frankham 1995), slow down re-invasion following adverse agricultural operations (Sherratt & Jepson 1993) and retard biological control where an ability to track pest infestations improves the chances of success (Murdoch & Briggs 1996).

Boundary habitats may also have positive roles, for example as stable habitats that are seasonally essential for aestivation or overwintering by some species (Sotherton 1984). Recolonization of the annually disturbed cropped areas follows when conditions become more suitable (Wissinger 1997). For species that only or predominantly inhabit boundary habitats, these features may act as corridors to movement and dispersal between non-agricultural areas. Non-crop habitats also fragment landscapes and understanding the ecology of these processes is critical if policies are to be based upon sound science and good management practices (Hunter 2002; Chackoff & Aizen 2005; Schweiger et al. 2005).

Quantifying the spatial distribution of an insect population and its relationship or otherwise with natural environmental factors, human-made features such as field units and field boundaries can reveal much ecological information of potential use to the management of the species. Traditional studies have quantified differences by taking a few random samples from respective habitats or treatments under comparison. However, these types of study are devoid of informative spatial data on the extent of species population distributions and associations with, often continuously varying, environmental factors.

Little is known about the temporal stability of spatial pattern (Hunter 2002), especially in agricultural habitats where disturbance and habitat type are constantly changing. However, the spatiotemporal dynamics of epigeal insect distributions have been examined recently using extensive and frequently sampled grids of traps in parts of fields, whole single fields or pairs of neighbouring fields, sometimes in conjunction with sampling other abiotic and biotic factors (Ericson 1978; Hengeveld 1979; Thomas, Parkinson & Marshall 1998; Holland, Perry & Winder 1999; Bohan et al. 2000; Brown 2000; Fernández-García, Griffiths & Thomas 2000; Thomas et al. 2001; Holland et al. 2004b; Pearce & Zalucki 2005). Most of these studies have been conducted in cereal crops and focused on Carabidae, one of the most important groups of farmland invertebrates, captured by pitfall trapping. At the spatial scales examined so far it has been shown that adult stages of carabids exhibit discrete and distinct spatial distributions. In some cases, patches of high population density have been shown to be quite spatially stable within years, with patches extending to several hectares. However, the true extent of population patches, and how non-crop habitats and their distribution in fragmented farmland affect the spatial dynamics of beneficial insects at scales beyond that of the field, have always been limited by the extent of the sampled area. Moreover, the stability of spatial distributions between years has not been investigated, although there is evidence that invertebrate numbers fluctuate within the same fields between years (Thomas, Holland & Brown 2002). In agricultural habitats substantial changes occur each year as a consequence of crop rotations, harvest and soil cultivations that may impact on invertebrate survival, leading to the creation of spatial pattern at farm level and beyond.

This paper reports findings from the 3D Farming Project, which began in 2000 with the overall aims of investigating ways to increase biodiversity on farmland and manipulating beneficial insects to provide more effective aphid control. One of the main components investigated the spatial–temporal dynamics of predatory insects in terms of the distribution, density and dispersal (3D) of populations, and aimed to extend earlier work to a spatial scale approaching that of a whole farm. Using an extensive grid of traps the study yielded population (activity–density) data in two spatial dimensions across a large area of a single farm. Three-dimensional (3-D) contour plots of population activity–density enabled us to address questions concerning the spatial distribution and temporal stability of populations and how they might be influenced by field boundaries and other factors. This approach allowed us to verify questions with respect to metapopulation theory and derive information important for the development of conservation biocontrol in arable crops. These included the following. To what extent and at what scale are epigeal insect distributions aggregated across a contiguous block of fields? How stable are aggregations within and between years? Do the field boundaries and cropping affect abundance and distribution within and across fields?

Materials and methods

The study site in Dorset, UK, comprised a 64-ha block of six arable fields separated from each other by various field boundary structures (Fig. 1). The block was surrounded by arable crops, except on one side where there was permanent pasture. There was a cereal–cereal–pea rotation over the 3 years of sampling. The soil was cultivated in the autumn prior to drilling of winter wheat but for spring-sown cereal and peas the stubble of the previous crop was retained through the winter and cultivations were conducted in late winter. The majority of field boundaries were hedges with a herbaceous–grass bank, although in many places the dominant plant species were Bromus sterilis L. and Urtica dioica L. A grid of 973 sampling locations was established across the study site, arranged in an offset grid pattern with 40-m spacing along the rows and 20-m spacing between rows within each field. The grid extended across the whole of the 64 ha, encompassing the entirety of three small fields (S1–3) and half the area of each of three large fields (L1–3). Each sample location was surveyed and located using the UK Ordnance Survey national grid reference using a differential global positioning system (Trimble Geoexplorer 3, Trimble Navigation Ltd, Sunnyvale, California, USA). At each sampling location two 6-cm diameter pitfall traps (half-filled with a 50% solution of ethylene glycol and detergent) were positioned 2 m apart and the data pooled for analyses.

Figure 1.

Study area and cropping each year.

To examine the distribution of epigeal beetles across the study area within the year, pitfall traps were opened for 1 week on four occasions in 2000 (2–9 May, 6–13 June, 28 June−5 July and 12–19 July). These data was then used to identify when beetle abundance and diversity were highest, allowing optimization of the sampling strategy in the following years. Trapping was conducted over two further periods in the succeeding years: 4–11 June and 9–16 July 2001; 10–17 June and 8–15 July 2002. The majority of the catch comprised carabid beetles (Coleoptera; Carabidae) and rove beetles (Coleoptera; Staphylinidae), which were identified to species. We present here the spatiotemporal distributions of three of the most important, common and abundant Carabidae in agricultural habitats, Pterostichus madidus Fbr., Pterostichus melanarius Ill. and Poecilus cupreus L., and one staphylinid species, Philonthus cognatus Steph. These species are all predatory and contribute to biological control (Good & Giller 1991; Sunderland 2002); however, the dietary range of the larger Pterostichus species is broader. Their biology differs to some extent, as described in Table 1.

Table 1.  Key biological features of the four beetle species
SpeciesOverwintering stage and siteBreeding periodEnvironmental conditionsMethod of dispersal
  • *

    Small proportion of the population survive through into following year.

  • Dimorphic, some individuals capable of flight.

Pterostichus madidus Larvae, field (adult, boundary)*AutumnHygrophilic, eurythermousWalking
Pterostichus melanarius Larvae, field (adult, boundary)*AutumnHygrophilic, eurythermousWalking
Poecilus cupreus Adult, boundary and field*SpringXerophilic, warm preferrentWalking
Philonthus cognatus Larvae, fieldAutumnUnknownWalking/flight

The spatial distribution of these beetles was analysed using SADIE (Spatial Analysis by Distance IndicEs; Perry et al. 1999), termed ‘red–blue’ analysis, for each sampling occasion. This calculates the degree of clustering in the form of (i) ‘patches’ of large counts, using the overall index and its associated probability Pi, or (ii) ‘gaps’ of small counts, using the overall index and its associated probability Pj (Perry et al. 1999). The null hypothesis of spatial randomness is indicated if, for a particular set, all of these indices have values around unity. A value of at least one index above unity suggests spatial non-randomness of some form; large positive values indicate patchiness (vi > c. 1·5) and large negative values (vj < c. −1·5) indicate membership of a gap (Perry et al. 1999). Distribution data are presented as two-dimensional contour maps of the cluster indices where vi > 1·5 and vj < −1·5, drawn using the package Surfer for Windows Version 6.04 (Golden Software Inc., Golden, Colorado, USA). Analyses were conducted on data from across the whole study area.

To test whether two data sets were spatially correlated, the correlation coefficient, X, between the clustering indices of each set was calculated according to the method described by Perry & Dixon (2002). Hence, if the indices of set one are denoted zi1, with mean q1, and those of set two zi2, with mean q2, then a measure of local spatial association for position i is given by:

χi = n(zi1 − q1)(zi2 − q2)/[Σi(zi1 − q1)2 Σi(zi2 − q2)2]1/2

The overall spatial association is the mean of these local values, X = Σi χi  n. The significance of X was tested against values Xrand from a randomization test that included a Dutilleul (1993) adjustment procedure to provide a probability value, PD. A positive correlation coefficient indicates that the two data sets are associated with patches coinciding spatially. A negative coefficient shows that they are disassociated with patches occurring in different locations to each other. This approach was used to test for association within years and between years for each species and for between species on each sampling occasion.


pterostichus madidus

During May 2000 Pterostichus madidus was present only in low numbers (Fig. 2), with the catches almost entirely composed of overwintered adults having emerged in the late spring. By 6 June, newly emerged adults (tenerals) of the next generation had started to appear in areas different from those where the overwintered adults had previously been captured (Fig. 3a–d). The distributions of beetles in these first two sampling occasions were therefore significantly disassociated (negative ) but were spatiotemporally stable (positive X   ) from 13 June until 19 July (Table 2). By 5 July (Fig. 3c) large numbers of tenerals were caught in three of the pea fields, S2, L3 and S3. These patches persisted in the same location until 19 July (Fig. 3d). The SADIE red–blue analysis of Pterostichus madidus revealed the presence of significant spatial structure on all sample dates and this increased between successive sampling occasions, although the numbers captured started to decline on the last sampling occasion (Fig. 2). The largest patches of Pterostichus madidus were found during the period 28 June−5 July within S3 (Fig. 3c), where complete coverage occurred, in S2, where a patch covered approximately 6 ha, and in L3, where patches covered three-quarters of the sampled area. Additional small patches were found within S1 and L2 on 19 July.

Figure 2.

Mean clustering indices (left y-axis) showing level of aggregation into patches of higher than average density () and gaps of lower than average density () and mean number (± 1 SE) per pair of pitfall traps (right y-axis) on each sampling occasion for Pterostichus madidus, Pterostichus melanarius, Poecilus cupreus and Philonthus cognatus.

Figure 3.

Spatial clustering for Pterostichus madidus across the study area for total beetles caught per pair of pitfall traps at each location for (a) 2–9 May 2000; (b) 6–13 June 2000; (c) 28 June−5 July 2000; (d) 12–19 July 2000; (e) 4–11 June 2001; (f) 9–16 July 2001; (g) 10–17 June 2002; (h) 8–15 July 2002. Maps indicate clusters of relatively large counts (denser contour pattern ‘patches’, for which vi > 1·5) and small counts (lighter contour pattern ‘gaps’, for which vj < −1·5).

Table 2.  Association indices comparing distribution of the insect species between sampling periods within years and between years for June and July ( *** P D  < 0·001 or > 0·999, ** P D  < 0·01 or > 0·99, * P D  < 0·05 or > 0·975)
  Pterostichus madidus Pterostichus melanarius Poecilus cupreus Philonthus cognatus
Within years
2–9 May 2000 −0·28 *** 0·57 ***  0·62 ***  0·56 ***
6–13 June 2000  0·68 *** 0·88 ***  0·61 *** −0·02 
28 June−5 July 2000  0·83 *** 0·91 ***  0·47 ***  0·28 ***
12–19 July 2000 −0·37 *** 0·52 ***  0·48 ***  0·14 **
4–11 June 2001  0·25 *** 0·70 ***  0·48 *** −0·10 
9–16 July 2001  0·72 *** 0·80 ***  0·07  0·02 
Between years
June 2000–01  0·43 *** 0·75 ***  0·28 *** −0·44 ***
June 2001–02  0·38 *** 0·77 ***  0·03  0·28 **
July 2000–01< 0·001 0·83 *** −0·18 **  0·09 
July 2001–02  0·65 *** 0·77 ***  0·15 *  0·13 **

Fewer Pterostichus madidus were caught in the following 2 years but their distribution was still strongly aggregated (Fig. 2), although this declined from June to July in each year. In June 2001 a large patch of Pterostichus madidus covered L3 and small patches occurred in all the other fields, but there were more in the wheat fields (Fig. 3e). By July their location had changed, with the two largest patches occurring in L3 (wheat) and L2 (peas) covering 15 ha, and parts of L1 (peas) (Fig. 3f). The patches persisted in L2 and L3 through 2002, covering 20 ha (Fig. 3g,h and Table 2). As the trapping grid covered only half of these fields, the patches may have extended further.

pterostichus melanarius

Fewer Pterostichus melanarius were captured than Pterostichus madidus, but the population size (activity–density) of the former reached its peak earlier in the summer (Fig. 2). This species showed a much more restricted distribution, with a 10-ha patch spreading across S1 and along one edge of S2 by June 2000 (Fig. 4a,b). The patches were found in the same location on all subsequent sampling occasions (see Appendix S1) and consequently had high spatiotemporal stability, even though the cropping changed (Table 2). The spatial statistics were highly significant on all sampling occasions.

Figure 4.

Spatial clustering for Pterostichus melanarius for (a) 2–9 May 2000; (b) 6–13 June 2000. Notation as in Fig. 3.

poecilus cupreus

Poecilus cupreus was the least numerous of the species but showed strong evidence of spatial pattern in the first 2 years (Fig. 2). The distribution was similar to that of Pterostichus melanarius during 2000 (Fig. 5a–d), with an 8-ha patch occurring along the edge of the three small pea fields, and these remained stable through the year (Table 2). By June 2001 Poecilus cupreus had become more widespread although still remained in proximity to the field boundaries (Fig. 5e). By July 2001 small patches were found only in those fields growing peas, L1 and L2. In June 2002 there was no significant spatial pattern but some had formed by July, although the patches were small (Fig. 5g) and there was only weak association with their location in 2001 (Table 2).

Figure 5.

Spatial clustering for Poecilus cupreus for all sampling occasions. Notation as in Fig. 3.

philonthus cognatus

Philonthus cognatus numbers were lower in July compared with June each year (Fig. 2), indicating loss of the species from the monitored area either through mortality or dispersal. The distribution of patches was more ephemeral than for the other species within 2000, with considerable variation in the level of association between sampling occasions (Table 2). Large patches were found in May 2000 covering the cereal fields L1 (13 ha) and half of L2 (9 ha) (Fig. 6a). By 5 July the patches were identified covering two of the pea fields, S1 (4 ha) and most (9 ha) of S2 (Fig. 6c), and on 19 July were in L1 and S2 (Fig. 6d). In June 2001 large patches covered most of S1, S2 and L3, but by July these were restricted to L1 (Fig. 6e,f). Patches occurred in L1, S1 and S2 in June 2002, although these only persisted through to July in L1. The greatest change in distribution occurred between 2000 and 2001, with disassociation for the pattern in June and no significant association for July (Table 2).

Figure 6.

Spatial clustering for Philonthus cognatus for all sampling occasions. Notation as in Fig. 3.

association between species

The distribution of Poecilus cupreus was positively associated with that of Philonthus cognatus and Pterostichus melanarius on most sampling occasions (Table 3). The latter two species were associated with each other on most sampling occasions, although during 2001 and 2002 there was also dissociation. Pterostichus madidus showed no clear pattern of association with Philonthus cognatus or Poecilus cupreus but was disassociated with Pterostichus melanarius.

Table 3.  Association indices comparing the distribution of the four insect species on each sampling occasion ( *** P D  < 0·001 or > 0·999, ** P D  < 0·01 or > 0·99, * P D  < 0·05 or > 0·975)
  Philonthus cognatus Poecilus cupreus Pterostichus madidus
2–9 May 2000
Poecilus cupreus −0·03  
Pterostichus madidus  0·35***−0·02 
Pterostichus melanarius  0·31*** 0·41*** 0·17*
6–13 June 2000
Poecilus cupreus  0·16*  
Pterostichus madidus −0·31*** 0·31*** 
Pterostichus melanarius  0·38*** 0·65*** 0·11
28 June−5 July 2000
Poecilus cupreus  0·42***  
Pterostichus madidus  0·36*** 0·53*** 
Pterostichus melanarius  0·32** 0·51*** 0·15
12–19 July 20/00
Poecilus cupreus  0·13*  
Pterostichus madidus  0·03 0·38*** 
Pterostichus melanarius  0·31*** 0·53*** 0·10
4–11 June 2001
Poecilus cupreus −0·34***  
Pterostichus madidus  0·53***−0·35*** 
Pterostichus melanarius  0·07 0·30***−0·17**
9–16 July 2001
Poecilus cupreus  0·37 ***  
Pterostichus madidus  0·22 ***−0·08 
Pterostichus melanarius −0·07 0·19 **−0·53 ***
10–17 June 2002
Poecilus cupreus  0·28***  
Pterostichus madidus −0·43***−0·09 
Pterostichus melanarius  0·47*** 0·16*−0·59***
8–15 July 2002
Poecilus cupreus  0·16**  
Pterostichus madidus  0·27*** 0·22* 
Pterostichus melanarius −0·20**−0·0739−0·60***


evidence of spatiotemporal pattern

The spatial scale, extent and duration of this study exceeded all previous investigations of this type and allowed us to answer some key questions regarding the spatiotemporal dynamics of epigeal insects while testing for the existence of metapopulations. The four epigeal beetles showed strong evidence of spatial pattern when examined across the contiguous block of six fields. However, the patch size within each field and across the site varied among the species; for example, the patches of Pterostichus melanarius covered approximately 10 ha whereas those of Pterostichus madidus were interlinked, extending across several fields, and covered 23 ha in July 2000. This is much larger than previously found for these species in arable fields (Thomas, Parkinson & Marshall 1998; Holland, Perry & Winder 1999; Brown 2000), reflecting the scale of this compared with previous studies.

Although occasional individuals could occur anywhere in the grid, the main aggregations of different species occupied different areas of the study site. This has been shown previously in smaller scale studies of some of these species. Pterostichus melanarius and Poecilus cupreus were found to be spatially separated, occurring in different fields (Thomas et al. 2001; Winder et al. 2001) or in different parts of the same field; Poecilus cupreus remained close to hedgerows, while the reverse was found for Pterostichus melanarius (Fournier & Loreau 1999; Winder et al. 2001). However, in a larger scale study patches of both species were found at the edge and in the centre of fields (Brown 2000). A number of putative mechanisms may cause this spatial differentiation, including preferences among species for slightly different environmental conditions; active avoidance among congenerics; competitive exclusion; mutual predation; differential effects of historical and current management within fields; or those abiotic and biotic factors influenced by such management, especially prey availability and soil moisture (Thiele 1977; Irmler 2003). Most probably a combination of positive and negative mechanisms will drive the spatial dynamics (Thomas, Holland & Brown 2002). Moreover, the apparent drivers may change according to the spatial resolution of the study. At the subfield level, crop cover determined the spatial pattern of Carabidae (Greenslade 1964; Best et al. 1981). At the field scale, soil moisture, prey abundance, crop type and weed cover were correlated with their spatial distribution (Hengeveld 1979; Holland, Perry & Winder 1999; Bohan et al. 2000; Winder et al. 2001; Thomas, Holland & Brown 2002). When distributions across the landscape were considered, moisture gradients, soil type, soil pH and field size were important (Judas & Schaefer 2002), as was connectivity of semi-natural elements (Schweiger et al. 2005). The scale, intensity and frequency of sampling may therefore determine whether relationships to factors responsible for spatial pattern are detected. Some adjustment may also be necessary according to the hypothesis being tested and the likely dynamics of any distribution. For most carabids a 30-m grid spacing was judged sufficient (Holland, Perry & Winder 1999), although for larger carabids that aggregate in larger patches this spacing can be increased, as in this study (Pearce & Zalucki 2005).

The four species exhibited stable distribution patterns within each year, with patches and gaps generally remaining in the same parts of the study area. This corroborates the findings of most other studies of within-year distributions (Thomas, Parkinson & Marshall 1998; Holland, Perry & Winder 1999; Fernández-García, Griffiths & Thomas 2000; Thomas et al. 2001). However, in other studies using grids smaller than a single field, Pterostichus melanarius has been shown to be more mobile within the year, moving in response to the distribution of slugs (Bohan et al. 2000) or aphids (Winder et al. 2001). Redistribution of carabids from June to August was likewise found using a grid-sampling approach that encompassed two pairs of fields (Brown 2000; Thomas, Holland & Brown 2002). In that case captures increased in a bean and declined in a wheat crop during July and August. The temporal variation in species distributions between June and July found in this study may be explained by differences in the numbers of adult overwinterers and emerging tenerals dominating the catch. For example, in June Pterostichus madidus was more abundant in wheat than peas but the reverse was found in July. In June the adults that had survived the winter dominated the catch, whereas by July most of the beetles caught were tenerals. This may also have held true for Philonthus cognatus.

In contrast to the spatial stability of distributions within a year, the location of patches between years varied among the four species. Distributions may change either as a consequence of differences in survival between locations or because of movement. Patches of Pterostichus melanarius remained in proximity to field boundaries in the smaller fields over the 3 years and mark–release–recapture studies conducted within the same fields confirmed that there was little movement outside these areas (Holland et al. 2004a). In contrast, patches of Poecilus cupreus changed between years even though this species also utilizes field boundaries as overwintering and summer foraging sites. Instead, they appeared to follow the pea crops in rotation. In 2002 Poecilus cupreus virtually disappeared when no peas were grown, suggesting that conditions directly related to the crop type were driving the distribution of this species.

The distribution of Pterostichus madidus was stable between years when the June data were compared but was unstable for July. This may occur if the same areas were used for overwintering each year, with the emergent beetles then dispersing. This species was found to be more dispersive than Pterostichus melanarius and was capable of crossing field boundaries (Holland et al. 2004a). Philonthus cognatus had a more ephemeral distribution within years compared with the carabid species. Although this species also overwinters within fields as larvae and is susceptible to intensive soil cultivations (Andersen 1999), it is highly dispersive, like most Staphylinidae, and will readily fly (Bohac 1999). Thus emergence within one area followed by dispersal may explain the distribution patterns found here.

factors determining spatiotemporal pattern

Crop rotations and the associated husbandry cause the greatest disruptions in farmland and affect insects through their impact on: the type and timing of cultivations; extent and timing of vegetation cover; abundance of prey; environmental conditions; agrochemical inputs and time of harvest (Kromp 1999; Holland & Luff 2000; Thorbek & Bilde 2004). Of the species studied here only Poecilus cupreus was linked to a particular crop. Evidence on the impact of crop type is contradictory for many species, in part because of the inadequacies of experimental design but also because different species may favour particular crops according to their respective phenologies, environmental requirements and diet. Such preferences ultimately affect species’ composition and dominance ratios (reviewed for Carabidae by Hance 2002).

Non-crop habitats have been identified as being important refuges and overwintering habitats for a large number of carabid species (Lee & Landis 2002). However, these field boundary features can also act as barriers to dispersal, especially for those species that predominantly inhabit field centres, thus affecting their spatial dynamics at the farm scale. In the present study, population patches of three species (Pterostichus madidus, Poecilus cupreus and Philonthus cognatus) spanned field boundaries, whereas those of Pterostichus melanarius were confined by the boundaries. Mark–release–recapture studies confirmed that the movement of Pterostichus melanarius was much more restricted than that of Pterostichus madidus (Holland et al. 2004a) and the field boundaries may have been restricting movement, as found in previous studies of Pterostichus melanarius and Poecilus cupreus (Thomas, Parkinson & Marshall 1998). However, there may simply have been no incentive to disperse. Pterostichus madidus had a more fragmented spatial pattern and, although a small proportion of marked individuals traversed field boundaries (Holland et al. 2004a), therefore movement was not considered to be the primary mechanism; instead factors determining survival were more likely to be responsible. The obstruction posed by different field boundary types is, however, likely to differ for each species (Duelli et al. 1990). In farmland landscapes where fields are divided by linear boundary features, the survival of metapopulations and their constituent local populations can theoretically depend upon the ability of species to re-invade fields following catastrophic disturbances causing population crashes, which, within intensively farmed fields, can occur relatively frequently. Barriers to dispersal can also restrict opportunities for gene flow between local populations (Frankham 1995) and may influence fitness (Reed & Frankham 2003), although only a few dispersers are necessary for significant gene flow and all fields have gateway access for machinery through which beetles can also pass.

relevance to farm management

The findings presented here and in the associated study (Holland et al. 2004a) therefore suggest that the patches of these large Carabidae found within each field are isolated to a large extent from each other by the field boundaries. Studies conducted in more homogeneous forest habitats, where there is less disturbance and fewer barriers to dispersal, revealed that patches of Pterostichus madidus and Pterostichus melanarius extended to 1·2 km2 (Judas, Dornieden & Strothmann 2002). Thus in farmland the impact of catastrophic disturbances is likely to be severe and long-term unless sufficient individuals survive in the surrounding non-crop habitats or are protected within the field to repopulate the field. Moreover, the rate of recolonization is likely to be positively correlated with the scale of the disturbance. In contrast, the more mobile species, such as Philonthus cognatus, may be better suited to the unstable conditions that occur in arable fields. Further investigations of recolonization are needed to identify the impact of farming practices on invertebrate distributions.

The extent to which the distribution patterns detected here could be extrapolated across the landscape needs further investigation before we can determine at what scale metapopulations of beetles exist in farmed landscapes. The detection of stable local patches extending across several fields supports the theory that metapopulations exist in farmland. However, some species were found to be quite ephemeral in their location across the study area and local populations may exist at much larger scales (e.g. Philonthus cognatus). Moreover, because of the annual fluctuations in abundance, studies based upon a single year of data may not reflect the true extent of a species’ distribution and should be treated with caution. This was evident when insect abundance across a 64-km2 block of arable farmland was measured annually for 30 years (Holland 2002). Landscape-scale investigations therefore need to take temporal changes into consideration. Regional distributions have been described (Luff 1998), and geographical position, local soil and crop conditions and year were the most influential factors (Luff 2002).

The species studied here are known to feed on crop pests but the fluctuations in their spatial dynamics indicate that the level of biocontrol offered by each species may be expected to vary within each field between years (Pearce & Zalucki 2005). This may not be important provided the total level of biocontrol remains unaffected. However, when the total number of predatory invertebrates across the study area was mapped there were similar fluctuations across the study area because the pattern detected reflected that of the dominant species, Pterostichus madidus (Holland et al. 2003). There is other evidence that beetle diversity has declined on farmland, with dominance by a few species (Croy 1987). Conservation biocontrol should therefore aim to encourage diversity as this will ensure that species are foraging for a greater proportion of the year and will be better able to withstand the impact of adverse farming operations. Farmland diversity may be achieved through manipulation of the non-crop areas that are utilized for overwintering and as a foraging resource in the summer. Reducing field size and thereby improving the boundary–field ratio is one approach; however, this may restrict the movement of epigeal invertebrates and restrict re-invasion after catastrophic events. Beetle abundance also varied considerably between years as a consequence of cropping, indicating that implementing a heterogeneous approach to crop distribution at farm and landscape scales would encourage biocontrol through the desynchronization of adverse husbandry practices. Conservation biocontrol should also aim to encourage the more mobile species that are better able to respond to the ephemeral environmental conditions and pest infestations that occur in arable fields. These approaches are also compatible with conservation strategies aimed at encouraging farmland biodiversity per se, and with recent changes to the Common Agricultural Policy and the widespread availability of agri-environment payments (Ormerod et al. 2003) they are more acceptable and achievable on farm.


The study was conducted as part of the 3D Farming Project, which was funded under the Sustainable Arable LINK Programme by the Department of the Environment, Food and Rural Affairs and Scottish Executive Environmental Rural Affairs Department, with additional financial support from Dow AgroSciences, Home-Grown Cereals Authority, Horticultural Development Council, Processors and Growers Research Organization, Tesco, Unilever, The Game Conservancy Trust, The Chadacre Agricultural Trust, The Dulverton Trust, The Manydown Company, The Worshipful Company of Farmers and The Yorkshire Agricultural Society. Sincere thanks to all those that helped with the study, especially Christina Reynolds, Sue Thomas, Barbara Smith, Sam Bishop, Vicky Carter and Catherine Holley. We gratefully thank Lord Cranborne for permission to use Cranborne farm and the staff of Cranborne Estates.