Environmental Pollution 145 (2007) 874e883

www.elsevier.com/locate/envpol

Type ‘A’ and ‘B’ recovery revisited: The role of field-edge

habitats for Collembola and macroarthropod community
recovery after insecticide treatment
Geoff K. Frampton a,*, Philip J.L. Gould a,1, Paul J. van den Brink b,c, Eleanor Hendy a
a
Ecology and Evolutionary Biology Group, School of Biological Sciences, University of Southampton,
Bassett Crescent East, Southampton SO16 7PX, UK
b
Alterra, Wageningen University and Research Centre, P.O. Box 47, 6700 AA Wageningen, The Netherlands
c
Department of Aquatic Ecology and Water Quality Management, Wageningen University, Wageningen University and Research Centre,
Wageningen, The Netherlands
Received 31 December 2005; received in revised form 30 April 2006; accepted 3 May 2006

Collembola recolonisation differs among species; effectiveness of unsprayed crop edges as sources
of arthropod recolonisation may depend on adjacent habitat.

Abstract

Previous work has identified two patterns of arthropod recovery after insecticide applications to arable crops: dispersal-mediated recolonisa-
tion from untreated areas (Type A) and recolonisation within treated areas assisted by reduced predation (Type B). In this study, connectivity
between field-edge habitats was manipulated using barriers to investigate whether a crop edge and adjacent hedgerow influence recolonisation of
an insecticide-treated crop by surface-active Collembola and other arthropods. Collembola recovery patterns differed among closely-related
taxa. Epigeic collembolan and macroarthropod communities were more diverse and abundant, and rates of artificial prey predation were higher,
in sprayed crop areas connected to both hedgerow and unsprayed crop edge than in sprayed areas connected to the unsprayed edge alone. These
findings indicate that effectiveness of unsprayed crop edges as sources of field recolonisation may depend on adjoining field margin habitats. An
assumption in risk assessment that unsprayed crop edges assist population recovery within treated areas is not supported.
Ó 2006 Elsevier Ltd. All rights reserved.

Keywords: Ecological effectiveness; Predation; Principal response curves; Recolonisation; Unsprayed crop edges

1. Introduction for sustaining such beneficial predator populations (e.g. Agusti

et al., 2003). Surface-active (epigeic) Collembola are highly
Collembola (springtails) are among the most abundant ar- sensitive to organophosphorus (OP) insecticides, with counts
thropods in agroecosystems and are preyed upon by a wide in sprayed areas often reduced locally to zero (Frampton,
range of specialist and generalist predators (Hopkin, 1997). 2002). Colonisation of cultivated fields by epigeic Collembola
Many generalist predators of Collembola (e.g. Carabidae, can be influenced strongly by an adjacent hedgerow (Alvarez
Staphylinidae and Linyphiidae) are also natural enemies of et al., 2000), but no studies have investigated whether hedge-
crop pests. Collembola could be an important prey resource rows or unsprayed crop edges contribute to Collembola re-
colonisation after pesticide applications. The only detailed
spatial information on recovery dynamics suggests that preda-
* Corresponding author. Tel.: þ44 2380 593112.
E-mail address: [email protected] (G.K. Frampton). tory arthropods recolonise sprayed areas slowly from un-
1
Present address: Plant and Invertebrate Ecology Group, Rothamsted sprayed refuges (‘Type A’ recovery), whereas Collembola
Research, Harpenden, Hertfordshire AL5 2JQ, UK. and other prey groups recolonise from within sprayed areas

0269-7491/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.envpol.2006.05.004
 G.K. Frampton et al. / Environmental Pollution 145 (2007) 874e883 875

(‘Type B’ recovery). Type B recovery may reflect a temporary unsprayed crop edge and adjacent hedgerow contribute to re-
reduction of predation pressure within sprayed areas (Duffield colonisation of an insecticide-treated crop by surface-active
and Baker, 1989; Duffield and Aebischer, 1994). Collembola and other arthropods.
Broad-spectrum insecticides may be prohibited from the
edges (outermost 6 m) of arable crops as a statutory practice 2. Materials and methods
to protect non-target arthropods (Campbell, 1995). Such pesti-
cide exclusion has been assumed not only to protect populations 2.1. Study site
within crop edges but also to assist recovery of populations in
The study was carried out during 2001 in a 35-ha rectangular field of winter
adjacent sprayed areas (Forster and Rothert, 1998). Seven pre- wheat Triticum aestivum cv. Maris Widgeon on a commercial farm in Hamp-
vious studies have investigated the effects of crop-edge pesti- shire, southern England (51  70 N, 1  20 W). Soil type was a light loam over chalk
cide exclusion on arthropods within adjacent sprayed areas (Andover Series, pH 7.6) comprising 8.4% organic matter, 8.9% clay, 30% silt,
(Hald et al., 1988; Hassall et al., 1992; Hawthorne, 1994; Mor- 18.3% fine sand and 10.5% coarse sand. The field was surrounded by hedgerows
of uniform structure, had been cropped under cereals in the preceding 10 years
eby, 1995; de Snoo and de Leeuw, 1996; Holland et al., 2000;
and was known to contain a diverse arthropod fauna. Two homogeneous hedge-
Tones et al., 2000). Due to a lack of effects, or lack of replication rows were chosen for study (Fig. 1). These were approximately 1.8 m in height
of the pesticide treatments, these studies do not provide evi- and 0.8e1.0 m wide with the woody species dominated by hawthorn Crataegus
dence that effects of pesticide exclusion (or selective use) ex- monogyna. Hedgerow bases were dominated by the grasses Poa trivialis and
tend beyond crop edges. Other detailed studies of arthropod Dactylis glomerata and contained up to 17 species of annual arable flora and
grasses; the composition and structure of the field boundary flora was similar
recolonisation dynamics have been carried out but these did
in each of the experimental study areas. Adjacent habitats were a field of herb-
not involve crop edges (Duffield and Baker, 1989; Jepson and age seed Lolium perenne cv. Ronja adjoining Hedgerow 1 and a mature apple
Thacker, 1990; Thomas et al., 1990; Thacker and Jepson, orchard (Malus domestica) adjoining Hedgerow 2.
1993; Duffield and Aebischer, 1994; Duffield et al., 1996;
Dunger et al., 2002). Overall, mainly due to limited research, 2.2. Study design
there appears to be no convincing evidence that unsprayed edges
An experimental approach using barriers to manipulate connectivity be-
assist recovery of populations within sprayed areas. tween the hedgerow, crop edge and sprayed field area was used to investigate
In this work, the connectivity between field-edge habitats the effect of a 6-m wide unsprayed crop edge and adjacent hedgerow on the
was manipulated using barriers to investigate whether an abundance and species composition of epigeic Collembola in the sprayed field

200m

6m

Hedgerow sprayed area unsprayed area

2 (total 200m × 398m (654m × 410m)
length
410m) 4
F F

3
80m
2 1

Hedgerow 1 (total length 860m)

wind direction

Treatment: C2 C1 C0

sprayed
13m s, p s, p s, p crop area

6m
unsprayed
3m s crop edge

0m
hedgerow

36 m

Fig. 1. (Top) The study site in a winter wheat field showing the locations of four blocks (1e4) of crop-edge experimental treatments and suction sampling positions
in the field interior (F). (Bottom) Arrangement of the experimental connectivity treatments (C0, C1, C2) within one of the blocks, showing the positions of barriers
(black lines) and the locations of suction (s) and pitfall (p) sampling.
876 G.K. Frampton et al. / Environmental Pollution 145 (2007) 874e883

area. The barriers impede the movement of ground-active Collembola and harvest; 70 days post application) and were preserved in methylated spirit im-
other arthropods (e.g. Gravesen and Toft, 1987; Holland, 1998) and have mediately after collection. Collembola were identified using binocular and
some advantages over other possible experimental approaches (Section 4). compound light microscopy (Hopkin, 2000). Due to the large number of Col-
Three experimental connectivity treatments were created using polythene lembola obtained, Isotoma viridis Bourlet and Isotoma anglicana Lubbock
barriers supported using ropes and wooden stakes. The barriers were erected were not identified separately and are recorded together as Isotoma viridis
on 23 May, 12 days in advance of the insecticide application, and extended ‘group’.
0.7 m above and 0.3 m below ground. The experimental treatments were: As a contingency in case wet weather precluded suction sampling, pitfall
C0, sprayed crop isolated from unsprayed crop edge and hedgerow; C1, traps were installed in each experimental treatment 13 m into the crop (Fig. 1).
sprayed crop connected to unsprayed edge but isolated from hedgerow; C2, In each treatment, five traps (9 cm diameter, 13 cm depth) were placed at 1 m
sprayed crop connected to both unsprayed crop edge and hedgerow. These spacing in a line parallel to the hedgerow. The traps were half-filled with water
treatments were created, respectively, by placing 12-m-long barriers between and a drop of detergent to break the surface tension so that captured arthropods
the sprayed crop and unsprayed crop edge (C0), between the unsprayed could not escape. Traps were operated for 7-day periods on 10 occasions after
crop edge and hedgerow (C1) or by using no barrier (C2) (Fig. 1). To prevent the insecticide application. After collection, the trap contents were rinsed with
movement of ground-active arthropods between adjacent treatments, barriers water (sieve mesh 150 mm) and preserved in methylated spirit.
were also installed perpendicular to the hedgerow, extending 18 m into the
crop. The treatments were repeated in four blocks to give four replicates;
two blocks were sited at each hedgerow (Fig. 1) at locations where field
2.6. Predation assessment
boundary structure and botanical composition were homogeneous. The spatial
order of treatments within each block was chosen randomly. Recolonisation patterns of predatory arthropods may affect rates of preda-
tion in insecticide-treated fields (Duffield et al., 1996). It was not feasible to
monitor individual predatoreprey interactions in the current work which in-
2.3. Insecticide application volved many predator and prey groups. Instead, an indication of the overall
predation pressure was obtained using Drosophila pupae (Diptera: Drosophi-
Chlorpyrifos (480 g litre1 EC, ‘Spannit’Ò; PBI) was applied in lidae) as indicator prey baits, in a similar approach to that of Duffield et al.
200 L ha1 water at 16:00 h BST on 5 June, 2001 during dry weather using (1996). Freeze-killed pupae were obtained from laboratory cultures and at-
a towed Chafer E-Series sprayer with low-drift 80  flat fan nozzles (TeejetÒ tached at regular spacing to 14  12 cm sections of waterproof card (me-
XR 80 03) at 0.5-m spacing. Boom height was 45 cm above the crop, operating dium-grade abrasive paper) using flour and water paste. Pupal baits were
pressure 2 bar and tractor speed 11 km h1. Wind speed was 8e10 km h1 and used on nine occasions (from 14 to 71 days after insecticide application),
the local screen temperature 18e19  C. During the application, the wind with 30 pupae per card on the first two dates and subsequently 15 per card.
direction (230  ) was towards the field interior, away from the two study On each sampling occasion three cards of pupae were placed in each replicate
hedgerows (which were oriented 30  /210  and 120  /300  ). An 8-ha area of of each treatment (C0, C1, C2), with each card sited 0.5 m from a pitfall trap
the field was sprayed, excluding 6-m-wide crop edges (Fig. 1). No other pes- (Fig. 1). The cards were placed with the pupae uppermost, to prevent pupae
ticide applications were made during the study. falling off, and were shielded from rainfall by 15-cm2 metal covers supported
ca. 3 cm above the ground on narrow legs to permit easy access of arthropods.
2.4. Insecticide deposition On each sampling occasion the bait cards remained in the field for 24-h pe-
riods, after which the number of damaged and removed pupae were counted,
Insecticide deposition was measured using 2.5  7.5 cm strips of Teejet excluding cards that had been visited by slugs.
water-and-oil sensitive paper, each laid horizontally, facing upwards, on a sep-
arate Petri dish lid to prevent contamination. Strips were placed at the crop 2.7. Data analysis
edge (0 m), and at distances of 3, 5.5 and 13 m into the crop in each replicate
of each experimental treatment. At each distance, strips were placed both on
As macroarthropod catches appeared to differ between the experimental
the ground surface and (supported on a wooden stake) at the top of the crop
treatments, pitfall catches of macroarthropods as well as suction catches of
foliage. The strips were positioned in the field immediately before the insec-
Collembola were analysed. The null hypothesis tested (H0) is that the experi-
ticide application. After the application, they were wrapped individually in
mental treatments did not influence arthropod abundance (i.e. C0 ¼ C1 ¼ C2).
foil and kept in a dry environment prior to examination. Insecticide deposition
Rejection of H0 would indicate that recolonisation occurred from the un-
on the strips was quantified in ml cm2 by examining scanned images of the
sprayed crop edge (Type A recovery) if C2  C1 > C0. Acceptance of H0
strips using computer image analysis software.
would indicate that the unsprayed crop edge did not on balance contribute
to arthropod recovery in the sprayed area. If recovery is promoted by low pred-
2.5. Arthropod sampling ator abundance (Type B recovery), the result could be either C0 ¼ C1 ¼ C2
(caused by generally low predator abundance in the sprayed area without
Epigeic Collembola were sampled using a petrol-driven Ryobi suction any influence of experimental treatments) or C0 > C1  C2 (caused by treat-
sampler (MacLeod et al., 1994) with an 11.5-cm-diameter nozzle containing ment-specific differences in predator recolonisation from the unsprayed edge
a muslin net (mesh < 100 mm). Each sample was obtained by placing the sam- and hedgerow). If H0 is accepted, comparisons of arthropod abundance in
pler nozzle on the ground surface between crop plants for 10 s at each of five the sprayed and unsprayed areas of the field and between the sprayed area
randomly-selected locations and pooling the catch. Samples (each 0.052 m2) and unsprayed crop edge can help to identify Type B recovery.
were taken 13 m into the crop in each experimental treatment (7 m from the Univariate analysis of variance (treatment fixed, block random; treat-
unsprayed crop edge into the sprayed area). To provide an indication of arthro- ment  block interactions included) was used to test H0 for individual arthro-
pod abundance within the unsprayed crop edge, samples were also taken at pod taxa and also the null hypothesis that rates of predation assessed with bait
3 m from the hedgerow in the sections of unsprayed edge without any barriers cards were independent of the connectivity treatment. Analyses were per-
(treatment C2); this was to avoid any possible influence the barriers might formed on normalised data (log (x þ 1) of arthropod counts x, and arcsine-
have on estimates of arthropod abundance in the unsprayed edge. Additional transformed proportions of pupae removed per bait card).
samples were taken from the interior of the field, 80 m from the field boundary The community response to the experimental treatments was investigated
and 30 m from the interface between sprayed and unsprayed areas (Fig. 1) to using principal response curves (PRC) analysis. PRC analysis is a multivariate
confirm the spatial extent of effects of the insecticide and to assist in the de- technique derived from redundancy analysis (RDA) that focuses on the propor-
tection of any recovery that might commence in situ within the sprayed area. tion of the variance in the data set that is explained by treatment and time. For
At each location, four samples were collected, at 2 m spacing, in a line parallel each species k the response Tdtk is modelled as a multiple (the species weight,
to the hedgerow. Samples were taken on 12 occasions, in dry weather, from 25 bk) of one basic response pattern, cdt, i.e. Tdtk ¼ bk  cdt (van den Brink and
May (11 days before insecticide application) to 14 August (1 week before crop Ter Braak, 1999). By plotting values of cdt, for each treatment t and time d,
 G.K. Frampton et al. / Environmental Pollution 145 (2007) 874e883 877

a PRC diagram is obtained that shows the temporal change in community both before and after the field received chlorpyrifos but in-
composition (the ‘principal response’) in each treatment relative to a reference creased after the insecticide application. The majority of Col-
treatment. Species weights indicate the affinity of each individual taxon with
the overall community response. The choice of reference treatment in PRC
lembola have high positive species weights, indicating higher
analysis (cdt ¼ 0) is arbitrary; for present purposes the zero-connectivity treat- abundance in the unsprayed crop edge than in the sprayed crop
ment (C0) is used. PRC analysis was performed using the software program area. Isotomurus spp., however, show the opposite pattern
CANOCO 4 (Ter Braak and Šmilauer, 1998). For each analysis the null (Fig. 3).
hypothesis that the PRC diagram does not display the treatment variance Differences in the fitted relative abundance (cdt) between
(Tdtk ¼ 0 for all t, d, k) was tested using an F-type statistic obtained by permut-
ing whole time series in the partial RDA from which the PRC was obtained
the experimental connectivity treatments were relatively small
(Ter Braak and Šmilauer, 1998). Permutation tests within sampling dates compared to the differences between the unsprayed crop edge
were also used to test the null hypothesis that the principal response cdt did and sprayed crop area. However, the highest relative abun-
not differ between treatments (van den Brink and Ter Braak, 1999). dance was consistently in the full-connectivity treatment
(C2) on 10 of the 12 sampling dates (Fig. 3). The null hypoth-
3. Results esis that cdt was similar for each connectivity treatment was
rejected for 7 of the 12 sampling dates. Fitted relative abun-
3.1. Insecticide deposition dance for treatment C2 was significantly higher than for C0
and C1 on six and four dates, respectively, whereas a signifi-
Very little insecticide reached the unsprayed crop edges, cant difference between treatments C0 and C1 occurred on
and none penetrated within 3 m of the hedgerow. Mean depo- one date (Fig. 3).
sition 5.5 m into the unsprayed edge (i.e. 0.5 m from the Total collembolan abundance (Fig. 4a) and taxonomic rich-
sprayed crop) was less than 0.1 ml cm2. In the sprayed area, ness (Fig. 4b) reflect the overall pattern shown by the PRC di-
13 m into the crop, mean chlorpyrifos deposition was agram, being higher in the unsprayed crop edge than in the
1.69 ml cm2 at crop height, which was twice that at the experimental barrier treatments in the sprayed area. Where
ground surface (Fig. 2). a difference occurred between the experimental treatments, to-
tal collembolan abundance and taxonomic richness were con-
sistently higher in C2 than C0 or C1 but there was no
3.2. Collembola communitiesdsuction sampling consistent difference between C0 and C1 (Fig. 4).
The overall collembolan community response reflects those
Suction sampling yielded ca. 99,000 Collembola in total, of the most abundant Collembola taxa, Lepidocyrtus spp.
representing 25 taxonomic groups (Appendix A). Of the total (Fig. 5a) and Isotoma viridis ‘group’ (Fig. 5b).These were con-
variance in the Collembola data set, 22% is explained by dif- sistently more abundant in C2 than C1 on several sampling
ferences between blocks, 41% by sampling date and 37% by occasions during July but did not exhibit consistent differences
experimental barrier treatment. The PRC diagram (Fig. 3) between C0 and C1. A less abundant species, Orchesella
summarises the overall community composition for the 25 villosa, also had clearly higher abundance in C2 than either
taxa and displays a significant proportion of the treatment ef- C0 or C1 at the end of the sampling period (Fig. 5c).
fects variance (85%; p ¼ 0.001). Negative effects of chlorpyrifos on Lepidocyrtus spp. and
The overall community composition differed significantly Isotoma viridis ‘group’ occurred close to the unsprayed field
( p ¼ 0.001) between the unsprayed crop edge (3 m from the edge and also at 80 m into the crop, with counts in the
hedgerow) and the sprayed crop (13 m from the hedgerow) sprayed area remaining close to zero throughout the sampling
on all sampling dates, with consistently higher relative abun- period (Fig. 6a,b). However, Pseudosinella spp. and Isotomurus
dance in the unsprayed crop edge; this difference was apparent spp. exhibited a different pattern, with the highest counts in
the field interior being in the sprayed area (Fig. 6c,d). These
latter taxa had relatively high spatial and temporal heteroge-
unsprayed crop edge sprayed crop neity of abundance with no consistent differences between
2.0
Estimated chlorpyrifos deposition

the connectivity treatments (data not shown; 95% confidence

limit (CL) overlapped for C0, C1 and C2 on all sampling
µl cm-2 (± 95 CL)

1.5 ground crop dates).

1.0 3.3. Macroarthropod communitiesdpitfall sampling

0.5 Pitfall sampling yielded 16,000 macroarthropods (i.e. ex-

cluding Collembola) in total, representing 54 taxonomic
groups (Appendix A). Of the total variance in the data set,
0.0
0m 3m 5.5m 13m 70% is explained by differences between blocks, 26% by
Distance from hedgerow (m) sampling date and 4% by experimental barrier treatment.
Fig. 2. Mean deposition (ml cm2  geometric 95% CL; n ¼ 16) of chlorpyr- The PRC diagram (Fig. 7) summarises the overall community
ifos estimated using water and oil sensitive paper at different locations in the composition and displays a significant proportion of the treat-
study field. ment effects variance (22%; p ¼ 0.004). Despite the relatively
878 G.K. Frampton et al. / Environmental Pollution 145 (2007) 874e883

unsprayed crop edge (3m from hedgerow)

sprayed crop area (13m from hedgerow):
2.5
connected to unsprayed crop edge + hedgerow (treatment C2)
connected to unsprayed crop edge only (treatment C1) bk
not connected (treatment C0) Isotoma viridis‘group’
2 Lepidocyrtus spp.
Orchesella villosa
a Heteromurus nitidus
2
a a a
a 1.5
insecticide
a a Entomobrya multifasciata
1.5
a a Sminthurinus elegans
a a
1
1
cdt

a
0.5 Deuterosminthurus spp.
0.5
b Sminthurididae
b b b b
b
Pseudosinella spp.
0 0
c Cyphoderus albidus
cd c c
Parisotoma notabilis
-0.5 Isotomurus spp.
0.5
-11 3 7 14 19 24 29 36 41 47 56 70
(25 May) (14 Aug)
Time (days after insecticide)

Fig. 3. PRC diagram showing the principal response (cdt) and species weights (bk) for a suction-sampled Collembola community (25 taxa) in an unsprayed crop
edge and at three locations (C0, C1, C2) in the sprayed crop differing in connectivity with the unsprayed edge and hedgerow. Treatment C0 is nominated as the
reference (cdt ¼ 0). Letter codes indicate significant differences in values of cdt ( p ¼ 0.001): a, between the unsprayed crop edge and all other treatments; b, be-
tween C0 and C2; c, between C1 and C2; d, between C0 and C1. For clarity, taxa with bk values close to zero are not shown.

(a) high heterogeneity of the data set compared to suction sam-

Total Collembola abundance pling, the fitted relative catch for the pitfall-sampled commu-
600 unsprayed crop edge (3m)
nity was consistently higher in treatment C2 than C0 or C1 on
Mean abundance per sample

C0
500 insecticide C1 sprayed crop (13m) most sampling dates, with differences between treatments sig-
(0.052 m2) ±95 CL

400
C2 nificant on three dates (Fig. 7). Three individual species of
Carabdae (Agonum dorsale, Harpalus rufipes and Pterostichus
300 madidus) as well as the total Carabidae and total macroarthro-
200
pod catch were affected significantly by the experimental
treatments, but effects were transient and limited to individual
100 sampling dates. In all cases the direction of treatment effects
0 was consistent (C2 > C1 and/or C2 > C0) (Table 1). Accumu-
-11 3 7 14 19 24 29 36 41 47 56 70 lated total catches of three coleopteran taxa also differed sig-
nificantly between the connectivity treatments, in all cases
(b) with higher catches in treatment C2 than C0 or C1 (Table 2).
Collembola richness (taxa with ≥ 5 individuals sample-1)
Mean number of taxa per sample

12
(0.052 m2) ± 95 CL

10

6 3.4. Predation
4
Predation rates indicated by the proportion of prey baits re-
2
moved were initially similar in treatments C0, C1, and C2 but
0 heterogeneity among the treatments increased with time. The
-11 3 7 14 19 24 29 36 41 47 56 70
(25 May) (14 Aug) proportion of prey removed was significantly higher in C2
Time (days after insecticide) than in C0 on two occasions at the end of the sampling period
but no other differences between treatments were statistically
Fig. 4. Mean (geometric 95% CL; n ¼ 16) abundance (a) and taxonomic
richness (b) of suction-sampled Collembola in an unsprayed crop edge and significant (Fig. 8). Higher predation pressure in C2 would be
at three locations (C0, C1, C2) in the sprayed crop differing in connectivity consistent with the higher overall abundance of predatory ar-
with the unsprayed edge and hedgerow (connectivity details are as in Fig. 3). thropods in this treatment.
 G.K. Frampton et al. / Environmental Pollution 145 (2007) 874e883 879

unsprayed crop edge (3m) sprayed area and hedgerow suggests that the effectiveness
C0 of an unsprayed crop edge as a source of recolonisation
C1 sprayed crop (13m) may depend on the presence of suitable field boundary
C2 habitat.
(a) Lepidocyrtus spp. Most of the Collembola taxa that were present in the un-
1000 insecticide
sprayed crop edges are affected adversely by chlorpyrifos
(Frampton, 1999, 2002). An in situ protective effect of exclud-
100 ing the insecticide is supported by the persistence of relatively
high Collembola densities (up to 9000 m2) in the unsprayed
10 crop edges. However, counts in the sprayed area at 13 m and
80 m from the hedgerow remained relatively low, with no ev-
1 idence that recovery reached the interior of the sprayed area
during the 70-day duration of the study. An exception is the
0.1 increase in counts of Isotomurus spp. and Pseudosinella spp.
CL

-11 3 7 14 19 24 29 36 41 47 56 70 in the sprayed area of the field interior whereas counts of these
taxa in the unsprayed area remained low. This pattern would
Mean abundance per sample (0.052 m2) ± 95

(b)
1000
Isotoma viridis ‘group’ be consistent with a resurgence of abundance following re-
lease from predation or other density-limiting biotic interac-
100 tions. Although data are scarce, there is evidence that
Isotomurus spp. and Pseudosinella spp. may in some situa-
10 tions be less adversely affected than other Collembola by
chlorpyrifos (Wiles and Frampton, 1996; Frampton, 1999).
1 (In other situations, Isotomurus spp. as a group may appear
highly susceptible to chlorpyrifos (G.K. Frampton, unpub-
0.1 lished data), perhaps reflecting different susceptibilities of
-11 3 7 14 19 24 29 36 41 47 56 70
individual Isotomurus species.) Although the mechanism
underlying the increased catches of Isotomurus and
(c)
Orchesella villosa Pseudosinella is not proven, Collembola recolonisation pat-
50
terns following an insecticide application clearly differ among
40 taxa, meaning that recovery patterns may not be interpreted
appropriately if the taxonomic resolution of monitoring is
30
too coarse.
20 Some Collembola (e.g. Lepidocyrtus spp. and Orchesella
villosa) appear to be capable of dispersing over relatively
10
large distances (at least tens of metres) by active locomotion
0 (Dunger et al., 2002). These taxa, together with Isotoma vir-
-11 3 7 14 19 24 29 36 41 47 56 70 idis ‘group’, may have originated from the hedgerow, indi-
(25 May) (14 Aug)
Time (days after insecticide)
cated by higher catches in treatment C2 than C0 or C1.
Such a pattern would be consistent with dispersal-mediated
Fig. 5. Mean (geometric 95% CL; n ¼ 16) abundance of three Collembola (Type A) recovery (Duffield and Aebischer, 1994). Although
taxa (aec) in an unsprayed crop edge and at three locations (C0, C1, C2) in hedgerow connectivity clearly had an impact on Collembola
the sprayed crop differing in connectivity with the unsprayed edge and hedge-
row (connectivity details and sampling dates are as in Fig. 3). Note that in (a)
abundance and species composition in the sprayed crop
the CL of treatment C1 are omitted for clarity; on all dates they overlapped the area, it is difficult to determine its general importance for
CL for treatments C1 and C2. recovery, as individual taxa exhibited different response pat-
terns. Hedgerow connectivity appears to have been more
important for O. villosa than for I. viridis ‘group’ and
4. Discussion Lepidocyrtus spp., as abundance of O. villosa progressively
increased in treatment C2 whereas for the other taxa the
There seems to be no robust evidence that leaving crop difference in abundance between treatments did not persist
edges unsprayed assists recovery of arthropod populations in to the end of the study. However, as O. villosa was not
adjacent pesticide-treated crop areas. This reflects a lack of present outside the crop edge prior to the insecticide appli-
reliable information rather than clear evidence for a lack of cation, population recovery in the sprayed area is difficult to
effects. In this work, the unsprayed crop edge alone did not define.
appear to act as a source of arthropod recovery for the sprayed The lower frequency of statistically significant treatment
crop area, as physically isolating it from the sprayed crop effects on predators compared to Collembola is not surprising,
made no difference to arthropod abundance or diversity in as the spatial scale of the study would have been inadequate
the sprayed area. The influence of connectivity between the for detecting effects of the experimental treatments on mobile
880 G.K. Frampton et al. / Environmental Pollution 145 (2007) 874e883

Crop unsprayed crop edge (3m) Field unsprayed crop area

edge sprayed crop, treatment C0 (13m) interior (80m) sprayed crop area
(a)
Lepidocyrtus spp.
400 400

300 insecticide 300 insecticide

200 200

100 100

0 0
-11 3 7 14 24 36 47 56 70 -11 3 7 14 24 36 47 56 70

(b)
Isotoma viridis ‘group’
300 300
CL

250 250
200 200
Mean abundance per sample (0.052m2) ± 95

150 150
100 100
50 50
0 0
-11 3 7 14 24 36 47 56 70 -11 3 7 14 24 36 47 56 70

(c)
Pseudosinella spp.
120 120
100 100
80 80
60 60
40 40
20 20
0 0
-11 3 7 14 24 36 47 56 70 -11 3 7 14 24 36 47 56 70

(d)
Isotomurus spp.
120 20
100
15
80
60 10
40
5
20
0 0
-11 3 7 14 24 36 47 56 70 -11 3 7 14 24 36 47 56 70
(25 May) (11 Aug) (25 May) (11 Aug)
Time (days after insecticide)

Fig. 6. Mean (geometric 95% CL) abundance of four Collembola taxa (aed) in sprayed and unsprayed areas of crop at the field edge (3 m and 16 m from the
hedgerow; n ¼ 16) and in the field interior (80 m from the hedgerow; n ¼ 4). Sampling dates are as in Fig. 3. Note differing y-axis scales in (d).

predatory arthropods such as Carabidae, Staphylinidae and A limitation is that many predators of Collembola (e.g. Liny-
Linyphiidae, some of which disperse aerially. The transient phiidae, some Carabidae and some Staphylinidae) only
significant effect of the treatments on Agonum dorsale (Table 1) hunt active prey, so it is unlikely that the predation rate es-
would be consistent with barriers in C0 and C1 interrupting timated using bait cards would reflect the predation pressure
the annual spring dispersal of this species from the field upon Collembola. However, the significantly lower predation
boundary into the crop (Jensen et al., 1989). The findings pressure in treatment C0 than in C2 suggests that the overall
are also consistent with previous work demonstrating that biocontrol potential of arthropods (for example as antago-
Harpalus rufipes may disperse from aggregations in crop nists of pests such as aphids) might be influenced by the
edges (Thomas et al., 2001) whilst Pterostichus madidus is type of field margin habitat where the unsprayed crop edge
able to disperse between fields across field boundaries is sited.
(Holland et al., 2004). To ensure that barriers did not influence the estimates of
Predation assessment using prey baits can only detect pre- arthropod abundance in the unsprayed crop edge, sampling
dation by arthropods that forage upon immobile prey. in the crop edge was carried out in treatment C2. This
 G.K. Frampton et al. / Environmental Pollution 145 (2007) 874e883 881

4.0 Harpalus rufipes

C0 3.5
C1 sprayed crop (13m) Lycosidae
bk 3.0
C2 Agonum dorsale
Bembidion obtusum
0.2 insecticide 2.5 Linyphiidae + Erigonidae
Pterostichus madidus
2.0
Trechus quadristriatus
1.5
Philonthus + Quedius spp.
0.1
Hydrophilidae
1.0
Silphidae
Loricera pilicornis
cdt Cryptophagidae
0.5
other Araneae
0 0
ab b a Tachyporus nitidulus
-0.5 Aleocharinae
Tachyporus hypnorum
-1.0 Anotylus spp.
-0.1 Curculionidae
-1.5 Tachyporus chrysomelinus

-2.0
Lathridiidae
-2.5
8 15 22 29 35 44 51 57 65 71
(13 Jun) (15 Aug)
Time (days after insecticide)

Fig. 7. PRC diagram showing the principal response (cdt) and species weights (bk) for a pitfall-sampled macroarthropod community (54 taxa) at three locations in
a sprayed crop (C0, C1, C2) differing in connectivity with the unsprayed crop edge and hedgerow (details are as in Fig. 3). Treatment C0 is nominated as the
reference (cdt ¼ 0). Letter codes indicate significant differences in values of cdt ( p < 0.005): a, between C0 and C2; b, between C1 and C2. For clarity, taxa
with bk values close to zero are not shown.

correlation between sampling locations is unlikely to affect could be carried out within homogeneous areas of crop
interpretation of the results, however, as differences in arthro- and hedgerow. These were important considerations as the
pod abundance between the unsprayed edge and the sprayed study was carried out on a commercial farm employing
crop area were considerably larger than the effects of the con- bird conservation management, which would be incompati-
nectivity manipulation treatments. Treatment C2 is an appro- ble with wide-scale use of broad-spectrum insecticides on
priate location for assessing abundance in the unsprayed crop crop edges. Organophosphorus insecticides and Collembola
edge, as the connectivity between the hedgerow and sprayed provide a useful model ecotoxicological system because
crop is realistic, whilst close proximity to the other treatments the insecticides cause clear and repeatable differences of
should minimise background spatial heterogeneity of arthro- abundance and diversity between treated and untreated areas
pod abundance. (e.g. Frampton, 1999, 2002). However, it is unlikely that re-
The use of barrier treatments to manipulate connectivity sults could be generalised to different insecticide modes of
between the sprayed crop, unsprayed crop edge and hedge- action. Studies using broad-spectrum organophosphate insec-
row permitted this study to be carried out at a relatively ticides might underestimate recovery that is assisted by re-
small spatial scale. Advantages of this approach are: (i) the duced predation because they have a broad activity against
area sprayed with chlorpyrifos could be kept to a minimum; both predators and prey, whereas the more widely-used syn-
(ii) spraying of crop edges was avoided; and (iii) the work thetic pyrethroid insecticides appear to be more harmful to

Table 1
Significant effects of the experimental treatments (C0, C1, C2) on pitfall catches of macroarthropods on individual sampling occasions
Days after insecticide Agonum dorsale Harpalus rufipes Pterostichus madidus Total Carabidae Total macro-arthropods
15 F2,48 ¼ 50.15 n.s. n.s. F2,48 ¼ 9.67 n.s.
p < 0.001 p ¼ 0.013
C2 > C0 ¼ C1 C2 > C1
22 n.s. F2,48 ¼ 50.15 n.s. n.s. n.s.
p ¼ 0.017
C2 > C0 ¼ C1
71 n.s. n.s. F2,48 ¼ 50.15 F2,48 ¼ 10.90 F2,48 ¼ 19.17
p ¼ 0.035 p ¼ 0.01 p ¼ 0.002
C2 > C0 ¼ C1 C2 > C0 ¼ C1 C2 ¼ C1 > C0
Each sampling occasion refers to the end of a 7-day trapping period. Effects were only statistically significant on those dates included in the table. Treatment effects
were independent of location (in all cases treatment  block p > 0.05). n.s., treatment effect not significant ( p > 0.05).
882 G.K. Frampton et al. / Environmental Pollution 145 (2007) 874e883

Table 2 Acknowledgements
Mean catches of Coleoptera per pitfall trap per 7 days averaged over 10 sam-
pling dates
This work was funded by the UK Department for Environ-
C0 C1 C2 F2,48 p ment, Food and Rural Affairs (DEFRA) Pesticides Safety Di-
Pterostichus madidus 1.9 a 1.9 a 3.2 b 7.71 0.005 rectorate as part of research project PN0934. We thank Ben
Total Carabidae 14.3 a 15.7 a 21.1 b 16.43 0.004 Gibbons and the Leckford Estate, Hampshire, UK, for kindly
Total Coleoptera 32.5 a 33.7 a 40.9 b 5.93 0.038
enabling use of the study site on a commercial farm and for
Treatments (C0, C1, C2) sharing the same letter (a, b) did not differ signifi- providing manpower and machinery for the insecticide
cantly (Tukey multiple range test). Treatment effects were independent of
location (in all cases treatment  block p > 0.05).
application.

predators than to Collembola (Frampton, 1999). Clarification

of arthropod recovery dynamics for other chemicals would Appendix A
Arthropod taxonomic groups identified in suction and pitfall samples
be prudent, given the widespread use of insecticides. It is
Sampling Arthropod group Number and identity of taxa monitored
also important that the role of unsprayed crop edges is clar-
method
ified with regard to whether they can be generally assumed
Suction Collembola: Arthropleona
to influence arthropod recovery in sprayed areas (Forster
Isotomidae 4 Isotoma viridis ‘group’, Parisotoma
and Rothert, 1998). notabilis, Isotomurus spp., other
Isotomidae
Hypogastruridae 1 Total Hypogastruridae
5. Conclusions Entomobryidae 11 Pseudosinella spp. (2), Lepidocyrtus
spp. (3), Entomobrya multifasciata,
Heteromurus nitidus, Orchesella
Unsprayed crop edges can provide in situ protection of ar-
spp. (2), Tomocerus spp., other
thropod populations from effects of insecticides but their ef- Entomobryidae
fectiveness as sources for arthropod recolonisation of Other Arthropleona 1 Total other Arthropleona
adjacent sprayed crops has not been adequately proven. Un- Collembola: Symphypleona
less more research is carried out to confirm the ecological Sminthuridae 6 Bourletiella hortensis,
function of unsprayed crop edges as source habitats, the as- Deuterosminthurus spp.,
sumption that unsprayed crop edges assist recovery of popu- Sminthurinus spp. (2),
lations within sprayed areas cannot be supported. Sminthurus viridis, other
Sminthuridae
Accordingly, in regulatory risk assessment, the protection Sminthurididae 1 Total Sminthurididae
goal of unsprayed crop edges should be clearly stated as being Other 1 Total other Symphypleona
restricted to those populations that occur in situ within the Symphypleona
crop edge. The taxonomic variation in recovery responses ob-
served in this study also advises against grouping Collembola Pitfall Coleoptera
Carabidae 24 Amara spp., Agonum spp. (2),
taxa together as a single response unit in ecotoxicological Asaphidion flavipes, Badister
studies. bipustulatus, Bembidion spp. (3),
Calathus fuscipes, Carabus
violaceus, Demetrias atricapillus,
Harpalus spp. (4), Leistus spp. (2),
C0 C1 C2 Loricera pilicornis, Nebria
F1,16= 33.19 F1,16= 59.8 brevicollis, Notiophilus biguttatus,
P= 0.010 P= 0.004 Pterostichus spp. (2), Trechus
Mean proportion of available

0.6 C2 > C0 C2 > C0 quadristriatus, other Carabidae

0.5 Staphylinidae 10 Aleocharinae, Anotylus spp.,
prey removed

0.4 Philonthus þ Quedius spp.,

Tachyporus spp. (5), Xantholininae,
0.3
other Staphylinidae
0.2 Other Coleoptera 17 Cantharidae, Clambidae,
0.1 families Coccinellidae, Colydiidae,
Chrysomelidae, Cryptophagidae,
0
14 22 30 35 44 52 59 65 71 Curculionidae, Elateridae,
(13 Jun) (15 Aug) Erotylidae, Histeridae,
Time (days after insecticide) Hydrophilidae, Lathridiidae,
Leiodidae, Nitidulidae,
Fig. 8. Predation rate (mean proportion of artificial prey removed) at three lo- Oedemeridae, Phalacridae,
cations (C0, C1, C2) in the sprayed crop differing in connectivity with the un- Silphidae
sprayed edge and hedgerow (connectivity details and sampling dates are as in Araneae 3 Linyphiidae, Lycosidae, other
Fig. 3). Treatment effects were independent of location (in all cases treat- Araneae
ment  block p > 0.05).
 G.K. Frampton et al. / Environmental Pollution 145 (2007) 874e883 883

References Holland, J.M., 1998. The effectiveness of exclusion barriers for polyphagous
predatory arthropods in wheat. Bull. Entomol. Res. 88, 305e310.
Holland, J.M., Winder, L., Perry, J.N., 2000. The impact of dimethoate on the
Agusti, N., Shayler, S.P., Harwood, J.D., Vaughan, I.P., Sunderland, K.D.,
spatial distribution of beneficial arthropods in winter wheat. Ann. Appl.
Symondson, W.O.C., 2003. Collembola as alternative prey sustaining spi-
Biol. 136, 93e105.
ders in arable ecosystems: prey detection within predators using molecular
Holland, J.M., Begbie, M., Birkett, T., Southway, S., Thomas, S.R.,
markers. Mol. Ecol. 12, 3467e3475.
Alexander, C.J., Thomas, C.F.G., 2004. The spatial dynamics and move-
Alvarez, T., Frampton, G.K., Goulson, D., 2000. The role of hedgerows in the
ment of Pterostichus melanarius and P. madidus (Carabidae) between
recolonisation of arable fields by epigeal Collembola. Pedobiologia 44,
and within arable Fields in the UK. Int. J. Ecol. Environ. Sci. 30, 35e53.
516e526.
Hopkin, S.P., 1997. Biology of the Springtails. Oxford University Press,
Campbell, P., 1995. Position document: Labelling and risk management strat-
Oxford.
egies for pesticides and terrestrial non-target arthropods: A UK proposal.
Hopkin, S.P., 2000. A Key to the Springtails (Insecta: Collembola) of Britain
Third UK Forum on Non-target Arthropods, Chesterford Park, 29e30
and Ireland, AIDGAP test version. Field Studies Council, Preston Mon-
March 1995. 10 pp.
tford, UK.
de Snoo, G.R., de Leeuw, J., 1996. Non-target insects in unsprayed cereal
Jensen, T.S., Dyring, L., Kristensen, B., Nielsen, B.O., Rasmussen, E.R., 1989.
edges and aphid dispersal to the adjacent crop. J. Appl. Entomol. 120,
Spring dispersal and summer habitat distribution of Agonum dorsale
501e504.
(Coleoptera, Carabidae). Pedobiologia 33, 155e165.
Duffield, S.J., Aebischer, N.J., 1994. The effect of spatial scale of treatment
Jepson, P.C., Thacker, J.R.M., 1990. Analysis of the spatial component of pes-
with dimethoate on invertebrate population recovery in winter wheat.
ticide side-effects on non-target invertebrate populations and its relevance
J. Appl. Ecol. 31, 263e281.
to hazard analysis. Funct. Ecol. 4, 349e355.
Duffield, S.J., Baker, S.E., 1989. Spatial and temporal effects of dimethoate
MacLeod, A., Wratten, S.D., Harwood, R.W.J., 1994. The efficiency of a new
use on populations of Carabidae and their prey in winter wheat. In:
lightweight suction sampler for sampling aphids and their predators in
Stork, N.E. (Ed.), The Role of Ground Beetles in Ecological and Environ-
arable land. Ann. Appl. Biol. 124, 11e17.
mental Studies. Intercept, Andover, UK, pp. 95e104.
Moreby, S.J., 1995. Heteroptera diversity and distribution within the cereal
Duffield, S.J., Jepson, P.C., Wratten, S.D., Sotherton, N.W., 1996. Spatial
ecosystem. In: McKinlay, R., Atkinson, D. (Eds.), Integrated Crop Protec-
changes in invertebrate predation rate in winter wheat following treatment
tion: Towards Sustainability? BCPC Symposium Proceedings No. 63. Brit-
with dimethoate. Entomol. Exp. Appl. 78, 9e17.
ish Crop Protection Council, Farnham, UK, pp. 151e158.
Dunger, W., Schultz, H.J., Zimdars, B., 2002. Colonization behaviour of Col-
Ter Braak, C.J.F., Šmilauer, P., 1998. CANOCO Reference Manual and User’s
lembola under different conditions of dispersal. Pedobiologia 46,
Guide to CANOCO for Windows: Software for Canonical Community
316e327.
Ordination (Version 4). Microcomputer Power, Ithaca, NY.
Frampton, G.K., 1999. Spatial variation in non-target effects of the insecticides
Thacker, J.R.M., Jepson, P.C., 1993. Pesticide risk assessment and non-tar-
chlorpyrifos, cypermethrin and pirimicarb on Collembola in winter wheat.
get invertebrates: integrating population depletion, population recovery
Pestic. Sci. 55, 875e886.
and experimental design. Bull. Environ. Contam. Toxicol. 51,
Frampton, G.K., 2002. Long-term impacts of an organophosphate-based
523e531.
regime of pesticides on field and field-edge Collembola communities.
Thomas, C.F.G., Hol, E.H.A., Everts, J.W., 1990. Modelling the diffusion
Pest Manage. Sci. 58, 991e1001.
component of dispersal during recovery of a population of linyphiid spi-
Forster, R., Rothert, H., 1998. The use of field buffer zones as a regulatory
ders from exposure to an insecticide. Funct. Ecol. 4, 357e368.
measure to reduce the risk to terrestrial non-target arthropods from pesti-
Thomas, C.F.G., Parkinson, L., Griffiths, G.J.K., Fernandez Garcia, A.,
cide use. Proceedings, 1998 Brighton ConferencedPests and Diseases.
Marshall, E.J.P., 2001. Aggregation and temporal stability of carabid beetle
pp. 931e938.
distributions in field and hedgerow habitats. J. Appl. Ecol. 38, 100e116.
Gravesen, E., Toft, S., 1987. Grass fields as reservoirs of polyphagous preda-
Tones, S.J., Ellis, S.E., Oakley, J.N., Powell, W., Stevenson, A., Walters, K.,
tors (Arthropoda) of aphids (Homopt., Aphididae). J. Appl. Entomol. 104,
2000. Use of unsprayed buffer zones in arable crops to protect terrestrial
461e473.
non-target arthropods in field margins from spray drift effects. MAFF Pro-
Hald, A.B., Overgaard Nielsen, B., Samsøe-Petersen, L., Hansen, K.,
ject PS0418. Department for Environment, Food and Rural Affairs
Elmegaard, N., Kjølholt, J., 1988. Sprøjtefri randzoner i kornmarker.
(DEFRA).
Miljøprojekt nr. 103. Miljøstyrelsen, Copenhagen, Denmark.
van den Brink, P.J., Ter Braak, C.J.F., 1999. Principal response curves: analysis
Hassall, M., Hawthorne, A., Maudsley, M., White, P., Cardwell, C., 1992.
of time-dependent multivariate responses of biological community to
Effects of headland management on invertebrate communities in cereal
stress. Environ. Toxicol. Chem. 18, 138e148.
fields. Agric. Ecosyst. Environ. 40, 155e178.
Wiles, J.A., Frampton, G.K., 1996. A field bioassay approach to assess the tox-
Hawthorne, A., 1994. Variation in the distribution of carabid beetles in cereal
icity of insecticide residues on soil to Collembola. Pestic. Sci. 47,
field headlands. PhD Thesis, University of East Anglia.
273e285.