Living where the food is: web location by linyphiid spiders in relation to prey availability in winter wheat

Authors


Summary

  • 1 Spiders form a major component of the generalist predator fauna, potentially able to restrict pest population growth, but their populations may be food-limited under current farming regimes. This study aimed to quantify food availability to spiders in winter wheat and to determine whether spider web locations are positively associated with available food resources.
  • 2 Mini-sticky traps (availability rate per 24 h, including prey falling from the crop) and mini-quadrats (instantaneous density on the ground by day) were used, in combination, to monitor the availability of potential prey to web-building species of money spider (Linyphiidae) in fields of winter wheat in Warwickshire, UK, 1997–98.
  • 3 These methods were applied to web sites of individual spiders and to non-web sites located randomly up to 30 cm away from each web. A total of 18 546 invertebrates were captured using these methods.
  • 4 Overall, significantly more potential prey were available in web sites than in non-web sites (both on sticky traps and in quadrats).
  • 5 Prey availability in May and July was about a third of that in June (both on sticky traps and in quadrats) and may have been below that known to be necessary for spiders to realize their maximum population growth rate.
  • 6 The peak rate of capture of linyphiid spiders on mini-sticky traps was 0·6 trap−1 day−1 at web sites, and approximately half this value at non-web sites. Numbers of spiders captured by mini-sticky traps and mini-quadrats increased exponentially as the season progressed. The high capture frequency in relation to population density, and the differential between web and non-web sites, points to a dynamic and aggregated distribution of spiders in winter wheat, which is consistent with what is known about mate-searching and web site abandonment rates by the Linyphiidae.
  • 7 The combination of techniques described here is recommended for monitoring prey availability in prey-enhancement programmes and may prove useful in quantitative studies of both intra- and interspecific interactions between spiders.

Introduction

In Britain, the area of cereal crops treated with pesticide has increased by 24% since 1994, with a concomitant 16% increase in the amount of active ingredients applied (Thomas, Garthwaite & Banham 1996). Although chemical pesticides are convenient to use, and often efficient and cost-effective in the short term, it is now appreciated that they are not a long-term option because of the associated cumulative problems (such as pest resistance and environmental pollution) that render pesticide-based agriculture unsustainable (Pimentel 1995). Biological control is a strong alternative option, and for the majority of low-value outdoor European crops this must be achieved through conservation of endemic biocontrol agents (i.e. ‘conservation biological control’; Ehler 1998) rather than by classical biological control or rear-and-release methods. Small soft-bodied pests (such as aphids) that are accessible on the vegetation and ground surface, are attacked by a range of natural enemies (pathogens, parasitoids, specialist and generalist natural enemies) that interact in complex ways (Sunderland et al. 1997). Generalist predators (e.g. spiders, carabid and staphylinid beetles) have the valuable attribute of being able to subsist on alternative non-pest prey. They can therefore either simply be present in the field before the pest arrives, performing a lying-in-wait strategy (Murdoch, Chesson & Chesson 1985; Chang & Kareiva 1999), or build up their populations on alternative foods early in the season and then impact on the pest population with a favourable predator : pest ratio during the early phase of pest population growth (Settle et al. 1996). Manipulative field experiments have demonstrated that generalist predators can often (Edwards, Sunderland & George 1979; Chiverton 1986; Duffield et al. 1996), but not always (Holland & Thomas 1997), contribute to commercially valuable reductions of aphid populations on wheat. Web-building species of money spider (Linyphiidae), such as Lepthyphantes tenuis (Blackwall) and Erigone atra (Blackwall), consume cereal aphids (Sunderland et al. 1987) and trap considerable numbers in their horizontal sheet webs (Sunderland, Fraser & Dixon 1986a; Alderweireldt 1994a), which can cover up to half of the surface area of a wheat field (Sunderland, Fraser & Dixon 1986b). Death of pests trapped in webs may contribute to pest control even if the spider does not consume them (Sunderland 1999).

Reviews of the literature have demonstrated that spider density is likely to be increased by within-crop habitat diversification (Samu & Sunderland 1999; Sunderland & Samu 2000). The mechanisms underlying this effect are not fully understood but it is probable that prey diversification is an important factor. Laboratory investigations suggest that the value of different prey types for supporting spider population growth varies greatly between major taxonomic groups (orders and families) of prey (Toft 1995; Sunderland et al. 1996; Beck & Toft 2000; Bilde, Axelsen & Toft 2000) and even between congeneric species [e.g. between the collembolans Isotoma anglicana Lubbock and Folsomia candida Willem (Toft & Nielsen 1997; K.D. Sunderland, J.D. Harwood & W.O.C. Symondson, unpublished data) and Isotoma tigrina Nicolet (T. Bilde & S. Toft, unpublished data)]. To guide the future development of improved practical techniques aimed at enhancing spider populations and increase their impact on pests, quantitative information (from the field or under simulated field conditions) will be needed on (i) the absolute and relative availability of different prey types, (ii) prey preferences, and (iii) diet-related spider population growth rates. In this paper we address the first of these aims and develop the methodology needed to quantify prey availability in the field. For this purpose it is necessary to quantify how many prey are available close to the webs of individual spiders, which are to be found on the ground or up to 10 cm above the ground (Sunderland, Fraser & Dixon 1986a). Sampling methods such as vacuum insect nets (Potts & Vickerman 1974; Moreby et al. 1994), pitfall traps (Nentwig 1982) and large sticky traps (Kajak 1965; Sunderland, Fraser & Dixon 1986b) fail to provide density estimates, or seriously underestimate density (Sunderland & Topping 1995; Sunderland et al. 1995), and lack the spatial precision needed for this study.

To be able ultimately to optimize the spatial distribution of within-field diversified prey resource it is necessary to know how efficiently and dynamically spiders are capable of relocating their webs in relation to temporal and spatial fluctuations of prey availability. Some web-building spiders are known to be efficient in this respect (McNett & Rypstra 1997) but others are relatively insensitive to prey availability (Schaefer 1978). There are, however, very few such studies of spiders in agricultural habitats (Sunderland & Samu 2000).

In this study we (i) quantified spatial and temporal variation in the availability of potential prey to web-building linyphiid spiders in fields of winter wheat, using a combination of ‘instantaneous’ and cumulative sampling techniques, and (ii) determined whether web location is correlated spatially with the abundance of potential prey. As far as we know, this is the first time that these topics have been investigated in an agricultural setting.

Methods

Description of study area

The study sites were winter wheat (cv. ‘Hereward’) fields of approximately 7 ha planted on a predominantly sandy loam soil at Horticulture Research International (HRI), Wellesbourne, Warwickshire, UK (52°12.18′ N, 1°36.00′ W). The fields were farmed according to standard farming practice and no insecticide applications were required during the period of investigation. The study sites were surrounded by fields of spring barley, winter barley, winter wheat and by hay meadows.

Sampling methodology

Sampling was carried out from late April until harvest (in late July/early August) in 1997 and 1998. The abundance of invertebrates (including potential prey) was determined by two different ground-based sampling methods: mini-sticky traps and mini-quadrats. These methods were designed to be small and precise enough to monitor potential prey in the immediate area of a web without the interference of ‘sampling noise’ from invertebrates that would never be encountered by the spiders. They also form a complementary pair of methods; the mini-sticky trap is a passive sampling technique that relies on activity of the prey, but can sample throughout 24 h and catch prey that fall or descend from higher strata, whereas collection from a mini-quadrat is an active sampling method, confined to the ground during a very limited time period (thus providing an ‘instantaneous’ density estimate). The use of mini-quadrats enables the collection of potential prey that are temporarily inactive and hiding under weeds, small stones and at the bases of cereal stems.

The plastic mini-sticky traps were 7·5 cm2 (1·5 cm × 5 cm, 2 mm thick), which is of comparable area to webs constructed by the common erigonid spiders E. atra and Erigone dentipalpis (Wider) (Sunderland, Fraser & Dixon 1986a; Alderweireldt 1994a). The traps were coloured with black acrylic paint (to minimize visual attraction by merging in colour with the ground surface) and were covered on the upper side with a thin acetate sheet coated with Oecotak A5, a non-toxic polybutene-based adhesive (Oecos, Kimpton, UK). The black traps were unlikely to absorb much additional heat from sunlight because they were placed on the ground surface, and were therefore shaded by the winter wheat. After traps were removed from the field, the detachable acetate sheet was placed in a bath of white spirit to dissolve the Oecotak, enabling recovery of trapped invertebrates into alcohol for storage and later identification. The mini-quadrats were circular sampling areas (diameter 10 cm, area approximately 78·5 cm2) defined by a template. These were placed on the ground, surrounding the web or non-web site. All invertebrates were collected from within the mini-quadrat by pooter, searching under loose soil and vegetation.

Sheet webs were located at random within the study fields, and linyphiid spiders present within these webs were captured and preserved for subsequent identification. Vacant webs were not categorized as web sites because spiders are known to leave their webs in active pursuit of prey (Alderweireldt 1994a; Schütt 1995) or abandon their web sites (Samu et al. 1996) in search of more profitable hunting grounds (Vollrath 1985; Gillespie & Caraco 1987). Therefore, only webs in which linyphiid spiders were present were included in the analysis. The webs were removed because spiders are known to be attracted by the presence of silk (Leborgne & Pasquet 1987; Hodge & Storfer-Isser 1997). This also minimized any potential interference with the traps. One mini-sticky trap was centred horizontally on the ground at the position from which a web was removed, and another placed at a random non-web site nearby (up to 30 cm away). Sampling was always carried out in pairs, to enable direct comparisons between each set of traps, which were left in situ for 24 h. Great care was taken not to disturb surrounding vegetation and dislodge arthropods onto the traps during placement and collection.

When sampling by mini-quadrat, the same pair-wise sampling procedure was used as described above for mini-sticky traps. The ground within the mini-quadrat was searched thoroughly, and all arthropods present were transferred immediately into alcohol. No attempt was made to collect arthropods from the crop vegetation above the mini-quadrat. Samples were taken during the daytime between 08.00 hours and 16.00 hours.

Comparisons between web-centred and non-web-centred mini-sticky traps were made from mid-June to mid-July in 1997 (n = 120 paired samples), and a more extensive monitoring programme performed in 1998 at regular intervals from late April until harvest (n = 250). Mini-quadrats were used to sample the abundance of potential prey during July 1997 (n = 52) and from late April until harvest in 1998 (n = 271). No comparisons were made for quadrat catches between the two years due to the different sampling dates.

Weather variables

Meteorological data (maximum and minimum air temperatures; soil temperature; rainfall; hours of sunshine; wind speed; rate of evaporation; relative humidity) were obtained from the weather station located at HRI Wellesbourne, which was within 1200 m of all field sites.

Sample data analysis

To stabilize variances, all sample data were transformed (log10 (x + 1)) prior to analyses. For the purposes of making direct comparisons between mini-sticky traps and mini-quadrats, data were converted into standard unit areas (per cm2). Analysis of variance ( anova) was used to analyse potential prey populations captured by mini-sticky traps and mini-quadrats. Catches of potential prey were pooled over time for comparisons of web vs. non-web sites, and mini-sticky traps verses mini-quadrats. A non-parametric Mann–Whitney U-test was used where the assumptions of anova could not be met, and paired sample t-tests were used to compare numbers of individual prey taxa captured by web and non-web traps. Where less common prey taxa were analysed, data were grouped into means per sampling session and the analysis was performed on mean numbers captured per session.

Potential prey items in the samples were separated from other invertebrates on the basis of size (the linyphiid species found in winter wheat are rarely able to capture and kill prey > 5 mm long) and in relation to information from the published literature for prey of linyphiids in European arable crops (Sunderland, Fraser & Dixon 1986a; De Keer & Maelfait 1987a, 1988; Nyffeler & Benz 1988; Alderweireldt 1994a; Jmhasly & Nentwig 1995; Toft 1995) and the results of laboratory prey acceptability trials (K.D. Sunderland, unpublished data).

Results

Data for web sites of all Linyphiidae were pooled and analysed collectively. Table 1 shows the numbers of each species/genus/subfamily of spider captured at web sites prior to sampling by mini-sticky traps and quadrats (i.e. the web owners). The total numbers of linyphiids recorded exceeded the number of web sites because, occasionally, more than one spider was present in a web.

Table 1. Number of each species/genus/subfamily caught in web sites prior to sampling by mini-sticky traps (MS) and mini-quadrats (MQ) during 1997 and 1998
Species/genus/subfamily1997 MS1997 MQ1998 MS1998 MQ
Erigone atra (Blackwall) 25 5 54 76
Erigone dentipalpis (Wider) 18 4 32 34
Milleriana inerrans (O.P.-Cambridge)  3 1  –  –
Oedothorax spp.  2 2  6 19
Lepthyphantes tenuis (Blackwall) 2525103113
Bathyphantes gracilis (Blackwall)  – 1 37 23
Meioneta rurestris (C.L. Koch) 39 6  8 11
Porrhomma errans (Blackwall)  – –  1  –
Micrargus subaequalis (Westring)  1 –  –  –
Pachygnatha degeeri Sundevall  1 – 1  –
Erigoninae juveniles 10 5 12 10
Linyphiinae juveniles  9 7  9  9
Total captured13356263295

Comparison of prey capture by sticky trap and quadrat

In 1997, web-centred mini-sticky traps (cumulative sampling over 24 h) caught more potential prey items, per cm2, than were found in mini-quadrats (instantaneous sampling during the day) (F1,340 = 5·67, P < 0·001). However, comparisons between individual prey taxa were not possible due to the small number of sampling dates (n = 4 for both mini-sticky traps and mini-quadrats).

Similarly in 1998, significantly more potential prey items were captured, per cm2, by web-centred mini-sticky traps than by web-centred mini-quadrats (F1,1038 = 1199·43, P < 0·001). There were also large differences between the two methods in the numbers of certain invertebrate taxa captured during both 1997 and 1998 (Fig. 1). Significantly more Collembola (F1,1038 = 532·05, P < 0·001), Diptera (Mann–Whitney, U = 0·0, n = 35, P < 0·001), Hymenoptera (U = 64·0, n = 35, P < 0·01) and Araneae (U = 77·5, n = 35, P < 0·05) were caught by web-centred mini-sticky traps than were sampled by web-centred mini-quadrats. There were large differences in the ratio of individual prey taxa captured, per cm2, between the two sampling methodologies (Fig. 1). The small dipterans that were classified as potential prey were especially diverse, with representatives from the families Cecidomyiidae, Lonchopteridae, Phoridae, Sciaridae, Mycetophilidae, Drosophilidae and Dolichopodidae.

Figure 1.

Mean number of potential prey, non-prey and spiders captured (per cm2) by mini-sticky traps and mini-quadrats within web sites during (a) 1997 and (b) 1998. Bars are ± SE. Numbers above columns are the ratios of numbers of prey captured by mini-sticky trap/mini-quadrat.

Location of potential prey and spiders

In 1997, significantly more potential prey items were captured in sites where Linyphiidae had constructed their webs than in non-web sites, as monitored by both mini-sticky traps (F1,238 = 6·80, P < 0·01) and mini-quadrats (F1,102 = 12·22, P < 0·01). Similar results were found in 1998, on different fields, again when monitored by both mini-sticky traps (F1,480 = 96·69, P < 0·001) and mini-quadrats (F1,520 = 75·51, P < 0·001). The temporal abundance of potential prey in 1998 was found to be highly variable (F9,480 = 28·45, P < 0·001) (Fig. 2a).

Figure 2.

Mean number (per trap per day) of (a) total potential prey, (b) all Collembola, (c) Isotoma anglicana (Collembola) and (d) Isotomurus palustris (Collembola) captured by web-centred (filled circles) and non-web-centred (empty circles) mini-sticky traps from May to July 1998. Bars are ± SE.

Collembola, which constituted > 60% of all potential prey items captured by web-centred mini-sticky traps during 1998, were significantly more abundant in web sites (mean per cm2 = 0·60 ± 0·04) than in non-web sites (mean per cm2 = 0·26 ± 0·02) (F1,480 = 109·02, P < 0·001). Similar distribution patterns were also evident when individual species of Collembola were analysed. The two most abundant isotomid collembolans, I. anglicana (sensu. Fjellberg 1980) and Isotomurus palustris Müller, were present in greater densities on web-centred, than on non-web, mini-sticky traps (I. anglicana: F1,480 = 39·69, P < 0·001; I. palustris: F1,18 = 6·41, P < 0·05) and in mini-quadrats (I. anglicana: F1,520 = 29·59, P < 0·001; I. palustris: F1,520 = 22·54, P < 0·001). The temporal capture frequencies of these species, and of Collembola as a whole, on mini-sticky traps varied significantly throughout the monitoring period (Fig. 2b–d) (all Collembola: F9,480 = 45·56, P < 0·001; I. anglicana: F9,480 = 55·60, P < 0·001; I. palustris: F9,480 = 17·22, P < 0·001). Similar significant differences between numbers captured over time were recorded by mini-quadrat sampling (all Collembola: F10,520 = 59·67, P < 0·001; I. anglicana: F10,520 = 99·30, P < 0·001; I. palustris: F10,520 = 77·06, P < 0·001). However, despite this temporal variability, more Collembola (total and isotomid) were captured on all dates at web-centred sites, regardless of sampling method.

The entomobryid collembolans Lepidocyrtus cyaneus Tullberg and Entomobrya multifasciata (Tullberg) constituted 25·4% of total potential prey items on mini-sticky traps and 57·6% in mini-quadrats in 1997, although this difference was not significant. They were relatively less dominant in 1998 (5·6% for mini-sticky traps and 14·3% for mini-quadrats), with no significant difference between the two sampling methods, although there was a significant difference in numbers captured by mini-sticky traps between the two years (U = 6198·5, n = 240, P < 0·05). No comparison was made for quadrat catches between the two years due to the different sampling dates. However, during 1998 the density of Entomobryidae increased logarithmically as the season progressed (Fig. 3). Analysis of covariance indicated that there was no significant difference between the slopes for mini-sticky traps and quadrats (F1,31 = 0·40, P > 0·05) or the y-axis intercepts (F1,31= 0·52, P > 0·05).

Figure 3.

Mean number of entomobryid collembolans (Lepidocyrtus cyaneus and Entomobrya multifasciata) captured per cm2 by web-centred mini-sticky traps and mini-quadrats during 1998. Points are log10 (means) for individual sampling occasions. Mini-sticky trap regression: log10 y = 0·00648x − 0·117, r2 = 0·80, t14 = 7·10, P < 0·001. Mini-quadrat regression: log10 y = 0·00862x− 0·164, r2 = 0·42, t19 = 3·64, P < 0·01.

Linyphiidae were significantly more abundant in web than non-web sites both on mini-sticky traps and in mini-quadrats, and Diptera, Hymenoptera and Coleoptera were more abundant in web than non-web sites within mini-quadrats (Table 2). The sex ratio of spiders captured also varied considerably. Male linyphiids were captured on significantly more occasions than females on mini-sticky traps in web sites (U = 56·5, n = 30, P < 0·01); 3·3 times more male than female spiders were also collected in non-web traps, although these differences were not statistically significant.

Table 2. Differences between web and non-web sites in numbers of potential prey (Diptera, Hymenoptera, Coleoptera, Aphididae) and spiders (Araneae) captured by (a) mini-sticky traps and (b) mini-quadrats, in winter wheat in 1998. The area of a mini-sticky trap (7·5 cm2) is approximately a tenth that of a mini-quadrat (78·5 cm2). Mean number of Araneae captured per web-centred mini-quadrat are presented as (x + 1) to account for the web owner
Variable t d.f.PMean per web site ± SEMean per non-web site ± SERatio web/ non-web
(a) Mini-sticky traps      
Collembola 7·5214< 0·001 4·48 ± 0·31 1·93 ± 0·182·32
Diptera 1·0214  0·326 0·29 ± 0·04 0·25 ± 0·031·16
Hymenoptera 1·3214  0·209 0·14 ± 0·03 0·08 ± 0·021·75
Coleoptera 2·0214  0·063 0·12 ± 0·02 0·06 ± 0·012·00
Aphididae 1·4914  0·157 0·20 ± 0·04 0·11 ± 0·021·81
Araneae 4·2614  0·001 0·30 ± 0·04 0·08 ± 0·023·75
(b) Mini-quadrats      
Collembola10·8519< 0·00127·66 ± 1·8113·66 ± 1·122·02
Diptera 3·2419  0·004 0·20 ± 0·03 0·08 ± 0·022·50
Hymenoptera 3·9419  0·001 0·08 ± 0·02 0·04 ± 0·012·00
Coleoptera 5·2519< 0·001 0·75 ± 0·07 0·34 ± 0·042·21
Aphididae 1·4019  0·117 0·21 ± 0·04 0·16 ± 0·031·31
Araneae32·4719< 0·001 1·78 ± 0·08 0·19 ± 0·049·37

Linyphiid density, monitored by mini-quadrat sampling, increased logarithmically with time (Fig. 4). Analysis of covariance indicated a significant difference between the slopes of the web and non-web regressions (F1,36 = 7·17, P < 0·05) and the y-axis intercepts (F1,36 = 76·51, P < 0·001), showing that linyphiid density was increasing more rapidly at web sites. The number of linyphiids captured by mini-sticky traps (both web and non-web) showed a similar pattern (Fig. 5) but analysis of covariance indicated no significant difference between the two slopes (F1,26 = 1·57, P > 0·05) or y-axis intercepts (F1,26 = 0·32, P > 0·05), but the two slopes are significantly different from zero (F1,26 = 14·28, P < 0·01). The log10 number of spiders captured by these methods within web sites also increased when regressed against maximum air temperature (mini-sticky traps: r2 = 0·27, t14 = 2·18, P < 0·05; mini-quadrats: r2 = 0·57, t19 = 4·91, P < 0·001) and soil temperature (mini-quadrats: r2 = 0·53, t19 = 4·47, P < 0·001).

Figure 4.

Mean number of Linyphiidae sampled by web and non-web-centred mini-quadrats during 1998. Data for web quadrats are presented as log10(x + 1) to account for the web owner. Web regression: log10y = 0·00621x − 0·080, r2 = 0·72, t19 = 6·78, P < 0·001. Non-web regression: log10y = 0·00330x − 0·360, r2 = 0·64, t19 = 5·63, P < 0·001.

Figure 5.

Mean number of Linyphiidae captured by web and non-web-centred mini-sticky traps during 1998. Web regression: log10y = 0·00388x − 0·395, r2 = 0·37, t14 = 2·73, P < 0·05. Non-web regression: log10y = 0·00195x − 0·250, r2 = 0·45, t14 = 3·25, P < 0·01.

A negative relationship was found to exist between the number of Linyphiidae and Collembola captured by web-centred mini-sticky traps (log10 Linyphiidae =−0·721 log10 Collembola + 0·636, r2 = 0·40, t14 = 2·97, P < 0·05). No such relationship was found using data from mini-quadrats.

Between-year variation in arthropod density

From the comparison of 4 consecutive weeks (mid-June to mid-July) of mini-sticky trap data collected during 1997 (mean number of total potential prey captured per web site = 3·12 ± 0·22, non-web site = 2·38 ± 0·20) with a data set for the same calendar period during 1998 (mean number per web site = 5·69 ± 0·50, non-web site = 3·95 ± 0·27), the mean number of arthropods captured during the second year was significantly greater (F1,476 = 20·55, P < 0·001).

Between-year variation in weather

Despite the large variation in the numbers of arthropods captured between years, there was no evidence to suggest that meteorological factors were responsible for such variability. Comparisons of data from 1997 and 1998 indicated that no category of meteorological data varied significantly in the 4 weeks prior to sampling, or during the months in which the study was undertaken (Table 3).

Table 3. Mean daily meteorological values (± SE) for climatic conditions in the months May, June and July during 1997 and 1998
 Mean per day
Weather variable19971998FP
Maximum air temperature (°C)18·23 ± 0·4017·52 ± 0·401·470·227
Minimum air temperature (°C) 7·76 ± 0·42 8·34 ± 0·351·070·302
Soil temperature at 10 cm (°C)14·85 ± 0·3414·22 ± 0·371·550·214
Rainfall (mm) 1·95 ± 0·34 2·17 ± 0·600·100·747
Sunshine (h) 5·85 ± 0·40 4·83 ± 0·363·650·057
Wind speed (m.p.h.) 8·50 ± 0·45 8·35 ± 0·390·070·795
Rate of evaporation 2·60 ± 0·16 2·32 ± 0·161·630·203
Relative humidity75·93 ± 1·0978·41 ± 0·992·830·094

Discussion

Quadrats (Bilsing 1920; Edgar 1969) and sticky traps (Kajak 1965; Sunderland, Fraser & Dixon 1986b; Riechert 1991; Bradley 1993; Gillespie & Tabashnik 1994; Marshall 1997) have been used previously to monitor the potential prey of spiders, but very few of these studies were in agricultural habitats. This appears to be the first study where both methods were applied simultaneously to individual web sites to obtain a balanced and spatially precise assessment of the availability of potential prey. Overall, it revealed that Collembola were the most abundant group of potential prey. Biomass of prey items was not recorded in this study, but it is likely that Diptera, Hymenoptera, Coleoptera and Aphididae would increase in relative importance if biomass were to be taken into account. Densities of potential prey (per cm2) were usually greater on mini-sticky traps than in mini-quadrats. This is, in part, because the latter offer an instantaneous measure of prey availability, whereas the former represent an availability rate (per 24 h in this case). It is, however, also likely that the cumulative differences are related to diel cycles in availability of prey to spiders in webs on and near the ground (Vickerman & Sunderland 1975; Leathwick & Winterbourn 1984; Sunderland 1996). Nocturnal prey availability could be an important aspect of the trophic biology of agricultural linyphiids because many species are particularly active at night (Thornhill 1983; De Keer & Maelfait 1987b; Alderweireldt 1994b). The ratio of aerial species, such as Diptera and Hymenoptera, sampled by mini-sticky traps in relation to mini-quadrats was particularly high, indicating the likelihood of underestimating the presence of such species if quadrat sampling is used alone. The ratio of Aphididae captured by mini-sticky traps in relation to quadrats was also relatively high, probably due to the presence of alate aphids descending into the crop, and the high frequency with which aphids are known to fall from the crop (Sunderland, Fraser & Dixon 1986b; Sopp, Sunderland & Coombes 1987). It is clear, therefore, that use of daytime mini-quadrats alone would have seriously underestimated the abundance and diversity of prey available to spiders.

There was considerable temporal variation in prey availability. Less than four prey items were available per mini-sticky trap per day (c. 0·5 cm−2) in web sites during May and July. This is approximately equivalent to four items per day entering the 8-cm2 web of an adult female Erigoninae spider (e.g. E. atra). If the 74-cm2 webs of adult female Linyphiinae spiders (e.g. L. tenuis) were on the ground they would receive a mean of 37 items per day, but as they are located up to 10 cm above ground they will receive only falling prey, and therefore fail to benefit from the traffic of ground-active prey. This does not necessarily mean that these linyphiids would receive less than this number of prey items, because their location may allow them to intercept and capture more flying species than they would at ground level. Solid horizontal sticky traps at 10 cm above ground caught half as many potential prey per unit area as traps placed on the ground (Sunderland, Fraser & Dixon 1986a), but solid traps might not be appropriate in an aerial location. Immature Linyphiidae, which constitute the vast majority of spiders during the growing season (Sunderland & Topping 1993; Topping & Sunderland 1998), have much smaller webs (16 cm2 for Linyphiinae and 3 cm2 for Erigoninae; Sunderland, Fraser & Dixon 1986a) and so will receive commensurately fewer prey. In laboratory studies, immature E. atra that received approximately one collembolan per day at 20 °C had markedly slower development rates than those that received an ad libitum supply (De Keer & Maelfait 1988). Immature E. atra in the current study are estimated to have received approximately 1·5 prey items per day in May and July, and so their development is likely to have been food-limited. De Keer & Maelfait (1988) also showed that adult female E. atra receiving less that three adult Drosophila melanogaster Meigen (Diptera) per day at 20 °C produced fewer eggs than those provisioned at higher rates. In May and July, adult female E. atra in our study would have received about four prey items per day, which were mostly small Collembola. The biomass of prey available was probably therefore less than the equivalent of three Drosophila and reproduction is likely to have been suboptimal. The situation was different in June, when prey availability increased to 12 prey items per trap per day (c. 1·6 cm−2). The seasonal pattern of prey availability recorded in this study matches the seasonal pattern of hunger assayed by Bilde & Toft (1998) for the linyphiid spider Oedothorax apicatus (Blackwall) in a Danish winter wheat field. They recorded hunger equivalent to 7 days of starvation at 20 °C in May and July, but a value of 3–4 days starvation during June. These results suggest that spiders are unlikely to exhibit prey preferences in May and July, when they may need to capture whatever they encounter in order to survive. In June, however, when prey availability is higher, it is possible that they will reject cereal aphids (which were found to be less-preferred prey in laboratory trials; Toft 1995; Beck & Toft 2000) in favour of alternative prey.

The number of potential prey caught per trap during a 4-week period was significantly greater in 1998 than in 1997 and this difference did not appear to be directly due to weather. However, different fields were investigated in the two years and so interfield variability may explain the observed differences. A full study of interfield variation of potential prey availability (using the methods developed here) would be worthwhile, as it might point to farming practices that are favourable and others that are deleterious in relation to fostering the alternative prey of generalist predators. This could include a comparison of winter and spring cereals, as plant taxa that are important food resources for arthropod herbivores (= potential prey for generalist predators) occur at greater densities in spring cereals (Hald 1999).

The linyphiid spiders at our study sites were not locating their webs randomly with respect to the spatial distribution of their food. This was true for all fields and sampling methods in 1997 and 1998. More potential prey were available in the web sites of spiders than in random non-web sites located up to 30 cm away from the web. We do not have information at this stage to determine whether Linyphiidae are responding directly to food availability or whether their aggregation in prey-rich patches is a secondary outcome of a response to other microhabitat variables such as high humidity and structural complexity of the vegetation. Further studies would be useful, matching web and non-web sites within crop rows and within spaces between rows, for example. It is possible that linyphiids were simply responding passively to a lack of prey, relocating at random until they reached areas of high prey availability, where further movements were arrested. In fact, if Linyphiidae were selecting sites on the basis of quality, the results suggest that their efficiency in selecting high-quality web sites improved with progression of season. Our results justify a full study of the determinants of microhabitat selection by linyphiid spiders in winter wheat. There are examples in the literature of microhabitat selection being primarily driven by food, or microclimate, or physical structure of the microhabitat, or avoidance of conspecifics and enemies; the prime determinant varies with habitat and species of spider (Samu & Sunderland 1999; Sunderland & Samu 2000). The fine-grain variation in the distribution of linyphiids and their prey described in this study indicates that mini-sampling techniques can be valuable for investigating spatial dynamics and predator–prey interactions in agro-ecosystems, especially for small ‘sit-and-wait’ strategist predators. There are a growing number of studies of the spatial dynamics of predators and prey in agro-ecosystems that employ nested sampling grids at a range of degrees of resolution, suitable for analysis by new techniques such as Spatial Analysis by Distribution Indices (Perry 1995, 1998). For large, highly mobile, carabid beetles, grid scales ranging from 16 m down to 0·25 m have been demonstrated to be appropriate (Bohan et al. 2000). Our study suggests that for smaller, less-mobile, predators important information will be lost unless finer scales are also included. Therefore, the scale should be chosen according to the species being investigated.

Numbers of both spiders and Entomobryidae (Collembola), captured by mini-sticky traps and mini-quadrats, increased logarithmically as the season progressed. At the same time, trap capture frequencies of isotomid Collembola (Fig. 2c,d) showed a sharp decline in July, when spider activity density was at a maximum. While it is possible that predation by spiders (and other natural enemies) was affecting population densities of isotomid Collembola (see below) this does not explain why capture frequencies of the Entomobryidae continued to increase. One possible explanation is that the Entomobryidae have powerful protection mechanisms against predators (Bauer & Pfeiffer 1991). Figure 3 clearly indicates that entomobryid activity must have been very low, with relatively few encountering webs, because the cumulative capture rates (mini-sticky traps) over time were not significantly different from the instantaneous densities (mini-quadrats). This too could contribute to lower levels of predation. Alternatively, Entomobryidae could be a non-preferred prey item, as the nutritive value of different species of Collembola is known to vary (Toft & Nielsen 1997). However, a number of other factors could be responsible for the differential trends in capture rates for the different groups of Collembola, including responses to changing climatic conditions, or some species of Collembola locating to different microhabitats such as the large cracks which appear in the soil late in the season. The negative relationship between numbers of spiders and Collembola captured by mini-sticky traps indicates that, as Collembola density decreases, spider activity density increases. Increases in predator activity, in response to reduced prey availability, have been found amongst other generalist predators such as carabid beetles (Chiverton 1984). However, increasing spider density may also be a factor. Although it is possible that the associations between Collembola and linyphiid numbers were in part a result of predation, quantitative immunological studies would be needed to confirm such a link in the field (Symondson et al. 1996; Bohan et al. 2000).

A surprisingly large number of linyphiid spiders was captured on the mini-sticky traps. These were not the original web owners returning to the web site because we removed and killed them as a routine part of the sampling protocol. Peak linyphiid densities (including immatures) in winter wheat are typically about 100 m−2 (Dinter & Poehling 1992; Sunderland & Topping 1993; Volkmar et al. 1994), which is equal to one spider per 100 cm2, or 0·08 spiders per trap. The actual capture rate per trap peaked at 0·6 linyphiids day−1 (web sites in July), which implies that either these spiders are extremely active throughout the field, or their activity is lower and concentrated in areas that they find attractive. Support for the latter hypothesis comes from the fact that peak capture rate on non-web mini-sticky traps was only 0·3 trap−1 day−1. Thus, the sampling programme has detected the aggregated and dynamic nature of spider distribution in winter wheat. The dynamic aspect is likely to be due to males searching for females, and females searching for new web sites. Male linyphiids are known to be, in general, more active than females, e.g. sex ratios in pitfall traps are strongly biased towards males (Sunderland 1987; Topping & Sunderland 1992), and they can be attracted to sex pheromones secreted onto web silk by the female (Watson 1986). Adult female linyphiids, such as L. tenuis, compete strongly for desirable web sites, a larger female being able to evict a smaller one. Web site abandonment from this and other causes results in the mean duration of an individual L. tenuis in a web site being less than 2 days (Samu et al. 1996). These mechanisms probably contributed to the high capture rate on mini-sticky traps and are also likely to have caused a fairly high encounter rate between individual spiders. Although, in general, spiders are highly cannibalistic (Edgar 1969; Wise 1993), it is known that web-weavers (including Linyphiidae) are much less so than hunters such as Lycosidae, Salticidae and Thomisidae (Nyffeler 1999), and so the encounters implied by the results from this study probably did not result in high rates of mortality. We suggest, however, that mini-sticky traps could, more generally, be a useful tool for quantitative arachnological investigations of intra- and interspecific interactions.

This study has shown that money spiders locate their webs in prey-rich areas of winter wheat fields. In spite of this behaviour, which enables them to maximize their access to food, it seems likely that prey availability in May and July is not sufficiently great for them to realize fully their potential population growth rate. Studies are needed to find practical modifications to current farming practice that will achieve enhanced prey availability, especially at times of year when prey densities are low. The mini-sampling techniques developed here, especially when used in combination, will enable prey availability to be monitored more efficiently in such prey-enhancement studies.

There is no certainty, however, that increased prey and spider density will always translate into improved pest control (Sunderland & Samu 2000). Indeed, the increased availability of high-quality alternative food could, to some extent, divert spiders from feeding on pests. Clearly, there is a need for manipulative field studies to investigate rigorously the interactions between spiders, pests and alternative prey before sound recommendations can be made to optimize the spider predation component of pest control.

Acknowledgements

We are very grateful to John Fenlon (HRI) for statistical advice, Sally Mann (HRI) for meteorological data, and Jørgen Axelsen (National Environmental Research Institute, Silkeborg, Denmark) for Collembola identification. J.D. Harwood was funded by a joint Cardiff University and HRI studentship, and K.D. Sunderland by the UK Ministry of Agriculture Fisheries and Food.

Received 9 October 1999; revision received 6 June 2000

Ancillary