Peter J. Mayhew, Department of Biology, University of York, York YO10 5YW, UK (e-mail firstname.lastname@example.org).
1Insect parasitoids comprise a large fraction of terrestrial biodiversity. Because of this diversity, species-level conservation of most parasitoid species is impractical and habitat conservation must substitute. However, habitat indicators of parasitoid abundance and diversity are poorly known.
2To identify such habitat indicators, parasitoid wasps in four ichneumonid subfamilies were sampled in the field herb layer of 15 woodlands in the Vale of York, UK, using Malaise traps. The catch was related to vegetation characteristics.
3A total of 1543 individuals in 60 species was recorded, representing 36% of UK species in the taxa sampled. Parasitoids tended to be more abundant and species rich in woodlands with a high broadleaf content and tree species richness. This pattern was observed in the ichneumonid subfamilies Pimplinae, Poemeniinae and Diacritinae.
4However, the ichneumonid subfamily Diplazontinae was found to vary in abundance and richness within rather than between woodlands and showed no association with measured habitat variables.
5Reserve selection analyses indicated that coniferous woodlands, and woodlands with a low abundance and richness of parasitoids, none the less can contribute to maximizing parasitoid diversity at the landscape scale.
6Synthesis and applications. At the individual woodland scale, broadleaved woodlands with high tree species richness appear best for conserving parasitoid abundance and diversity. At a landscape scale however, a variety of woodland habitat types can maximize diversity of all parasitoid taxa. We hypothesize that the degree of association between parasitoid abundance and diversity, and characteristics of the vegetation within habitats will decrease with an increase in the number of trophic links that separate them.
Little studied but ecologically important taxa make up much of the planet's biodiversity (May 1988). In most cases neither the time nor resources will exist to plan conservation strategies for each individual species belonging to such taxa (Hughes, Daily & Ehrlich 2000). An alternative to species-level conservation is a habitat-based approach, where priority habitats for target taxa are identified and conserved, therefore protecting whole communities.
Parasitoids are insects, mainly wasps (Hymenoptera), that develop to maturity by feeding on the body of another host arthropod, eventually killing it. The parasitoid wasps are extremely species rich and essential to the maintenance of species diversity in other organisms (LaSalle & Gauld 1991). Yet, despite their ecological influence, relatively little is known about their biology, distribution and diversity. For large-scale conservation of parasitoids, habitat conservation may be the only practical approach (Hochberg 2000). However, there are few studies of parasitoid diversity at the habitat scale on which to base conservation recommendations.
Greater plant architectural complexity would be expected to increase the species richness of herbivores, which form the hosts of many parasitoids (Strong, Lawton & Southwood 1984). One could therefore predict that habitats rich in plant species and with complex plant architecture would be richer in parasitoids. However, host specificity amongst parasitoids is known to vary across different habitats (Price 1991) and, as a result, findings from studies at the host level do not easily translate into predictions within whole habitats (Menalled et al. 1999). Parasitoid diversity at the habitat level will be an emerging property derived from numerous biological processes (Price 1991), not all of which will operate in the same way across taxa.
Much of what is known concerning the diversity of parasitoid communities comes from literature-based studies of parasitoids reared from their individual insect hosts (Hawkins 1994). Of the little work concerning the effect of habitat characteristics, most has concentrated on agro-ecosystems as a result of interest in the potential use of parasitoids for biological control (Altieri, Cure & Garcia 1993). These agricultural studies may not simply translate to other systems and thus there is a need for studies at the habitat level to identify general patterns and, ultimately, the mechanisms behind them.
Woodland habitats are among the most stable elements of human-dominated landscapes, which now cover a substantial proportion of the Earth's surface. As such, they play an important role in the maintenance of biodiversity (Petit & Usher 1998). We compared the parasitoid faunas of 15 woodlands in the Vale of York, UK. We tested the null hypothesis that parasitoid abundance and diversity are the same across woodland types. If woodlands do differ in parasitoid abundance and diversity, then it may be possible to identify habitat indicators that explain a significant component of the variation. We tested the null hypothesis that parasitoid abundance and diversity show no association with woodland vegetation variables. If vegetation indicators of parasitoids can be used to identify woodlands that are particularly valuable for parasitoid abundance and diversity, it does not necessarily follow that, at a landscape scale, it is best to conserve only those woodlands with the highest abundance and diversity of parasitoids. To address this scale issue, we tested the null hypothesis that conserving only one or a variety of woodland types in the landscape has no effect on the inclusion of parasitoid species.
Materials and methods
The study was conducted in the Vale of York, UK. To reduce variation in factors such as topography, weather and soil type, the area within which woodlands were chosen was limited. Woodlands were all larger than 2 ha as smaller patches of habitat may not be capable of supporting insect communities distinct from surrounding habitats (Levenson 1981). Although no maximum size was determined for the selection of woodlands, selection was limited to farm woodlands that were relatively small, with none of those used exceeding 20 ha.
Fifteen farm woodlands were chosen to include a range of tree species types giving different woodland habitats (broadleaved, mixed and coniferous habitats; Table 1). Broad classification of woodland habitats in Britain is based on canopy type (JNCC 1993) and this information is also displayed on UK Ordnance Survey maps, so these habitat definitions are directly relevant to conservation.
Table 1. Location, area and habitat type of the study woodlands. Habitat is as given on 1:25 000 Ordnance Survey maps
SE 562 450
SE 629 417
SE 660 501
SE 563 449
SE 629 413
SE 644 408
SE 749 433
Many Gates Plantation
SE 693 537
SE 609 438
New Drive Plantation
SE 753 427
SE 732 442
SE 733 445
SE 603 443
SE 696 539
SE 644 453
sampling the parasitoid hymenoptera
Four closely related subfamilies of the Ichneumonidae were chosen for study: Diplazontinae, Pimplinae, Diacritinae and Poemeniinae. These subfamilies have useable species-level keys (Beirne 1941; Fitton, Shaw & Gauld 1988), are known to be abundant in many habitats and have a wide range of hosts. All species in these four subfamilies are winged parasitoids.
The Diplazontinae is a relatively small subfamily, with 55 British species in 12 genera (Broad 2005). All species are thought to be endoparasitoids of aphidophagous Syrphidae (Diptera), with host records available for approximately 50% of species. Oviposition is into the host egg or larva and emergence of the adult is from the host puparium (Fitton & Rotheray 1982). The characteristic of allowing the host to develop after parasitization defines them as koinobionts, as opposed to idiobionts (see below). Adult females commonly occur near aphid colonies, searching for hosts and feeding on syrphid eggs and larvae, or on aphid honeydew (Rotheray 1981a). Males of some species have been found to form swarms beneath large tree canopies and others are found at aphid colonies and flowers (Rotheray 1981b).
The subfamily Pimplinae exhibits a wider range of biologies and hosts than any other subfamily of the Ichneumonidae (Fitton et al. 1988). It is also probably the most extensively studied subfamily, largely because many species are parasitoids of economically important pests. In the British Isles there are 103 species in 30 genera (Broad 2005). The subfamily is currently divided into three monophyletic tribes, the Delomeristini, Ephialtini and Pimplini, which demonstrate distinct ecologies (Gauld, Wahl & Broad 2002; Broad 2005). In the British Isles the Delomeristini consists of nine species in two genera. Host information for this tribe is poor but some species in the genus Delomerista appear to parasitize cocoons of sawflies and other ichneumonids, while those in the genus Perithous appear to be associated most frequently with aculeate Hymenoptera (Fitton et al. 1988). The Ephialtini consists of 75 species in 24 genera. Host groups for this tribe are varied and are found across the orders Coleoptera, Hymenoptera, Diptera and Arachnida. The majority of species are ectoparasitoids of holometabolous insect larvae, pre-pupae and pupae (Fitton et al. 1988). The Pimplini consists of 20 species in three genera. All are chiefly idiobiont endoparasitoids of the pupae of Lepidoptera (Fitton et al. 1988). Idiobiont parasitoids permanently paralyse or kill the host before the parasitoid egg hatches. The host is consumed in the location and state in which it was attacked (Askew & Shaw 1986).
The subfamilies Poemeniinae and Diacritinae were previously grouped within the Pimplinae (Fitton et al. 1988) and were included in this study for that reason, although they are now recognized as distinct subfamilies (Wahl & Gauld 1998; Gauld et al. 2002). The Poemeniinae contains six species in Britain (Broad 2005). Members of the Poemeniinae develop as idiobiont ectoparasitoids and at least some of these species are most often collected in association with dead and standing timber (Fitton et al. 1988). Diacritinae is one of the few subfamilies for which the larvae are completely unknown (Wahl & Gauld 1998). In Europe only one species is known, Diacritus aciculatus (Fitton et al. 1988).
Two Malaise traps were set up in each woodland during July and August 2003, the main Hymenoptera flight season. The two-trap design was used to investigate the influence of trap siting on catch and to provide a measure of within-woodland variability. The following rules were used to locate trapping areas: to control for aspect we sampled two areas 10 m either side of the mid-point of the southern edge of the woodland. To account for edge effects, these areas were located 10 m in from the edge of the woodland habitat. In each of these areas we marked out a 20 × 10-m quadrat within which we set a Malaise trap. Traps were open for 2 consecutive weeks in every 4, 1–15 July and 29 July−12 August The trap bottle was changed after 1 week to give two samples of 1-week duration in each month.
In order to identify possible habitat indicators of parasitoid abundance and diversity, a suite of habitat variables was measured (see Table S1 in the supplementary material).
Vegetation was sampled in late July/early August at the site of the Malaise trap and then more widely across each woodland, using quadrats at two scales: 20 × 20 m for the canopy trees and shrub layer and 2 × 2 m for the field and herb layer. Each Malaise trap was at the centre of a 20-m quadrat to give a detailed record of the vegetation present around the trap. At random coordinates within this quadrat, five 2-m quadrats were surveyed. Within the woodland as a whole, two more 20-m quadrats were surveyed, their location being determined by generating random coordinates within the north-east and north-west quarters of the woodland and using these as the south-west corner of a quadrat. Again, at random coordinates within the larger quadrats, five 2-m quadrats were surveyed.
In the 20-m quadrats all tree and woody shrub species taller than 1 m were counted and identified to species. All woody shrubs less than 1 m in height and herbs within the 2-m quadrats were identified to species. Ferns, fungi, [grasses + sedges] and [mosses + lichens + liverworts] were not identified to species but were grouped thus. A visual estimate of the percentage total vegetation cover for the herb layer was made. An estimate of canopy cover was taken visually from the south-west corner of each 2-m quadrat using a gridded acetate. The acetate was held at arm's length towards the canopy and the number of grid squares in which canopy cover was seen were counted. This number was then divided by the total number of squares on the grid.
Plant height diversity and plant architectural diversity were measured within the field–herb layer using the method of Southwood, Brown & Reader (1979). A 2-m high sampling pin was marked at height intervals of 5 cm, 10 cm and successive 20 cm until 1 m and 25-cm intervals thereafter. The total number of touches in each height category was recorded and used to provide a measure of plant height diversity. Plant architectural diversity was measured by recording the number and types of plant structures that were touching the pin (see Table S2 in the supplementary material). Five samples were taken using the pin at random coordinates within each 2-m quadrat. The diversity of both plant height and plant architectural diversity was estimated using the log series diversity index α, following Southwood et al. (1979).
Species accumulation curves were calculated to indicate the extent to which the regional fauna was sampled, and how many species each extra woodland would add. Curves were generated by running 15 sequences of accumulations, one with each of the 15 woodlands at the start and then adding one woodland at random until all 15 were used.
Comparison of woodlands
To test the null hypothesis that parasitoid abundance and diversity are the same across different woodlands, a mixed-model nested analysis of variance (anova) was carried out on the data for individual traps in each of the 4 sampling weeks. Factors in the analysis were trap, woodland, habitat and week (time). Habitat was a fixed effect classifying woods as either predominantly broadleaved (> 50%) or predominantly coniferous (> 50%). Traps (a random effect) were nested within woods (a random effect) and woods within habitat. This was a repeated-measures design so week was included in the analysis to account for the problem of pseudoreplication by repeat visits to the same trap sites. Non-significant factors and interaction terms were sequentially removed from the model to leave only those explaining significant proportions of variation in the data. Abundance data were log10(x + 1) transformed to reduce right skew.
In addition to exploring patterns in abundance and richness, the Simpson's index (D) was used to explore patterns in the evenness of parasitoid assemblages in relation to habitat variables. Simpson's index is recommended as a good estimate of diversity for relatively small sample sizes (Magurran 2004). The index is heavily weighted towards the most abundant species in the sample while being less sensitive to species richness. The index is expressed here as the reciprocal (1/D) and the value of the measure will rise as the assemblage becomes more even.
To test the null hypothesis that woodland vegetation variables are not associated with parasitoid abundance and diversity, parasitoid abundance, richness and 1/D were used as response variables in a backwards stepwise regression analysis. Explanatory variables were the habitat variables measured from each woodland. The stepwise procedure was allowed to select the predictors using the criteria probability of F to enter ≤ 0·05 and probability of F to remove > 0·05.
The landscape scale
To test whether a variety of woodland types in the landscape maximizes the conservation of parasitoid species, we conducted two tests. In the first test we asked which was the best selection of woods for maximizing parasitoid diversity. Two simple reserve selection algorithms were used. In the first algorithm, based on species richness, the woodland with the highest species richness was identified. Then all the species that were found in that woodland were removed from the data set and the woodland with the highest remaining richness was identified. This process was run five times to select the five most important woodlands for conservation. To determine whether the woodlands selected by this analysis performed better than a random collection of woodlands, we ran 1000 repetitions of the data for each higher taxon to calculate the mean number of species conserved by randomly selecting five woodlands. In the second algorithm, where selecting woodlands in order of their contribution of rare species to the data set, the system was the same as for species richness except species were weighted by the number of woodlands in which they occurred. Species found in all woodlands were worth 1/15, those found in two woodlands 2/15, and so on until those found in just one were worth 15/15. For both algorithms we then simply examined post-hoc what woodland types were represented in the optimal reserve selections.
As the results of this test suggested that a mixture of woodland types, both coniferous and broadleaved, would be best at preserving landscape-level diversity, we conducted a further reserve selection experiment to test the generality of this finding. In this further test, we selected five woodlands at random from the data and counted the number of species included. However, as well as choosing woodlands from the data set as a whole, which allowed a mixture of woodland types, we also restricted choice to either only coniferous woods or only deciduous woods. We ran 105 repetitions for each set of woodland types.
parasitoid species and individuals
The 30 Malaise traps captured a total of 1543 individuals and 60 species over the 4 weeks (see Table S3 in the supplementary material; the full data set is available from P. J. Mayhew).
The species accumulation curve for all species (Fig. 1) and sites did not reach an asymptote. The accumulation curves for Diplazontinae and Poemeniinae (Fig. 1) reached an asymptote relatively rapidly, suggesting that the results of analyses for these taxa were robust. For the Pimplinae the curve did not plateau after sampling at 15 woodlands (Fig. 1). Notably, the rank order of species richness of pimplines and diplazontines changed as more woods were sampled: if only a few woods were sampled, more diplazontine species were found than pimplines. However, if many woods were sampled, more pimpline species were found than diplazontines (Fig. 1).
comparison of woodlands
Woodlands differed in parasitoid abundance and richness but different taxa were found to differ in the scale of their response (Table 2). Preliminary analysis of data for the pimpline tribes, Pimplini and Ephialtini, highlighted differences between them, therefore separate analyses were conducted for these taxa (only two individuals of the third tribe, Delomeristini, were caught, therefore they were not considered separately). Woodlands differed significantly in the abundance of all species combined and Diacritus aciculatus, and in the abundance and richness of the Pimplinae, Pimplini and Poemeniinae. For all species combined, Diacritus aciculatus, the Pimplinae and Pimplini, woodland habitats (broadleaved or coniferous habitat) differed significantly in parasitoid abundance and/or richness (Table 2). No significant differences between woodlands or habitat were found for the Diplazontinae and Ephialtini, although there were significant differences for parasitoid abundance and richness between traps, within woodlands, for these taxa (Table 2). For all taxa, abundance and richness varied significantly between weeks (Table 2).
Table 2. Results of the mixed-model nested anovas. Non-significant factors were sequentially removed from the initial model to leave only those factors explaining significant variation in the data. Habitat was a fixed effect, wood and trap were random effects. Wood(habitat) indicates that the factor wood is nested within the factor habitat. A non-significant result is signified as NS; *P < 0·05 **P < 0·01 ***P < 0·001
Overall, the best indicators of a high parasitoid abundance, richness and diversity were tree/shrub species richness and broadleaf content (Table 3 and Fig. 2). These habitat variables showed a consistently positive relationship with parasitoid abundance, richness and 1/D. Canopy cover was found to show a consistent negative relationship with a number of parasitoid variables. The Diplazontinae were found to show little association with the vegetation.
Table 3. Results of backwards stepwise regression. Parasitoid data used were those combined over the 4 trapping weeks and two traps for each woodland. The values of the standardized coefficients (Beta) are given with the significance of the variables in the final model. Independent variables are listed in order of decreasing importance (measured by number of asterisks; ***P < 0·001, **P < 0·01, *P < 0·05)
Where selecting woodlands in order of priority based on either their contribution to species richness or rarity, some coniferous woodlands and woodlands with low species richness (e.g. NDPL and HAPL; see Table S3 in the supplementary material) were included in the best selection of five (Table 4). Only four runs of the reserve selection analysis were needed to ensure a set of woodlands was attained that contained all the species of Diplazontinae present in the data set (Table 4). Comparing the performance of the reserve selection algorithm for species richness against the selection of five random woodlands, the algorithm performed significantly better for each taxon (Table 5).
Table 4. Results from reserve selection algorithms based on richness and rarity. Richness scores are the number of species conserved by the addition of that woodland. Rarity scores represent the sum of the values for all species present in the woodland and not already represented in a previously selected woodland. Habitat categories for each woodland are given in parentheses; B, broadleaf; C, coniferous
Table 5. A comparison of species richness conserved through the selection of 5 woodlands by either a reserve selection algorithm or random selection
The mean (2·5 and 97·5 percentiles) number of species included in five randomly selected woods was 39·42 (38, 41) if only coniferous woods were selected, 42·73 (40, 46) if only deciduous woods were selected, and 42·48 (38, 47) if there was no restriction on choice of woods. Both the best (50 parasitoid species) and worst (36 parasitoid species) selection of five woods were mixtures of broadleaved and coniferous woods.
In this study woodlands were found to differ in their parasitoid abundance and richness, but parasitoid taxa were found to differ in their response to habitat variables. For certain taxa, tree/shrub species richness and broadleaf content were identified as possible indicators of high parasitoid abundance and diversity. However, the selection of woodlands that maximizes parasitoid diversity at a landscape scale includes both deciduous and coniferous woods.
Despite the fact that parasitoids are always at least two trophic levels above the vegetation, it appears that, for some taxa at least, vegetation characteristics may predict parasitoid abundance and diversity. This is important because it would generally be impractical to use the hosts as indicators. Often we have little or no host–parasitoid information and the potential host range of a single parasitoid species, let alone whole taxa, makes sampling all hosts impractical.
In this study, although broadleaf content was found to be a useful indicator of abundance and diversity in some taxa, tree/shrub species richness was found to be the most important indicator of high parasitoid diversity. This result is in agreement with Sperber et al. (2004), who found an increase in the number of parasitoid families with tree species richness, which was ascribed to an increased heterogeneity and availability of resources (Sperber et al. 2004). Sääksjärvi et al. (2006) also found a positive association between plant species richness and ichneumonid species richness in Amazonian forest. There was no relationship in this study, or that of Sperber et al. (2004), to the species richness of the ground vegetation. This is in contrast with results from cropping systems (Risch 1979; Altieri 1984; Letourneau 1987) and suggests that results from such simplified agricultural systems may not translate to more complex ecosystems.
Although this study was not designed to answer questions regarding the biological mechanisms underlying the resulting patterns, we can hypothesize as to what they may be. Notably, those taxa whose abundance and diversity differed between woodlands and which were associated with the vegetation, particularly the pimpline tribe Pimplini, are parasitoids of herbivores. The most abundant species of Pimplini caught in this study were Pimpla insignatoria (Gravenhorst) (171 individuals captured), Pimpla contemplator (Müller) (142) and Pimpla flavicoxis Thomson (51). Together, these three species accounted for 67% of all the Pimplinae sampled and 89% of all the Pimplini. All are parasitoids of macrolepidoptera and are thought to be, at least in part, arboreal (Fitton et al. 1988). Such host associations may to some extent be driving the patterns we see in the Pimplinae in relation to broadleaved habitats. Habitat characteristics appear less important for the tribe Ephialtini, for which the host groups are more varied and are often predators. These hosts are found across the orders Coleoptera, Hymenoptera, Diptera and Arachnida (Fitton et al. 1988). Fraser (2005) conducted a second study using the same parasitoid taxa, trapping techniques and two of the same woodlands used here, in 2004. Patterns of species abundance were found to be consistent across 2003 and 2004 giving a degree of temporal robustness to our findings.
The abundance and diversity of Diplazontinae were not significantly different between woodlands or associated with woodland habitat characteristics, and the species accumulation curve (Fig. 1) suggests woodlands are generally very similar in Diplazontine composition. The aphidophagous Syrphidae hosts (Fitton & Rotheray 1982) are themselves already two trophic levels above the vegetation. Furthermore, adult syrphids are known to be highly mobile. When evaluating the potential for Syrphidae as environmental indicators, Sommaggio (1999) suggested they are useful for assessing landscape diversity. It may be therefore that patterns in diplazontine diversity and community composition are also evident at the landscape scale rather than the woodland scale.
These differences across parasitoid taxa suggest the following hypothesis: the degree of association between parasitoid abundance and diversity and characteristics of the vegetation within habitats will decrease with an increase in the number of trophic links that separate them. Understanding the mechanisms driving associations between habitat variables and patterns in wasp assemblages may provide a basis for understanding factors influencing the regulation of arthropod assemblages by wasps in the landscape (Lassau & Hochuli 2005).
implications for sampling
The results of this study provide important implications for sampling regimes designed to provide information regarding parasitoid abundance and diversity. The high variation between weeks for all taxa indicates that any sampling regime must continue over several consecutive weeks to account for this temporal variation in parasitoid communities. Also, as illustrated by the variability between traps for the Diplazontinae, samples can vary at small spatial scales and therefore replication of traps within a single habitat patch is necessary.
The species accumulation curve (Fig. 1) for the Pimplinae shows no sign of an asymptote, in common with the findings of Sääksjärvi et al. (2004). Most of the additional species after nine woodlands are represented by just single individuals, showing the typically long abundance ‘tail’ of rare species (Gaston & Blackburn 2000). The 16 Pimplinae species that occur as singletons in our samples were spread over 12 woodlands, suggesting that additional sampling would be unlikely to alter the overall patterns we have found. None the less, a more complete species list would require further sampling of woodlands within the landscape.
implications for conservation
This work has suggested that woodlands may play an important role in the maintenance of biodiversity in agricultural landscapes. For the two larger subfamilies sampled, the Diplazontinae and Pimplinae, 38% and 33%, respectively, of the entire UK species lists were recorded in the 15 woodlands sampled. This is a high proportion in comparison with comparable British woodland samples of other taxa, e.g. macrolepidoptera 24% (Usher & Keiller 1998), birds 23% (Mason 2001), Syrphidae 20% and Carabidae 12% (Humphrey et al. 1999). Woodland habitats provide parasitoids with resources, such as alternate hosts, food for adults, overwintering habitats and appropriate microclimates (van Emden 1965; Powell 1986; van Emden 1990; Dyer & Landis 1997), and enhance landscape complexity in agricultural landscapes, which has been shown to enhance diversity in other taxa in surrounding fields (Schmidt et al. 2005). A notable species that was found in four of the farm woodlands in this study was the Pimpline Zatypota albicoxa (Walker). Only four individuals of this species have been found previously in the UK, in three localities. This parasitoid has been reared from Achaearanea simulans (Thorell), a spider listed as nationally notable (http://www.britishspiders.org.uk, accessed 25 December 2004) and used as an ancient woodland indicator (M. R. Shaw, personal communication). Its presence suggests that farm woodlands can provide small patches of high-quality habitat allowing woodland species to persist in agricultural landscapes.
At the whole woodland scale, woodlands with high broadleaf content and high tree/shrub species richness would appear to be most beneficial to taxa such as the Pimplinae and Diacritinae. For the Diplazontinae, the type of woodland would appear relatively unimportant at this scale. However, the optimal reserve selection of five woodlands from our list included both coniferous woods and woods of low species richness, suggesting that such woods can contribute to parasitoid diversity at the landscape scale. Selecting only coniferous woodlands at random resulted in significantly lower average species richness than only selecting deciduous woodlands or selecting both types. Mixtures of both woodland types on average were no worse or better than only deciduous woods, but could result in fewer species than coniferous woods only or more species than deciduous woods only. Clearly, care will also have to be taken with respect to other habitat variables, such as tree species richness, if parasitoid species richness at the landscape scale is to be maximized.
We are grateful to M. Shaw (Pimplinae, Poemeniinae and Diacritinae), G. Rotheray (Diplazontinae) and E. Diller (Diplazontinae) for help with species identification, and H. Edwards and R. Shortridge for assistance with fieldwork. We thank the many landowners for permission to establish traps in their woodlands. This work was funded by a Natural Environment Research Council studentship to S. E. M. Fraser.