Contributed Paper

Effectiveness of Vegetation Surrogates for ParasitoidWasps in Reserve SelectionSALLY E. M. FRASER, ALISON E. BERESFORD, JENNIFER PETERS, JOHN W. REDHEAD, ALASTAIRJ. WELCH, PETER J. MAYHEW,∗ AND CALVIN DYTHAMDepartment of Biology, University of York, P.O. Box 373, York YO10 5YW, United Kingdom

Abstract: Selecting suitable nature reserves is a continuing challenge in conservation, particularly for target

groups that are time-consuming to survey, species rich, and extinction prone. One such group is the parasitoid

Hymenoptera, which have been excluded from conservation planning. If basic characteristics of habitats or

vegetation could be used as reliable surrogates of specific target taxa, this would greatly facilitate appropriate

reserve selection. We identified a range of potential habitat indicators of the species richness of pimpline

parasitoid communities (Hymenoptera: Ichneumonidae: Pimplinae, Diacritinae, Poemeniinae) and tested

their efficiency at capturing the observed diversity in a group of small woodlands in the agricultural landscape

of the Vale of York (United Kingdom). Eight of the 18 vegetation-based reserve-selection strategies were

significantly better at parasitoid species inclusion than random selection of areas. The best strategy maximized

richness of tree species over the entire reserve network through complementarity. This strategy omitted only

2–3 species more (out of 38 captured in the landscape as a whole) than selections derived from the parasitoid

survey data. In general, strategies worked equally well at capturing species richness and rarity. Our results

suggest that vegetation data as a surrogate for species richness could prove an informative tool in parasitoid

conservation, but further work is needed to test how broadly applicable these indicators may be.

Keywords: complementarity, indicator species, insect conservation, nature reserve selection, parasitoid Hy-menoptera, reserve network, species richness surrogates

Efectividad de loa Sustitutos de Vegetacion para Avispas Parasitoides en la Seleccion de Reservas

Resumen: La seleccion de reservas naturales adecuadas es un reto permanente en la conservacion, par-

ticularmente para grupos cuyos muestreos consumen mucho tiempo, que son ricos en especies y que estan

propensos a la extincion. Uno de esos grupos es el de los Hymenoptera parasitoides, que han sido excluidos de

la planificacion de la conservacion. Si las caracterısticas basicas de los habitats o de la vegetacion pudieran

ser utilizadas como sustitutos confiables de taxa especıficos, la seleccion de reservas adecuadas se facilitarıa

enormemente. Identificamos una gama de indicadores potenciales de habitat para la riqueza de especies

de comunidades de parasitoides (Hymenoptera: Ichneumonidae: Pimplinae, Diacritinae, Poemeniinae) y

probamos su eficiencia en la captura de la diversidad observada en un grupo de bosques pequenos en el

paisaje agrıcola del Vale de York (R. U.). Ocho de las 18 estrategias para la seleccion de reservas basadas en

vegetacion fueron significativamente mejores para la inclusion de especies de parasitoides que la seleccion

aleatoria de areas. La mejor estrategia maximizo la riqueza de especies de arboles en toda la red de reservas

mediante la complementariedad. Esta estrategia omitio solo 2–3 especies mas (de 38 registradas en todo el

paisaje) que las selecciones derivadas de los datos del muestreo de parasitoides. En general, las estrategias

funcionaron bien tanto para la captura de especies como de rareza. Nuestros resultados sugieren que los

datos de vegetacion como un sustituto de la riqueza de especies podrıan ser una herramienta informativa

para la conservacion de parasitoides, pero se requiere mas trabajo para probar la amplitud de la aplicacion

de estos indicadores.

∗Address correspondence to P. J. Mayhew, email pjm19@york.ac.ukPaper submitted March 18, 2008; revised manuscript accepted May 26, 2008.

142Conservation Biology, Volume 23, No. 1, 142–150C©2008 Society for Conservation BiologyDOI: 10.1111/j.1523-1739.2008.01069.x

Fraser et al. 143

Palabras Clave: complementariedad, conservacion de insectos, especies indicadoras, Hymenoptera parasitoide,red de reservas, reserva natural, sustitutos de la riqueza de especies

Introduction

One of the most common tools in conservation is theestablishment of nature reserves. The aim of many re-serve networks is to represent as many species as possible(Williams et al. 2000). Ideally, one would select reservesfrom a range of potential sites by simple and rapid analy-sis of a few variables that are reliable surrogates for totalspecies richness or particular target taxa (Oliver et al.1998).

Nevertheless, little is known about how to predict thediversity of taxa that are hard to survey, biodiverse, andextinction prone (Samways 2005). One such group is theinsect parasitoids (Shaw & Hochberg 2001). Parasitoidsare insects, mainly wasps (Hymenoptera), that developto maturity by feeding on the still-living body of anotherarthropod and eventually killing it (Godfray 1994). InBritain around 6000 species of parasitoid Hymenopterahave been described, and they comprise at least 25% ofthe total insect fauna (Shaw & Hochberg 2001). Largeproportions of parasitoids remain undescribed, and iden-tifying described species is a specialized occupation: acase of taxonomic impediment (Ward & Lariviere 2004).As for many other organisms, parasitoids are rare in termsof local population size and site occupancy (Fraser et al.2008b) and therefore are potentially subject to doublejeopardy from both sources of extinction risk (e.g., Gas-ton 1999). Because of their high degree of specializationand trophic status, parasitoids may be especially vulner-able to extinction threats such as habitat fragmentation(Kruess & Tscharntke 1994) and climate change (Stire-man et al. 2005).

One way in which taxonomic impediment can be re-duced is through the use of surrogates to indirectly surveydiversity (Favreau et al. 2006; Rodrigues & Brooks 2007).A broad range of possible surrogates for insect diversityhas been investigated, including biotic and abiotic vari-ables (Favreau et al. 2006; Rodrigues & Brooks 2007).Nevertheless, these studies have failed to produce a con-sensus because many are applicable only to specific cases(Lovell et al. 2007). Thus, it is essential that detailed, andideally mechanistic, studies of the congruence betweenpossible surrogates and the target group be undertaken.

Vegetation characteristics are the most promising ofthe indicators of parasitoid diversity at individual sitesthat have been investigated (e.g., Lassau & Hochuli 2005,2007; Saaksjarvi et al. 2006; Fraser et al. 2007). We soughtto assess whether vegetation characteristics can be usedto identify an effective network of reserves for a group ofparasitoid Hymenoptera, the pimplines, in small wood-

lands set in an agricultural landscape in the Vale of York(United Kingdom). We based potential reserve-selectionstrategies on vegetation criteria and assessed the effec-tiveness of these strategies at representing total speciesrichness and rarity.

Methods

Landscape and Land Use

The Vale of York is a lowland agricultural landscapein northeastern England lying between the upland dis-tricts of the Pennines to the west and the North YorkMoors to the east. The area has undergone changes dur-ing the last century that are typical of lowland Britain,including the abandonment of traditional farming prac-tices such as haymaking, adoption of new crops such aswinter wheat, land drainage, an increase in farm size andfield size with consequent removal of hedgerows, andincreased use of chemical inputs. All these changes arestrongly linked with declines in species richness (Marren2002).

Relative to most other European countries, the UnitedKingdom has little forest land, although recent policy hasbeen to increase this (Forestry Commission 2002). Exist-ing woods serve a variety of purposes including recre-ation, commercial timber production (mostly conifers),and nature conservation (Rackham 2006). We examinedwoodlands for this study because they play an importantrole in the maintenance of species richness in agriculturallandscapes (e.g., Petit & Usher 1998). Woodlands are animportant habitat for parasitoids because many of theirhosts feed on trees, shrubs, and ground vegetation or boreinto dead wood or other structural features that wood-lands provide (Fitton et al. 1988). Changes in woodlandmanagement affect species richness and are often drivenby economics or national policy (Forestry Commission2002, 2004). A great many protected areas in the UnitedKingdom consist of sites similar in size to those consid-ered here (Hopkinson et al. 2000); hence, we consideredrelatively realistic sites as potential nature reserves and arepresentative landscape for lowland Britain.

Pimplines in the United Kingdom

We studied 3 closely related subfamilies of the Ichneu-monidae: Pimplinae, Diacritinae, and Poemeniinae, col-lectively referred to here as pimplines because speciesin the latter 2 subfamilies were previously groupedwithin the former. These subfamilies have keys to U.K.

Conservation Biology

Volume 23, No. 1, 2009

144 Effectiveness of Vegetation Surrogates

species (Fitton et al. 1988; Shaw 2006), are abundant inwoodlands, and have a wide range of hosts. They havealso been used as indicators of species richness in diversegeographic locations (Gaston & Gauld 1993; Saaksjarvi etal. 2004; Fraser et al. 2007).

The Pimplinae exhibit a wider range of life historiesand hosts than any other subfamily of the Ichneumonidae(Fitton et al. 1988; Shaw 2006). In the British Isles thereare 103 species in 30 genera (Broad 2005). The subfam-ily is divided into 3 monophyletic tribes (Delomeristini,Ephialtini, and Pimplini), each with distinct ecologies(Gauld et al. 2002; Broad 2005). In the British Isles theDelomeristini consists of 9 species in 2 genera. Somespecies parasitize cocoons of sawflies and ichneumonids,whereas others attack aculeate Hymenoptera (Fitton etal. 1988). The Ephialtini consists of 75 U.K. species in 24genera. Host groups for this tribe are found across the or-ders Coleoptera, Hymenoptera, Diptera, and Arachnida.The majority of species are ectoparasitoids of insect lar-vae or pupae (Fitton et al. 1988). The Pimplini consistsof 20 U.K. species in 3 genera. They are endoparasitoidsof Lepidoptera pupae (Fitton et al. 1988).

Six species in the Poemeniinae occur in Britain (Broad2005). They are ectoparasitoids and are associated withbeetles or sphecid wasps in dead and standing timber(Fitton et al. 1988). Diacritinae is one of the few subfam-ilies for which the hosts are completely unknown (Wahl& Gauld 1998). In Europe only one species is known,Diacritus aciculatus (Fitton et al. 1988).

WASP COLLECTION

Pimpline wasps were collected from 2 Malaise traps ineach of 15 woods south and east of York over 4 weeksin July and August 2003, the main Hymenoptera flightseason (Fraser et al. 2007). To control for variation in as-pect and for edge effects, we placed traps on either sideof the midpoint of the wood’s southern edge, 20–40 mapart, and 10–20 m from the woodland edge. In 2004 inintensive sampling of 2 woods, we found evidence foredge effects on community composition within woodsand on richness and abundance of 1 subfamily. Edges didnot, however, have an overall effect on species richness(Fraser et al. 2008a). Nothing specific is known aboutpimpline movement distances, but given the scales in-volved, we think it is highly unlikely that a trapped waspin one wood might otherwise have been trapped in an-other wood, so to this extent woods were independentsites. We also treated our woods as independent sitesbecause conservation would probably work at the en-tire site scale; hence, each wood could be considered aconservation unit. We based our choice of woodlands ontheir small size (<20 ha) and the inclusion of both pre-dominantly broadleaved or coniferous woodland types(see Supporting Information).

Collection of Habitat Data

In each woodland we sampled vegetation in 20-mquadrats surrounding a Malaise trap and in 2 other 20-mquadrats at random locations within the northeast andnorthwest quarters of the wood (Fraser et al. 2007).Within these 4 quadrats all tree and shrub species morethan 1 m tall were identified. Trees and shrubs providehosts for many of the studied pimpline species (Fittonet al. 1988). We sampled the herb layer in five 2-mquadrats within each 20-m quadrat. Every species ofground vegetation less than 1 m in height was identi-fied (fungi and bryophytes were not identified to specieslevel). Some pimpline species exclusively use hosts asso-ciated with low-growing plants (Fitton et al. 1988). Wemeasured canopy cover at the southeast corner of eachsmall quadrat (percentage of cells on a gridded sheetof acetate, held at arm’s length, through which foliagecould be seen). A visual estimate of the percentage ofground cover was also made in each of these quadrats.To measure plant architectural diversity, we placed a 2-msampling pin (marked in divisions of 5, 10, and 20 cm) atrandom locations in each small quadrat. Touches by plantmaterial against the pin in each height bracket were cat-egorized and translated to a diversity index value (α),following Southwood et al. (1979). Plant architectural di-versity is associated with the species richness of insectherbivores (Strong et al. 1984) and hence may indicatehost diversity.

Reserve-Selection Strategies

We first explored simple rank-based criteria (Table 1) forreserve selection. Each wood was ranked in terms of thehabitat variable in question, and reserves were chosenfrom the highest ranked wood first, progressively includ-ing lower ranked woods. We used 11 habitat variables(Supporting Information). We used tree species richness(total sampled in the four 20-m quadrats) as a variablebecause it correlates with parasitoid richness across in-dividual sites (Sperber et al. 2004; Saaksjarvi et al. 2006;Fraser et al. 2007). We also used tree diversity (see Lassau& Hochuli 2007) and measured it with 1-D (D is Simp-son’s diversity index) because it is robust to small samplesize (Magurran 2004). Highest tree density (20/m2) andlowest canopy cover were also used. Both factors areassociated with the parasitoid community across sites(Sperber et al. 2004; Lassau & Hochuli 2005; Fraser etal. 2007). We used the presence of native tree speciesand common tree species as variables because they re-spectively support larger invertebrate communities thanintroduced species and more parasitoids than rare trees(Kennedy & Southwood 1984). We selected woodlandswith the highest proportion of native tree species andwith the highest abundance of common tree species. Forthe latter, a commonness score for each tree species wascalculated as the proportion of the total number of trees

Conservation Biology

Volume 23, No. 1, 2009

Fraser et al. 145

Table 1. Summary of reserve-selection strategies tested to find surrogates for parasitoids.∗

Strategy Type Algorithm

Tree species richness simple ranking sites ranked on basis of named variable, top-rankedsites chosen first

Simpson’s diversity for trees (1-D) as above as aboveTree density (20/m2) as above as aboveLowest canopy cover (%) as above as aboveProportion of native trees as above as aboveProportion of common trees as above as aboveAbundance of signal tree species as above as aboveSimpson’s diversity for ground vegetation (1-D) as above as abovePlant architectural diversity (α) as above as abovePercentage ground cover as above as aboveSignal plant abundance as above as aboveWoodland area (ha) as above as aboveLowest woodland isolation (m) as above as aboveTree species richness with complementarity ranking with

complementarityfirst site chosen is highest ranked for that variable,

other sites then ranked on basis of focal variablespecies not already included

Ground species richness with complementarity as above as abovePimpline species richness with complementarity as above as abovePimpline rarity with complementarity as above as aboveTree species richness (broadleaf and coniferous) ranking with secondary

criterionsites ranked on basis of primary variable, and the

highest ranked sites selected (If >1/3 of finalselection is already of the next-highest rankedwoodland type, woods of that type are ignored.)

Simpson’s diversity for trees (broadleaf andconiferous)

as above as above

Tree density (broadleaf and coniferous) as above as aboveSpecies richness of ground signal plants and plant

architectural diversityas above sites ranked on basis of primary variable (If sites are

tied, secondary variable used as tie-breaker.)Species richness of ground signal plants and

percentage ground coveras above as above

∗For calculation of variables, see Methods.

sampled, over all woods combined, represented by thatspecies. We then used this score to weight the abundanceof common species in each wood (commonness score ×abundance in a wood). The score for a wood was the sumof the weighted abundances of all its tree species.

Diversity of ground vegetation (1-D), ground-vegetation species richness, percent ground cover, ar-chitectural diversity, and abundance of “signal species”(plant species with which particular parasitoid speciesare associated) were also used as surrogates. The former2 are associated with the parasitoid community (Fraser etal. 2007). Although plant architectural diversity is a poorpredictor of high parasitoid diversity across individualwoodlands (Fraser et al. 2007), we included it here be-cause some researchers have found it increases species in-clusion at a landscape scale (e.g., Woodcock et al. 2007).Correlations among parasitoids and plant species rich-ness, diversity, or structure are generally thought to re-veal a greater range of potential host niches (Saaksjarvi etal. 2006).

Signal plant species in the herb layer were Rubus spp.,Urtica spp., Juncus spp. + grasses (treated collectively).To estimate relative abundance of signal species, the num-ber of 2-m quadrats in which the 3 groups appeared was

summed for each wood. Rubus spp. conceal the spiderhosts of Zaglyptus spp. but are probably also importantfor a number of species that attack hosts more gener-ally in the field layer or bushes (e.g., Pimpla rufipes,Scambus brevicornis, both parasitoids of Lepidoptera)(Fitton et al. 1988). Juncus and grasses are associatedwith Endromopoda, which attack stem-boring or gall-forming hosts (e.g., sawflies) in monocotyledons (Fittonet al. 1988). Likewise they provide a number of hostsfor the more generalist species attacking Lepidoptera,and grass tussocks are an aestivation site for Itoplectis

maculator, which attacks mainly Tortricidae (Fitton etal. 1988). Urtica spp. are the host plants of many Lepi-doptera species, some of which (e.g., Aglais urticae) areattacked by Pimpla spp. (Pyornila 1977).

In our study signal tree species show associations withpimpline species by providing food for hosts (Fitton etal. 1988). For each tree species we calculated a weight-ing score by summing the number of pimpline speciescollected that associate with that tree. Association wasdetermined from Fitton et al. (1988) and Shaw (2006).The abundance score for each wood was the numberof trees sampled of each species times their weightingscore, summed over all trees, and then divided by the

Conservation Biology

Volume 23, No. 1, 2009

146 Effectiveness of Vegetation Surrogates

total number of trees sampled in the wood. The weight-ing scores were Quercus spp., 5; Pinus spp., 3; Betula

spp., 2; Alnus spp., 1; Salix spp., 1; Corylus avellana, 2;Acer pseudoplatanus, 1; Sambucus nigra, 2; and Sorbus

aucuparia, 1.In addition to these vegetation variables, we included 2

other spatial variables: woodland area and distance to thenext-nearest wood (estimated from 1:25,000 ordnancesurvey map). The species–area relationship is a well-supported phenomenon (Gaston & Blackburn 2000), andwe tested the efficiency of area in case reserve sizes wereconfounding other selection criteria. Isolation may alsoaffect parasitoid communities (e.g., Kruess & Tscharntke1994).

Additionally, we investigated more complicatedreserve-selection criteria (Table 1). One set of strate-gies involved complementarity (e.g., Pressey et al. 1993;Williams et al. 1996, 2000), whereby consecutive re-serves are selected on the basis of species not already in-cluded. Thus, complementarity helps maximize a variableacross the entire reserve network. We implemented thesestrategies in a linear selection program in C++ (availableon request from C.D.), designed to exclude woodlandsfrom the data set once they had been selected and recal-culate the values of remaining woodlands accordingly.Complementarity was applied with tree species richness,which proved a relatively efficient surrogate in simplerank-based selection and in predicting diversity acrosssites. Ground species richness was also used as a surro-gate in complementarity selection for comparison. Wealso ran control complementarity selections with pim-pline species richness and rarity. These selections pro-vided a hypothetical ideal for comparison with nonsurro-gate variables. When applying complementarity to theserichness and rarity criteria, reserves were selected on thehighest richness or rarity of tree or pimpline species notalready included in selected woods.

A final class of strategies involved the use of 2 variablessimultaneously (Table 1). Percentage of broadleaf treeswas used as a second variable in strategies involving treespecies richness, tree species diversity, and tree densitybecause a previous analysis (Fraser et al. 2007) showedthis was a good predictor of parasitoid abundance and di-versity across individual sites and that the best selectionof 5 woods contained a mixture of broadleaved and conif-erous sites (Fraser et al. 2007). In each case selectionswere required to contain a mixture of at least one-thirdprimarily broadleaved sites and one-third primarily conif-erous sites. Ranking of woods was performed as for theprimary variables (e.g., tree species richness), but whenthe next wood selected was of a type that had already metits quota (e.g., broadleaved), that wood was passed over;the next wood included was then the highest rankedwood of the other type (e.g., coniferous). In additionto these strategies, we used signal species richness toinitially rank the woods. We broke ties by selecting the

wood with the highest architectural diversity or percentground cover. This use of signal species as a coarse fil-ter before applying more specific selection criteria hasbeen used successfully to predict invertebrate diversity(Panzer & Schwartz 1998).

Unsurprisingly, many of the habitat variables con-sidered above were intercorrelated (Supporting Infor-mation). Fairly strong correlations occurred betweenbroadleaf content and canopy cover (positive), broadleafcontent and ground plant diversity (negative), and treediversity and plant architectural diversity (positive) (Sup-porting Information).

Strategy Assessment

We used 3 criteria to compare the effectiveness of the se-lection strategies including the observed parasitoid com-munity: parasitoid species richness; rarity of occupancyof species across sites; and Simpson’s index of diversity.Results with Simpson’s index were very similar for allstrategies and are not reported. We scored rarity by as-signing each parasitoid species a rank calculated by sub-tracting the number of woods in which the species wasfound from 16, then dividing the result by 15 (Fraser et al.2007). Therefore, a species occurring in 15 woodlandsscored 1/15, whereas a species found in only 1 woodscored 15/15, giving additional weight to rare speciesbecause of their relatively low occupancy. These resultsfrom each selection strategy were compared against 1000random selections for each of 3, 5, and 7 woodlands.These sizes of reserve areas were chosen to test the effectof reserve size: 7 was the maximum number of woodlandsused because larger networks included most species andhence did not discriminate among selection strategies.Statistical significance was obtained when a selection’sresulting values lay on or above the 95th percentile ofthose gained from the relevant random selection. Wechose a best-possible selection by examining the high-est richness and rarity achieved in the thousand randomselections.

To assess whether strategies performed well at bothrichness and rarity and at all sizes of reserve networks,we performed a Pearson’s correlation between pairs ofthe efficiency measures (e.g., species richness of 3 woodselections against species richness of 5 wood selections),with the 22 different strategies tested as replicates.

Results

The best area selections in the 1000 random choices ofareas produced 27, 32, and 34 species of pimplines forreserve networks of 3, 5, and 7 reserves. Over all 15woods, 38 pimpline species were collected (Table 2).

Rarity scores were 17.8, 22.7, and 24.6, respectively. Innetworks derived from pimpline species richness and rar-ity via the complementarity technique, performance was

Conservation Biology

Volume 23, No. 1, 2009

Fraser et al. 147

Table 2. Pimpline species richness and rarity scores for reserve networks of 3, 5, and 7 reserves, chosen with different reserve-selectionstrategies.∗

No. of reserves in network

3 5 7 3 5 7

species richness rarity score

Random (mean) 19.7 24.2 27.9 10.9 15.2 18.8Random (upper 95 percentile) 24 29 32 15.1 19.8 22.7Best (out of 1000 random selections) 27 31 34 17.8 21.8 24.7Worst (out of 1000 random selections) 13 18 21 5.27 9.27 13Tree species richness 24∗ 28 31 15.0 18.9 21.9Simpson’s diversity for trees (1-D) 21 29∗ 33∗ 13.0 19.7 23.7∗

Tree density (20 m−2) 22 25 31 13.1 15.9 21.7Lowest canopy cover (%) 18 21 23 9.3 12.6 14.3Proportion of native trees 18 23 29 9.3 14.0 19.8Proportion of common trees 19 21 26 10.1 11.9 16.8Abundance of signal tree species 15 19 24 7.0 10.3 15.1Simpson’s diversity for ground vegetation (1-D) 18 22 29 9.5 13.3 19.8Plant architectural diversity (α) 21 29∗ 32∗ 12.1 19.9∗ 22.9∗

Percentage ground cover 18 23 29 8.9 14.1 19.1Signal plant abundance 14 22 31 6.1 13.3 21.9Woodland area (ha) 19 25 28 10.5 15.9 18.7Lowest woodland isolation (m) 21 23 31 12.1 13.9 20.2Tree species richness with complementarity 25∗ 29∗ 32∗ 15.9∗ 19.8∗ 22.7∗

Ground species richness with complementarity 18 27 30 9.9 17.9 20.7Pimpline species richness with complementarity 27∗ 32∗ 34∗ 17.8∗ 22.7∗ 24.6∗

Pimpline rarity with complementarity 27∗ 32∗ 34∗ 17.8∗ 22.7∗ 24.6∗

Tree species richness (broadleaf and coniferous) 22 29∗ 33∗ 13.9 19.7 23.7∗

Simpson’s diversity for trees (broadleaf and coniferous) 20 29∗ 33∗ 12.1 19.7 23.7∗

Tree density (broadleaf and coniferous) 23 26 28 13.9 16.9 18.8Species richness of ground signal plants 21 28 33∗ 11.6 18.1 22.8∗

and plant architectural diversitySpecies richness of ground signal plants 19 27 33∗ 10.5 18.0 23.7∗

and percentage ground cover

∗Results significantly better than random.

almost exactly the same as for the best possible selectionsfrom the random trials. This suggests our complemen-tarity algorithm was very efficient. The 95th percentileof the randomly selected areas outperformed the meanby about 4, but the complementarity algorithm outper-formed the mean by another 2–3 units.

Only 3 of the simple ranked-based criteria performedsignificantly better than random, and in no case was thisso for all sizes of networks (Table 2). Tree species rich-ness performed significantly better than random for net-works of 3 woods; they omitted only 3 species more thanselections derived from the parasitoid data (Table 2). Treediversity performed similarly well for networks of 5 and7 areas, sometimes only omitting 1 species more than se-lections determined on the basis of parasitoids. Neverthe-less, tree diversity did not attain significantly better thanrandom results for networks of 3 wooded areas. Plant ar-chitectural diversity performed similarly to tree diversity.Signal ground plants and trees performed rather poorlyas surrogates in networks of 3 areas, which suggests theyneed to be abundant before they raise parasitoid speciesrichness.

Tree species richness with complementarity as a basisfor area selection was significantly better than randomfor all network sizes investigated and for species richnessand rarity. It performed better than the mean of randomselections by about 5 for richness and rarity and omit-ted only 2–3 species more than selections made on thebasis of presence of the parasitoids. Richness of groundspecies in contrast did not perform significantly betterthan random at any network size (Table 2).

For pimpline species richness and rarity in the largerreserve networks, selection of areas on the basis of treespecies richness, including both broadleaved and conif-erous sites, performed significantly better than randomselection of sites (Table 2). Use of this variable omittedonly one more species than selections derived from par-asitoid data. In 4 of 6 cases, it performed better than treespecies richness on its own. Simpson’s diversity for treesincluding broadleaved and coniferous woods performedwell, but generally just as well as Simpson’s diversity fortrees on its own. Tree density, including broadleavedand coniferous woods, did not perform significantly bet-ter than random in any case, and in general did not do

Conservation Biology

Volume 23, No. 1, 2009

148 Effectiveness of Vegetation Surrogates

better than tree density on its own. Species richness ofground signal plants in combination with 2 tie-breakingvariables performed significantly better than random fornetworks of 7 wooded areas (Table 2), again omittingonly 1 species more than selections derived from thepresence of parasitoids.

Strategies that returned high richness scores also re-turned high rarity scores and vice versa (Table 3). Strate-gies that returned high richness or rarity scores for smallnetworks also did so for larger networks, and vice versa(Table 3).

Discussion

Our major finding was that simple measures of vegetationcan act as efficient surrogates for parasitoid richness inreserve selection. When these surrogates were used toselect conservation sites, the sites included more speciesoverall and more rare species than sites selected on arandom basis. Most reserves selected on the basis of veg-etation indicators contained a mixture of common andless common species, giving a similar rank–abundanceprofile. Hence the results with Simpson’s diversity as anefficiency score were similar across strategies. The bestsurrogate, maximization of tree species richness acrossthe reserve network, performed nearly as well as the bestrandom selection. Our site selections considered richnessand rarity representation, but not the costs of reserve es-tablishment, management, or the present or future vul-nerability of sites. Hence, selected sites should not beconsidered “optimal” in a more general sense.

Previous researchers discovered that parasitoid rich-ness at individual sites is associated with a number ofhabitat variables, including forest type (Hilszczansk et al.2005; Fraser et al. 2007), structural diversity (Hilszczansket al. 2005; Fraser et al. 2007), tree species richness (Sper-ber et al. 2004; Saaksjarvi et al. 2006; Fraser et al. 2007),density of vegetation or canopy cover (Sperber et al.2004; Fraser et al. 2007), and overall measures of ecolog-

Table 3. Pearson’s correlation matrix (r) for the efficiency measuresacross the 22 reserve-selection strategies in Table 1.∗

No. ofSpecies richness Rarity score

woodlandVariable sites 3 5 7 3 5 7

Species 3 – 0.799 0.572 0.993 0.797 0.571richness

5 – – 0.838 0.832 0.998 0.8607 – – – 0.605 0.827 0.990

Rarity 3 – – – – 0.832 0.612score

5 – – – – – 0.8557 – – – – – –

∗All coefficients are significant at p < 0.01.

ical complexity (e.g., canopy cover, ground flora height,and ground cover) (Lassau & Hochuli 2005, 2007). In ourstudy, only 3 simple, ranked-based strategies of site se-lection performed significantly better than random siteselection: tree species richness, tree species diversity,and plant architectural diversity. Nevertheless, in noneof these cases was performance significantly better thanrandom for all sizes of reserve networks. Some of thesestrategies sometimes performed worse than the mean ofrandom site selections, which suggests they may evenpreclude selection of the best reserve sites, at least ifthey are used in this situation and for this taxon. Thesestrategies included the use of the highest ranked siteson the basis of proportion of native trees and on thebasis of ground vegetation diversity, 2 features often con-sidered by foresters to be beneficial for conservation ofspecies richness (e.g., Forestry Commission 2002, 2004),the former receiving particularly heavy emphasis. Thus,although results obtained via our snapshot approach maynot extrapolate to temporal changes, it may be possibleto improve U.K. management of forest biodiversity withrespect to some parasitoid taxa across regions.

The principle of complementarity has been empha-sized as important in selection of reserve networks be-cause it helps ensure that different reserves capture differ-ent elements of biodiversity present in the landscape andhence helps maximize elements across the reserve net-work as a whole (e.g., Pressey et al. 1993; Williams et al.1996, 2000). Four of the reserve-selection strategies wetested were based on complementarity. Ground speciesrichness as a surrogate performed relatively poorly. Nev-ertheless, reserve selection determined on the basis oftree species richness was the best of all the vegetation-based strategies and performed almost as well as thebest possible strategies identified from random sampling.This result is consistent with previous findings that treespecies richness is a good predictor of abundance and di-versity of pimplines at individual sites in the same data set(Fraser et al. 2007). Pimpline richness is likely enhancedby tree richness because many of the species we sampledare associated with hosts that themselves are associatedwith particular tree species (Fitton et al. 1988); henceproviding more tree species provides hosts for more po-tential parasitoid species. Because tree species richnessis relatively easy to assess, it may be relatively easy toincorporate parasitoids into conservation planning.

One might intuitively expect foresters to maximizetree species richness to support total regional speciesrichness. The U.K. Forestry Standard (Forestry Commi-sion 2004) makes several explicit references to main-taining or enhancing tree diversity, including enhancingthe richness of large plantation stands and the use ofspecies diversity in identifying woodlands of high conser-vation value. Nevertheless, the majority of emphasis onenhancing species richness is on broadleaf and native tree

Conservation Biology

Volume 23, No. 1, 2009

Fraser et al. 149

planting, use of local tree provenances (Forestry Com-mission 2004), recreating natural or seminatural habitattypes (see also Rodwell & Patterson 1994), and improv-ing structural diversity, for example by creating forestedge environments in glades or rides (Forestry Commis-sion 2004). Many of these strategies may benefit otherelements of biodiversity, and even sometimes parasitoids(e.g., Fraser et al. 2008a). These management strategiesmay also inadvertently create sites with high tree speciesrichness, but not necessarily. Floral richness and “diver-sity of woodland within a stand” are 2 of the U.K. indica-tors of sustainable forestry (Forestry Commission 2002),although neither is directly equivalent to tree species rich-ness. Our data suggest that it might be useful to considertree species richness more explicitly or with greater em-phasis in forest biodiversity management. Nevertheless,again we caution that our snapshot approach, comparingseveral woods, may not extrapolate to temporal changeswithin individual woods.

This raises the further issue of whether conserving par-asitoid richness will effectively conserve other taxonomicgroups. Data on the congruence between parasitoidsand other animal taxa are limited. Hughes et al. (2000)showed that Hymenoptera (including many parasitoids)and Diptera richness are positively associated across sitesand tend to have the same rank richness across habitatsas other insect orders. More generally, congruence variesbetween Hymenoptera richness (parasitoids normally ex-cluded) and that of other taxonomic groups (e.g., Oliveret al. 1998; Lovell et al. 2007). In a recent review Favreauet al. (2006) came to no conclusions about the situationsin which species richness surrogates generally work. In asimilar review of the effectiveness of surrogates in com-plementarity representation, of relevance to our recom-mendations, a general weak but positive surrogacy powerwas found that was more pronounced among terrestrialtaxa (Rodrigues & Brooks 2007). Unsurprisingly, noneof the studies reviewed in Rodrigues and Brooks (2007)included parasitoids, but 2 included woody plants as sur-rogates for other taxa. Chiarucci et al. (2005) showedthat woody plant species in Italy were rather poor surro-gates for fungi. Nevertheless, Kati et al. (2004) found thatwoody plants in northern Greece adequately conservedall other taxa examined except orchids. This suggeststhat selecting reserves for parasitoids, on the basis of treespecies richness with complementarity, may also be ef-fective for some, but not all, other taxa. More specifictesting is required.

Our results show, for the first time, that simplevegetation-based surrogates can significantly improve theinclusion of parasitoid species in reserve networks. Thissuggests the means to improve reserve site selection forand management of this species-rich and cryptic, but vul-nerable, taxon. The challenge now is to test the robust-ness of these findings for different habitat types, geo-graphic locations, and taxa.

Acknowledgments

We are grateful to M. Shaw for help with species identi-fication and H. Edwards and R. Shortridge for fieldworkassistance. We thank the many woodland owners for per-mission to establish traps in their woodlands. This workwas supported by a Natural Environment Research Coun-cil Studentship to S.E.M.F.

Supporting Information

Characteristics of each wood (Appendix S1) and corre-lations between those characteristics (Appendix S2) areavailable as part of the online article. The author is respon-sible for the content and functionality of these materials.Queries (other than absence of the material) should bedirected to the corresponding author.

Literature Cited

Broad, G. R. 2005. Checklist of British and Irish Ichneu-monidae (Hymenoptera). Biological Records Centre, MonksWood, county. Available from http://www.brc.ac.uk/downloads/Ichneumonidae checklist.pdf (accessed April 2005).

Chiarucci, A., F. D’Auria, V. De Dominicis, A. Lagana, C. Perini, andE. Salerni. 2005. Using vascular plants as a surrogate taxon to maxi-mize fungal species richness in reserve design. Conservation Biology19:1644–1652.

Favreau, J., C. Drew, G. Hess, M. Rubino, F. Koch, and K. A. Eschel-bach. 2006. Recommendations for assessing the effectiveness of sur-rogate species approaches. Biodiversity and Conservation 15:3949–3969.

Fitton, M. G., M. R. Shaw, and I. D. Gauld. 1988. Pimpline Ichneumon-flies, Hymenoptera, Ichneumonidae, Pimplinae. Handbooks for theidentification of British insects. Volume 7, part 1. Royal Entomolog-ical Society, London.

Forestry Commission. 2002. U.K. indicators of sustainable forestry.Forestry Commission, Edinburgh.

Forestry Commission. 2004. The U.K. forestry standard. Forestry Com-mission, Edinburgh.

Fraser, S. E. M., C. Dytham, and P. J. Mayhew. 2007. Determinants ofparasitoid abundance and diversity in woodland habitats. Journal ofApplied Ecology 44:352–361.

Fraser, S. E. M., C. Dytham, and P. J. Mayhew. 2008a. The effectivenessand optimal use of Malaise traps for monitoring parasitoid wasps.Insect Conservation and Diversity 1:22–31.

Fraser, S. E. M., C. Dytham, and P. J. Mayhew. 2008b. Patterns in theabundance and distribution of ichneumonid parasitoids within andacross habitat patches. Ecological Entomology 33:473–483.

Gaston, K. J. 1999. Implications of interspecific and intraspecificabundance-occupancy relationships. Oikos 86:195–207.

Gaston, K. J., and T. M. Blackburn. 2000. Pattern and process in macroe-cology. Blackwell, Oxford, United Kingdom.

Gaston, K. J., and I. D. Gauld. 1993. How many species of Pimplines(Hymenoptera: Ichneumonidae) are there in Costa Rica? Journal ofTropical Ecology 9:491–499.

Gauld, I. D., D. B. Wahl, and G. Broad. 2002. The suprageneric groupsof the Pimplinae (Hymenoptera: Ichneumonidae): a cladistic re-evaluation and evolutionary biological study. Zoological Journal ofthe Linnean Society 136:412–485.

Godfray, H. C. J. 1994. Parasitoids, behavioural and evolutionary ecol-ogy. Princeton University Press, Princeton, New Jersey.

Conservation Biology

Volume 23, No. 1, 2009

150 Effectiveness of Vegetation Surrogates

Hilszczanski, J., H. Gibb, J. Hjalten, O. Atlegrim, T. Johansson, R. B. Pet-tersson, J. P. Ball, and K. Danell. 2005. Parasitoids (Hymenoptera,Ichneumonidae) of saproxylic beetles are affected by forest succes-sional stage and dead wood characteristics in boreal spruce forest.Biological Conservation 126:456–464.

Hopkinson, P., J. M. J. Travis, J. R. Prendergast, J. Evans, R. D. Gregory,M. G. Telfer, and P. H. Williams. 2000. A preliminary assessment ofthe contribution of nature reserves to biodiversity conservation inGreat Britain. Animal Conservation 3:311–320.

Hughes, J. B., G. C. Daily, and P. R. Ehrlich. 2000. Conservation of insectdiversity: a habitat approach. Conservation Biology 14:1788–1797.

Kati, V., P. Devillers, M. Dufrene, A. Legakis, D. Voukou, and P. Lebrun.2004. Testing the value of six taxonomic groups as biodiversityindicators at a local scale. Conservation Biology 18:667–675.

Kennedy, C. E. J., and T. R. E. Southwood. 1984. The number of speciesof insects associated with British trees—a re-analysis. Journal ofAnimal Ecology 53:455–478.

Kruess, A., and T. Tscharntke. 1994. Habitat fragmentation, species loss,and biological control. Science 264:1581–1584.

Lassau, S. A., and D. F. Hochuli. 2005. Wasp community responsesto habitat complexity in Sydney sandstone forests. Austral Ecology30:179–187.

Lassau, S. A., and D. F. Hochuli. 2007. Associations between waspcommunities and forest structure: Do strong local patterns holdacross landscapes? Austral Ecology 32:656–662.

Lovell, S., M. Hamer, R. Slotow, and D. Herbert. 2007. Assessmentof congruency across invertebrate taxa and taxonomic levels toidentify potential surrogates. Biological Conservation 139:113–125.

Magurran, A. E. 2004. Measuring biological diversity. Blackwell, Oxford,United Kingdom.

Marren, P. 2002. Nature conservation. Harper Collins, London.Oliver, I., A. J. Beattie, and A. York. 1998. Spatial fidelity of plant,

vertebrate, and invertebrate assemblages in multiple-use forest ineastern Australia. Conservation Biology 12:822–835.

Panzer, R., and M. W. Schwartz. 1998. Effectiveness of a vegetation-based approach to insect conservation. Conservation Biology12:693–702.

Petit, S., and M. B. Usher. 1998. Biodiversity in agricultural landscapes:the ground beetle communities of woody uncultivated habitats.Biodiversity and Conservation 7:1549–1561.

Pressey, R. L., C. J. Humphries, C. R. Margules, R. I. Vane-Wright, and P.H. Williams. 1993. Beyond opportunism: key principles for system-atic reserve selection. Trends in Ecology & Evolution 8:124–128.

Pyornila, M. 1977. Parasitism in Aglais urticae (L.) (Lep., Nymphalidae).IV. Pupal parasitoids. Annales Entomologici Fennici 43:21–27.

Rackham, O. 2006. Woodlands. Harper Collins, London.Rodriques, T. M., and T. M. Brooks. 2007. Shortcuts for biodiversity con-

servation planning: the effectiveness of surrogates. Annual Reviewof Ecology, Evolution and Systematics 38:13–37.

Rodwell, J., and G. Patterson. 1994. Creating native woodlands. ForestryCommission bulletin 112. Her Majesty’s Stationery Office, London.

Saaksjarvi, I. E., S. Haataja, S. Neuvonen, I. D. Gauld, R. Jussila, J. Salo,and A. Marmol Burgos. 2004. High local species richness of parasiticwasps (Hymenoptera: Icheneumonidae; Pimplinae and Rhyssinae)from the lowland rainforests of Peruvian Amazonia. Ecological En-tomology 29:735–743.

Saaksjarvi, I. E., K. Ruokaolainen, H. Tuomisto, S. Haataja, P. V. A. Fine,G. Carenas, I. Mesones, and V. Vargas. 2006. Comparing composi-tion and diversity of parasitoid wasps and plants in an Amazonianrain-forest mosaic. Journal of Tropical Ecology 22:167–176.

Samways, M. J. 2005. Insect diversity conservation. Cambridge Univer-sity Press, Cambridge, United Kingdom.

Shaw, M. R. 2006. Notes on British Pimplinae and Poemeniinae (Hy-menoptera: Ichneumonidae), with additions to the British list.British Journal of Entomology and Natural History 19:217–238.

Shaw, M. R., and M. E. Hochberg. 2001. The neglect of parasitic Hy-menoptera in insect conservation strategies: the British fauna as aprime example. Journal of Insect Conservation 5:253–263.

Southwood, T. R. E., V. K. Brown, and P. M. Reader. 1979. The relation-ships of plant and insect diversities in succession. Biological Journalof the Linnean Society 12:327–348.

Sperber, C. F., K. Nakayama, M. J. Valverde, and F. D. Neves. 2004.Tree species richness and density affect parasitoid diversity in cacaoagroforestry. Basic and Applied Ecology 5:241–251.

Stireman, J. O., III, et al. 2005. Climatic unpredictability and para-sitism of caterpillars: implications of global warming. Proceedingsof the National Academy of Sciences of the United States of America102:17384–17387.

Strong, D. R., J. H. Lawton, and T. R. E. Southwood. 1984. Insects onplants. Blackwell, Oxford, United Kingdom.

Wahl, D. B., and I. D. Gauld. 1998. The cladistics and higher classifica-tion of the Pimpliformes (Hymenoptera: Ichneumonidae). System-atic Entomology 23:265–298.

Ward, D. F., and M.-C. Lariviere. 2004. Terrestrial invertebrate surveysand rapid biodiversity assessment in New Zealand: lessons fromAustralia. New Zealand Journal of Ecology 28:151–159.

Williams, P., D. Gibbons, C. Margules, A. Rebelo, C. Humphries, and R.Pressey. 1996. A comparison of richness hotspots, rarity hotspotsand complementary areas for conserving diversity of British birds.Conservation Biology 10:155–174.

Williams, P. H., N. D. Burgess, and C. Rahbek. 2000. Flagship species,ecological complementarity and conserving the diversity of mam-mals and birds in sub-Saharan Africa. Animal Conservation 3:249–260.

Woodcock, B. A., S. G. Potts, D. B. Westbury, A. J. Ramsay, M. Lambert,S. J. Harris, and V. K. Brown. 2007. The importance of sward ar-chitectural complexity in structuring predatory and phytophagousinvertebrate assemblages. Ecological Entomology 32:302–311.

Conservation Biology

Volume 23, No. 1, 2009