Ecological Entomology (2003) 28,299–308 Consequences of removing a keystone herbivore for the abundance and diversity of arthropods associated with a cruciferous shrub ADELA GONZA´ LEZ-MEGI´ A S and J O S E´ M. GO´ MEZ Departamento de Biologı´ a Animal y Ecologı´ a,Facultad de Ciencias,Universidad de Granada,Spain Abstract. 1. The effect of the removal of Timarcha lugens (Chrysomelidae),one of the main herbivores of Hormathophylla spinosa (Cruciferae),on the abundance of co-occurring phytophagous insects,the abundance of non-phytophagous arthropods (detritivores,predators,and parasitoids),and the structure and diver- sity of the entire arthropod community,was studied for 3 years (1999–2001). 2. There was competition between T. lugens and co-occurring herbivores; the removal of T. lugens was correlated with an increase in the abundance of sap- suckers,flower-feeders,and,above all,folivores. 3. Timarcha lugens also had an indirect effect on arthropods belonging to other trophic levels; the abundance of predators increased significantly after the removal of T. lugens. 4. Community composition was affected by the experimental removal. In addi- tion,the diversity of the overall community increased after removal of T. lugens. 5. The study demonstrated experimentally that T. lugens has a significant effect not only on other species belonging to the same trophic level,but also on the abundance of species belonging to higher trophic levels,and,consequently,on the entire structure and diversity of the complex community in which it is immersed. Key words. Chrysomelidae,community structure and diversity,exploitative competition,insect communities, Timarcha lugens. Introduction strong effects on each other (Hunter,1992; Stewart,1996). Indeed,temporally and spatially separated competitive The overall importance of interspecific competition in struc- effects mediated by the host plant appear to be common turing insect communities is still open to debate (Damman, among phytophagous insects (Faeth & Wilson,1997; 1993; Denno et al.,1995; Stewart,1996). Whereas competi- Masters & Brown,1997). Furthermore,some guilds comprise tion is intense and frequent in some insect communities (e.g. taxonomically and ecologically distant species,which Harrison et al.,1995; Hudson & Stiling,1997; Waltz & may also be engaged in competition for shared resources Whitham,1997; Denno et al.,2000; Fisher et al.,2000),little (Davidson et al.,1984,1985; Hochberg & Lawton,1990; evidence for competition has been found in other commu- Hunter,1992; Tscharntke,1997; Suominen et al.,1999a,b; nities (e.g. Evans,1992; Faeth,1992; Marquis & Whelan, Go´ mez & Gonza´ lez-Megı´ as,2002). To obtain an accurate 1994; Cornell & Hawkins,1995; Schmitz,1998). Although idea of the actual role that competition plays in insect ecological theory assumes that competition occurs mainly communities,it may be necessary to consider the existence between members of the same guild,recent studies have of interactions between dissimilar organisms (Go´ mez & demonstrated that herbivores in different guilds can exert Gonza´ lez-Megı´ as,2002). Competition can affect not only individual- and population-level traits,such as behaviour and abundance, Correspondence: Adela Gonza´ lez-Megı´ as,School of Biology, but also community-level parameters,such as diversity, University of Leeds,Leeds LS2 9JT,U.K. E-mail: bgyagm@ composition,and structure,and sometimes the effects can leeds.ac.uk differ between levels (Connell,1983; Faeth & Wilson,1997; # 2003 The Royal Entomological Society 299 300 Adela Gonza´lez-Megı´as and Jose´ M. Go´mez Denno et al.,2000). Nevertheless,the differential effect of season,both new adult and older individuals are found competition at different organisation levels has seldom been throughout the summer. There are two peaks of adult studied (Denno et al.,1995). Indeed,the role of interspecific emergence,once soon after snowmelt,the other during the competition in structuring insect communities has mostly first week of August been inferred from population-level studies (Tscharntke, The study site was in the Sierra Nevada (Granada prov- 1997; Speight et al.,1999); however population dynamics ince,Spain),in an 3 ha area located at 2828 m a.s.l. and community organisation are influenced greatly by (37404500N,3 2202500W). The area is an open high- interactions within and between all trophic levels,and ana- mountain scrubland (6.4 Æ 1.4% shrub cover,15.8 Æ 12.1 lyses of pairwise interactions often fail to explain patterns of plants/20 m2) dominated by H. spinosa (95% of total plant co-existence and abundance at the community level cover),with a few individuals belonging to two other (Tscharntke,1997). These species with a pervasive influence shrubs, Reseda complicata and Sideritis glacialis,and some on the overall community composition are considered key- scattered perennial herbs. Individual shrubs occur as dis- stone species (Hunter,1992),and their removal produces a crete units,surrounded by open interspaces of bare soil and dramatic change in the associated community. schist. The work reported here investigated experimentally the effects of a potential keystone species,the beetle Timarcha lugens Rosenhauer (Chrysomelidae),on the phytophagous Experimental design and non-phytophagous arthropod community associ- ated with the host shrub Hormathophylla spinosa (L.) The effect of T. lugens on the insect community was Ku¨ pfer (Cruciferae). Timarcha lugens is monophagous on studied by means of a removal experiment using a com- H. spinosa,feeding on its flowers,fruits,and vegetative pletely randomised design in each of 3 years (1999,2000,and tissues(Gonza´ lez-Megı´ as&Go´ mez,2001).Thisbeetlespecies 2001). In early June 1999,40 shrubs of similar size,at the consumes large quantities of plant tissue (4.5 Æ 0.6 mg per same phenological stage and early flowering,were selected. individual per day; Gonza´ lez-Megı´ as & Go´ mez,2001), Shrubs were assigned randomly to one of the following decreasing in some cases the fruit set of the plant by more treatments: (1) Timarcha exclusion: T. lugens were excluded than 30% (Go´ mez & Gonza´ lez-Megı´ as,2002). The specific selectively from 20 shrubs by hand removal of all the beetles objective was to assess the effect of the removal of T. lugens (see Floyd,1996; Waltz & Whitham,1997,for a similar on the abundance of co-occurring phytophagous insects procedure) every 5 days throughout the experiment. Other living on the same or different parts of the shrubs,the invertebrates living in the shrubs were not disturbed by abundance of non-phytophagous arthropods (detritivores, removing the beetles using forceps. This method was used predators,and parasitoids),and the structure and diversity instead of,for example,the application of soil insecticide or of the entire insect community. Tanglefoot1 to the base of the shrubs (see,for example, Hudson & Stiling,1997),because many other herbivorous and predatory insects living in H. spinosa foliage and Materials and methods flowers are also apterous,and would thus also be excluded from experimental shrubs together with the T. lugens indivi- Study system duals. This method excluded the beetles efficiently from these plants; after the third removal period,no T. lugens remained Hormathophylla spinosa is a long-lived stunted shrub (Go´ mez & Gonza´ lez-Megı´ as,2002). (2) Twenty plants were occurring in high mountains of the western Mediterranean, not subjected to removal of T. lugens to serve as controls. from southern France to North Africa. This thorny mass- The experimental shrubs did not differ among treatments in flowering shrub is typically hemispherical in shape when either size (F ¼ 1.992,d.f. ¼ 1,38, P ¼ NS) or distance to the reproductive,bearing from 480 to over 75 000 flowers per nearest conspecific (F ¼ 2.188,d.f. ¼ 1,38, P ¼ NS). The year,arranged in inflorescences,which outgrow its surface. abundance of T. lugens during the experimental period was Timarcha lugens is a high-altitude apterous,medium- 8.9 Æ 5.6 individuals per shrub in 1999,8.9 Æ 5.5 in 2000, sized (43.1 Æ 2.07 mg dry weight, n ¼ 130) beetle endemic and 9.9 Æ 6.1 in 2001 (repeated-measures ANOVA (rm- to the Sierra Nevada mountains (Spain),occurring from ANOVA), F ¼ 0.33,d.f. ¼ 2,20, P ¼ NS). 2400 to 3200 m a.s.l. It starts to feed on H. spinosa soon after snowmelt (late June at the study site),and is active until the end of September,feeding by chewing leaves as Data collection well as flower and fruits (Go´ mez & Zamora,2000; Gonza´ lez-Megı´ as & Go´ mez,2001). Larvae emerge at the The arthropod fauna living in the foliage was examined beginning of the season,consuming young leaves. Larvae each year using the beating method (Sutherland,1996). and adults can be found sharing the same plant. Flowering Each shrub was tapped for 20 s with a wooden stick,and occurs after larvae have buried under the plant to pupate dislodged invertebrates were caught in a 20 Â 10 cm beating (2 weeks after the emergence period),so it is only occasion- tray held beneath the shrubs. Because this method is ally possible to observe larvae eating buds (Gonza´ lez- destructive (Sutherland,1996),the arthropod fauna was Megı´ as,2001). As adults live longer than one breeding sampled twice (mid July and mid August) each year. Both # 2003 The Royal Entomological Society, Ecological Entomology, 28,299–308 Interactions between T. lugens and co-occurring arthropods 301 periods coincided with peaks of activity of T. lugens and The effect of each response variable was studied separately with the maximum abundance of arthropods in the study for each year using one-way ANOVAs (Proc GLM; SAS, area (Gonza´ lez-Megı´ as,2001). In addition,this method has 1997). The proportional contribution of each trophic presumably negligible effects on herbivore populations, group to the community was contrasted between treatments because less than 20% of the shrub canopy was sampled for each year using one-way contingency analyses (Proc each time.