vol. 152, no. 5 the american naturalist november 1998 Experimental Tests of the Dependence of Arthropod Diversity on Plant Diversity Evan Siemann,1,* David Tilman,1 John Haarstad,1 and Mark Ritchie2 1. Department of Ecology, Evolution, and Behavior, University of of resources should support a greater number of con- Minnesota, St. Paul, Minnesota 55108 sumer species, most models predict (e.g., Lotka 1925 2. Department of Fisheries and Wildlife, Utah State University, Volterra 1926 Gause 1934 MacArthur 1972 Whittaker Logan, Utah 84322 1975 Tilman 1986 Rosenzweig 1995) and some correla- tive studies (e.g., Murdoch et al. 1972 Nagel 1979 Submitted November 17, 1997 Accepted May 29, 1998 Southwood et al. 1979 Prendergast et al. 1993 Par- menter et al. 1995 Niemela et al. 1996) and experimental studies (e.g., Pimentel 1961 Root 1973 Altieri and Le- tourneau 1982 Lawton 1983 Altieri 1984 Siemann abstract: Because a diversity of resources should support a di- versity of consumers, most models predict that increasing plant di- 1998) have found that increasing plant diversity increases versity increases animal diversity. We report results of a direct ex- arthropod herbivore diversity. However, these studies are perimental test of the dependence of animal diversity on plant confounded by changes in plant community composition diversity. We sampled arthropods in a well-replicated grassland ex- that correlate with changes in plant diversity. The rela- periment in which plant species richness and plant functional rich- tionship between plant diversity and herbivore diversity ness were directly manipulated. In simple regressions, both the may be nonlinear because herbivore loads are often lower number of species planted (log2 transformed) and the number of in polycultures due to differences in the ability of herbi- functional groups planted significantly increased arthropod species richness but not arthropod abundance. However, the number of vores to locate host plants in mixed stands, suitability of species planted was the only significant predictor of arthropod smaller patches of host plants, and/or differences in the species richness when both predictor variables were included in effects of parasites and predators (reviewed in Andow ANOVAs or a MANOVA. Although highly significant, arthropod 1991). species richness regressions had low R2 values, high intercepts (24 Increases in arthropod herbivore diversity could po- arthropod species in monocultures), and shallow slopes. Analyses tentially cascade up to higher trophic levels, leading to a of relations among plants and arthropod trophic groups indicated greater diversity of parasites and predators (Hunter and that herbivore diversity was influenced by plant, parasite, and predator diversity. Furthermore, herbivore diversity was more Price 1992 Siemann 1998). Increasing plant diversity strongly correlated with parasite and predator diversity than with could also increase the diversity of higher trophic levels plant diversity. Together with regression results, this suggests that, directly by increasing the diversity of floral resources that although increasing plant diversity significantly increased arthro- many arthropod parasites and predators utilize or require pod diversity, local herbivore diversity is also maintained by, and (e.g., Sweetman 1936 Clausen 1940 Price et al. 1980 in turn maintains, a diversity of parasites and predators. Powell 1986 Jervis et al. 1993). Changing plant diversity Keywords: arthropod diversity, plant diversity, grasslands, preda- may also influence the interactions between herbivores tors, parasites, herbivores. and their predators and parasites, for example, by chang- ing parasite and predator foraging efficiency (e.g., Pimen- tel 1961 Strong et al. 1984 Russell 1989 Andow and Prokym 1990 Coll and Bottrell 1996). These interaction It has been suggested that plant diversity should be im- portant in determining animal diversity (e.g., Hutchinson modifications (sensu Wooton 1993, 1994) may appear as direct effects of plants on parasite and predator diversity 1959 Southwood 1978 Erwin 1982 Tilman 1982 May 1990 Hunter and Price 1992). Because a greater number in statistical analyses. Because many arthropods consume or forage on cer- * To whom correspondence should be addressed. Present address: Department of tain types of plants���such as forbs versus grasses in Ecology and Evolutionary Biology, Rice University, Houston, Texas 77005 E- grasslands (e.g., Hansen and Ueckert 1970 Knutson and mail: siemann@ruf.rice.edu. Campbell 1974 Boutton et al. 1978 Evans 1984 Belov- Am. Nat. 1998. Vol. 152, pp. 738���750. ��� 1998 by The University of Chicago. 0003-0147/98/5205-0008$03.00. All rights reserved. sky 1986 Porter and Redak 1997) or bushes versus trees

Plant and Arthropod Diversity 739 in forests (MacArthur and MacArthur 1961 MacArthur diversity on ecosystem functioning (Tilman et al. 1997). The experimental design and setup are reported in detail 1965)���adding more types of plants (i.e., functional groups) may be as effective in increasing animal diversity elsewhere (Tilman et al. 1994). In August 1993, the field was sprayed with a general as adding plant species per se. Correlative studies have shown that the architectural or structural diversity of herbicide (Round-Up, Monsanto, St. Louis) and burned after the vegetation was dead and dry. The upper 6���8 cm plants, which is likely correlated with both plant species and functional diversity, may be an important determi- of sod and soil were then removed to reduce the seed bank. The remaining soil was plowed and repeatedly nant of arthropod diversity (reviewed in Lawton 1983). Because increasing plant species diversity (Naeem et al. disked. In spring 1994, the field was disked again and smoothed. The field was divided into 342 plots (13 m 3 1995 Tilman et al. 1996) or plant functional diversity (Hooper and Vitousek 1997 Tilman et al. 1997) can in- 13 m with only the inner 11 m 3 11 m sampled) sepa- rated by walkways and roadways that were kept bare. In crease productivity, increasing plant diversity may also increase arthropod diversity indirectly by increasing total May 1994, plots were seeded with a constant mass of seeds added to each plot (divided equally among the spe- arthropod abundance and, thus, allowing rarer species to persist locally (Hutchinson 1959 Connell and Orias 1964 cies). Plots were reseeded in May 1995. As plants grew, some plots were sprayed with suitable herbicides to elim- MacArthur 1969 Brown 1981 Abrams 1995 Rosenzweig 1995 Siemann 1998). inate weeds. All plots were hand weeded two to four times every year of the experiment, and some plots were Herbivore diversity may also be influenced by the di- versity of higher trophic levels. If herbivores have appro- also spot sprayed with herbicides in the first and second years of the experiment. The experiment was composed priate trade-offs between predator and parasite suscepti- bility versus competitive ability in the absence of of three subexperiments. Two of these subexperiments were described elsewhere (Tilman et al. 1997) and were predation and parasitism, then predators and parasites may allow a greater diversity of herbivores to coexist used in this study. In order to test for the effects of plant taxonomic di- (e.g., Cramer and May 1972 Roughgarden and Feldman 1975 Levin et al. 1977 Tilman 1986 Holt et al. 1994 versity on arthropod diversity, the plant species composi- tion of each of 163 13 3 13-m plots was determined by Leibold 1996). There have been ample demonstrations that certain predators maintain herbivore diversity (e.g., random draws of one, two, four, eight, or 16 perennial plant species drawn from a core pool of 18 species (four Paine 1966 Connell 1972 Menge and Sutherland 1976 Power et al. 1996) and that herbivore abundance in gen- species each of four functional groups���C3 grasses, C4 grasses, legumes, and nonlegume forbs���and two species eral (e.g., Paine 1966 Connell 1972 Hairston and Hair- ston 1993) and arthropod herbivores in particular (e.g., from the woody functional group). There were 34, 35, 29, 30, and 35 replicates, respectively, at each level of di- Andrzejewska et al. 1967 Strong et al. 1984 Schmitz 1993 Denno et al. 1995) can be strongly limited by pred- versity. In this subexperiment, plant diversity and plant composition are uncorrelated. ators and parasites. Chain modeling (Cox and Wermuth 1993) is a tool that can be used to sort out direct and To better distinguish between the responses of arthro- pod diversity to plant taxonomic and functional diver- indirect responses of animal diversity to changes in plant diversity. sity, 79 additional plots were assigned combinations of one, two, or three functional groups and two, four, or In order to test whether increasing plant species di- versity and/or increasing plant functional diversity in- eight species. Species compositions of these plots were chosen by random draws of functional groups followed creases animal diversity, we directly manipulated plant species diversity and plant functional diversity in a well- by random draws of species within these functional groups. When needed, we used a pool of 16 additional replicated grassland experiment and measured arthropod diversity and abundance. In order to investigate direct species (four species in each of the four nonwoody func- tional groups). Another 46 plots were created with 32 of and indirect responses of arthropods to plant diversity, we examined the relationships among number of these 34 species. The 288 plots (from pooling these two types of plots with the random species draw subexperi- species planted and the diversities of different arthropod trophic groups using chain modeling and regression. ment) uncouple species diversity, functional diversity, and functional composition but have a weak correlation between these and species composition. In order to esti- Material and Methods mate the number of arthropods that might be present as Experimental Setup aerial plankton, an additional two plots were kept bare. A complete list of the plant species used in the experi- This experiment was established in a 10-ha ������brome field������ at Cedar Creek, Minnesota (,50 km north of Minne- ment is given in appendix A, and the number of plots of each treatment is given in appendix B. apolis/St. Paul), primarily to study the effects of plant