LETTER Non-neutral patterns of species abundance in grassland communities W. Stanley Harpole* and David Tilman Department of Ecology, Evolution, and Behavior, University of Minnesota, St Paul, MN, USA *Correspondence: E-mail: wharpole@uci.edu Abstract Although the distribution of plant species abundance in a Minnesota grassland was consistent with neutral theory, niche but not neutral mechanisms were supported by the ability of species traits to predict species abundances in three experimental grassland communities. In particular, data from 27 species grown in monoculture showed that species differed in a trait, R*, which is the level to which each species reduced the concentration of soil nitrate, the limiting soil nutrient and which is predicted to be inversely associated with competitive ability for nitrogen (N). In these N-limited habitats, species abundance ranks correlated with their predicted competitive ranks: low R* species, on average dominated. These correlations were significantly different than expected for neutral theory, which assumes the exchangeability of species traits. Additionally, we found that changes in relative abundance after environmental change (N-addition or disturbance) were not neutral but also were significantly associated with R*. Keywords Cedar Creek, Konza, neutral theory, niche theory, nitrogen addition, nitrogen competition, R*, species abundance, succession, traits. Ecology Letters (2006) 9: 15���23 I N T R O D U C T I O N Niche theory and neutral theory make diametrical assump- tions about the necessity and importance of species traits in determining species abundance and diversity patterns (Hub- bell 2001 Chave et al. 2002 Clark & McLachlan 2003). Niche theory assumes that species traits represent evolu- tionary adaptations to the physical and biotic environment (Ackerly et al. 2000 Gillespie 2004), that species face unavoidable tradeoffs, and that such tradeoffs are an essential mechanism allowing interacting species to coexist and determine the relative species abundances (Tilman 1982, 1988 Chesson 2000 Rees et al. 2001 Chase & Leibold 2003 Reich et al. 2003). Neutral theory has been proposed to be a simple yet sufficient explanation for observed relative species abundance patterns (Bell 2001 Hubbell 2001). The essence of neutral models of biodiversity is that all individuals are assumed to be functionally equivalent (Bell 2001 Hubbell 2001). Hubbell (2005) has asserted that functional equival- ence does not require that species be identical in their traits, just that differences in traits do not lead to any differences in their per capita demographic rates. Because niche mecha- nisms and neutral processes are not mutually exclusive, the question then is whether niche mechanisms are significantly more important than a null expectation of neutrality. Here, we test whether patterns of species abundances in midwes- tern USA prairie communities show a niche assembly signature beyond what would be expected from neutral processes. In addition we explore whether species compo- sition changes in response to environmental change might be consistent with neutral expectations, an issue that neutral theory currently fails to address (Chase 2005). Many tests of neutral theory have focused on attempts to distinguish differences between the goodness-of-fit of alternative expected distributions to observed species abundance distributions. There have been recent advances leading to analytical solutions to Hubbell���s neutral model (Volkov et al. 2003 Etienne 2005) and superior methods for model comparison that can allow better discrimination between the potential importance of dispersal vs. niche mechanisms in structuring ecological communities (Etienne & Olff 2005). However, there have been few ��strong tests�� of neutral theory, for example, tests using species traits to predict the identity of dominant vs. rare species (McGill 2003 Wootton 2005). Here, we emphasize this latter approach, although as a logical point of departure we first Ecology Letters, (2006) 9: 15���23 doi: 10.1111/j.1461-0248.2005.00836.x ��2005 Blackwell Publishing Ltd/CNRS

ask: is the distribution of species abundances in these prairies consistent with neutral theory? Second, species abundance distributions might give the impression of neutrality that may or may not be supported by stronger tests asking what is the relationship between species traits and abundance ��� are species equivalent? We test this using data from three experiments at Cedar Creek, Minnesota (CDR) and Konza Prairie, Kansas (KNZ), each with very different assembly histories. Third, we ask do species abundance patterns change in a neutral manner in response to environmental change? We test this using data from a: (i) nitrogen (N) gradient and (ii) post-disturbance succes- sional gradient. Because soil N is the major limiting resource in these grasslands (Tilman 1984a), we use an index of competitive ability for N as the primary plant trait we consider. In particular, we use R* (Tilman 1982) which is predicted to be inversely related to competitive ability for N, where R* is the level to which a species reduces soil nitrate when it is grown to equilibrium in monoculture. R* summarizes and results from the full suite of morphological and physiological traits related to acquiring, transforming, conserving and losing resources (Tilman 1988, 1990). As an integrative index of a species�� competitive ability, R* correlates with multiple species traits associated with a low N environment (Craine et al. 2002), and has been shown to predict species dominance in terrestrial and aquatic systems (Tilman 1982 Wedin & Tilman 1993 Grover 1997 Fox 2002). This R* rule has been shown to be robust even when strict model assumptions are not met (Fox 2002). We used species R* values from 27 perennial forb and grass species planted in monoculture to test whether species abundance patterns in multiple experiments at CDR and KNZ might be consistent with neutral theory. Specifically, if species are assumed to be competitively equivalent then there should be no relationship between a species trait such as R* and species abundance. To test this hypothesis, we used species abundance data from three grassland experi- ments that were similar in that they were all N-limited communities and approximately equilibrial, but they differed greatly in how they were assembled: (i) old field grasslands at CDR that have undergone c. 60 years of succession following their abandonment from agriculture (ii) experi- mentally assembled grassland communities at CDR in which all species were established at equal initial abundances thus allowing local interactions in the absence of regional dispersal processes to determine abundances over 5 years and (iii) experimental grasslands at KNZ that have never been ploughed and thus may represent relict fragments of the tall-grass prairie ecosystem. KNZ is environmentally similar but spatially distant and isolated from CDR. We tested two corollary hypotheses. (i) If neutrality holds and species are essentially functionally equivalent, then species should respond equivalently to environmental changes such as N-deposition there should be no direc- tional change in the relative abundance of species with increasing levels of added N. Alternatively, if changes in relative abundance over time or space are because of deterministic sorting according to species traits (Clark & McLachlan 2003) then the importance of N-competition traits will decrease as N becomes less limiting. (ii) If species are functionally equivalent with respect to their traits, then successional changes in their relative abundances following disturbance should be stochastic. For instance, in succes- sional old fields, after abandonment from agriculture, under the assumptions of neutrality, relative abundances of colonizing species should reflect their relative abundance in the regional species pool and migration changes thereafter should be due to stochastic drift. Alternatively, immediately after abandonment from agriculture, fields would be dominated by a mixture of annuals and perennials associated with agriculture, mostly of Eurasian origin. If there are niche differences based on the N competitive abilities of these species, with native prairie perennials having lower R* values (Tilman & Wedin 1991), then the Eurasian species should be replaced by native prairie perennials that have low abundance in the agricultural landscape, that may be slow colonists, but are superior N competitors. M E T H O D S Study sites We analysed data from multiple experiments conducted at the CDR Long-Term Ecological Research site (LTER) in East Bethel, Minnesota and at the KNZ LTER in Manhattan, Kansas. CDR soils derive from a glacial outwash sand plain (Grigal et al. 1974) for which N is the primary limiting resource (Tilman 1984b). Vegetation at CDR is a mix of successional and prairie-like oldfield grasslands, savanna, woodland and wetland (Pierce 1954). KNZ soils range from upland cherty silt loam to lowland silt clay loam (Knapp et al. 1998). Productivity at KNZ is also limited by N (Knapp et al. 1998). Grassland communities at KNZ, in contrast to CDR, have never been ploughed. Grassland vegetation at both sites is characterized as tall grass prairie (Knapp et al. 1998). Monoculture gardens and species traits To quantify species in terms of an index for competitive ability, we used species-level data from monoculture plots of common prairie plant species at CDR (Craine et al. 2002). These experimental plots were established in 1992 (CDR LTER experiment E111). Species, typically were planted in 16 W. S. Harpole and D. Tilman ��2005 Blackwell Publishing Ltd/CNRS