Biodiversity: population versus ecosystem stability

Ecology, March, 1996 by David Tilman

INTRODUCTION

Fragmentation, grazing, forestry, and nutrient deposition are decreasing the biological diversity of many of the earth's remaining semi-natural ecosystems (e.g., Ehrlich and Ehrlich 1981, Aerts and Berendse 1988, Wilson 1988, 1992, Ehrlich and Daily 1993). Elton (1958:145-153) suggested that decreased diversity would lead to decreased ecological stability and functioning, and presented several arguments in support of this. However, there has been continuing debate about the diversity-stability hypothesis (e.g., Gardner and Ashby 1970, May 1973, 1974, McNaughton 1977, 1985, Pimm 1979, 1984, Lawton and Brown 1993, Schultze and Mooney 1993, Vitousek and Hooper 1993, Givnish 1994, Tilman and Downing 1994). For instance, in a simple model of multispecies competition, May (1973) showed that population dynamics were progressively less stable as the number of competing species increased, and concluded that there need not be any relationship between diversity and stability. Gardner and Ashby (1970), DeAngelis (1975), Gilpin (1975), and Pimm (1979) reached similar conclusions using different models. In contrast, McNaughton (1977) presented data on plant productivity in the Serengeti that supported Elton. King and Pimm (1983) modeled systems like McNaughton's and found that higher plant diversity generally led to greater biomass stability in response to changes in herbivory, just as McNaughton had observed. Similarly, Vitousek and Hooper (1993) suggested that the rates of many ecosystem processes could be increasing but saturating functions of species diversity.

Despite this controversy, there have been surprisingly few studies of the relationships between diversity and stability (Mellinger and McNaughton 1975, McNaughton 1977, 1985, Leps et al. 1982, Ewel 1986, Berish and Ewel 1988, Ewel et al. 1991, Frank and McNaughton 1991, Tilman and Downing 1994). In this paper I report on a 13-yr study of plant species abundances, diversity, and production in 207 permanent Minnesota grassland plots that experienced great climatic variability, including a major drought in 1987 and 1988. I analyze the relationships between plant species richness and year-to-year variability in the abundances of both individual species and total plant community biomass. These analyses help reconcile the seemingly divergent predicted effects of biodiversity on population vs. ecosystem stability (Elton 1958, May 1973, McNaughton 1977). I also further analyze the relationships between plant species richness and drought resistance and resilience presented in Tilman and Downing (1994). The resilience analyses are expanded to control for the confounding effects of the size of the perturbation on recovery (Pimm 1984).

Because plant species richness was not directly controlled in this experiment, but rather differed among plots and fields and in response to nitrogen addition (Tilman 1993), detailed information on other factors that covaried with plant species richness is presented. This is important because correlations among variables in a multi-causal system, such as an ecosystem, are open to alternative interpretations. Such problems are minimized, but not eliminated, by multiple regressions that statistically control for variables that covary with species richness.

METHODS

In 1982, four fields at Cedar Creek Natural History Area were chosen for an experimental study of the effects of nitrogen on grassland productivity, species dynamics, and biodiversity. Methods are in Tilman (1987, 1993) and are only summarized here. Fields A, B, and C were unburned successional fields that had been abandoned from row crop farming for 14, 25, and 48 yr, respectively, in 1982. Field D was a prairie opening in native, never-farmed savanna and has been burned in the spring for 2 of every 3 yr starting in 1966 (Tester 1989, Faber-Langendoen and Tester 1993). In each field, an area fenced to reduce densities of mammalian herbivores (which otherwise increase after nutrient addition; Tilman 1983) was divided into plots, and plots randomly assigned to one of nine treatments. Fields A, B, and C each had 54 plots, each 4 x 4 m, with six replicates per treatment. Field D had 45 plots, each 2 x 4 m, with five replicates per treatment. Plots were separated by 1-m walkways. Experimental treatments (Tilman 1987) were (1) no nutrient addition (controls), (2) addition of all nutrients but N (i.e., of P, K, Ca, Mg, S, Na, and trace metals), and (3-9) addition of all nutrients but N in conjunction with one of seven different experimental rates of N addition (1, 2, 3.4, 5.4, 9.5, 17, or 27.2 g[center dot][m.sup.-2][center dot][yr.sup.-1] of N as N[H.sub.4]N[O.sub.3]). Nutrients were added annually in equal amounts in early spring (1-7 May) and early summer (23-30 June). In addition, all plots received [approximately equal to]1 g[center dot][m.sup.-2][center dot][yr.sup.-1] of N via atmospheric deposition. The rates of N addition used in the remainder of this paper are the sum of experimental and atmospheric N addition. Fine-ground agricultural lime was added, as needed, to prevent changes in soil pH associated with N[H.sub.4]N[O.sub.3] addition.