Descriptions of ecosystems normally use highly aggregated units, such as trophic levels or guilds. Those aggregated units are used as "black boxes," irrespective of the identity and number of species within them. However, since the early days of ecology, the question "what is the good of so many species?" has attracted interest. This question was initially discussed is relation to ecosystem "stability." Goodman's (1976) influential review of the sparse evidence in favor of the "diversity-stability hypothesis" discredited it as a kind of myth creation and showed that the term "stability" was defined so loosely that a critical test of the hypothesis was practically impossible. After Goodman's intervention, the topic was almost forgotten for nearly two decades, but it experienced a revival during the mid 1990s and even extended its scope. The question was extended from diversity-stability relationships to the importance of biodiversity for ecosystem functioning in general (note that the now-fashionable term "biodiversity" has in principle no meaning other than "diversity" as defined in Section 7.4).
The revival of the topic was partially indebted to more rigorous problem definitions and experimental designs, but even more importantly to the growing concern about the current wave of species extinctions. According to some researchers, the current rate of species extinctions is c.100-1000 times higher than preindustrial background levels (Chapin et al. 2000, Pimm et al. 1995). In such a situation, how many species are needed for ecosystems to function as we are used to see them function becomes a burning question. Although there is still much scientific debate on the topic, some consensus has been reached (summarized in Hooper et al. 2005). Ecosystem functions that relate to diversity are either defined as stability properties (persistence, constancy, resistance, or elasticity to perturbations) and biogeochemical properties (productivity, nutrient cycling, and retention), or in a more anthropo-centric way as "ecosystem goods and services" (in the case of fresh water this may mean supply of drinking water, fisheries, aesthetic and hygienic qualities for recreation and tourism, fishing, etc.).
In principle, studies on the effects of biodiversity on ecosystem functions could be based on correlations of field data or on experiments. However, correlative studies have one difficulty: differences in local diversity are in most cases dependent on various environmental conditions (e.g., solar energy input, nutrient richness, disturbance regime, regional species pool and dispersal) which in themselves influence ecosystem functions. It may become very hard or almost impossible to find the appropriate statistical techniques to quantify the extra influence of diversity on ecosystem functions. The same applies to experiments where diversity differences between the different experimental units have been manipulated indirectly via environmental factors. Therefore, most of the reliable studies are based on artificially composed communities, but even those are not without methodological problems.
• Species identity versus richness: Early studies have been criticized because it was not clear whether the enhanced performance of species richer treatments was due to species richness per se or to the specific properties of the additional species in the richer treatments (Huston 1997). This problem can be overcome by combining two levels of replication: same species number/same species and same species number/different species (Downing and Leibold 2002). Obviously, this makes such experiments extremely laborious and restricts the experimental approach to small and easily cultured organisms.
• Species versus functional diversity: Conventional measures of diversity and species richness (see Section 7.4.1) make no distinction between mixtures of species consisting of similar or different species. On the other hand, one would intuitively expect that a mixture of three equal-sized diatoms would produce a less pronounced diversity effect than a mixture of a diatom, a flagellate and a nitrogen-fixing cyanobacterium. This problem can be solved by lumping species into predefined (but subjective) functional groups and treating functional groups as quasi-species when calculating diversity. More sophisticated methods are borrowed from numerical taxonomy. Species are ordered in an n-dimensional space of n relevant and quantifiable traits; distances between species are calculated and used to construct a cluster diagram whose attributes can be used as diversity measures (Petchey and Gaston 2006).
• Non-random species loss: Natural communities differ from artificially assembled experimental communities by having a history of competitive exclusions, successful invasions, and unsuccessful invasion attempts. This means that they represent the most stable subset of the total of all possible species combinations, irrespective of their diversity. Similarly, the expected loss of species by environmental stress is by no means random. Thus, a statistically desirable attribute of experimental communities (random assemblage) makes them unnatural at the same time. The problem is acknowledged, but the scientific community is still waiting for convincing experimental designs.
Sampling effect versus complementarity Diversity effects can fall into two broad categories: the sampling effect and the complementarity effect. The sampling effect means that in species-richer communities there is a higher probability of finding particularly productive or stress-resistant species, or species better adapted to changed environmental conditions. The complementarity effect depends on niche differences between the coexisting species and/or positive interactions (symbiosis, facilitation). Niche differences would for example result in a more complete usage of the resource base and, thus, higher productivity. Some authors (Doak et al. 1998) have called the sampling effect a statistical inevitability without ecological mechanism behind it, but even so, it would be an ecological reality with potentially important implications for conservation issues. Attempts to separate the sampling form the complementarity effect in experimental data are based on the concept of overyielding (Hector et al. 2002). This means that a species mixture performs better in the parameter of interest than (i) the species best performing in monoculture, (ii) the dominant species in monoculture, or (iii) a weighted average of the species mixture calculated from single-species performances in monoculture. Obviously, each of the three standards of comparison can give different results, and the debate about the most appropriate choice is still going on.
The insurance hypothesis (Yachi and Loreau 1999) is closely related to the sampling effect. It rests on the assumption of redundancy among species— that functionally similar species can replace each other if disturbances eliminate one species, or if the environmental changes lead to the disadvantage of one and the advantage of another species. Intuitively, one would expect the insurance to become "safer" the more species there are.
with periphyton (Matthiessen and Hillebrand 2006). The aquaria had a biofilm at the bottom consisting of 15 algal species, which comprised the regional species pool. In each aquarium, there were plastic tubes ("local communities") with initial species numbers of 2, 4, and 8 species. The plastic tubes were open at the top and totally submerged, thus permitting a low background level of dispersal. This background dispersal was enhanced by stirring up the bottom biofilm 0, 1, 2, 7, 14, and 28 times during the 28 day experimental period. Final species richness in the local communities was unrelated to initial species richness of the local communities, but increased strongly with the frequency of stirring events up to 7 events;a further increase of stirring frequency did not increase species richness any further (Fig. 8.27).
Most biodiversity-ecosystem effect experiments rely on closed species assemblages, but it is obvious that the insurance effect can also be achieved by immigration, provided that the immigrating species have easily transportable propagules or that powerful transport vectors exist, as is the case in the recol-onization of disturbed river reaches from upstream (see Section 5.8).
The importance of the regional species pool was demonstrated in simple aquarium experiments
0.0 0.2 0.4 0.6 0.8 1.0 Dispersal frequency (in 28 d)
0.0 0.2 0.4 0.6 0.8 1.0 Dispersal frequency (in 28 d)
Was this article helpful?