The huge amount of work on the effects of species richness on ecosystem function have generally shown that with greater plant species richness, you tend to have increased primary productivity, nutrient uptake and greater stability to disturbances. This has been interpreted as meaning that maintaining ecosystem functions at certain thresholds is dependant upon species richness, and decreases in species richness reduce productivity as well as ecosystem stability and multifunctionality.
However, the shape of this relationship depends on how functions react to losing species.
Most evidence for the effects of species richness on ecosystem function has come from controlled experiments which manipulate plant species richness randomly. Random manipulation of the plant communities is equivalent to supposing that extinction risk is the same for all species.
This is obviously not true in nature, where particular characteristics of a species can determine the likelihood of extinction (e.g. body size in mammals; dispersal mechanism, seed size and reproductive strategies in plants). Indeed, the studies that have examined the effects of realistic extinction patterns on ecosystem function have tended to show quite different results to those which assume random extinction.
Researchers would argue that up until now they have tried to look at the general patterns of ecosystem function with changes in richness. This is fine, and more of this kind of work needs to be done, particularly outside of grasslands. However, we also need to understand what characteristics determine extinction probabilities and how these link to traits which might influence ecosystem function. Conceptually this has been characterised as the linkages between response traits, those traits which determine the establishment or persistance of a species, and effect traits, those traits which influence ecosystem properties and functions.
To explore this idea I will use the example of primary productivity. If traits which promote primary productivity are positively correlated with those that encourage persistence following disturbance, then primary productivity may be relatively robust to species loss. However, if there is an negative relationship, then there would be a disproportionate effect on productivity. Finally, if there is no obvious link between traits then the assumptions of biodiversity ecosystem function experiments and random extinction would hold true.
The situation is obviously more complex than I have described here with abundance playing a massive role in determining ecosystem functions. However, I think investigating the linkages between traits is likely to be best hope we have of forming generalisations about how species extinctions can alter ecosystem functioning in the real world.
This is a massive task and up until now I have seen little work which attempts to deal with this directly, rather than as a purely conceptual issue (though there has been some very good work on non-random extinctions). As such these ideas remain largely untested. However, with the surge in interest in ecosystem services it is vital we try to work out how the functions that support services that we depend upon work. We are living in a time of unprecedented alteration of biodiversity by humans and these changes could have dramatic effects on our life support system. Extinctions form part of this problem, but we have little idea about the consequences of real extinctions on ecosystem functions and services.