related to groundwater levels and stream flows. While it is extremely difficult to predict the consequences of an introduction, certain types of species are clearly higher risk than others. Weedy ("r"-selected) species that have large numbers of offspring, rapid growth, and high dispersal abilities, are more likely to become established in new habitats than "K"-selected species with lower population growth rates.

While the removal of a keystone species will, by definition, have profound impacts on a community and the services it provides, the removal or addition of some species in communities appears to have little demonstrable effect either on other species or on an ecosystem process. In these situations, other species compensate for the absence of the target species, at least over the short term. (It is not known for sure if all functions of the species in question are compensated for; in fact, it is rarely understood what the full range of functions is for each species.) As is the case with keystone species, it is not possible to predict based on general information about the biology of a species whether its loss would have little effect on the ecosystem. However, the one group of species whose loss would tend to have relatively little impact contains species that are currently in danger of extinction, simply because the primary impact of their loss would already have been felt when the species was originally reduced in population size to the point of endangerment.

Thresholds may sometimes exist in the response of ecosystem services to changes in abundance or distribution of species. In many of the world's biomes, ecosystems can exist in two or more alternative states that differ widely in rates of productivity or some other ecosystem services. Semiarid grassland can remain productive as long as grazing density is modest, but when density rises above a threshold the ecosystem shifts rapidly to a state of low productivity. Often, large and sustained reductions in exploitation are necessary to return ecosystems to high-productivity states, and in some cases, recovery from the undesirable state is not possible. At present, scientific knowledge is insufficient to be able to predict either the presence or the level of thresholds in ecosystem behavior.

We currently have insufficient knowledge to make accurate assessments of the potential consequences of human-induced climate change on ecological systems. A few general predictions can be made: species found only high on mountains will be threatened with loss; protected areas with significant elevational or latitudinal gradients will be more secure than those without; species restricted to coastal regions will face significant threats from sea level rise, etc. By virtue of more complete knowledge of the requirements of crop plants, our ability to make predictions of impacts is somewhat better for agro-ecosystems than for natural systems. But our predictive ability is seriously constrained due to the high regional uncertainty of predictions from climate models and the lack of detailed autecological information (information on the interaction of a species and its physical environment) on many important species.