from past experience; and (iii) based on the feedback to managers, communities, and other constituencies, management interventions are then refined.

Recognition of the inherent variability in natural systems has also led to greater emphasis being placed on the need for long-term monitoring. The recent detection of a dramatic drop in phytoplankton abundance off the coast of Southern California was made possible only due to a monitoring program initiated in 1951 by the California Cooperative Fisheries Investigations (CalCOFI) (Roemmich and McGowan 1995). In contrast, the causes of coral bleaching episodes in the Caribbean remain controversial, in part because of the lack of long-term monitoring of water temperatures in the region. Without long-term records, it is difficult to distinguish human-caused changes from natural changes in most systems. Even with long-term records, however, conclusive evidence of human-caused influences typically also requires some form of experimental verification.

Finally, because of the predictive limits implicit in the fact that ecological communities are not tightly co-evolved assemblages of species existing in an equilibrium state, a premium is placed on site-specific information about the identity and biology of the species that occur in a given area. Resource managers are typically constrained by the absence of inventories of species in most regions of the United States.

To some extent, the need for inventories could be diminished if we had better knowledge of the complementarity of patterns of distribution of various groups of species. It has long been assumed, for example, that the pattern of diversity of plants would be a good predictor of the pattern of diversity in other groups of terrestrial organisms, because of the importance of the physical structure provided by plants for the diversity of other taxa in the community. We know that the distribution of plant diversity does not completely mirror other groups (Prendergast et al. 1993), but there are relatively few studies that seek to determine just how coincident patterns actually are.

The greatest need for improved ecological knowledge currently relates to its role in setting environmental goals. We are in the midst of a fundamental rethinking of environmental goals, but while our science is good enough to call into question some of our past goals, we cannot yet be certain of their replacements. In the past we set narrowly defined goals for species protection, wetlands conservation, or sustainable harvests. Yet progress toward these narrow goals did not always achieve appropriate balance among the many benefits we received from living systems and were often in conflict. Worse still, many of our environmental management goals were designed to be reactive to threats to the environment and,