is an emergent property with profound importance to humans. Flexibility is the ecosystem-level equivalent of the ability of species to ''adapt" in response to environmental changes in ways that lead to persistence. For the individual species that flexibility is embedded in its genetic diversity; the same idea applies at the ecosystem level, where biotic diversity is equally important to resilience. Management for sustainability means preservation of biodiversity both for its own sake and because of its importance in maintaining ecosystem resilience.
Marine ecosystems are often defined by geographic boundaries, such as the Bering Sea. They can be large and can overlap geographic and political boundaries; their own boundaries are often not easy to delineate or define (Alexander 1990). In this discussion, we use the concept of the large marine ecosystem described by Sherman (1990) as "relatively large regions of the world, generally on the order of 200,000 km2 [or more], characterized by unique bathymetry, hydrography, and productivity…" The geographic scale of most marine ecosystems is larger than the scale of the local human communities that depend on them. Furthermore, marine ecosystems are frequently include discontinuous habitats with substantial inputs (e.g., nutrients, sediments, energy, invaders) from other kinds of habitats in the ecosystem or from other ecosystems. Because of this coupling of habitats in different areas, management schemes at the local level must include the effect of local decisions on the larger ecosystem and over long times.
Fishes and other components of marine food webs have complex life histories with different habitat requirements at each of several stages. Differences among species in spawning grounds and dispersal ability have important implications for management. Knowledge of habitat requirements and of timing of settlement of larval dispersal stages is needed to understand the effects of localized changes in environment and to predict the strength of interactions among species with complex life histories.
Recent approaches to resource management place humans and their institutions squarely within ecosystems (Pickett and Ostfield 1995) and recognize the need for constructive and broad-based participation of the public in policy making and implementation (Kessler and Salwasser 1995). As experience with comanagement, virtual communities, participatory research, ITQs, and other institutional innovations increases, so does the importance of designing them appropriately. If it is important to match institutional scales of complexity with biological ones (Ostrom in press, NRC 1996a, 1996b), more work is needed to determine how this can be done in general and in particular cases. The development of comanagement and community-based management raises many issues, including ones of balancing interests of local and interest groups with those of larger publics and longer-term ecological systems and finding ways to gain the benefits