Scientific understanding of the structure and function of ecological systems has advanced tremendously over the past three decades. During this time, ecology and evolutionary biology have changed from descriptive disciplines into experimental and theoretical sciences with an ever-growing capability for prediction. As a result, ecological knowledge has become more and more useful to policy-makers. The utility of ecological information has been further enhanced by advances in research tools and technologies—particularly the development of sophisticated mathematical models of ecological systems and the growing use of geographical information systems—that help to provide information to decision-makers in forms tailored to the decisions that they face.

Still, the limits of scientific understanding are obvious. Even where general trends, such as wildlife population declines or changing stream quality, are clear, scientists are often unable to determine the impact of a specific action on those trends with any precision (or even whether the trends are a consequence of previous human actions or are natural). Problems of cumulative effects, lack of site-specific ecological knowledge, and the natural variability of ecological systems conspire to add substantial uncertainty to almost all uses of scientific knowledge in environmental decision-making. As a consequence, we must place as much emphasis today on techniques and policies for coping with uncertainty as we do on efforts to reduce that uncertainty.

In this paper, I assess the influence of recent advances in ecological knowledge on environmental policy decision-making, the current status of policy-relevant ecological knowledge, and key opportunities where advances in knowledge (or technologies) would improve decision-making.

Ecological knowledge influences three general aspects of environmental policy. First, choices about basic societal goals are made based on knowledge—or assumptions—about how the world works. Consequently the overarching goals we set for environmental management are based, in part, on our knowledge of ecological systems. For example, it was long assumed that biological communities were highly co-evolved equilibrium systems of organisms. Based on that assumption, a common goal of resource management has been to maintain certain systems in their ''natural" state. But research that has found that biological communities are neither highly co-evolved nor regulated around equilibrium calls into question the very meaning of a natural state. We can't objectively define a natural state in the absence of equilibrium states because we have no way of knowing what the structure and function of a system would have been if humans had never intervened. While we might set a goal of minimizing human intervention in certain systems, our current knowledge of ecology suggests that we should not set a goal of maintaining a system in a "natural state."

Second, ecological knowledge undergirds our ability to predict the ecological—and economic and social—consequences of human impacts, such as filling a wetland, changing stream flows, or introducing chemicals into the environment. Until recently, virtually all environmental management has been reactive rather