indicators for the impacts of agriculture (OECD, 1997). The OECD program seeks to improve our understanding of the agricultural-environmental relationships affecting sustainability in agriculture so that managers can better develop, monitor, and assess agricultural programs that address the environmental problems.

The OECD's selection criteria for agricultural-environmental indicators include: (1) policy relevance, (2) analytical validity, (3) measurability, and (4) level of aggregation (i.e., scale issues, including both space and time). These criteria could be especially relevant to watershed management if the OECD definitions of various units of scale (i.e., field, farm, watershed, regional, national, and global) were modified so that watersheds were addressed.

In many areas in the American West, rangelands are critical components of watersheds. The National Research Council has defined rangeland health as "the degree to which the integrity of the soil and the ecological processes of rangelands are sustained" (NRC, 1994). The Natural Resources Conservation Service (NRCS), Bureau of Land Management (BLM), and other agencies and organizations charged with management of some 740 million acres (300 million hectares) of western rangelands (about a third of all U.S. land) have initiated programs to identify attributes comprising indicators for rangeland health, focusing on soil stability and watershed function. Research organizations such as the U.S. Department of Agriculture Agricultural Research Service (ARS) and the Land Grant Colleges in the West are also developing research programs to quantify watershed health and its indicators across a range of watershed scales and environments. Efforts to better quantify watershed health are under way in many areas, and one example is presented in Box 5.1.

One ingredient that is built into successful watershed management is the use of "good science" in decisionmaking, but even the best science is inadequate to remove all uncertainty when dealing with environmental, economic, and social systems. The manager would prefer that the scientist or engineer provide hard and fast numbers with clearly understood implications, but despite the use of the best available models, data, and ideas, experts cannot offer precise understanding about the way. watersheds and their components work. Risk assessment is one method of improving the usefulness of science for the decisionmaker because it provides improved understanding of the degree and types of uncertainty in management applications (NRC, 1993). Ecological risk assessments differ from environmental impact analysis and hazards assessments. The issue of uncertainty is of special importance to watershed managers because they often must work with incomplete information, and because watershed processes exhibit random, or stochastic, behavior. Environmental risk assessment is a scientific procedure that can augment the tool kit of the manager in dealing with an uncertain world where