The worsening degradation of natural resources urgently requires the adoption of more sustainable management practices. This need has led to growing interest and investment in monitoring systems for tracking the condition of natural resources. This study is concerned with the design of monitoring systems that have direct relevance for the management of natural resources. We call these Policy Relevant Monitoring Systems (PRMS). Such systems have several key characteristics. They provide: a) a decision framework for selecting resource problems to monitor that offer potentially large social payoffs relative to the costs of monitoring, b) timely, including early warning information on emerging problems, c) a means of identifying the causes of an emerging problem, d) an analytical framework for identifying options for corrective action, e) an institutional framework for achieving ownership among key stakeholders (the resource users and those affected by the resource use) and agreement about emerging problems, the corrective actions to take, and effective implementation, and f) a built-in mechanism for learning from past experience to improve the performance of the monitoring system over time. The approach is developed and illustrated through detailed examination of the Arenal-Tempisque watershed in Costa Rica. This watershed exhibits classic multiple user and externality problems: deforestation by dairy and cattle farmers in the upper watershed leads to soil erosion and siltation of the various reservoirs that feed an important hydro-electric power generation system, and agro-chemical use by irrigated farmers has adverse impacts on a highly valued wetlands park and on wildlife and fishing in the lower reaches of the watershed.