Numerous site-specific studies have quantified the form and magnitude of P loss in agricultural runoff, but in order to meet the requirements of end-users who need to implement cost-effective control strategies, this research needs to be incorporated into generic models of nonpoint source pollutant loss from land. Such models need to be based on the expert knowledge and should be simple to use and easy to apply. End-users such as government agencies and water utilities are largely concerned with assessing the relative risk, or gaining rough estimates of the amounts, of agricultural nonpoint P pollution from different land use practices. Data-hungry process-based models, while elegant and all-encompassing, may not be suitable for the simple decision support frameworks. A number of risk assessment approaches aimed at predicting P loss from nonpoint sources are being developed in Europe, North America and Australia. There are many similarities in these independently-developed indices and models but also some important differences reflecting different strategic priorities for both research and land management needs. Similarly, more complex process-based models calculating P loads from land to water have been developed. These models are mostly used by research communities to get a more accurate and dynamic understanding of P loads and to develop sets of best management practices, which are often site-specific. This chapter critically reviews current risk assessment approaches for managing nonpoint P loss from agricultural land, and compares the criteria used to set risk-based targets. Examples of process-based models that can be used together with risk assessment approaches to ensure success in agricultural P management are also presented.