In recent years, improvements in point-source depuration technologies have highlighted the problems regarding agricultural nonpoint (diffuse) sources, and this issue has become highly relevant from the environmental point of view. The considerable extension of the areas responsible for this kind of pollution, together with the scarcity of funds available to local managers, make minimizing the impacts of nonpoint sources on a whole basin a virtually impossible task. This article presents the results of a study intended to pinpoint those agricultural areas, within a basin, that contribute most to water pollution, so that operations aimed at preventing and/or reducing this kind of pollution can be focused on them. With this aim, an innovative approach is presented that integrates a field-scale management model, a simple regression model, and a geographic information system (GIS). The Lake Vico basin, where recent studies highlighted a considerable increase in the trophic state, mainly caused by phosphorus (P) compounds deriving principally from the intensive cultivation of hazelnut trees in the lake basin, was chosen as the study site. Using the management model Groundwater Loading Effects of Agricultural Management Systems (GLEAMS), the consequences, in terms of sediment yield and phosphorus export, of hazelnut tree cultivation were estimated on different areas of the basin with and without the application of a best management practice (BMP) that consists of growing meadow under the trees. The GLEAMS results were successively extended to basin scale thanks to the application of a purposely designed regression model and of a GIS. The main conclusions can be summarized as follows: The effectiveness of the above-mentioned BMP is always greater for erosion reduction than for particulate P reduction, whatever the slope value considered; moreover, the effectiveness with reference to both particulate P and sediment yield production decreases as the slope increases. The proposed approach, being completely distributed, represents a considerable step ahead compared to the semidistributed or lumped approaches, which are traditionally employed in research into tools to support the decision-making process for land-use planning aimed at water pollution control.