Correct identification of P sources in rural watersheds is critical for the development of cost-effective measures to combat agriculturally-driven eutrophication. The chemical composition of various storm runoff types (field surface runoff, field drain outfalls, roads, farmyards, and septic tanks) and the receiving streams in three micro (<10 km2) watersheds of varying agricultural intensity were monitored over a 2-yr period. Mean weekly stream soluble reactive phosphorus (SRP) and total phosphorus (TP) concentrations increased from 29 and 69 μg L-1, respectively in the watershed with the lowest intensity agriculture to 382 and 503 μg L-1, respectively in the watershed with high intensity agriculture and a village sewage treatment works. Concentrations of TP in storm runoff varied by up to two orders of magnitude reflecting the complex origins, routing, and composition of contributing source areas. Application of the DESPRAL test suggested field runoff TP concentrations were influenced by both P and organic matter in soil. However, runoff from impervious surfaces (farmyard and roads), and/or influenced by septic tank discharges, was significantly more concentrated (0.08-16 mg TP L-1, mean >>1 mg L-1) than surface and subsurface runoff from cultivated land and pasture (0.02-3.6 mg TP L-1, mean <1 mg L-1), and/or contained a significantly greater proportion (>50% vs. <50%) of P in dissolved forms. It is concluded that P sources associated with the functioning of rural communities (impervious surfaces, detergents, and wastewater) may be more ecologically relevant than those associated with agriculture and should be better quantified and controlled to avoid localized eutrophication impacts.