The National Phosphorus Project rainfall simulator was used to quantify overland flow and P transport from nine sites distributed throughout the watershed of a New York City Watershed Agriculture Program collaborating dairy farm. Observed concentrations of total dissolved phosphorus (TDP) were low (0.007-0.12 mg L(-1)) in flow from deciduous forest, extensively managed pasture, and hillside seeps; moderate (0.18-0.64 mg L(-1)) in flow from intensively managed pastures, a hayfield, and a cow path; and extremely high (11.6 mg L(-1)) in flow from a manured barnyard. Concentrations of TDP from sites without fresh manure were strongly correlated with soil test P (TDP [mg L(-1)] = 0.0056 + 0.0180 x Morgan's soil test phosphorus [STP, mg kg(-1)]; R2 = 84%). Observed concentrations of suspended solids were low (16-137 mg L(-1)) in flow from vegetated sites, but were higher (375-615 mg L(-1)) in flow from sites with little ground cover (barnyard, cow path, plowed field). Under dry summer conditions the time to observed overland flow was shorter (<18 min) for nonfield areas (seeps, barnyard, cow path) than for field and forest areas (27-93 min), indicating that hydrologically active nonfield areas of minor spatial extent but with high soil P (e.g., cow paths and barnyards) can play a significant role in summertime P loading. When soils started from field capacity (second-day) time to overland flow was uniformly less than 23 min, indicating that under wet watershed conditions low-P source areas can dilute overland flow from concentrated sources.