Nonpoint source pollution from cropland has been identified as the primary source of nitrogen and sediment, and a significant source of phosphorus in the Chesapeake Bay. These pollutants, whether from point or nonpoint sources, have been found to be the primary cause of declining water quality in the Bay. Numerous studies have indicated that, for many watersheds, a few critical areas are responsible for a disproportionate amount of the nutrient and sediment yield. Consequently, if pollution control activities are concentrated in these critical areas, then a far greater improvement in downstream water quality can be expected with limited funds. In this research a phosphorus transport model is incorporated into ANSWERS, a distributed parameter watershed model. The version of ANSWERS used has an extended sediment transport model which is capable of simulating the transport of individual particle classes in a sediment mixture during the overland flow process. The phosphorus model uses a nonequilibrium desorption equation to account for the desorption of phosphorus from the soil surface into surface runoff. The sediment-bound phosphorus is modeled as a function of the specific surface area of the soil and transported sediment. The equilibrium between the soluble and sediment-bound phosphorus is modeled using a Langmuir isotherm. The extended ANSWERS model was verified using water quality data collected from rainfall simulator plot studies conducted on the Prices Fork Research Farm in Blacksburg, Virginia. The plots consisted of four 5.5 m wide by 18.3 m long strips with average slopes ranging from 6.2 to 11 percent. Two of the plots were tilled conventionally, and the remaining two were no-till. Simulated rainfall at an intensity of 5 cm/h was applied to the plots and runoff samples were analysed for sediment and phosphorus. The model was then verified by comparing the simulated responce with the observed data. The results of the verification runs ranged from satisfactory to excellent. Also developed is a technique for selecting a design storm for ANSWERS. The technique creates an n-year recurrence interval storm with a duration equal to the time of concentration of the watershed. The intensity pattern is simulated on a ten-minute interval using a first-order Markov model with a lognormal distribution. Using a two-year recurrence interval design storm, the use of the model is demonstrated for evaluating the application of conservation practices to critical areas on a Virginia watershed. Application of BMP's to critical areas is shown to be substantially more cost effective in terms of pollutant reduction than nonselective placement of BMP's if cost sharing funds are involved. / M.S.