Soil erosion is a problem that causes both the loss of valuable soil and water quality degradation by sediment and attached nutrients and pollutants. Soil loss and sediment discharge to waterbodies can be prevented and mitigated by best management practices (BMPs). The USDA-NRCS reported that BMPs such as filter strips and grassed waterways can reduce up to 50 to 70% of the runoff and sediment leaving agricultural land. This study includes measured data analyses and representation of BMPs such as row cropping, filter strips and grassed waterways in a small experimental watershed (2.09 ha) at Nelson Farm, MS. Row cropping was conducted throughout a study period from 1990 to 1995 (6 years). Filter strips and grassed waterways were implemented in October 1991 and August 1994, respectively. A total of three phases were established according to the BMP implementation schedule. Phase 1 started from December 1990 to October 1991 with row cropping only, phase 2 started from October 1991 to August 1994 with row cropping and filter strips, and phase 3 started from September 1994 to December 1995 with row cropping, filter strips, and grassed waterway. The WEPP (Water Erosion Prediction Project) model and its parameters including physically based parameters are used and adjusted for the representation of BMPs. Input data includes weather data measured by the National Sedimentation Laboratory researchers and staff in Oxford, Mississippi. Management, crop yield, and BMP information in previous studies by Dabney and Meyer were used. Watershed characteristics were derived from a 2 foot contour map using GIS. These watershed characteristics were also found in previous studies. The parameter setting for the vegetative practices were calibrated and validated to agree with observed data at the location of the practices in each phase. There are two WEPP projects depending on two different flow paths, one is based on flow along the cropping row and the other is based on flow over the crop rows when the rows are collapsed by large runoff events after tillage. The Soil and Water Assessment Tool (SWAT) was run with the same watershed properties and parameter settings although there were some parameters not available in SWAT. In general, SWAT estimated larger runoff and sediment yields although the difference was small. A possible explanation is that SWAT uses one representative slope, which is a linear slope, while WEPP uses dynamic slope steepness, which includes convex and concave slope shapes. Another reason is that there is no spatial location for filter strips and no representation for the multiple filter strips is possible in SWAT. The BMP impact analyses (using WEPP) estimate the reduction of runoff and sediment yield by each BMP in each phase. The results of these analyses show that row cropping reduced 11.1% of runoff and 15.8% of sediment yield in phase (1) Row cropping and filter strips reduced 14.6% of runoff and 41.9% of sediment yield in phase (2) Row cropping, filter strips, and grassed waterway reduced 20.5% of runoff and 65.7% of sediment yield in phase (3) Total reduction by all BMPs installed according to their schedule during the study period was 14.3% of runoff and 38.5% of sediment yield. Various scenario analyses have been conducted. These include the failure of BMPs, the effectiveness of a combination of BMPs, the long-term impacts of BMPs, and the spatially distributed estimation of the effectiveness of filter strips. BMP failure analysis was conducted for filter strips and grassed waterway in each phase based on the parameter value adjustment by 25%, 50%, 75%, and 100% from the range of possible parameter values. In phase 2, with filter strips, the results of filter strips failure analysis show the reduction of runoff ranges from -4.2% (best) to +0.6% (no function) and that of sediment yield ranges from -29.6% (best) to +14.0% (no function). There would be more runoff and sediment yield generated than observed when the filter strips are not maintained because there is no vegetation on filter strips. In phase 3, with filter strips and a grassed waterway, the results of failure analysis show the reduction of runoff ranges from -2.1% (best) to +12.1% (no function) and that of sediment yield ranges from -14.3% (best) to +138.1% (no function). For the same reason, the runoff and sediment yield were larger in the 'no function' scenario than observed. (Abstract shortened by UMI.).