This study was conducted to improve understanding and prediction of sediment delivery through agricultural watersheds, with emphasis on the pathways from edge-of-field to receiving waters. The study was focused on agricultural watersheds within the University of Wisconsin (UW) - Platteville Pioneer Farm and one of the UW Discovery Farms located in southwestern Wisconsin. Artificial neural network (ANN) models were developed to predict runoff and sediment yield from agricultural watersheds that employ best management practices (BMPs). Results showed that input parameters representing BMPs were important for accurately simulating runoff and sediment yield from these watersheds. The study also showed that ANN models were able to successfully simulate runoff and sediment yield during training, validation and testing phases. Sediment eroded from upland source areas is often carried to the watershed outlet via grassed waterways. Critical shear stress of the soil is often estimated to determine the potential for soil to be detached. Previous studies suggest that critical shear stress may vary with antecedent moisture content. The dynamic nature of critical shear stress in an upland agricultural field and grassed waterway of a nested watershed was investigated at Pioneer Farm by measuring critical shear stress over a range of antecedent soil moisture conditions. Results showed that critical shear stress in both the grassed waterway and the agricultural field increased as soil moisture increased until the soil moisture content reached the plastic limit. Above the plastic limit, critical shear stress of the soil decreased significantly and was relatively constant, ultimately rendering the soil more susceptible to erosion. Finally, the process-based Water Erosion Prediction Project (WEPP) model was used to develop regressions equations that use channel, watershed and storm characteristics to estimate sediment delivery ratios (SDRs) for grassed waterways draining upland agricultural fields. Upland agricultural management scenarios considered included: (i) corn-oat-alfalfa crop rotation, chisel plow tillage, and terraces, and (ii) corn-oat-alfalfa crop rotation, chisel plow tillage, and no-terraces. Better R2 values resulted from equations developed for non-terraced fields compared to terraced fields suggested that channel and storm parameters were better able to explain the variation in SDR for grassed waterways draining from non-terraced fields.