A deterministic structure imitating a sediment yield model should mathematically approximate the behavior of two distinct phases of the phenomenon, the upland phase and the lowland channel phase. For upland erosion, research is most needed in quantifying gully erosion, whereas for lowland streams it is most needed to explain the influence of meanders on bed material transport and to develop a floodplain accounting component. In both phases, research is needed to explain the effects of unsteadiness and flow nonuniformities on transport. Sensitivity analysis studies of sediment yield models are needed to illustrate how well the models can be calibrated and what output precision can be expected from input data with known statistical characteristics. Present estimates indicate that although a regression of annual sediment discharge on annual water discharge might be expected to give a prediction within 20% of the observed value, even a well-calibrated digital model might give an error greater than 40% for a single storm on a small watershed.