A 1.9-mile reach of the Big Lost River, between the Idaho National Engineering and Environmental Laboratory (INEEL) diversion dam and the Pioneer diversion structures, was investigated to evaluate the effects of streambed erosion and bedrock constrictions on model predictions of water-surface elevations. Two one-dimensional (1-D) models, a fixed-bed surface-water flow model (HEC-RAS) and a movable-bed surface-water flow and sediment-transport model (HEC-6), were used to evaluate these effects. The results of these models were compared to the results of a two-dimensional (2-D) fixed-bed model [Transient Inundation 2-Dimensional (TRIM2D)] that had previously been used to predict water-surface elevations for peak flows with sufficient stage and stream power to erode floodplain terrain features (Holocene inset terraces referred to as BLR#6 and BLR#8) dated at 300 to 500 years old, and an unmodified Pleistocene surface (referred to as the saddle area) dated at 10,000 years old; and to extend the period of record at the Big Lost River streamflow-gaging station near Arco for flood-frequency analyses. The extended record was used to estimate the magnitude of the 100-year flood and the magnitude of floods with return periods as long as 10,000 years. In most cases, the fixed-bed TRIM2D model simulated higher water-surface elevations, shallower flow depths, higher flow velocities, and higher stream powers than the fixed-bed HEC-RAS and movable-bed HEC-6 models for the same peak flows. The HEC-RAS model required flow increases of 83 percent [100 to 183 cubic meters per second (m3/s)], and 45 percent (100 to 145 m3/s) to match TRIM2D simulations of water-surface elevations at two paleoindicator sites that were used to determine peak flows (100 m3/s) with an estimated return period of 300 to 500 years; and an increase of 13 percent (150 to 169 m3/s) to match TRIM2D water-surface elevations at the saddle area that was used to establish the peak flow (150 m3/s) of a paleoflood with a return period of 10,000 years. A field survey of the saddle area, however, indicated that the elevation of the lowest point on the saddle area was 1.2 feet higher than indicated on the 2-ft contour map that was used in the TRIM2D model. Because of this elevation discrepancy, HEC-RAS model simulations indicated that a peak flow of at least 210 m3/s would be needed to initiate flow across the 10,000-year old Pleistocene surface. HEC-6 modeling results indicated that to compensate for the effects of streambed scour, additional flow increases would be needed to match HEC-RAS and TRIM2D water-surface elevations along the upper and middle reaches of the river, and to compensate for sediment deposition, a slight decrease in flows would be needed to match HEC-RAS water-surface elevations along the lower reach of the river. Differences in simulated water-surface elevations between the TRIM2D and the HEC-RAS and HEC-6 models are attributed primarily to differences in topographic relief and to differences in the channel and floodplain geometries used in these models. Topographic differences were sufficiently large that it was not possible to isolate the effects of these differences on simulated water-surface elevations from those attributable to the effects of supercritical flow, streambed scour, and sediment deposition.