Sediment-transport investigations of the upper Yellowstone River, Montana, 1999 through 2001: Data collection, analysis, and simulation of sediment transport

The upper Yellowstone River in Montana is an important State and national water resource, providing recreational, agricultural, and commercial benefits. Floods in 1996 and 1997, with recorded peak discharges having recurrence intervals close to 100 years, caused substantial streambank erosion and hill- slope mass wasting. Large quantities of sand-, gravel-, and cobble-sized material entrained by the flood flows became flood-bar deposits, creating a source of sediment available for transport during future floods. The flood damage and resulting sedimentation raised concerns about potential streambank-stabilization projects and how the river and riparian corridor might be managed in the future. The U.S. Geological Survey, in cooperation with the Park Conservation District, the Montana Department of Transportation, and the U.S. Army Corps of Engineers, investigated sediment transport in the upper Yellowstone River near Livingston from 1999 through 2001 as part of a cumulative effects study to provide a scientific basis for future river management decisions. The purpose of this report is to present the results of data collection, analysis, and simulation of sediment transport for the upper Yellowstone River.

The study area included a 13.5-mile study reach of the upper Yellowstone River where substantial sediment transport occurred in 1996 and 1997. In this study area, the upper Yellowstone River is a high gradient, coarse-bed stream having a slope of about 0.0028 foot per foot or more than 14 feet per mile. The study area drains about 3,551 square miles, and runoff results primarily from snowmelt during the spring and summer months. As part of sediment-transport investigations, the U.S. Geological Survey surveyed river cross sections, characterized streambed-material particle size using particle counts and sieve analyses, and collected bedload- and suspended-sediment data during three runoff seasons (1999-2001). Data were collected for stream discharges that ranged from 2,220 cubic feet per second (typical of pre- and post-runoff discharge) to 25,100 cubic feet per second (about 125 percent of bankfull discharge).

The distribution of streambed-material particle size was determined, and sediment-transport curves for bedload discharge, suspended-sediment discharge, and total-sediment discharge were developed. The threshold values of streamflow and average stream velocity needed for initiation of bedload transport for selected sediment-size classes showed that little to no bedload was transported for an average stream velocity below about 3 feet per second, and the only particle size transported as bedload at that velocity was sand. Over the range of stream discharges sampled and with silt- and finer-sized particles excluded, bedload discharge averaged about 18 percent of the total-sediment discharge, equal to bedload discharge plus suspended-sediment discharge. At the lowest and highest stream discharges sampled, bedload was, respectively, less than about 2 percent and about 30 percent of the total-sediment discharge. Over the range of stream discharges sampled, the sand-sized part of the total suspended-sediment discharge averaged about 48 percent, where the total suspended-sediment discharge included sand-, silt- and finer-sized particles. At the lowest and highest stream discharges sampled, the sand-sized part of the total suspended-sediment discharge was, respectively, less than about 16 percent and about 50 percent of the total suspended-sediment discharge. The sediment-transport curves were compared to curves for selected sites in the western United States having drainage areas ranging from 21 square miles to over 20,000 square miles. Daily sediment loads transported at bankfull discharge were calculated for each site and results were plotted in relation to drainage area. Results based on the 1999-2001 data-collection period indicate that the estimated daily bedload transported at bankfull discharge in the upper Yellowstone River exceeded the envelope line that bounds the upper end of the data for other selected sites in the Northern Rocky Mountains and is similar in magnitude to that for selected sites in Alaska having braided channels and glacial and snowmelt runoff. Similar comparisons for suspended sediment indicate that daily suspended-sediment load at bankfull discharge is relatively high in the upper Yellowstone River, plotting slightly above the envelope line that bounds the upper end of the data for other selected sites in the Northern Rocky Mountains.

Sediment data were used to develop individual transport equations for seven size classes of sediment ranging from small cobbles to very fine sand. A step-wise regression procedure relating sediment discharge to important hydraulic variables showed that average stream velocity was the only significant variable at the 95-percent confidence level. Bedload and suspended-sediment data and equations indicate that more sand is transported for a given velocity than any other particle size, and very little sand-size sediment load is transported below an average stream velocity of about 2.5 feet per second. Transport of coarser-sized sediment (limited to bedload) becomes very little for an average velocity less than about 3.5 feet per second. Results for the 1999-2001 data-collection period indicate that sediment transport in the upper Yellowstone River tends to be limited more by the transport capacity of the stream (capacity or transport limited), than to the availability of sediment in the watershed (supply limited).

Sediment data collected and analyzed were used to simulate sediment transport in the study reach using the BRIdge Stream Tube model for Alluvial River Simulation, or BRI-STARS computer model. The model was calibrated and verified using selected data from historical runoff periods. Simulated total-sediment loads, on a reach-averaged basis, were in good agreement with the total-sediment loads determined from the transport curve for the 2-year flood hydrograph but were considerably smaller for the total-sediment loads determined from the transport curve for the 50-, 100-, and 500-year flood hydrographs. The differences probably were largely due to the inability of the model to simulate streambank erosion, hillslope mass-wasting, and other channel-widening processes, which had supplied substantial quantities of sediment to the channel during the 1996 and 1997 floods, and probably continued to contribute to the sediment load in the subsequent years (1999-2001) when the data were collected. Furthermore, the transport curve was applied beyond the measured data for the highest discharges, and may thus be unreliable. Also, the transport curve derived from only limited data may not apply over the full duration of the hydrograph and sediment might be transported over only a portion of the hydrograph, especially for rivers like the upper Yellowstone where snowmelt runoff predominates. The true sediment discharge is, therefore, unknown and might be closer to the simulated values than to the values calculated from the transport curve.