Geomorphic response of the Sandy River, Oregon, to removal of Marmot Dam

Abstract

The October 2007 breaching of a temporary cofferdam constructed during removal of the 15-meter (m)-tall Marmot Dam on the Sandy River, Oregon, triggered a rapid sequence of fluvial responses as ~730,000 cubic meters (m 3 ) of sand and gravel filling the former reservoir became available to a high-gradient river. Using direct measurements of sediment transport, photogrammetry, airborne light detection and ranging (lidar) surveys, and, between transport events, repeat ground surveys of the reservoir reach and channel downstream, we monitored the erosion, transport, and deposition of this sediment in the hours, days, and months following breaching of the cofferdam. Rapid erosion of reservoir sediment led to exceptional suspended-sediment and bedload-sediment transport rates near the dam site, as well as to elevated transport rates at downstream measurement sites in the weeks and months after breaching. Measurements of sediment transport 0.4 kilometers (km) downstream of the dam site during and following breaching show a spike in the transport of fine suspended sediment within minutes after breaching, followed by high rates of suspended-load and bedload transport of sand. Significant transport of gravel bedload past the measurement site did not begin until 18 to 20 hours after breaching. For at least 7 months after breaching, bedload transport rates just below the dam site during high flows remained as much as 10 times above rates measured upstream of the dam site and farther downstream. The elevated sediment load was derived from eroded reservoir sediment, which began eroding when a meters-tall knickpoint migrated about 200 m upstream in the first hour after breaching. Rapid knickpoint migration triggered vertical incision and bank collapse in unconsolidated sand and gravel, leading to rapid channel widening. Over the following days and months, the knickpoint migrated upstream more slowly, simultaneously decreasing in height and becoming less distinct. Within 7 months, the knickpoint had migrated 2 km upstream from the dam site and became a riffle-like feature approximately 1 m high and a few tens of meters long. Knickpoint migration, vertical incision, and lateral erosion evacuated about 15 percent of the initial reservoir volume (125,000 m 3 ) within 60 hours following breaching, and by the end of the high flows in May 2008, about 50 percent of the volume had been evacuated. Large stormflows in November 2008 and January 2009 eroded another 6 percent of the original volume of impounded sediment. Little additional sediment eroded during the remainder of the second year following breaching. The rapid erosion of sediment by the modest flow that accompanied dam breaching was driven mainly by the steep hydraulic gradient associated with the abrupt change of base level and knickpoint formation and was aided by the unconsolidated and cohesionless character of the reservoir sediment. In the ensuing months, transport competence diminished as channel geometry evolved and the river gradient through the reservoir reach diminished. Changes in profile gradient in conjunction with channel coarsening and widening led to a rapid slowing of the rate of reservoir erosion. Sediment transport and deposition were strongly controlled by channel-gradient discontinuities and valley morphology downstream of the dam site. Those influences led to a strong divergence of sand and gravel transport and to deposition of a sediment wedge, as much as 4 m thick, that tapered to the preremoval channel bed 1.3 km downstream of the dam site. After 2 years, that deposit contained about 25 percent of the total volume of sediment eroded from the reservoir. The balance was distributed among pools within the Sandy River gorge, a narrow bedrock canyon extending 2 to 9 km downstream of the dam site, and along the channel farther downstream. A two-fraction sediment budget for the first year following breaching indicates that most of the gravel eroded from the reservoir reach was deposited within the sediment wedge and within the gorge, whereas eroded sand largely passed through the gorge and was broadly dispersed farther downstream. The sequence of transporting flows affected the specific trajectory of reservoir erosion and downstream sediment transport during the 2 years following breaching. However, because the overall erosion was largely a consequence of knickpoint retreat and channel widening, which in the 2 years after removal had affected most of the reservoir reach, it is unlikely that the specific sequence of flows significantly affected the overall outcome. Because the knickpoint had largely passed through the reservoir within 2 years, and the remaining reservoir sediment is mostly isolated high above armored or bedrock banks, it is unlikely that substantial additional sediment from the reservoir site will enter the system unless very large flows occur. Continued channel evolution downstream of the dam site is probable as deposits formed in the first 2 years are episodically mobilized. Below the Sandy River gorge, detection of effects related to release of reservoir sediment is challenging, especially in areas of sand deposition, because of the high background supply of sand in the river and substantial channel dynamism.