Scientific Investigations Report 2007–5008
U.S. GEOLOGICAL SURVEY
Scientific Investigations Report 2007–5008
Back to Table of Contents
The calibrated Detroit Lake model was used to examine several scenarios. Results from the following applications are presented herein:
Determining the amount of sediment that exits the reservoir and the amount of sediment that is retained by the reservoir is important in determining the depletion of reservoir storage, making decisions on reservoir operation, and assessing downstream effects. Reservoir sediment trap efficiency is defined as the percentage of sediment input retained by the lake. Methods such as the Brune ratio, calculated from ratios of reservoir capacity to annual water inflow, have been used to calculate this quantity (Brune, 1953; Snyder and others, 2004). However, calculations that use a constant reservoir capacity or residence time are not recommended for reservoirs such as Detroit Lake that vary greatly in surface elevation and volume over the course of a year. That change in volume also changes the residence time and the corresponding amount of time available for suspended sediment to settle (Ambers, 2001). The calibrated Detroit Lake model keeps track of suspended sediment deposition, allowing trap efficiency to be calculated directly with the model. In calendar year 2002, the trap efficiency of Detroit Lake was calculated to be 93 percent, in 2003 it was calculated as 90 percent, and during the 2005–06 storms, it was calculated as 94 percent (table 5). Because of residence time, the sediment that exits the reservoir in one year may have entered the reservoir in a previous year; similarly, suspended sediment that enters the reservoir in 2002 may not exit until a future year. Most of the suspended sediment that enters the lake is trapped; only 10 percent or less exited in the modeled time periods.
If sediment grains with a 2.7 g/cm3 density are assumed to have a 30 percent pore space upon deposition, leading to sediment deposits with 1.89 g/cm3 dry bulk density, then approximately 14,300 m3 (11.6 acre-ft) of sediment was deposited in the lake in calendar year 2002, 11,820 m3 (9.6 acre-ft) in 2003, and 34,900 m3 (28.3 acre-ft) from December 1, 2005, to February 1, 2006. The mass of sediment entering and deposited in the reservoir in the 2-month period during the 2005–06 storms was more than double the mass of sediment deposited in the entire calendar years of either 2002 or 2003, emphasizing the importance of storm flows to reservoir sediment budgets. These volumes of sediment are all less than 0.01 percent of Detroit Lake’s full pool volume of 561 million m3 (455,000 acre-ft), indicating that Detroit Lake is not likely to fill up anytime soon. The actual density of the deposited sediment can vary depending on factors such as composition, age, and wetting and drying, which can lead to compaction (Vanoni, 1975).
The model results indicated that most of the sediment was deposited in the upper reaches of the reservoir and areas near major inflows (fig. 22). The velocity of the inflows, and their sediment carrying capacity, decreases substantially as they enter the pooled reservoir. The predicted deposition rates can be substantial during a large storm inflow. Area-averaged deposition rates in the North Santiam arm of the lake during the 2005–06 storms, for example, were simulated to be as high as 20 kg/m2 for the 2-month simulation period (fig. 22). Assuming a sediment bulk density of 1.89 g/cm3, this deposition could have reached an average thickness of 1.06 cm. This is an area-averaged value; some areas would have less deposition, and others would have more. Deposited sediment can be resuspended and redistributed during other storm events, particularly if the stage of the lake decreases enough to convert that area from a pooled reach to a riverine reach. The Detroit Lake model currently does not include algorithms to fully simulate the effects of scour and sediment resuspension, which may be important processes that redistribute sediment further into the reservoir.
The calibrated model also was used to investigate the size of the sediment flowing into and exiting the reservoir. In 2002, suspended sand- and silt-sized sediment made up 85 percent of the inflow sediment mass. It made up 83 percent of the inflow mass in 2003, and 92 percent during the storm events between December 1, 2005, and February 1, 2006. In contrast, most of the sediment that exited the reservoir downstream was suspended clay. That clay-sized sediment made up 91 percent of the outflow mass of sediment in 2002, 93 percent of the outflow sediment mass in 2003, and 84 percent of the outflow sediment mass for the 2005–06 storms. Table 5 summarizes the masses of suspended sediment in the inflows and outflow, as well as the reservoir trap efficiencies, for each of the three modeled time periods.
The calibrated model was used to examine the relative contribution of each of Detroit Lake’s tributaries to the suspended sediment in the reservoir outflow for 2003. Each suspended sand and silt and suspended clay group for each inflow, including the distributed tributaries (the small ungaged tributary groups), was modeled as a separate sediment group for this analysis. The settling rates for each group were the same as those in the calibrated model, so that the sediment behavior overall was unchanged. The only difference was the ability to track various sources of sediment through the system. In order to determine the initial distribution of sediment from various sediment sources in the reservoir to start the year, an entire year’s simulation was run with estimated initial conditions; then the distribution of sediment from various sources in the reservoir on the last day of the year was used to reinitialize the sediment distribution at the start of the year, before running the simulation a final time.
The results from this scenario show that all inflows to Detroit Lake contributed to the suspended sediment in the reservoir outflow (fig. 23, table 6). The contributions were generally greatest from the North Santiam and Breitenbush Rivers, the largest tributaries, throughout the year. The contribution from the North Santiam River was unique in that it contributed suspended sediment to the outflow from September through November, whereas the other tributaries had ceased to deliver much sediment in that period. The North Santiam River is the only Detroit Lake tributary sourced in a glacial basin, and the reasons for the different characteristics of sediment in the headwaters of this river are still under investigation by the USGS.
Model results indicated that sediment that exited the reservoir in 2003 was mostly suspended clay (93 percent of the total mass). The largest contribution of this clay-sized sediment came from the North Santiam River, followed by Breitenbush River, Blowout Creek, the distributed tributaries (small ungaged tributaries), Kinney, French, and Box Canyon Creeks. The minor amount of suspended sand and silt discharged from the reservoir came predominantly from Kinney Creek, Blowout Creek, and the ungaged small tributaries. Tributaries that enter farther upstream contributed less suspended sand and silt compared to those that enter closer to the dam, probably because the suspended sand and silt settles more rapidly. Table 6 quantifies each tributary’s export efficiency for both suspended sand and silt and suspended clay. The North Santiam River, entering about 8 km upstream, contributed no suspended sand and silt to the reservoir outflow. The distributed tributaries represent an estimate of sediment associated with the ungaged fraction of inflow to the reservoir. The model apportions the distributed tributaries’ inflows with their associated sediment to every segment in the model, including those segments closest to the dam. In the model, sediment introduced into the reservoir near the dam does not have much time to settle before it reaches the dam outflow. This is the reason that the distributed tributary represents a relatively large portion of suspended sediment in the reservoir outflow. Some small ungaged inflows do enter Detroit Lake near the dam (fig. 1), and they are likely to contribute disproportionately to the coarse sediment load in the reservoir outflow.
The water temperature in the North Santiam River at the location of the dam, without the influence of the reservoir, was estimated by calculating a volume-weighted mix of Detroit Lake model inflows for 2003 (fig. 24). This is only an estimate, as it does not include warming or cooling that might occur between each gage location and the dam. For the 2003 inflow mix, maximum water temperatures occurred in July, whereas the modeled reservoir outflow reached annual maximum temperatures later in the year, in early October. The presence of the reservoir, with its midlevel lake outlet, produced cooler temperatures in summer and warmer temperatures in autumn compared to the river in the absence of the reservoir. This effect has been observed for many other storage reservoirs in which the outflow draws water from relatively deep in the water column (for instance Hagg Lake, Sullivan and Rounds, 2006). Besides differences in seasonal temperature patterns, the reservoir outflow also had smaller 24-hour temperature variations compared to the inflow-mix.
In addition to the fact that the reservoir outflow can exceed Oregon’s maximum water temperature criterion in autumn, the cool summer temperatures of the outflow also may be undesirable (Larson, 2000). In an attempt to address this issue, the USACE experimented with mixing the cool waters released from Detroit Dam’s power-generating outlet with warmer surface waters released from the spillway in June and July of 1979. Further study by USACE led to the recommendation that a selective withdrawal tower be installed as part of Detroit Dam (Larson, 2000 referring to two USACE reports). This type of device would blend water from different levels in the water column in order to release water with a more natural seasonal temperature pattern. The USACE recently completed a selective withdrawal tower at Cougar Reservoir, another large storage reservoir in the Willamette Basin, to address downstream temperature issues (Resource Management Associates, 2003).
The calibrated Detroit Lake model was used to assess the degree to which a more natural seasonal temperature pattern could be achieved downstream of the dam by using such a device. The hypothetical selective withdrawal device was set up to withdraw water from the penstocks at their current elevation, and from a sliding gate outlet that could be adjusted once per day vertically through the water column to select an elevation, with a corresponding water temperature, that could be blended into the penstocks to meet a temperature target. The elevation of the sliding-gate outlet was determined automatically by the model in an attempt to match a user-specified downstream temperature target. In this run, the model was given a downstream temperature target based on the inflow-mix water temperature and “filtered” twice with a 7-day running average (fig. 24B). These running averages were used simply to smooth the target temperature time series and remove daily variations and some effects of short-term weather variations. The resulting outflow water temperature from this simulation is shown in figure 24B along with the temperature target. The selective withdrawal device allowed the reservoir outflow to approximate a more-natural seasonal temperature pattern for most of the year. However, the reservoir did not have a sufficient stored volume of cold water in autumn to meet the target, leading to a small temperature peak at that time of year. The peak was small, and was less than the water temperature criterion. A comparison to the reservoir in 2003 without selective withdrawal (fig. 25A), shows that selective withdrawal (fig. 25B) led to a shallower thermocline, as warm water was withdrawn from near the lake surface in summer. Further work may examine the effect of such an altered thermal structure of the lake on the transport and fate of suspended sediment in the lake. Additional scenarios also could be run to simulate different outlet characteristics, various temperature targets, altered flows, changes in reservoir operation, and/or climate shifts.
Back to Table of Contents