In 1963, the U.S. Department of the Interior’s Bureau of Reclamation finished building Glen Canyon Dam on the Colorado River in northern Arizona, 25 kilometers upstream from Grand Canyon National Park. The dam impounded 300 kilometers of the Colorado River, creating Lake Powell, the nation’s second largest reservoir.
By 1974, scientists found that the downstream river’s alluvial sandbars were eroding because the reservoir trapped the fine sediment that replenished the deposits during annual floods. These sandbars are important structures for many kinds of life in and along the river.
Now, by implementing a new strategy that calls for repeated releases of large volumes of water from the dam, the U.S. Department of the Interior (DOI) seeks to increase the size and number of these sandbars. Three years into the “high-flow experiment” (HFE) protocol [U.S. Department of the Interior, 2012], the releases appear to be achieving the desired effect. Many sandbars have increased in size following each controlled flood, and the cumulative results of the first three releases suggest that sandbar declines may be reversed if controlled floods can be implemented frequently enough.
Harnessing the Water and Sediment of the Colorado River
The 220-meter-high and 480-meter-wide Glen Canyon Dam has dramatically altered the 425-kilometer segment of the Colorado River that runs from the dam to Lake Mead, the nation’s largest reservoir, at the western end of Grand Canyon National Park (Figure 1). Within the Grand Canyon, and especially its upstream end, known as Marble Canyon, the dam has eliminated fine sediment once supplied from the upstream Colorado River basin, decreased peak flow volumes and magnitudes, increased low-flow magnitudes, and caused daily discharge fluctuations that generate hydroelectric power.
Collectively, these changes reduce both the size and number of the river’s sandbars [Dolan et al., 1974]. Sandbars, which occur primarily in eddies downstream from rapids, provide flat ground for camping and for backwater habitat used by native and introduced fish. They also support vegetation and supply fresh sand to dune fields that bury and protect important archaeological sites.
In 1996, scientists started experiments to learn how best to rebuild eroded sandbars. Many of these experiments involved releasing controlled floods through the hydroelectric turbines and facilities that bypass water around the turbines. These releases, known as HFEs, are about half the magnitude of the average predam spring flood and last 3 to 8 days, which nevertheless amounts to 2 to 3 times the amount of water normally released from the dam over a given period of time.
The first controlled flood occurred in 1996. This release demonstrated that sandbars grow rapidly during the first few days of a flood and that much of the deposition is eroded within 6 to 12 months by normal dam operations [Webb et al., 1999]. The most important scientific finding of the 1996 flood was that sand supplied to the Colorado River by tributaries experienced short residence times, as evidenced by declining sand concentrations during the release [Rubin et al., 2002]. These findings revealed the importance of timing controlled floods to occur shortly after flash floods from major tributaries deliver sand to the Colorado River.
Scientists and resource managers tested this paradigm by releasing controlled floods in 2004 and 2008 after seasonal thunderstorms triggered a series of large sand inputs via tributaries. These floods elevated suspended sand concentrations in the Colorado River and deposited large eddy sandbars.
The evidence from these experiments indicated that releases timed to follow sand inputs, as suggested by Rubin et al. , are, in fact, an effective sandbar-building strategy [Schmidt and Grams, 2011].
Toward a Flood Protocol
Insights gained from the 1996, 2004, and 2008 controlled floods [Schmidt and Grams, 2011] and from scientists’ understanding of the river’s sand budget [Wright et al., 2008] allowed scientists and resource managers to develop the current HFE protocol. DOI now schedules controlled floods depending on sand accumulation in Marble Canyon, the 100-kilometer river segment downstream from the Paria River. This river is the first and most important sand-contributing tributary downstream from Glen Canyon Dam.
Scientists compute sand accumulation as the difference between tributary sand input and sand export from Marble Canyon. Using this computation, resource managers aim to release controlled floods to redistribute the accumulated sand from the bed of the river to sandbars along the banks.
Although the protocol is conceptually simple, implementing it involves several science and policy challenges:
- tracking sand influx from tributaries and transport downstream by the Colorado River
- computing in-channel sand storage
- predicting sand flux for potential controlled floods
- scheduling potential controlled floods in a way that does not disrupt regional water supply and hydropower demands
- evaluating the effects of each controlled flood before triggering the next one
Wright and Kennedy  proposed a strategy to meet these challenges by coupling field measurements with stream flow and sediment routing models. The protocol roughly follows this strategy.
Timing Floods to Maximize Bar Building
To time controlled floods for maximum sandbar-building effect, U.S. Geological Survey scientists continuously monitor sand flux at the downstream end of Marble Canyon using acoustic instruments calibrated to conventional sampling of suspended sediment [Griffiths et al., 2012]. Scientists also make initial estimates of sand inputs from Paria River floods within 24 hours of each event using a model based on observed correlations among discharge, sand concentration, and sand grain size [Topping, 1997].
Scientists adjust model estimates and reduce uncertainty by measuring sand concentrations and particle sizes in water samples collected in the field. They use this combined modeling-sampling approach because sand concentrations in the Paria River are too high for other techniques, such as acoustics, to work reliably. The data and tools that implement the model to compute Paria River sand inputs are available to water managers, stakeholders, and the public.
As the sediment input season progresses, engineers from the Bureau of Reclamation use the sand routing model of Wright et al.  to predict sand export from Marble Canyon for different possible controlled flood scenarios. Scientists have calibrated and verified this empirical model specifically for Marble Canyon on the basis of water discharge and sediment grain size.
The modeling process is iterated to identify when, how much, and for how long water must be released in a controlled flood to export approximately the same amount of sand as was supplied by tributaries during the input season. The goals of this strategy are to build sandbars by mobilizing the recently accumulated sand and to avoid eroding older sand deposits.
Dam operators then schedule the release of a controlled flood from Glen Canyon Dam to match the flow magnitude and duration identified in the modeling process.
Scientists and managers use a combination of photographs and repeat topographic surveys to evaluate the short- and long-term effects of the controlled floods on sandbars. A network of remote time-lapse cameras at 43 sites distributed throughout Grand Canyon captures daily changes in sandbars that show the immediate effects of each controlled flood.
Scientists also use the photographs to semiquantitatively categorize sandbars into groupings of sites that show deposition, erosion, or no significant change. They provide these data to managers shortly after each release.
Researchers survey the topography at many of the same study sites annually to quantitatively estimate trends in sandbar area and volume [Hazel et al., 2010]. Thus, the topographic surveys provide precise measures of sandbar response, and the photographic monitoring provides timely data before and after every controlled flood.
Limiting the Flow
The Colorado River ecosystem downstream of Glen Canyon Dam is not the only resource relying on Lake Powell’s water. The dam produces a significant amount of hydroelectric power, and releasing long-duration, large-magnitude floods could affect the capacity to generate power later.
DOI caps the controlled flood volume at the maximum flow rate that can be attained with all of Glen Canyon Dam’s power plant turbines and the full capacity of the dam’s bypass tubes (1274 cubic meters per second). DOI also caps the duration of releases greater than power plant capacity (~890 cubic meters per second) at 96 hours to limit the loss of potential future power generation associated with bypassing water around the turbines.
With sufficient sediment, higher releases would likely build larger sandbars, but such releases would require using emergency spillways and are outside the scope of the HFE protocol. Whenever river managers implement controlled floods, they also reduce reservoir releases in other months to keep the annual volume of water released from Lake Powell consistent with the agreements established among the users of Colorado River water. Controlled floods can be released following either summer/fall or winter sand-supplying floods in tributaries.
Testing the Protocol
In the period from July to October 2012, just 3 months after DOI adopted its HFE protocol, Paria River floods delivered 690,000 ± 117,000 metric tons of sand to the Colorado River. In October, managers made a preliminary estimate of that sand flux and used it in the sand routing model to develop and schedule a controlled flood (see Figure 2). Paria River sand inputs in 2013 and 2014 were 2.8 and 1.8 times larger, respectively, than those in 2012, allowing managers to release additional controlled floods.
Sand inputs were more than enough to support the release of the maximum discharge and duration allowed under the protocol, but operators could not release more than 1050 cubic meters per second because some turbines were shut down for maintenance. Because these infrastructure issues limited the maximum release, these controlled floods exported from Marble Canyon less than 60% of the sand delivered in 2012 and less than 30% of the amount of sand delivered in 2013 and 2014. In each case, the Wright et al.  model has been a valuable tool for managers because it has provided a rational basis for designing controlled floods that make efficient use of the sand supplied by tributaries.
Time-lapse images showed that at least half the monitored sandbars increased in size following each controlled flood (Figure 3). However, the response of downstream sandbars to floods does not seem to vary systematically, consistent with observations from previous controlled floods [Hazel et al., 2010]. Researchers think that the amount of sand individual sandbars accumulate varies as a function of local velocities in eddies, which themselves vary due to differences in channel morphology [Grams et al., 2013].
Although resource managers have not yet established quantitative goals for sandbar rebuilding, they consider the 2012–2014 results encouraging.
The Colorado River’s Uncertain Future
The success of these initial controlled floods does not guarantee that sandbars will continue to grow. Sandbars erode between each controlled flood. Thus, the long-term effects of the HFE protocol depend on how Colorado River runoff, operational decisions about releases from Lake Powell, and seasonal precipitation in the Paria River and other tributary watersheds affect the ability of dam operators to continue implementing controlled floods. Future sediment inflows from tributaries are highly uncertain because they depend heavily on flash floods triggered by rainfall associated with intense seasonal thunderstorms, which deliver a large fraction of the Southwest’s rain.
Current climate change models cannot reliably predict how seasonal thunderstorm activity will change in the future. A succession of high-snowpack years or operational decisions to transfer water storage from Lake Powell to Lake Mead could also result in large releases of clear water that typically cause sandbar erosion; indeed, such releases occurred from 1996 to 2000 [Mueller et al., 2014] and in 2011. In these conditions, sufficient sand accumulation to trigger controlled floods is unlikely.
However, almost all climate change projections predict increases in temperature and decreases in Colorado River runoff [Vano et al., 2014]. With recent annual releases equal to or lower than releases from 2000 to 2010, the HFE protocol is likely to increase sandbar size and allow more sand to be retained in Marble Canyon, as anticipated by Wright et al. .
In this uncertain future, balancing ecosystem goals with growing needs for water and power will continue to be a challenge for society. The HFE protocol is one approach to meet some of those challenges.
Through the incorporation of scientific research, technological advances in monitoring capabilities, and the best available models, scientists and resource managers have developed a strategy that is both flexible and coupled with ecosystem drivers such as runoff and sediment delivery. Although long-term success cannot be predicted, the early results of HFE attempts to maintain the Grand Canyon’s sandbars show promise.
U.S. Geological Survey sediment transport and sandbar monitoring in the Grand Canyon is supported by the Glen Canyon Dam Adaptive Management Program administered by the U.S. Department of the Interior Bureau of Reclamation. The authors acknowledge the contributions of the many scientists, field technicians, and river guides who have dedicated themselves to the study of the Colorado River in the Grand Canyon. Robert Tusso and Joseph E. Hazel Jr. assisted with analysis of photographs and sandbar data.