Between 1999 and 2005, drought in the western United States led to a >44 m fall in the level of Lake Powell (Arizona-Utah), the nation's second-largest reservoir. River discharges to the reservoir were halved, yet the rivers still incised the tops of deltas left exposed along the rim of the reservoir by the lake-level fall. Erosion of the deltas enriched the rivers in sediment such that upon entering the reservoir they discharged plunging subaqueous gravity flows, one of which was imaged acoustically. Repeat bathymetric surveys of the reservoir show that the gravity flows overtopped rockfalls and formed small subaqueous fans, locally raising sediment accumulation rates 10–100-fold. The timing of deep-basin deposition differed regionally across the reservoir with respect to lake-level change. Total mass of sediment transferred from the lake perimeter to its bottom equates to ~22 yr of river input.

It is apparent that before emplacement of the dam gully degradation in terraces was restored by periodic alluvial deposition from river floods, but perhaps even more important is the redistribution of flood sands onto higher terraces by wind. Thus, we propose the term "restorative base-level hypothesis" to emphasize the dynamic equilibrium between gully erosion and renewed deposition, a process that remains active in Cataract Canyon but is disrupted in Grand Canyon by the presence and operation of the dam.

We developed type geomorphic settings to develop a conceptual process model for the diverse small-catchment geomorphic system in Grand Canyon. Research findings explain how streams are able to cross broad, flat terraces given a rainfall event and how they become progressively more integrated with the river. The primary channelization processes are ponding and overflow, alluvial fan progradation, and infiltration and piping, all of which contribute to nickpoint migration. An understanding of these processes was essential to building the geomorphic model.

The predictive mathematical model quantifies erosional vulnerability by applying a hypothetical rainfall event of 25 mm/hour onto a catchment above a "pristine" terrace sequence. The principal driving factor for erosion is basin area. The principal resisting factor for erosion is terrace diffusion capacity, which is a function of terrace sand cross-sectional area and infiltration capacity. Several important modifying factors are applied to the basic model to determine relative vulnerability of each terrace to gully erosion. Vulnerability of the top terrace at each catchment is plotted against the measured amount of gully erosion in that terrace, providing a base line against which progressive changes in gully depth can be easily monitored in the future.

Field studies and research show that: (1) gully erosion of terraces has been severe during the past 20 years in Grand Canyon due to unusually high precipitation; and (2) sediment deprivation coupled with the lack of large annual floods has caused a reduction in restorative (depositional) factors. Continued measurement and documentation of geomorphic processes in catchments, particularly at type geomorphic settings, will further refine and verify the predictability of the model. We conclude that beach-habitat-building flows are essential for initiating natural restorative processes and that one of the most important processes in gully mitigation may be eolian reworking of newly deposited flood sands onto higher terraces. Prior to the construction of Glen Canyon Dam, gully-deepening and river/wind depositional processes were in dynamic equilibrium, allowing the preservation of ancient cultural sites for the past several thousand years.