Numerical Modeling of Lateral Erosion During Reservoir Drawdown

The drawdown of reservoirs behind dams is an important management strategy (e.g. for decommissioning or maintaining aging infrastructure, flushing of fish and sediment), as well as a compelling opportunity to study individual and coupled erosional processes. In particular, lateral erosion in reservoirs during drawdown is recognized as an important mechanism for sediment evacuation but not as well-studied or modeled as knickpoint migration.

A numerical model was thus developed to represent retrogressive bank erosion with reservoir drawdown and grain size, with the goal of evaluating factors controlling the rate and volume of, and mechanisms associated with, lateral erosion during drawdown. Key modeled processes included hydraulic toe erosion, drawdown of groundwater within the bank, geotechnical failure via limit equilibrium analysis, and retrogression that couples downdrag and normal stresses between failure blocks. Validation of the model was not possible due to a lack of high frequency surveys for reservoir drawdowns. Instead, field observations from the Elwha dam removal were used to provide representative pre-drawdown reservoir geometry, soil characteristics, and drawdown operations. Field measurements were coupled with five drawdown scenarios: Staged, Staged with Short drawdown increments, Staged with Long drawdown increments, Slow, and Rapid drawdown.

From a process perspective, results highlight the importance of including retrogression as an avenue for sequential block failures as all fine-grained scenarios exhibited at least one time step with more than one failure. The drawdown increment was found to be a first-order control slope instability via the development of drained or undrained conditions. A majority of failures occurred under undrained conditions, and results highlighted the importance of effective internal friction under drained conditions. From a management perspective, results highlighted how the design of drawdown, and consequently the processes driving the timing of erosion, created a tradeoff between the amount of impact created and when the impact is produced. The study also articulated the need for, and challenges of, coupling models and field observations, particularly for systems that are rapidly changing and where not all variables can be accurately measured.