Uncertainty of Sediment Transport Modeling Through Two Run-of-River Mega-Dams On The Madeira River

The Madeira River supplies ~15% of the Amazon River’s total flow but transports up to half of the Amazon’s overall sediment load. Until recently, basin-wide sediment dynamics and river geomorphology have not been well quantified due to data scarcity of the remote watershed. Understanding sediment dynamics is increasingly important as the Madeira region continues to rapidly develop. In the last decade, two mega (>1000 MW), hydroelectric “run-of-river” (ROR) dams were installed along the Madeira River mainstem, often together referred to as the Madeira Hydroelectric Complex (MHC). By definition, ROR reservoirs have little water regulation capacity or ability to store water, especially during high flow periods. Related to sediment transport, ROR reservoirs are expected to fill with sediments to a dynamic equilibrium relatively quickly because of the preserved hydraulic energy for sediment movement through these systems. However, this depositional expectation may not hold along large, tropical rivers with extremely high sediment loads. Deterministically determining the time to dynamic equilibrium and predicted maintenance through modeling is highly uncertain due to the sparseness of long-term sediment monitoring before dam planning. In this work, the uncertainty of present and future MHC sedimentation was reassessed using the post-dam monitoring data (2013-2020), data collected on an independent field campaign in 2018, and a one-dimensional sediment transport model. First, the recent field data were synthesized with relevant sediment and geomorphological data from the literature. Modeling of reservoir sedimentation was then conducted for both present and projected future climatic conditions using the HEC-RAS sediment transport model. Uncertainties attributed to both input data and model parameters were tracked by randomly sampling observed sediment characteristics and all realistic model parameter values. Total likelihood of parameter sets and marginal parameter probability distributions were established from present-day model runs (2013-2020) and subsequently used to track uncertainty of sediment input data on future projections of sediment transport under different synthetic hydrologic regimes. Since dam construction, monitoring of control cross-sections show major sedimentation near the foot of the dams and on the now flooded, pre-dam floodplains (in some places as high as 10-15m), much more rapidly than previously modeled. In particular, the 2013-2014 flood (estimated as a 300-year return interval flood – 0.33% AEP) caused significant sedimentation above turbine intake elevation at both dams, illustrating the value in considering a range of synthetic, non-stationary hydro-sedimentological future scenarios. Reducing uncertainty of future dam effects on Madeira River sediment transport can be achieved through continued sediment monitoring, model uncertainty analyses, and leveraging improved non-stationary predictions of future hydrology in this globally important basin.