Quantifying uncertainty and variability in sediment yield estimates in Le Sueur River Basin
Friday, 19 December 2014
The Soil and Water Assessment Tool (SWAT) is commonly used to estimate suspended sediment loads and predict changes in loads as a result of changes in management or conservation practices at large watershed (>2000 km2) scales. However, several simplifications and omissions of processes render the model susceptible to systematic biases. For example, net gains or losses of sediment during channel routing are estimated by very simplified means that are difficult to validate. Also, the model does not account for channel processes such as bluff erosion, which is implicated as a major contributor of sediment in the Le Sueur River Basin (LRB) in south central Minnesota. As sediment yield is directly proportional to channel (peak or mean) velocity or flow depth for all four formulations of sediment yield models utilized by SWAT, it is critical to correctly quantify the water available and the distribution time of its arrival. In that context, proper quantification of the extent, distribution and efficiency of subsurface tile drainage is important in agricultural landscapes. SWAT provides a mechanism to account for sub-surface tile drainage, but in practice actual data to constrain this flow pathway are typically not available. For example, although extensive use of subsurface drainage in LRB has been documented, the location, density and extents of the tile drainage are unknown, leading to uncertainty in the flow residence time estimates. In this study we have explored the uncertainties related to subsurface drainage and channel-margin sediment gains and losses. The base SWAT model is modified to account for changes in geometry resulting from the simultaneous occurrence of deposition and erosion. Furthermore, we demonstrate that multiple parameter combinations can result in statistically similar sediment yields at sub-basin outlets and suggest approaches to quantify and/or circumvent such problems of equifinality.