Estimating Bedload Sediment Delivery To The Great Lakes From Sixty Michigan Rivers

This research was completed in partial fulfillment of a Doctorate in Civil Engineering and involved the development of an empirical equation using regression analysis to predict bedload sediment delivery to the river outlet of 60 Michigan rivers and five sub-watersheds. Watershed sediment delivery is the total amount of sediment generated within a watershed and delivered to the river outlet over a particular timeframe. The total sediment load transported by a river to the river outlet consists of dissolved load, wash load (silts and clays), and bed material load. Bed material load consists of suspended load and bed load. Suspended load is the portion of the bed material load that is lifted by turbulence to travel within the water column above the river bed at elevations greater than a few sediment grain diameters. Prediction of bed load sediment delivery at the river outlet was the focus of this research and is the portion of the bed material load that travels within a few grain diameters of the river bed and moves by rolling, sliding, and saltating along the bed of the river. With respect to the regression analysis, the dependent variable was the measured watershed sediment delivery estimates based on (1) analysis of U.S. Army Corps of Engineers dredging data at federally maintained Harbors and navigation channels at the outlets of 12 Michigan rivers, and (2) watershed sediment delivery estimates based on the results of 137Cs and 210Pb radiometric dating of sediment cores collected from five Michigan reservoirs. Eighteen characteristics of the fluvial system and watershed were evaluated using step-wise regression analysis and include: watershed area; the mean annual flow and/or recurrence interval flows of the river draining the watershed at the river outlet; characteristics of the watershed such as land use as expressed by the watershed Runoff Curve Number and the average and maximum elevation of the watershed relative to the receiving water elevation; and the percentage of the watershed covered in depositional areas such as natural surface water bodies, aquatic and upland wetlands, and manmade dams and reservoirs. Eighty-seven regressions were completed using both non-transformed and natural log transformed dependent and independent variables. Based on the natural log normal regression analyses of dependent and independent variables, Regression 3-36 provided the best balance of significance (0.014), R2 (0.538), and relative low P-values for the following three predictor variables: • 1.5 year recurrence interval flow (P-value: 0.002), • percent of watershed covered in upland and aquatic wetlands (P-value: 0.149), • percent of the watershed covered in manmade reservoirs (P-value: 0.387). Review of the predicted watershed sediment delivery estimates using Regression 3-36 in comparison to the estimated watershed sediment delivery estimates based on USACE dredging data and radiometric dating are within +/- 70% for 13 of the 17 watersheds. The largest differences (based on total metric tonnes) between the predicted sediment delivery using regression 3-36 and the watershed sediment delivery estimates based on either USACE dredging data or radiometric dating were noted for the Saginaw River, Grand River, St. Joseph River and the Menominee River.