Predicting Channel Mobility in an Altered River using Remote Sensing and Statistical Models

Located in the southern Great Plains region, the Canadian River is a braided, sand-bed river that has been significantly altered by anthropogenic activities. Historically, the Canadian River maintained a wide, shallow channel that experienced periodic destructive flooding that laterally adjusted the channel planform. After the construction of large upstream hydropower dams in the 1960s, the natural flow and sediment regimes changed, narrowing the channel and disconnecting the river to its floodplain. The reduced overbank flooding enabled vegetation encroachment and disabled the mobility of the channel. However, the extent to which channel mobility is affected differs longitudinally as tributary inputs, vegetation establishment, and other smaller scale processes can have differing prominence between reaches.

In this study, a data-driven, statistical model was developed as a decision-support tool to predict the longitudinal variation in channel mobility. The model uses historical streamflow data, vegetation presence, sinuosity, and slope data to predict whether narrowing or widening will occur at 100-m transects (n = 683) for a 70-km long section of the Canadian River near Norman, Oklahoma. The longitudinal variation in channel mobility was assessed using remotely-sensed, active channel width measurements of cross-sections spaced 100-m apart. A preliminary multinomial logistic model was developed using data from 2010 to 2017 and demonstrated high accuracy (90%) to categorize the cross-section narrowing or widening. Current work includes expanding the temporal extent of the data from 2003 to 2021 to include higher magnitude flooding periods and increase the predictive range of the statistical model. Impacts of this study include the development of a decision-support tool for river restoration and identifying future at-risk areas for decreasing channel mobility.