Trevor Stull

Graduation Semester and Year




Document Type


Degree Name

Master of Science in Civil Engineering


Civil Engineering

First Advisor

Habib Ahmari


Suspended sediment transport in river basins is important for many water management planning activities to maintain safe drinking water for the community and maintenance of water quality and waterways for the ecosystem. Currently, the traditional way to measure suspended sediment effectively and reliably is by collecting field samples in the river body, which is very time consuming and only provide a point value of suspended sediment within the waterbody at the instant the sample was taken. This thesis focuses on developing models that estimate suspended sediment concentrations for the lower Brazos River using satellite imagery from publicly available data and machine learning methods. The use of optical properties such as satellite imagery and turbidity measurements have been gaining support recently and provide a more continuous record of suspended sediment concentrations and in the case of satellite imagery a spatial relationship once a model is developed. Historical samples of suspended sediment concentrations from the United States Geological Survey and Texas Commission on Environmental Quality and satellite imagery from Landsat Missions and Sentinel Mission 2 were utilized to develop models to estimate suspended sediment concentrations for the lower Brazos River. Models used in this thesis to accomplish this goal include support vector machines, artificial neural networks, extreme learning machines, and exponential relationships. In addition, flow and depth measurements from the United States Geological Survey were used to develop rating curves to estimate suspended sediment concentrations for the Brazos River as a baseline comparison of the models that used satellite imagery to estimate suspended sediment concentrations. Models were evaluated using the Taylor Diagram analysis on the test data set developed for the Brazos River data. Sixteen of the models using satellite imagery as inputs that were developed for this thesis performed with a coefficient of determination R2 above 0.69 with the three best performing models having an R2 of 0.83 to 0.85. One of the best performing models was then applied estimate suspended sediment concentrations before, during, and after Hurricane Harvey to evaluate Hurricane Harvey’s impact to the sediment dynamics along the lower Brazos River and the model’s ability to achieve this goal. The models that used satellite imagery developed for this thesis were also evaluated on the San Bernard River to test their ability outside of the Brazos River; however, all models achieved an R2 below 0.04. In addition, the importance of input variable to estimate suspended sediment concentrations were evaluated using the principal component analysis which determined that the NIR, red, and green bands were significant to achieve this goal. This was confirmed with the best performing models developed incorporating the Red-Green Ratio as an input.


Machine learning, Remote sensing, Suspended sediment concentration


Civil and Environmental Engineering | Civil Engineering | Engineering


Degree granted by The University of Texas at Arlington

Available for download on Thursday, August 01, 2024