Robert Chance

Date of Award

January 2025

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Geology

First Advisor

Jaakko Putkonen

Second Advisor

Taufique Mahmood

Abstract

Hillslopes are one of the most pervasive landforms on Earth yet the details of the processes that drive their evolution are not completely understood. Excluding boulders, the majority of mass and volume in soil-mantled hillslopes is composed of sand-sized and finer grains. Understanding the transport rates and governing transport mechanisms of these grains is therefore key to advancing our knowledge of hillslope evolution.This research investigates sand-sized sediment transport through field experiments at 20 sites tracking surface sediment motion and 11 sites measuring depth-dependent transport. These empirical results are compared against outputs from a novel grain-based hillslope diffusion simulation (GBHDS), which provides theoretical estimates of sediment flux. Additionally, a physics-informed machine learning surrogate model was trained on GBHDS outputs and embedded within a genetic algorithm to optimize model parameters more efficiently than traditional trial-and-error approaches. Results show that in semiarid climates, sediment transport predominantly occurs intermittently within a thin surface layer composed of medium and fine sands. Sediment grains of this size are overrepresented in the grain size distribution of soil mantled hillslopes and are likely transported by radial dispersion processes such as bioturbation or rain splash during dry years and sheetwash during wet years. This study also reveals how sediment transport rates depend on grain size, hillslope angle, and transport process. Lastly, this study demonstrated that embedding a physics-informed surrogate machine learning model within a genetic algorithm yields rapid, effective parameter optimization for complex geophysical models. These findings can help better predict how hillslopes evolve over time and refine models of sediment movement toward streams and valleys. These insights can refine erosion models used in land management, infrastructure stability assessments, and conservation planning in semi-arid environments.

Download

Share

COinS