UC Berkeley

Hydrological modeling is essential for a variety of water management challenges, from natural stream restoration to urban flood control. The widespread availability of high-frequency and/or high-resolution remotely sensed data provides an opportunity to improve hydrological modeling. While incorporating large amounts of data is computationally intensive, tools that distill information from remote sensing into useful empirical relationships, simple tools, or summary parameters allow for greater uptake in situations when resources are limited or practitioners prefer a simpler or quicker method.

This dissertation approaches surface water modeling using simple process-based analytical and empirical models and remotely sensed data. High-resolution imagery is used to examine empirical relationships between in-channel wood length and channel width and characterize the differences in the role of large wood in spring-fed and runoff-fed streams. A model of overland flow on convergent/divergent surfaces is presented and verified via numerical and laboratory techniques. This method is then applied to examine the limits of applicability of the Rational Method, one of the most widely-recommended tools for predicting peak flows, in terms of the fundamental mathematical assumptions. This work identifies landscape properties that cause the Rational Method to perform poorly under idealized conditions and presents a framework for applying the Rational Method as an optimization.