Electronic Theses and Dissertations Archive

Date

2026

Document Type

Dissertation

Degree Name

Doctor of Philosophy

Department

Civil Engineering

Committee Chair

Claudio Meier

Committee Member

Daniel Wright

Committee Member

Dorian Burnette

Committee Member

Farhad Jazaei

Abstract

Understanding short-duration (sub-daily) extreme precipitation (SDEP) is crucial for urban hydrology. These events, often driven by highly localized convective storms, exhibiting high spatial heterogeneity and temporal intermittency, are major drivers of flash floods and urban pluvial flooding. Owing to these characteristics, SDEP events are often inadequately sampled by sparse, low temporal-resolution rain-gauge networks, such as those in the United States. This limitation introduces two major challenges: (i) structural uncertainty in depth-duration-frequency values (DDFs) and (ii) low signal-to-noise (S/N) ratios in at-a-station trend analyses, both of which remain insufficiently explored. Using Monte Carlo simulations, Bayesian inference, partial duration series (PDSs), regional frequency analyses (RFAs), trend analyses, and spatial statistics, this dissertation addresses these challenges, delivering several methodological advances. To address the first challenge, we used hierarchical bootstrapping and Bayesian frameworks, finding that low-density rain-gauge networks introduce positive biases in DDFs at low recurrence intervals (≤ 10 years), negative biases at high recurrence intervals (≥ 50 years), and inflated uncertainty across all recurrence intervals. For recurrence intervals ≥ 5 years, uncertainty inflation scales linearly with network thinning at a rate of ~ 9 - 20% per station lost per 1000 km2. To address the second challenge, we advance a neighborhood-based trend detection approach. Experiments using synthetic and observational datasets demonstrate improved S/N ratios and greater spatial coherence in detected trends, compared to at-a-station analyses. Methodological contributions advance the statistical foundations of RFA. We show that practitioners need not worry about the choice of minimum interevent time (MIT) when estimating short-duration DDFs, as the RFA - L-moment framework yields robust DDFs across MIT values ranging from 1 to 8 hours. This robustness arises from the preservation of tail-event identities: the magnitudes and ranks of the largest exceedances in the PDSs remain invariant to MIT choice. A novel dual-criterion filtering algorithm is proposed to extract spatially independent regional PDSs, reducing dependence-induced bias in DDFs. In addition, a new metric is introduced to characterize the influence of spatial dependence on DDFs. Collectively, this dissertation advances the state of the art by quantifying overlooked sources of uncertainty and developing methods that support more reliable hydrologic design.

Comments

Data is provided by the student.

Library Comment

Dissertation or thesis originally submitted to ProQuest/Clarivate.

Notes

Embargoed until 01-28-2027

Download

Available for download on Thursday, January 28, 2027

Share

COinS