Electronic Thesis and Dissertation Repository

Thesis Format

Integrated Article

Degree

Doctor of Philosophy

Program

Civil and Environmental Engineering

Supervisor

Kopp, Gregory A.

2nd Supervisor

Sills, David L.

Co-Supervisor

Abstract

Wind loading is a critical component of design for engineered structures. In particular, the design of spatial structures like electricity transmission line systems and long-span bridges is not only dependent on the statistics of design wind speeds, but also the nature of loading events. The locality of loading events can result in special differential loading cases across multiple spans of the structure that, in many cases, can govern the design. While the statistics of design wind speeds are widely developed and methodological, the locality (scale) of high-intensity wind events has yet to be explored in a generalized manner to be linked to design wind speeds and incorporated into the design of spatial structures.

The main challenge with determining the scale of high-intensity wind events is instrumentation. Design wind speeds are historically estimated from anemometer records. Anemometer networks have sufficient temporal resolution to capture peak event velocities but lack the spatial coverage to estimate the scale of high-intensity wind events.

Accordingly, the current study aims at estimating the scale of high-intensity wind events by utilizing data from weather radars. Archived radar data have the potential of estimating the scale of wind events, but such task comes with challenges related to temporal and spatial resolution of radar archives, as well as numerical issues related to the retrieval process. The study first explores the retrieval process and how it compares to design wind speeds estimated using anemometer records. After developing a relative degree of confidence in the ability of radar retrievals to estimate design wind speeds, the study proceeds to propose a Machine Learning technique that utilizes radar data to estimate the scale of high-intensity wind events based on 60-min anemometer records. The proposed technique is then applied on contiguous United States and design wind speed statistics are compared for the cases of compiled datasets, and datasets segregated based on scales of events. The results show that for the Northeast coast of the US, or for the case of return periods higher than 100 years anywhere else in contiguous US, events of the smallest classified scale were found to have higher design wind speeds than the compiled dataset. Therefore, smaller scale events need to be considered separately when dealing with spatial structures under the circumstances of high return periods, or Northeast coast of US.

Summary for Lay Audience

Understanding how wind affects structures is crucial for designing things like power lines and long bridges. It's not just about knowing how strong the wind can get on average; it's also about understanding how different sizes wind events can impact structures.

Currently, we estimate wind speeds using data from anemometers, which measure wind at specific points. However, anemometers can tell us about wind speed but not how widespread intense wind events are. To get a clearer picture of this, we’re using weather radar data, which help us estimate the size of these high-intensity wind events.

This study uses radar data to get a more accurate idea of the size of wind events. It then trains a Machine Learning technique to find the scale of wind events using anemometer records. We then compare wind speed statistics across different regions in the US, focusing on how different sizes of wind events affect the estimated design speeds.

The results reveal that for regions like the Northeast coast or for rare, high-intensity wind events with return periods of over 100 years, smaller-scale wind events tend to have higher design wind speeds compared to the overall data. This means that smaller-scale, intense wind events should be considered separately when designing structures in these areas.

Available for download on Thursday, January 01, 2026

Share

COinS