Degree

Doctor of Philosophy (PhD)

Department

Civil and Environmental Engineering

Document Type

Dissertation

Abstract

Natural hazards generate tremendous amounts of debris that negatively impact communities and overwhelm waste management infrastructure. The challenge in disaster debris planning and management stems from inconsistent and scarce post-disaster debris quantity data due to the chaotic nature of recovery operations. This leads to a limited understanding of key factors influencing disaster debris generation across hazards and regions, hindering the accuracy and efficiency of modeling. With the increasing availability of comprehensive post-disaster waste datasets, the capacity to understand debris generation is expanding. This dissertation aims to fill knowledge gaps on disaster debris generation as follows: (1) investigate existing technologies for disaster debris field data collection, (2) evaluate the state-of-practice for debris prediction, (3) identify debris drivers using machine learning techniques, and (4) explore relationships between hazard type, intensity, and debris quantities in multi-hazard events. To address objective (1), multiple remote sensing technologies available for quantifying debris are compared using post-disaster data collected following Hurricane Ida. This study found that remote sensing tools have diverse applications for debris quantification and the guidance provided can support disaster waste managers. To address objective (2), a post-disaster waste dataset from Hurricane Harvey was used to investigate the amount of debris produced, better understand controlling factors of debris quantities, and compare actual debris tonnages to predictive models. This study found that type of flooding had the highest influence on debris quantities and existing prediction methods can be adapted using post-disaster data. To address objective (3), machine learning algorithms were applied to the aforementioned dataset to better understand the drivers influencing flood debris. This study found that unsupervised algorithms can uncover hidden relationships and that first-floor elevation is a key predictor of flood debris but is not considered in modeling efforts. Lastly, to address objective (4), hazard intensity and type are correlated with debris quantities using post-disaster waste data collected after Hurricane Ian. This study found that compound hazards generate disproportionately high quantities of debris and that models should integrate multi-hazard interactions in regions susceptible to compound hazards. This dissertation demonstrates how disaster debris characterization and quantification can be advanced using data-driven approaches, furthering the state-of-practice.

Date

10-22-2024

Committee Chair

Jafari, Navid H.

Download

Share

COinS