Integration of Physically-Based and Data-Driven Modeling Approaches for Compound Coastal Flood Hazard Assessment Under Uncertainties

Thumbnail Image
Journal Title
Journal ISSN
Volume Title
University of Alabama Libraries

Flood hazard assessment is an essential component of risk and disaster management that helps identify areas exposed to flooding as well as support decision making, and emergency response. Floods can result from isolated, concurrent, or successive drivers of (non-) extreme origin (e.g., fluvial, pluvial, and oceanic) and so put society and the environment at constant risk. Specifically, a combination of either concurrent or successive flood drivers with potential impacts larger than those from isolated drivers is defined as compound flooding (CF). Contemporary studies in compound flood hazard assessment (CFHA) and modeling have focused on simulating inundation extent, water depth, and velocities at local or regional scale. However, those studies often neglect inherent uncertainties associated with forcing data, observations, model parameters, and model structure. A comprehensive analysis of these uncertainty sources is thus imperative, but it requires advanced statistical techniques such as data assimilation (DA) to adequately account for error propagation in compound flood modeling. Chapters 1 to 4 present previous peer-review studies oriented towards a better characterization of uncertainty in CFHA. Those studies include the following research topics: (i) analysis of wetland elevation error and correction of coastal digital elevation models, (ii) compound effects of wetland elevation error and uncertainty from flood drivers, (iii) effects of model selection and model structure error on total water level prediction, and (iv) long-term wetland dynamics associated with urbanization, sea level rise, and hurricane impacts. Chapter 5 presents a cost-effective approach based on deep learning (DL) and data fusion (DF) techniques that enables efficient estimation of exposure to compound coastal flooding at regional scale. Chapter 6 presents a DA scheme based on the Ensemble Kalman Filter (EnKF) technique and hydrodynamic modeling to improve water level (WL) predictions and CFHA in coastal to inland transition zones where pluvial, fluvial, and coastal processes interact. The last section of this dissertation summarizes the main findings of these studies and discusses future research areas that are worth exploring in the context of CFHA.

Electronic Thesis or Dissertation
Compound flooding, Data assimilation, Flood hazard assessment, Hurricanes, Machine learning, Uncertainties