Abstract

Accurate and timely flood inundation forecasts are crucial for mitigating the devastating impacts of floods, which affect 1.47 billion people globally and cause significant economic losses. While traditional flood forecasting methods, such as hydrologic and hydrodynamic models, have limitations in providing comprehensive and spatially explicit inundation information, recent research has focused on leveraging Earth observation data and machine learning for enhanced flood prediction. This dissertation advances the state-of-the-art in data-driven flood inundation forecasting by exploring the capabilities and scalability of the Forecasting Inundation Extents using Rotated Empirical Orthogonal Function (FIER) framework. This dissertation focuses on investigating different aspects of FIER that contribute to operational use of the tool. The research presents 1) a novel data fusion methodology for optical and Synthetic Aperture Radar (SAR) imagery to generate dense time series of surface water maps that can be used as inputs into FIER, 2) a cloud-native dataset for the National Water Model (NWM) and application programming interface (API) for efficient access to the operational NWM hydrologic model data, and 3) an assessment of FIER's scalability for large geographic domains. Multiple case studies are presented for each topic illustrating the applicability for use with FIER and operational use. The results of this research demonstrate the required components and potential of the FIER framework for operational large-scale flood inundation forecasting. The developed data fusion methodology enhances the temporal resolution and completeness of input data for FIER, while the NWM API improves access to and usability of essential operational hydrologic model data. The investigation of scaling FIER to large geographic domains demonstrates the framework's ability and potential for application over regional to continental scales. These contributions pave the way for developing robust, computationally efficient flood inundation forecasting systems, particularly in data-scarce regions where traditional hydrodynamic models are often infeasible. By providing accurate and timely flood inundation forecasts, this research contributes to ongoing research and application activities related to water resources and disaster risk reduction in the US and internationally, empowering communities and decision-makers with valuable information for decision making and mitigating the devastating impacts of floods.

Degree

PhD

College and Department

Ira A. Fulton College of Engineering; Civil and Environmental Engineering

Rights

https://lib.byu.edu/about/copyright/

Date Submitted

2025-03-18

Document Type

Dissertation

Handle

http://hdl.lib.byu.edu/1877/etd13541

Keywords

flood inundation, forecasting, remote sensing, hydrologic model, data fusion, forecasting inundation extents using REOF, FIER, API, cloud processing

Language

english

Download

Included in

Engineering Commons

Share

COinS