Doctor of Philosophy (PhD)
Geography and Anthropology
Floods represent one of the most severe natural disasters worldwide. It is crucial to accurately assess flood hazards to mitigate the impact and enhance community resilience. This study focuses on three key components of flood hazard management, including (1) flood frequency estimation, (2) inundation delineation, and (3) flood susceptibility assessment. The first project presents a Bayesian Log-Pearson Type III model with Spatial Priors (BLP3-SP) that utilizes spatial regression priors to enhance the accuracy of flood frequency estimation for sites with limited data. The method is compared with two other algorithms with different priors. The results indicate that BLP3-SP performs the best and reduces the uncertainty in flood frequency estimation by almost 50% compared to using only local data. The second project explores the potential and problems of using Sentinel-1 SAR imagery for inundation mapping with the Louisiana 2016 August Flood as a case study. The study finds that flood changes the SAR backscatter intensity in different ways for different land covers, which affects the accuracy of inundation mapping. Four different approaches, including a U-Net deep learning approach, a subpixel abundance-based approach, a HAND-assisted approach, and an InSAR-based approach, were applied to the study. The HAND-assisted approach shows the most accurate results and InSAR-based approach shows potential in distinguishing flood areas in regions with high temporal consistency. The third project aims to develop a flood susceptibility map for the Lake Maurepas Watershed using deep-learning methods. Two novel convolutional neural network algorithms, CNN-1D and CNN-2D, are applied to predict the spatial pattern of flood hazard in the study area, considering 25 environmental factors. The CNN-2D model achieves the highest prediction accuracy and is used to generate the flood susceptibility map, which shows that high and very high flood susceptibility is widely distributed in the southern part of the study area. A scenario analysis shows that the flood susceptibility map can be updated for different hydrological scenarios and used to support flood management and raise public awareness about flood risks.
Tian, Dan, "Mapping, Modeling, and Predicting Extreme Flood Events with Cloud Computing" (2023). LSU Doctoral Dissertations. 6095.
Available for download on Friday, April 03, 2026