Floods are among the most devastating natural hazards across the globe, and it isexpected to be escalated in the future mostly due to the warming climate. Depside tremendousnational governments' protection efforts, they still result in the loss of lives and properties.Understanding flood dynamics and developing fast and reliable flood forecasting systems areessential for mitigating the associated risks and implementing proactive risk managementstrategies. This dissertation is an attempt to characterize flood hazard and enhance flood riskassessment by leveraging enhanced statistical approaches and recent physical models. The firstfour chapters present previous peer-reviewed studies and are as follows:The first chapter is on assessing flash flood characteristics including frequency, duration,and intensity in addition to their associated property damages. The second chapter presents asystematic framework that considers a variety of features explaining different components ofrisk and examines multiple machine learning methods to predict flash flood damage. In chapterthree, we assess the sensitivity of the HEC-RAS 2D to its configuration factors and parameters.In chapter four, we are developing a new tropical cyclone scaling system that uses Copulas forcategorizing Tropical Cyclones (TCs) based on the likelihood of a given set of severity forrainfall, surge, and wind speed. Finally, in chapter five, we present a systematic framework thatuses Deep Learning (DL) algorithms, and hydrodynamic models to generate probabilistic floodwater levels at different locations along coastal rivers.