Multi-mechanism floods (MMFs) are flood events caused by the simultaneous occurrence of multiple flood mechanisms such as storm surge, precipitation, waves, and tides. The term compound floods, which is a broader term frequently used in the current literature, includes MMFs as a subset. MMF events can have more severe impacts on communities and the built environment than single-mechanism floods. Therefore, a realistic probabilistic assessment of the frequency and severity of flood hazards requires the inclusion of the hazard contribution of MMFs. This dissertation addresses four objectives related to the probabilistic evaluation of MMFs. First, this dissertation develops a lexicon and framework for discussing a broad range of MMFs and then defines the gaps and shortcomings of the current literature. Second, this dissertation develops a Bayesian approach (BA) for performing a probabilistic assessment of a specific type of MMF hazard, namely tropical cyclone-induced increase in river discharge arising from multiple flood mechanisms. The Bayesian model is built using a Bayesian Network (BN). Five computationally justifiable predictive "placeholder" models are developed in this approach to estimate conditional probability tables in the BN. Third, the performance of the BN is assessed for "reasonableness" using three historical storms that affected the study area. Fourth, the capability of the BN for information updating is demonstrated by setting information related to historical observations as evidence in the developed BN and conducting forward and backward inferences. Finally, this study concludes with a summary and synthesis of the gaps and weaknesses of current literature and practices in addressing compound flood hazards. This study further highlights the capabilities and challenges of the developed Bayesian approach and outlines proposed next steps to address these challenges.