The 100-year floodplain serves as the primary communicator of flood risk, but this delineation and its corresponding maps have been shown to be inadequate indicators of flood risk and poor predictors of flood damage, especially in urban areas. To be fair, most 100-year floodplain maps were never intended to convey flood risk: the boundaries of the floodplain were drawn to set insurance rates. They do not and never were designed to convey information related to depth or duration of inundation, water flow, historical damage, or susceptibility to pluvial flooding. In addition, the current flood maps do not inform investments in flood control and other infrastructure investments that positively or negatively influence urban flood risk. In short, decisions based off these maps are working on limited information often resulting in misguided efforts to increase flood adaptation and resilience. Although traditional hydrologic and hydraulic (H&H) models such as HEC-RAS could be used to provide more accurate flood hazard information, they are limited in terms of their computational loads, time to execute, and expense making them infeasible to run over large areas especially in regions with limited H&H data. To address this limitation this study uses a machine learning (ML) algorithm to estimate parcel level flood hazard along the southeastern Texas coast using a long-term record of parcel level historical flood damage. The purpose was to create improved flood hazard maps that not only better captures where flooding may occur, but to also enhance risk communication. The rationale for creating these new flood risk maps is not to replace the existing FEMA regulatory floodplain, but to compliment it in such a way that increases the ability of decision makers and residents to make decisions that increase their flood resilience.