Floods are one of the most frequent types of destructive natural disaster that causes serious safety risks and economic losses. Flood simulation has been an effective way to model the behavior and dynamics of flood and support the analysis like flood disaster assessment. Currently, most 3D flood simulations often take heterogenous data (e.g. flood measurements, terrain data) as input for calculation, but lack a unified representation of these data, and the environment objects are normally modeled in a simplified manner with only geometric information while neglecting their semantics. To address this gap, this study developed a CityGML Flood Application Domain Extension (Flood ADE) for representing and storing flood information along with semantic information of 3D urban objects. We presented the methodology for developing the CityGML Flood ADE and described its thematic modules. The effectiveness of the developed Flood ADE is demonstrated by using 3D building data and flood data in the Lundamo area of the Sokna River, Trondheim, Norway. This work will not only support flood prevention decisions, analysis, and post-disaster assessments, but also promote cross-departmental collaborations.