Urban drainage systems (UDS) are essential for managing stormwater runoff, but limited sensing infrastructure often hinders timely flood detection and real-time response. This study proposes a multi-scale optimization framework to enhance flood monitoring in urban areas under sensor deployment constraints. The framework integrates 1D–2D flood simulations using EPA-SWMM and the KICT-UF model to simulate both subsurface and surface flow dynamics under various extreme rainfall scenarios. To quantify the information value of each monitoring location, information entropy theory is applied to each node in the 1D system and each grid cell in the 2D domain. Using these entropy values, a multi-objective optimization is performed with the NSGA-II algorithm to select sensor placements that balance high data value and spatial coverage. Additionally, a matrix completion-based approach is used to evaluate the accuracy of reconstructing unmeasured water levels from the selected sensor locations. The framework is applied to a real-world testbed in City S, South Korea, covering a 2.06 km² area with 547 drainage nodes. Results show that even with limited sensors, high-resolution flood monitoring can be achieved when guided by spatial information theory. This framework offers a scalable, data-driven solution for sensor network design in flood-resilient urban infrastructure planning.