Virtual Reality (VR) concepts have been widely adapted to various disciplines and industries. The Augmented Reality Sandbox (AR-Sandbox) is an intuitive terrain visualization facility that realizes digital content projection onto the sand or clay surface and interactive operation. In this study, we developed an assistance tool to enhance AR-Sandbox's utility through the integration of the multi-objective optimization algorithm, enabling users to customize objectives including maximizing flow path length, minimizing maximum runoff velocity, and minimizing earthworks. Based on the non-dominated sorting genetic algorithm-Ⅱ (NSGA-Ⅱ) and multi-scenario plan references, this method aims to provide a more quantitative and efficient way than trial and error-based terrain-design processes.