Coastal cities like Hong Kong face escalating threats from climate intensified extreme weather, as evidenced by Super Typhoons Hato (2017), Mangkhut (2018), and Saola (2023), which caused billions in damages and widespread flooding. These events underscore the urgent need for an integrated prediction platform capable of modeling complex climate-coast-human interactions to enhance resilience. Developing such a digital twin coast system, however, entails significant scientific challenges: capturing multiscale wave-coast dynamics, integrating real time sensing with physics based simulation, and translating high fidelity predictions into intuitive decision support. To address these challenges, we are developing a next generation digital twin for Hong Kong that synergizes adaptive resolution digital terrain modeling, physics enabled multiscale simulation of wave-structure-soil interactions, artificial intelligence, and high performance computing. The coastal digital twins help to move beyond isolated simulation toward an actionable, multiscale decision support system that unifies physics, AI, and immersive visualization, providing a scalable blueprint for enhancing the safety and resilience of coastal megacities under a changing climate. A key recent advance is a physical-visual dual fidelity framework that couples an AI-based Earth system model (Aurora) [1] for tropical cyclone and wave forecasting with high-resolution Smoothed Particle Hydrodynamics (SPH) [2] simulations of nonlinear wave overtopping. To bridge the gap between numerical output and human comprehension, we integrate 3D Gaussian Splatting for photorealistic coastal rendering and fluid visualization, creating an intuitive virtual replica. Validated against Super Typhoon Mangkhut's impact on Tseung Kwan O, the twin accurately reproduced documented overtopping events and demonstrated practical utility by optimizing temporary flood barrier placement to significantly reduce inundation risk.