In order to fully utilize the energy resources of different offshore islands and improve the power reliability, establishing island microgrid (IMG) clusters has become a promising way for marine energy systems. However, the relatively harsh and variable weather condition poses challenges to the planning of IMGs. This paper proposes a three-stage distributionally robust optimization (DRO) planning model to configure IMG clusters considering marine environmental uncertainties. Firstly, an energy management framework for IMG clusters is established, which includes diesel generator, renewable energy generation, and fixed battery energy storage as the island energy supply components, and utilizes the submarine cables (SCs) as well as mobile energy storage vessels (MESVs) as the energy interconnections among different islands. Then, the ambiguity set for marine environmental uncertainties, which include the wind speed, solar irradiance, ocean current, and wave, is constructed based on Wasserstein distance. On this basis, the proposed planning model is established to configure island energy supply and link components. Specifically, the first stage determines the configuration of SCs and the capacity of energy storage systems and MESVs. The second stage determines the operational decisions and the third stage solves the real-time power dispatch of IMGs. Finally, the two-layer column constraint generation algorithm is utilized to solve the proposed planning model. A realistic island cluster in the South China Sea is selected for case study. The results show that the proposed planning model can greatly enhance the power reliability and reduce 30.8% operational costs of the IMG. In a real typhoon scenario, the expected energy not supplied (EENS) can be reduced by 59.3% through the adoption of the proposed planning model.