(1) Simulations performed by a global ice-ocean model with high resolution in northern high latitudes in order to improve the reproducibility of the Arctic Ocean in global climate models. Two-way nested grid version of ice-ocean model COCO (Kurogi et al., 2013; Kurogi et al., 2016) is used. The inner model domain covers roughly the area to the north of 61N and is nested into the outer model domain, which covers the whole globe. The horizontal resolution of the inner and outer models are approximately 2.5 km and 0.25 degree, respectively. (2) To make the initial condition, only the outer model is spun up with restoring temperature and salinity below 50 m depth to PHCv3 climatology (Steele et al., 2001). The result is then interpolated to the inner model domain. (3) The atmospheric dataset JRA55-do (Tsujino et al., 2018) is used to drive the nested model. When a simulation is performed under a future climate condition, the data is modified depending on the difference between future projection and present atmospheric condition. (4) Total data size will be 20 TB per case (10 years). Grid Information Dimension Size (slice number of the dimension) row: 7200(inner), 1440(outer) column: 1280(inner), 1280(outer) vertical: 62 Resolution Unit row; approx. 2.5 km (inner), 0.25 deg. (outer) column: approx. 2.5 km (inner), 0.25 deg. (outer) vertical: 10 in the uppermost 50 m, 31 in the uppermost 500 m.