A suite of model simulations is used to investigate the spatiotemporal variability of the Arctic Ocean circulation and the observing systems required to capture it. A comparison with sea level observations shows that all model runs realistically simulate inter-annual sea level varia-bility, but the simulated seasonal sea level variability and underlying changes in the model salinity need to be further improved. At periods <30 days, sea level variability is equivalent barotropic and strongly captured by bottom pressure observations. At the seasonal period, both barotropic and baroclinic processes contribute involving variations in the mass and densi-ty fields. Over the entire Arctic Ocean, steric height variability is dominated by halosteric effects in the upper layer. The salinity changes are related to sea ice processes, river runoff, and redistribution of the freshwater. At decadal timescales, sea level variations in the Canadi-an Basin relate to halosteric effects in the upper and intermediate layers. An adjoint sensitivi-ty analysis reveals that the decadal salinity variability is caused by anticyclonic/cyclonic wind stress, which accumulate/release freshwater in the upper layer and enhance/reduce geostrophic currents in the intermediate layer. The adjoint model simulations identify the importance of moorings and satellite altimetry on monitoring the Arctic salinity and circulation changes: while moorings capture more local salinity changes, the satellite altimetry may capture the influence of freshwater originating from the Bering Strait and the Eurasian Basin. Our study suggests that to capture basin-wide salinity changes, we need to deploy moorings in different positions spreading across the Arctic Ocean.